THE AXON GUIDE

A Guide to Electrophysiology &

Biophysics Laboratory Techniques

The Axon Guide — 2500-0102 Rev. C

 Molecular Devices (now part of MDS Analytical Technologies)
The Axon Guide
A Guide to Electrophysiology & Biophysics Laboratory Techniques

Copyright
Third Edition © Copyright 2008, MDS Analytical Technologies. All rights reserved. No part of this
publication may be reproduced, transmitted, transcribed, stored in a retrieval system, or translated
into any language or computer language, in any form or by any means, electronic, mechanical,
magnetic, optical, chemical, manual, or otherwise, without the prior written permission of MDS
Analytical Technologies, 1311 Orleans Drive, Sunnyvale, California, 94089, United States of
America.

Trademarks
CYBERAMP, DIGIDATA, GENECLAMP and PCLAMP are registered trademarks and
AXOPATCH, AXOCLAMP and MULTICLAMP are trademarks of MDS Analytical
Technologies.
These trademarks are not to be used in any type of promotion or advertising without written
permission from MDS Analytical Technologies.
All other trademarks or registered trademarks are the property of their respective owners.

Disclaimer
MDS Analytical Technologies reserves the right to change its products and services at any time to
incorporate technological developments. This publication is subject to change without notice.
Although this manual has been prepared with every precaution to ensure accuracy, MDS Analytical
Technologies assumes no liability for any errors or omissions, nor for any damages resulting from
the application or use of this information.

Questions?
Phone: +1-800-635-5577
+1-408-747-1700
Fax: +1-408-747-3603
Web: www.moleculardevices.com/support

For additional offices, please see the contact information on the back cover of this publication.

2 The Axon Guide — 2500-0102 Rev. C

 Contents
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .ix
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .xiii

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii
Editorial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .xix
Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .xxi
1. Bioelectricity
Electrical Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Electrical Currents. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Resistors and Conductors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Ohm’s Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
The Voltage Divider . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Perfect and Real Electrical Instruments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Ions in Solutions and Electrodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Capacitors and their Electrical Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Currents Through Capacitors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Current Clamp and Voltage Clamp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Glass Microelectrodes and Tight Seals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2. The Laboratory Setup

The In Vitro Extracellular Recording Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
The Single-Channel Patch Clamping Setup. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Vibration Isolation Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Electrical Isolation Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Radiative Electrical Pickup. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Magnetically-Induced Pickup. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Ground-Loop Noise. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Equipment Placement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
List of Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

The Axon Guide — 2500-0102 Rev. C i

 3. Instrumentation for Measuring Bioeletric Signals from Cells
Extracellular Recording . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Single-Cell Recording . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Multiple-Cell Recording . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Intracellular Recording Current Clamp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Voltage Follower . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Bridge Balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Junction Potentials. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Track . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Current Monitor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Headstage Current Gain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Capacitance Compensation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Leakage Current. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Headstages for Ion-Sensitive Microelectrodes. . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Bath Error Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Cell Penetration: Mechanical Vibration, Buzz and Clear . . . . . . . . . . . . . . . . . . . 45
Command Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Intracellular Recording—Voltage Clamp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
The Ideal Voltage Clamp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Real Voltage Clamps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Large Cells—Two-Electrode Voltage Clamp . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Small Cells—Discontinuous Single-Electrode Voltage Clamp . . . . . . . . . . . . . . . 52
Discontinuous Current Clamp (DCC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Continuous Single-Electrode Voltage Clamp (Whole-cell Patch Clamp) . . . . . . . 58
Series-Resistance Compensation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Pipette-Capacitance Compensation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Whole-Cell Capacitance Compensation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Rupturing the Patch. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
Which One Should You Use: dSEVC or cSEVC? . . . . . . . . . . . . . . . . . . . . . . . . 69
Space Clamp Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Single-Channel Patch Clamp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
The Importance of a Good Seal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
Resistor Feedback Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
Capacitor Feedback Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
Special Considerations for Bilayer Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . 77
How Fast is “Fast”? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Measurement of Changes in Membrane Capacitance . . . . . . . . . . . . . . . . . . . . . . 78
Seal and Pipette Resistance Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Micropipette Holders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Current Conventions and Voltage Conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

ii The Axon Guide — 2500-0102 Rev. C

 Definitions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
Whole-Cell Voltage and Current Clamp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Patch Clamp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Patch Clamp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Two-Electrode Voltage Clamp. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Single-Electrode Voltage Clamp. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Space-Clamp Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Other . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

4. Microelectrodes and Micropipettes

Electrodes, Microelectrodes, Pipettes, Micropipettes and Pipette Solutions . . . . . . . . 85
Fabrication of Patch Pipettes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Pipette Properties for Single-Channel vs. Whole-Cell Recording . . . . . . . . . . . . . . . . 89
Types of Glasses and their Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Thermal Properties of Glasses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
Noise Properties of Glasses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Leachable Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

5. Advanced Methods in Electrophysiology

Recording from Xenopus Oocytes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
What is a Xenopus Oocyte? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
Two-Electrode Voltage Clamping of Oocytes . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Patch Clamping Xenopus Oocytes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
Patch-Clamp Recording in Brain Slices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
The Cleaning Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
The Blind Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
Advantages and Disadvantages of the Two Methods of Patch Clamping
Brain Slices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
Macropatch and Loose-Patch Recording . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
Gigaseal-Macropatch Voltage Clamp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Loose-Patch Voltage Clamp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
The Giant Excised Membrane Patch Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Pre-Treatment of Muscle Cells. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
Pipette Fabrication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

The Axon Guide — 2500-0102 Rev. C iii

 Seal Formation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
Recording from Perforated Patches and Perforated Vesicles . . . . . . . . . . . . . . . . . . . 121
Properties of Amphotericin B and Nystatin . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Stock Solutions and Pipette Filling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Properties of Antibiotic Partitioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
The Advantages of the Perforated-Patch Technique . . . . . . . . . . . . . . . . . . . . . . 123
The Limitations of the Perforated-Patch Technique. . . . . . . . . . . . . . . . . . . . . . 123
Suggested Ways to Minimize the Access Resistance . . . . . . . . . . . . . . . . . . . . . . 124
Other Uses for Perforated Patches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
Enhanced Planar Bilayer Techniques for Single-Channel Recording . . . . . . . . . . . . 129
Solving the Problems of High Resolution and Voltage Steps . . . . . . . . . . . . . . . 130
Assembling a Bilayer Setup for High-Resolution Recordings . . . . . . . . . . . . . . . 136
Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

6. Signal Conditioning and Signal conditioners

Why Should Signals Be Filtered? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
Fundamentals of Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
-3 dB Frequency. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
Type: High-pass, Low-pass, Band-pass or Band-reject (notch) . . . . . . . . . . . . . 144
Order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
Implementation: Active, Passive or Digital. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
Filter Function. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
Filter Terminology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
- 3 dB Frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
Attenuation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Pass Band. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Stop Band . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Phase Shift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Overshoot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Octave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Decade. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Decibels (dB) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
10-90% Rise Time. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Filtering for Time-Domain Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Filtering for Frequency-Domain Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Sampling Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
Filtering Patch-Clamp Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

iv The Axon Guide — 2500-0102 Rev. C

 Digital Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
Preparing Signals for A/D Conversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
Where to Amplify . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
Pre-Filter vs. Post-Filter Gain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
Offset Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
AC Coupling and Autozeroing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
Time Constant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
Saturation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
Overload Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
Averaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Line-Frequency Pick-Up (Hum) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Peak-to-Peak and rms Noise Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Blanking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
Audio Monitor—Friend or Foe? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
Electrode Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
Common-Mode Rejection Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

7. Transducers
Temperature Transducers for Physiological Temperature Measurement . . . . . . . . . 167
Thermistors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
IC Temperature Transducers that Produce an Output Current Proportional
to Absolute Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
IC Temperature Transducers that Produce an Output Voltage Proportional
to Absolute Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
Temperature Transducers for Extended Temperature Ranges . . . . . . . . . . . . . . . . . . 169
Thermocouples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
Resistance Temperature Detectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
Electrode Resistance and Cabling Affect Noise and Bandwidth . . . . . . . . . . . . . . . 171
High Electrode Impedance Can Produce Signal Attenuation . . . . . . . . . . . . . . . . . 171
Unmatched Electrode Impedances Increase Background Noise and Crosstalk . . . . . 172
High Electrode Impedance Contributes to the Thermal Noise of the System . . . . . 173
Cable Capacitance Filters Out the High-Frequency Component of the Signal. . . . . 174
EMG, EEG, EKG and Neural Recording . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
EMG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
EKG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
EEG. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
Nerve Cuffs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175

The Axon Guide — 2500-0102 Rev. C v

 Metal Microelectrodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
Bridge Design for Pressure and Force Measurements . . . . . . . . . . . . . . . . . . . . . . . 176
Pressure Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
Force Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Acceleration Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Length Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Implantable Length Gauges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Linear Potentiometers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
Linear Variable Differential Transformers (LVDT) . . . . . . . . . . . . . . . . . . . . . . 178
Self-Heating Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
Isolation Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
Insulation Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179

8. Laboratory Computer Issues and Considerations

Select the Software First . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
How Much Computer Do You Need? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
The Machine Spectrum: Capability vs. Price . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Hard Drive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Graphics Card . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Peripherals and Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Data Backup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Recommended Computer Configurations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183

9. Acquisition Hardware
Fundamentals of Data Conversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
Quantization Error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
Choosing the Sampling Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Converter Buzzwords . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
Gain Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
Linearity Error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
Differential Nonlinearity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
Least Significant Bit (LSB) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
Monotonicity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
Deglitched DAC Outputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
Timers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
Digital I/O . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194

vi The Axon Guide — 2500-0102 Rev. C

 Optical Isolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
Operating Under Multi-Tasking Operating Systems . . . . . . . . . . . . . . . . . . . . . . . . 195
Software Support. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195

10. Data Analysis

Choosing Appropriate Acquisition Parameters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Gain and Offset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Sampling Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
Filtering at Analysis Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
Integrals and Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
Single-Channel Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
Goals and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
Sampling at Acquisition Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
Filtering at Acquisition Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
Analysis-Time Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
Generating the Events List . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
Setting the Threshold for a Transition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
Baseline Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
Missed Events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
False Events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
Multiple Channels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
Analyzing the Events List . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
Histograms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
Histogram Abscissa Scaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
Histogram Ordinate Scaling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
Errors Resulting from Histogramming Events Data . . . . . . . . . . . . . . . . . . . . . . 205
Amplitude Histogram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
Fitting to Histograms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
Fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
Reasons for Fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
Statistical Aspects of Fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
Methods of Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212

11. Noise in Electrophysiological Measurements

Thermal Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
Shot Noise. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
Dielectric Noise. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
Excess Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220

The Axon Guide — 2500-0102 Rev. C vii

 Amplifier Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Electrode Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
Electrode Noise in Single-Channel Patch Voltage Clamping . . . . . . . . . . . . . . . 223
Seal Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
Noise in Whole-Cell Voltage Clamping. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
External Noise Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
Digitization Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
Aliasing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Filtering. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
Summary of Patch-Clamp Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
Headstage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
Holder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
Electrode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
Seal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
Limits of Patch-Clamp Noise Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246

A. Appendix: Guide to Interpreting Specifications

General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Thermal Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
RMS Versus Peak-to-Peak Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
Bandwidth—Time Constant—Rise Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
Filters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
Microelectrode Amplifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
Voltage-Clamp Noise. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
10–90% Rise Time (t10–90) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
1% Settling Time (t1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
Input Capacitance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
Input Leakage Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
Patch-Clamp Amplifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
Bandwidth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250

Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253

viii The Axon Guide — 2500-0102 Rev. C

 List of Figures
Figure 1.1: Conservation of current. Current is conserved at a branch point.. . . . . . . . 2
Figure 1.2: A typical electrical circuit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Figure 1.3: Summation of conductance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Figure 1.4: Equivalent circuit for a single-membrane channel.. . . . . . . . . . . . . . . . . . . 4
Figure 1.5: Ohm’s law. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Figure 1.6: IR drop.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Figure 1.7: A voltage divider. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Figure 1.8: Representative voltmeter with infinite resistance. . . . . . . . . . . . . . . . . . . . . 8
Figure 1.9: The silver/silver chloride electrode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Figure 1.10: The platinum electrode.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Figure 1.11: Capacitance. A charge Q is stored in a capacitor of value C held
at a potential ΔV.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Figure 1.12: Capacitors in parallel add their values. . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Figure 1.13: Membrane behavior compared with an electrical current. . . . . . . . . . . . 13
Figure 1.14: RC parallel circuit response.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Figure 1.15: Typical voltage-clamp experiment.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Figure 1.16: Intracellular electrode measurement. . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Figure 1.17: Good and bad seals.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Figure 3.1: An ideal micropipette buffer amplifier. . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Figure 3.2: A high-quality current source. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Figure 3.3: The “Bridge Balance” technique. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Figure 3.4: Illustration of bridge balancing while micropipette is extracellular. . . . . . 32
Figure 3.5: Series current measurement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Figure 3.6: Virtual-ground current measurement. . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Figure 3.7: Bootstrapped power supplies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Figure 3.8: Capacitance neutralization circuit.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Figure 3.9: Two-electrode virtual-ground circuit.. . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Figure 3.10: The ideal voltage clamp. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Figure 3.11: Conventional two-electrode voltage clamp. . . . . . . . . . . . . . . . . . . . . . . 48
Figure 3.12: Block diagram and timing waveforms. . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Figure 3.13: Biphasic voltage response of a high-resistance micropipette.. . . . . . . . . . 55
Figure 3.14: Voltage and temporal errors caused by the presence of Ra. . . . . . . . . . . . 59
Figure 3.15: Series resistance correction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

The Axon Guide — 2500-0102 Rev. C ix

 Figure 3.16: Prediction implemented empirically by the computer. . . . . . . . . . . . . . . 62
Figure 3.17: Implementation of prediction based on the knowledge of cell
parameters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Figure 3.18: Pipette capacitance compensation circuit . . . . . . . . . . . . . . . . . . . . . . . . 65
Figure 3.19: Whole-cell capacitance compensation circuit. . . . . . . . . . . . . . . . . . . . . 67
Figure 3.20: Using the injection capacitor to charge the membrane capacitance. . . . . 68
Figure 3.21: Resistive headstage. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
Figure 3.22: Capacitive feedback headstage. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
Figure 3.23: Signal handling during resets in the capacitor-feedback headstage.. . . . . 76
Figure 3.24: Typical current noise in bilayer experiments. . . . . . . . . . . . . . . . . . . . . . 78
Figure 4.1: Noise properties of glasses.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Figure 5.1: Stage V or VI oocytes as found in an ovarian lobe.. . . . . . . . . . . . . . . . . . 98
Figure 5.2: The Cleaning technique. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Figure 5.3: The Blind technique.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Figure 5.4: Combining whole-cell voltage clamping with loose-patch current
recording. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Figure 5.5: Pipette fabrication for giant excised patch method. . . . . . . . . . . . . . . . . 119
Figure 5.6: Forming a cell-attached perforated patch and a perforated vesicle. . . . . . 127
Figure 5.7: Aperture formation in a plastic cup (not drawn to scale).. . . . . . . . . . . . 137
Figure 5.8: Voltage-step activation of a K-channel from squid giant axon axoplasm. 140
Figure 6.1: Frequency response of 4th-order Bessel and Butterworth filters. . . . . . . 145
Figure 6.2: Illustration of filter terminology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Figure 6.3: Difference between a 4th- and 8th-order transfer function. . . . . . . . . . . 149
Figure 6.4: Step response comparison between bessel and butterworth filters. . . . . . 150
Figure 6.5: Use of a notch filter: inappropriately and appropriately.. . . . . . . . . . . . . 151
Figure 6.6: Distortion of signal caused by high amplification prior to filtering. . . . . 157
Figure 6.7: Distortion of signal caused by AC-coupling at high frequencies. . . . . . . 158
Figure 6.8: Comparison of autozeroing to AC-coupling. . . . . . . . . . . . . . . . . . . . . . 159
Figure 7.1: Electrode resistance and cabling can degrade a signal. . . . . . . . . . . . . . . 171
Figure 7.2: High electrode resistance can attenuate a signal and increase crosstalk. . 172
Figure 7.3: Wheatstone Bridge circuit with amplifier. . . . . . . . . . . . . . . . . . . . . . . . 176
Figure 9.1: Analog-to-digital conversion.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
Figure 10.1: Histogram scaling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
Figure 10.2: Sampling error. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
Figure 10.3: The likelihood function. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208

x The Axon Guide — 2500-0102 Rev. C

Figure 10.4: Confidence limits. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Figure 11.1: Noise equivalent circuits of a resistor.. . . . . . . . . . . . . . . . . . . . . . . . . . 216
Figure 11.2: Operational amplifier noise model. . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Figure 11.3: Noise model of a simplifier current-to-voltage converter. . . . . . . . . . . . 221
Figure 11.4: Noise in whole-cell voltage clamping. . . . . . . . . . . . . . . . . . . . . . . . . . 235
Figure 11.5: Folding of the frequency axis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
Figure 11.6: Squared transfer function of a Gaussian filter. . . . . . . . . . . . . . . . . . . . 243

The Axon Guide — 2500-0102 Rev. C xi

xii The Axon Guide — 2500-0102 Rev. C
 List of Tables
Table 3.1: Correction vs. Prediction.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Table 4.1: Electrical And Thermal Properties of Different Glasses. . . . . . . . . . . . . . . 89
Table 4.2: Chemical Compositions of Different Glasses.. . . . . . . . . . . . . . . . . . . . . . 91
Table 5.1: Brain Slice Techniques. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
Table 7.1: Common Thermocouple Materials. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
Table 9.1: Example Temperature Data.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
Table 11.1: Thermal Noise of Resistors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
Table 11.2: Shot Noise.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
Table A.1: Noise of a 10 MW Resistor in Microvolts rms. . . . . . . . . . . . . . . . . . . . 247
Table A.2: rms Current in Picoamps. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250

The Axon Guide — 2500-0102 Rev. C xiii

xiv The Axon Guide — 2500-0102 Rev. C
 Preface
MDS Analytical Technologies is pleased to present you with the third edition of The
Axon Guide, a laboratory guide to electrophysiology and biophysics. The purpose of this
guide is to serve as an information and data resource for electrophysiologists. It covers a
broad scope of topics ranging from the biological basis of bioelectricity and a description
of the basic experimental setup to a discussion of mechanisms of noise and data analysis.

The Axon Guide third edition is a tool benefitting both the novice and the expert electro-
physiologist. Newcomers to electrophysiology will gain an appreciation of the intricacies
of electrophysiological measurements and the requirements for setting up a complete
recording and analysis system. For experienced electrophysiologists, we include in-depth
discussions of selected topics, such as advanced methods in electrophysiology and noise.

This edition is the first major revision of the Axon Guide since the original edition was
published in 1993. While the fundamentals of electrophysiology have not changed in that
time, changes in instrumentation and computer technology made a number of the origi-
nal chapters interesting historical documents rather than helpful guides. This third edi-
tion is up-to-date with current developments in technology and instrumentation.

This guide was the product of a collaborative effort of many researchers in the field of
electrophysiology and of MDS Analytical Technologies’ staff. We are deeply grateful to
these individuals for sharing their knowledge and devoting significant time and effort to
this endeavor.

David Yamane
Director, Electrophysiology Marketing
MDS Analytical Technologies

November 2007

The Axon Guide — 2500-0102 Rev. C xv

xvi The Axon Guide — 2500-0102 Rev. C
 Introduction
Axon Instruments, Inc., was founded in 1983 to design and manufacture instrumentation
and software for electrophysiology and biophysics. Its products were distinguished by the
company’s innovative design capability, high-quality products, and expert technical sup-
port. Today, the Axon Instruments products are part of MDS Analytical Technologies’
portfolio of life science and drug discovery products. The Axon brand of microelectrode
amplifiers, digitizers, and data acquisition and analysis software provides researchers with
ready-to-use, technologically advanced products which allows them more time to pursue
their primary research goals.
Furthermore, to ensure continued success with our products, we have staffed our Techni-
cal Support group with experienced Ph.D. electrophysiologists.
In addition to the Axon product suite, MDS Analytical Technologies has developed three
automated electrophysiology platforms. Together with the Axon Cellular Neuroscience
products, MDS Analytical Technologies spans the entire drug discovery process from
screening to safety assessment.
In recognition of the continuing excitement in ion channel research, as evidenced by the
influx of molecular biologists, biochemists, and pharmacologists into this field,
MDS Analytical Technologies is proud to support the pursuit of electrophysiological and
biophysical research with this laboratory techniques workbook.

ACKNOWLEDGMENT TO MDS ANALYTICAL

TECHNOLOGIES CONSULTANTS AND CUSTOMERS
MDS Analytical Technologies employs a talented team of engineers and scientists dedi-
cated to designing instruments and software incorporating the most advanced technology
and the highest quality. Nevertheless, it would not be possible for us to enhance our prod-
ucts without close collaborations with members of the scientific community. These col-
laborations take many forms.
Some scientists assist MDS Analytical Technologies on a regular basis, sharing their
insights on current needs and future directions of scientific research. Others assist us by
virtue of a direct collaboration in the design of certain products. Many scientists help us
by reviewing our instrument designs and the development versions of various software
products. We are grateful to these scientists for their assistance. We also receive a signifi-
cant number of excellent suggestions from the customers we meet at scientific confer-
ences. To all of you who have visited us at our booths and shared your thoughts, we
extend our sincere thanks. Another source of feedback for us is the information that we

The Axon Guide — 2500-0102 Rev. C xvii

 receive from the conveners of the many excellent summer courses and workshops that we
support with equipment loans. Our gratitude is extended to them for the written assess-
ments they often send us outlining the strengths and weaknesses of Axon Cellular
Neuroscience products from MDS Analytical Technologies.

xviii The Axon Guide — 2500-0102 Rev. C

 Editorial
Editor
Rivka Sherman-Gold, Ph.D.

Editorial Committee
Michael J. Delay, Ph.D.

Alan S. Finkel, Ph.D.

Henry A. Lester, Ph.D.1

W. Geoffrey Powell, MBA

Rivka Sherman-Gold, Ph.D.

Layout and Editorial Assistance

Jay Kurtz

Artwork

Elizabeth Brzeski

Editor – 3rd Edition

Warburton H. Maertz

Editorial Committee – 3rd Edition

Scott V. Adams, Ph.D.

S. Clare Chung, Ph.D.

Toni Figl, Ph.D.

Warburton H. Maertz

Damian J. Verdnik, Ph.D.

1.Lester is at the California Institute of Technology, Pasadena, CA. The other editorial staff are from MDS Analytical Technologies.

The Axon Guide — 2500-0102 Rev. C xix

 Layout and Editorial Assistance – 3rd Edition
Simone Andrews

Eveline Pajares

Peter Valenzuela

Damian J. Verdnik, Ph.D.

Alfred Walter, Ph.D.

Acknowledgements
The valuable inputs and insightful comments of Dr. Bertil Hille, Dr. Joe Immel and
Dr. Stephen Redman are much appreciated.

xx The Axon Guide — 2500-0102 Rev. C

 Contributors
The Axon Guide is the product of a collaborative effort of MDS Analytical Technologies
and researchers from several institutions whose contributions to the Guide are gratefully
acknowledged.

John M. Bekkers, Ph.D., Division of Neuroscience, John Curtin School of Medical

Research, Australian National University, Canberra, A.C.T. Australia.

Richard J. Bookman, Ph.D., Department of Molecular & Cellular Pharmacology, Miller

School of Medicine, University of Miami, Miami, Florida.

Michael J. Delay, Ph.D.

Alan S. Finkel, Ph.D.

Aaron Fox, Ph.D., Department of Neurobiology, Pharmacology and Physiology, Univer-

sity of Chicago, Chicago, Illinois.

David Gallegos.

Robert I. Griffiths, Ph.D., Monash Institute of Medical Research, Faculty of Medicine,

Monash University, Clayton, Victoria, Australia.

Donald Hilgemann, Ph.D., Graduate School of Biomedical Sciences, Southwestern Med-

ical School, Dallas, Texas.

Richard H. Kramer, Department of Molecular & Cell Biology, University of California

Berkeley, Berkeley, California.

Henry A. Lester, Ph.D., Division of Biology, California Institute of Technology, Pasadena,

California.

Richard A. Levis, Ph.D., Department of Molecular Biophysics & Physiology, Rush-Pres-

byterian-St. Luke’s Medical College, Chicago, Illinois.

Edwin S. Levitan, Ph.D., Department of Pharmacology, School of Medicine, University

of Pittsburgh, Pittsburgh, Pennsylvania.

M. Craig McKay, Ph.D., Department of Molecular Endocrinology, GlaxoSmithKline,

Inc., Research Triangle Park, North Carolina.

David J. Perkel, Department of Biology, University of Washington, Seattle, Washington.

The Axon Guide — 2500-0102 Rev. C xxi

 Stuart H. Thompson, Hopkins Marine Station, Stanford University, Pacific Grove, Cali-
fornia.

James L. Rae, Ph.D., Department of Physiology and Biomedical Engineering, Mayo

Clinic, Rochester, Minnesota.

Michael M. White, Ph.D., Department of Pharmacology & Physiology, Drexel University

College of Medicine, Philadelphia, Pennsylvania.

William F. Wonderlin, Ph.D., Department of Biochemistry & Molecular Phamacology,

West Virginia University, Morgantown, West Virginia.

xxii The Axon Guide — 2500-0102 Rev. C

 1. Bioelectricity
Chapter 1 introduces the basic concepts used in making electrical measurements from
cells and the techniques used to make these measurements.

1.1. ELECTRICAL POTENTIALS

A cell derives its electrical properties mostly from the electrical properties of its mem-
brane. A membrane, in turn, acquires its properties from its lipids and proteins, such as
ion channels and transporters. An electrical potential difference exists between the interior
and exterior of cells. A charged object (ion) gains or loses energy as it moves between
places of different electrical potential, just as an object with mass moves “up” or “down”
between points of different gravitational potential. Electrical potential differences are usu-
ally denoted as V or ΔV and measured in volts; therefore, potential is also termed voltage.
The potential difference across a cell relates the potential of the cell’s interior to that of the
external solution, which, according to the commonly accepted convention, is zero.

The potential difference across a lipid cellular membrane (“transmembrane potential”) is

generated by the “pump” proteins that harness chemical energy to move ions across the
cell membrane. This separation of charge creates the transmembrane potential. Because
the lipid membrane is a good insulator, the transmembrane potential is maintained in the
absence of open pores or channels that can conduct ions.

Typical transmembrane potentials amount to less than 0.1 V, usually 30 to 90 mV in most

animal cells, but can be as much as 200 mV in plant cells. Because the salt-rich solutions
of the cytoplasm and extracellular milieu are fairly good conductors, there are usually very
small differences at steady state (rarely more than a few millivolts) between any two points
within a cell’s cytoplasm or within the extracellular solution. Electrophysiological equip-
ment enables researchers to measure potential (voltage) differences in biological systems.

1.2. ELECTRICAL CURRENTS

Electrophysiological equipment can also measure current, which is the flow of electrical
charge passing a point per unit of time. Current (I) is measured in amperes (A). Usually,
currents measured by electrophysiological equipment range from picoamperes to micro-
amperes. For instance, typically, 104 Na+ ions cross the membrane each millisecond that a
single Na+ channel is open. This current equals 1.6 pA (1.6 x 10-19 C/ion x 104 ions/ms x
103 ms/s; recall that 1 A is equal to 1 coulomb (C)/s).

The Axon Guide — 2500-0102 Rev. C 1

 1. Bioelectricity

Two handy rules about currents often help to understand electrophysiological phenom-
ena: 1) current is conserved at a branch point (Figure 1.1); and 2) current always flows in
a complete circuit (Figure 1.2). In electrophysiological measurements, currents can flow
through capacitors, resistors, ion channels, amplifiers, electrodes and other entities, but
they always flow in complete circuits.

I total

I1 I2

I total = I 1 + I 2

Figure 1.1: Conservation of current. Current is conserved at a branch point.

I
I

Microelectrode
Battery

I = I 1+ I 2

Cell
I1 I2

Capacitor

I
ELECTRONIC
INSTRUMENT
I

Figure 1.2: A typical electrical circuit.

Example of an electrical circuit with various parts. Current always flows in a complete
circuit.

2 The Axon Guide — 2500-0102 Rev. C

 1.3. Resistors and Conductors

1.3. RESISTORS AND CONDUCTORS

Currents flow through resistors or conductors. The two terms actually complement one
another—the former emphasizes the barriers to current flow, while the latter emphasizes
the pathways for flow. In quantitative terms, resistance R (units: ohms (Ω)) is the inverse
of conductance G (units: siemens (S)); thus, infinite resistance is zero conductance. In
electrophysiology, it is convenient to discuss currents in terms of conductance because
side-by-side (“parallel”) conductances simply sum (Figure 1.3). The most important
application of the parallel conductances involves ion channels. When several ion channels
in a membrane are open simultaneously, the total conductance is simply the sum of the
conductances of the individual open channels.

G G Gtotal = 2G

γ γ

G total = 2 γ

ion lipid
channel bilayer

Figure 1.3: Summation of conductance.

Conductances in parallel summate together, whether they are resistors or channels.

A more accurate representation of an ion channel is a conductor in series with two addi-
tional circuit elements (Figure 1.4):

1 A switch that represents the gate of the channel, which would be in its conducting
position when the gate is open, and

The Axon Guide — 2500-0102 Rev. C 3

 1. Bioelectricity

2 A battery that represents the reversal potential of the ionic current for that channel.
The reversal potential is defined operationally as the voltage at which the current
changes its direction. For a perfectly selective channel (i.e., a channel through which
only a single type of ion can pass), the reversal potential equals the Nernst potential
for the permeant ion. The Nernst potential for ion A, Ea, can be calculated by the
Nernst equation:

E A = ( RT / z A F )1n{[ A ] o / [ A ] i } = 2.303( RT / z A F ) log10 {[ A ] o / [ A ] i } ( units: volts ) (1)

where R is the gas constant (8.314 V C K-1 mol-1), T is the absolute temperature
(T = 273° + °C), zA is the charge of ion A, F is Faraday’s constant (9.648x104 C mol-1),
and [A]o and [A]i are the concentrations of ion A outside the cell and inside the cell,
respectively. At 20 °C (“room temperature”), 2.303(RT/zAF) log10{[A]o/[A]i} =
58 mV log10{[A]o/[A]i}for a univalent ion.

E reversal

Figure 1.4: Equivalent circuit for a single-membrane channel.

A more realistic equivalent circuit for a single-membrane channel.

For instance, at room temperature, a Na+ channel facing intracellular Na+ concentration
that is ten-fold lower than the extracellular concentration of this ion would be represented
by a battery of +58 mV. A K+ channel, for which the concentration gradient is usually
reversed, would be represented by a battery of -58 mV.

Reversal potentials are not easily predicted for channels that are permeable to more than
one ion. Nonspecific cation channels, such as nicotinic acetylcholine receptors, usually
have reversal potentials near zero millivolts. Furthermore, many open channels have a
nonlinear relation between current and voltage. Consequently, representing channels as
resistors is only an approximation. Considerable biophysical research has been devoted to
understanding the current-voltage relations of ion channels and how they are affected by
the properties and concentrations of permeant ions.

4 The Axon Guide — 2500-0102 Rev. C

 1.4. Ohm’s Law

The transmembrane potential is defined as the potential at the inner side of the mem-
brane relative to the potential at the outer side of the membrane. The resting membrane
potential (Erp) describes a steady-state condition with no net flow of electrical current
across the membrane. The resting membrane potential is determined by the intracellular
and extracellular concentrations of ions to which the membrane is permeable and on their
permeabilities. If one ionic conductance is dominant, the resting potential is near the
Nernst potential for that ion. Since a typical cell membrane at rest has a much higher per-
meability to potassium (PK) than to sodium, calcium or chloride (PNa, PCa and PCl,
respectively), the resting membrane potential is very close to EK, the potassium reversal
potential.

1.4. OHM’S LAW

For electrophysiology, perhaps the most important law of electricity is Ohm’s law. The
potential difference between two points linked by a current path with a conductance G
and a current I (Figure 1.5) is:

ΔV = IR = I/G (units: volts) (2)

G = I/R Δ V=IR

Figure 1.5: Ohm’s law.

The Axon Guide — 2500-0102 Rev. C 5

 1. Bioelectricity

This concept applies to any electrophysiological measurement, as illustrated by the two

following examples:

1 In an extracellular recording experiment: the current I that flows between parts of a

cell through the external resistance R produces a potential difference ΔV, which is
usually less than 1 mV (Figure 1.6). As the impulse propagates, I changes and,
therefore, ΔV changes as well.

I
V

Figure 1.6: IR drop.

In extracellular recording, current I that flows between points of a cell is measured as

the potential difference (“IR drop”) across the resistance R of the fluid between the two
electrodes.

2 In a voltage-clamp experiment: when N channels, each of conductance γ, are open,

the total conductance is Nγ. The electrochemical driving force ΔV (membrane
potential minus reversal potential) produces a current NγΔV. As channels open and
close, N changes and so does the voltage-clamp current I. Hence, the voltage-clamp
current is simply proportional to the number of open channels at any given time.
Each channel can be considered as a γ conductance increment.

6 The Axon Guide — 2500-0102 Rev. C

 1.5. The Voltage Divider

1.5. THE VOLTAGE DIVIDER

Figure 1.7 describes a simple circuit called a voltage divider in which two resistors are con-
nected in series:

R1 R1
V1 = E
E R 1+ R 2

R2 R2
V2 = E
R 1+ R 2

Figure 1.7: A voltage divider.

The total potential difference provided by the battery is E; a portion of this voltage
appears across each resistor:

R1 R2
ΔV1 = E ; ΔV 2 = E (2a)
R1 + R 2 R1 + R 2

When two resistors are connected in series, the same current passes through each of them.
Therefore the circuit is described by:

ΔV1 + ΔV 2 = E (2b)

where E is the value of the battery, which equals the total potential difference across both
resistors. As a result, the potential difference is divided in proportion to the two resistance
values.

1.6. PERFECT AND REAL ELECTRICAL INSTRUMENTS

Electrophysiological measurements should satisfy two requirements: 1) They should accu-
rately measure the parameter of interest, and 2) they should produce no perturbation of
the parameter. The first requirement can be discussed in terms of a voltage divider. The
second point will be discussed after addressing electrodes.

The Axon Guide — 2500-0102 Rev. C 7

 1. Bioelectricity

The best way to measure an electrical potential difference is to use a voltmeter with infi-
nite resistance. To illustrate this point, consider the arrangement of Figure 1.8 A, which
can be reduced to the equivalent circuit of Figure 1.8 B.

Micropipette Electrode with Resistance Re

A

MEASURING
CIRCUIT

Rm Em
Membrane Resistance

Bath Electrode

B EQUIVALENT CIRCUIT
Re
Resting
Potential
Em Rm Rin
V

R in
V = ER + R
in e

Figure 1.8: Representative voltmeter with infinite resistance.

Instruments used to measure potentials must have a very high input resistance Rin.

Before making the measurement, the cell has a resting potential of Erp, which is to be
measured with an intracellular electrode of resistance Re. To understand the effect of the
measuring circuit on the measured parameter, we will pretend that our instrument is a
“perfect” voltmeter (i.e., with an infinite resistance) in parallel with a finite resistance Rin,
which represents the resistance of a real voltmeter or the measuring circuit. The combina-
tion Re and Rin forms a voltage divider, so that only a fraction of Erp appears at the input
of the “perfect” voltmeter; this fraction equals ErpRin/(Rin + Re). The larger Rin, the closer

8 The Axon Guide — 2500-0102 Rev. C

 1.6. Perfect and Real Electrical Instruments

V is to Erp. Clearly the problem gets more serious as the electrode resistance Re increases,
but the best solution is to make Rin as large as possible.

On the other hand, the best way to measure current is to open the path and insert an
ammeter. If the ammeter has zero resistance, it will not perturb the circuit since there is no
IR-drop across it.

1.6.1. IONS IN SOLUTIONS AND ELECTRODES

Ohm’s law—the linear relation between potential difference and current flow applies to
aqueous ionic solutions, such as blood, cytoplasm and sea water. Complications are intro-
duced by two factors:

1 The current is carried by at least two types of ions (one anion and one cation) and
often by many more. For each ion, current flow in the bulk solution is proportional to
the potential difference. For a first approximation, the conductance of the whole
solution is simply the sum of the conductances contributed by each ionic species.
When the current flows through ion channels, it is carried selectively by only a subset
of the ions in the solution.
2 At the electrodes, current must be transformed smoothly from a flow of electrons in
the copper wire to a flow of ions in solution. Many sources of errors (artifacts) are
possible. Several types of electrodes are used in electrophysiological measurements; the
most common is a silver/silver chloride (Ag/AgCl) interface, which is a silver wire
coated with silver chloride (Figure 1.9). If electrons flow from the copper wire
through the silver wire to the electrode AgCl pellet, they convert the AgCl to Ag
atoms and the Cl- ions become hydrated and enter the solution. If electrons flow in
the reverse direction, Ag atoms in the silver wire that is coated with AgCl give up their
electrons (one electron per atom) and combine with Cl- ions that are in the solution
to make insoluble AgCl. This is, therefore, a reversible electrode, i.e., current can flow
in both directions. There are several points to remember about Ag/AgCl electrodes:
a The Ag/AgCl electrode performs well only in solutions containing chloride ions.
b Because current must flow in a complete circuit, two electrodes are needed. If the
two electrodes face different Cl- concentrations (for instance, 3 M KCl inside a
micropipette1 and 120 mM NaCl in a bathing solution surrounding the cell), there
will be a difference in the half-cell potentials (the potential difference between the
solution and the electrode) at the two electrodes, resulting in a large steady
potential difference in the two wires attached to the electrodes. This steady
potential difference, termed liquid junction potential, can be subtracted
electronically and poses few problems as long as the electrode is used within its
reversible limits.
c If the AgCl is exhausted by the current flow, bare silver could come in contact with
the solution. Silver ions leaking from the wire can poison many proteins. Also, the

1. A micropipette is a pulled capillary glass into which the Ag/AgC1 electrode is inserted (see Chapter 4).

The Axon Guide — 2500-0102 Rev. C 9

 1. Bioelectricity

half-cell potentials now become dominated by unpredictable, poorly reversible

surface reactions due to other ions in the solution and trace impurities in the silver,
causing electrode polarization. However, used properly, Ag/AgCl electrodes possess
the advantages of little polarization and predictable junction potential.

copper wire silver wire

electron (e- ) flow

_
Cl AgCl Coating
Ag

Electrode reaction:
Ag + Cl- AgCl + electron (e- )
This reaction can also be presented by:
Ag + Cl-
+
AgCl

+e- -e-
Ag

Figure 1.9: The silver/silver chloride electrode.

The silver/silver chloride electrode is reversible but exhaustible.

Another type of electrode, made of platinum (Pt) (Figure 1.10), is irreversible but not
exhaustible. At its surface, Pt catalyzes the electrolysis of water. The gaseous H2 or O2
produced, depending on the direction of current flow, leaves the surface of the electrode.
If both electrodes are Pt electrodes, the hydroxyl ions and protons are produced in equal
numbers; however, local pH changes can still occur.

10 The Axon Guide — 2500-0102 Rev. C

 1.7. Capacitors and their Electrical Fields

copper wire platinum wire

_
electron (e ) flow

Pt
H 2 or O2 bubbles

1
e - + H 2O OH - + 2 H 2 or H 2O
+ 1
2H + 2 O2 + e-

Figure 1.10: The platinum electrode.

A platinum electrode is irreversible but inexhaustible.

1.7. CAPACITORS AND THEIR ELECTRICAL FIELDS

The electrical field is a property of each point in space and is defined as proportional to
the force experienced by a charge placed at that point. The greater the potential difference
between two points fixed in space, the greater the field at each point between them. For-
mally, the electrical field is a vector defined as the negative of the spatial derivative of the
potential.

The concept of the electrical field is important for understanding membrane function.
Biological membranes are typically less than 10 nm thick. Consequently, a transmem-
brane resting potential of about 100 mV produces a very sizable electrical field in the
membrane of about 105 V/cm. This is close to the value at which most insulators break
down irreversibly because their atoms become ionized. Of course, typical electrophysio-
logical equipment cannot measure these fields directly. However, changes in these fields
are presumably sensed by the gating domains of voltage-sensitive ion channels, which
determine the opening and closing of channels, and so the electrical fields underlie the
electrical excitability of membranes.

Another consequence of the membrane’s thinness is that it makes an excellent capacitor.

Capacitance (C; measured in farads, F) is the ability to store charge Q when a voltage ΔV
occurs across the two “ends,” so that:

Q = CΔV (4)
The formal symbol for a capacitor is two parallel lines (Figure 1.12). This symbol arose
because the most effective capacitors are parallel conducting plates of large area separated
by a thin sheet of insulator (Figure 1.11) an excellent approximation of the lipid bilayer.

The Axon Guide — 2500-0102 Rev. C 11

 1. Bioelectricity

The capacitance C is proportional to the area and inversely proportional to the distance
separating the two conducting sheets.

+ + + +
C ΔV
- - - -

Figure 1.11: Capacitance. A charge Q is stored in a capacitor of value C held at a potential ΔV.

When multiple capacitors are connected in parallel, this is electronically equivalent to a

single large capacitor; that is, the total capacitance is the sum of their individual capaci-
tance values (Figure 1.12). Thus, membrane capacitance increases with cell size. Mem-
brane capacitance is usually expressed as value per unit area; nearly all lipid bilayer
membranes of cells have a capacitance of 1 μF/cm2 (0.01 pF/μm2).

C C Ctotal = 2C

C C 2C

Figure 1.12: Capacitors in parallel add their values.

1.8. CURRENTS THROUGH CAPACITORS

Equation 4 shows that charge is stored in a capacitor only when there is a change in the
voltage across the capacitor. Therefore, the current flowing through capacitance C is pro-
portional to the voltage change with time:

12 The Axon Guide — 2500-0102 Rev. C

 1.8. Currents Through Capacitors

ΔV
I =C (5)
Δt

Until now, we have been discussing circuits whose properties do not change with time. As
long as the voltage across a membrane remains constant, one can ignore the effect of the
membrane capacitance on the currents flowing across the membrane through ion chan-
nels. While the voltage changes, there are transient capacitive currents in addition to the
steady-state currents through conductive channels. These capacitive currents constitute
one of the two major influences on the time-dependent electrical properties of cells (the
other is the kinetics of channel gating). On Axon Cellular Neuroscience voltage- or patch-
clamp amplifiers, several controls are devoted to handle these capacitive currents. There-
fore it is worth obtaining some intuitive “feel” for their behavior.

The stored charge on the membrane capacitance accompanies the resting potential, and
any change in the voltage across the membrane is accompanied by a change in this stored
charge. Indeed, if a current is applied to the membrane, either by channels elsewhere in
the cell or by current from the electrode, this current first satisfies the requirement for
charging the membrane capacitance, then it changes the membrane voltage. Formally, this
can be shown by representing the membrane as a resistor of value R in parallel with capac-
itance C (Figure 1.13).

C R = I/G

Capacitance Conductance

Figure 1.13: Membrane behavior compared with an electrical current.

A membrane behaves electrically like a capacitance in parallel with a resistance.

Now, if we apply a pulse of current to the circuit, the current first charges up the capaci-
tance, then changes the voltage (Figure 1.14).

The Axon Guide — 2500-0102 Rev. C 13

 1. Bioelectricity

I
time

Figure 1.14: RC parallel circuit response.

Response of an RC parallel circuit to a step of current.

The voltage V(t) approaches steady state along an exponential time course:

V (t ) = Vin f (1 − e −t / τ )

The steady-state value Vinƒ (also called the infinite-time or equilibrium value) does not
depend on the capacitance; it is simply determined by the current I and the membrane
resistance R:

V inf = IR (6)

This is just Ohm’s law, of course; but when the membrane capacitance is in the circuit, the
voltage is not reached immediately. Instead, it is approached with the time constant τ,
given by:

τ = RC (7)

Thus, the charging time constant increases when either the membrane capacitance or the
resistance increases. Consequently, large cells, such as Xenopus oocytes that are frequently
used for expression of genes encoding ion-channel proteins, and cells with extensive mem-
brane invigorations, such as the T-system in skeletal muscle, have a long charging phase.

1.9. CURRENT CLAMP AND VOLTAGE CLAMP

In a current-clamp experiment, one applies a known constant or time-varying current and
measures the change in membrane potential caused by the applied current. This type of
experiment mimics the current produced by a synaptic input.

In a voltage clamp experiment one controls the membrane voltage and measures the trans-
membrane current required to maintain that voltage. Despite the fact that voltage clamp

14 The Axon Guide — 2500-0102 Rev. C

 1.9. Current Clamp and Voltage Clamp

does not mimic a process found in nature, there are three reasons to do such an experi-
ment:

1 Clamping the voltage eliminates the capacitive current, except for a brief time
following a step to a new voltage (Figure 1.15). The brevity of the capacitive current
depends on many factors that are discussed in following chapters.
2 Except for the brief charging time, the currents that flow are proportional only to the
membrane conductance, i.e., to the number of open channels.
3 If channel gating is determined by the transmembrane voltage alone (and is insensitive
to other parameters such as the current and the history of the voltage), voltage clamp
offers control over the key variable that determines the opening and closing of ion
channels.

V1

capacitive transient

Idt = CV1
I = V1 /R

Figure 1.15: Typical voltage-clamp experiment.

A voltage-clamp experiment on the circuit of Figure 1.13.

The patch clamp is a special voltage clamp that allows one to resolve currents flowing
through single ion channels. It also simplified the measurement of currents flowing
through the whole-cell membrane, particularly in small cells that cannot be easily pene-
trated with electrodes. The characteristics of a patch clamp are dictated by two facts:

1 The currents measured are very small, on the order of picoamperes in single-channel
recording and usually up to several nanoamperes in whole-cell recording. Due to the
small currents, particularly in single-channel recording, the electrode polarizations
and nonlinearities are negligible and the Ag/AgCl electrode can record voltage
accurately even while passing current.
2 The electronic ammeter must be carefully designed to avoid adding appreciable noise
to the currents it measures.

The Axon Guide — 2500-0102 Rev. C 15

 1. Bioelectricity

1.10. GLASS MICROELECTRODES AND TIGHT SEALS

Successful electrophysiological measurements depend on two technologies: the design of
electronic instrumentation and the properties and fabrication of glass micropipettes. Glass
pipettes are used both for intracellular recording and for patch recording; quartz pipettes
have been used for ultra low-noise single-channel recording (for a detailed discussion of
electrode glasses see Chapter 4). Successful patch recording requires a tight seal between
the pipette and the membrane. Although there is not yet a satisfactory molecular descrip-
tion of this seal, we can describe its electrical characteristics.

Requirement 2) above (see Section 1.6., “Perfect and Real Electrical Instruments”) states
that the quality of the measurement depends on minimizing perturbation of the cells. For
the case of voltage recording, this point can be presented with the voltage divider circuit
(Figure 1.16).

seal resistance R S

EQUIVALENT CIRCUIT

V intracellular
RS
V = EK
EK RS + RK
RS V
RK

extracellular

Figure 1.16: Intracellular electrode measurement.

This intracellular electrode is measuring the resting potential of a cell whose membrane
contains only open K+ channels. As the seal resistance Rs increases, the measurement
approaches the value of EK.

For the case of patch recording, currents through the seal do not distort the measured
voltage or current, but they do add to the current noise. Current noise can be analyzed
either in terms of the Johnson noise of a conductor, which is the thermal noise that
increases with the conductance (see Chapter 7 and Chapter 11), or in terms of simple sta-
tistics. The latter goes as follows: If a current of N ions/ms passes through an open chan-
nel, then the current will fluctuate from one millisecond to the next with a standard
deviation of √N. These fluctuations produce noise on the single-channel recorded traces.
If an additional current is flowing in parallel through the seal (Figure 1.17), it causes an
increase in the standard deviations. For instance, if the current through the seal is ten-fold
larger than through the channel, then the statistical fluctuations in current flow produced
by the seal are √10 (316%) larger than they would be for a “perfect” seal.

16 The Axon Guide — 2500-0102 Rev. C

 1.11. Further Reading

AMPLIFIER

Iseal

Ichannel

I=0
good seal

bad seal

single-channel event

Figure 1.17: Good and bad seals.

In a patch recording, currents through the seal also flow through the measuring cir-
cuit, increasing the noise on the measured current.

1.11. FURTHER READING

Brown, K. T., Flaming, D. G. Advanced Micropipette Techniques for Cell Physiology.
John Wiley & Sons, New York, NY, 1986.
Hille, B., Ionic Channels in Excitable Membranes. Second edition. Sinauer Associates,
Sunderland, MA, 1991.
Horowitz, P., Hill, W. The Art of Electronics. Second edition. Cambridge University
Press, Cambridge, UK, 1989.
Miller, C., Ion Channel Reconstitution. C. Miller, Ed., Plenum Press, New York, NY,
1986.
Sakmann, B., Neher, E. Eds., Single-Channel Recording. Plenum Press, New York, NY,
1983.
Smith, T. G., Lecar, H., Redman, S. J., Gage, P. W., Eds. Voltage and Patch Clamping
with Microelectrodes. American Physiological Society, Bethesda, MD, 1985.
Standen, N. B., Gray, P. T. A., Whitaker, M. J., Eds. Microelectrode Techniques: The Ply-
mouth Workshop Handbook. The Company of Biologists Limited, Cambridge, England,
1987.

The Axon Guide — 2500-0102 Rev. C 17

 1. Bioelectricity

18 The Axon Guide — 2500-0102 Rev. C

 2. The Laboratory Setup
Each laboratory setup is different, reflecting the requirements of the experiment or the
foibles of the experimenter. Chapter 2 describes components and considerations that are
common to all setups dedicated to measure electrical activity in cells.

An electrophysiological setup has four main requirements:

1 Environment: the means of keeping the preparation healthy;

2 Optics: a means of visualizing the preparation;
3 Mechanics: a means of stably positioning the microelectrode; and
4 Electronics: a means of amplifying and recording the signal.
This Guide focuses mainly on the electronics of the electrophysiological laboratory setup.

To illustrate the practical implications of these requirements, two kinds of “typical” setups
are briefly described, one for in vitro extracellular recording, the other for single-channel
patch clamping.

2.1. THE IN VITRO EXTRACELLULAR RECORDING SETUP

This setup is mainly used for recording field potentials in brain slices. The general objec-
tive is to hold a relatively coarse electrode in the extracellular space of the tissue while
mimicking as closely as possible the environment the tissue experiences in vivo. Thus, a
rather complex chamber that warms, oxygenates and perfuses the tissue is required. On
the other hand, the optical and mechanical requirements are fairly simple. A low-power
dissecting microscope with at least 15 cm working distance (to allow near-vertical place-
ment of manipulators) is usually adequate to see laminae or gross morphological features.
Since neither hand vibration during positioning nor exact placement of electrodes is criti-
cal, the micromanipulators can be of the coarse mechanical type. However, the microma-
nipulators should not drift or vibrate appreciably during recording. Finally, the electronic
requirements are limited to low-noise voltage amplification. One is interested in measur-
ing voltage excursions in the 10 μV to 10 mV range; thus, a low-noise voltage amplifier
with a gain of at least 1,000 is required.

2.2. THE SINGLE-CHANNEL PATCH CLAMPING SETUP

The standard patch clamping setup is in many ways the converse of that for extracellular
recording. Usually very little environmental control is necessary: experiments are often

The Axon Guide — 2500-0102 Rev. C 19

 2. The Laboratory Setup

done in an unperfused culture dish at room temperature. On the other hand, the optical
and mechanical requirements are dictated by the need to accurately place a patch elec-
trode on a particular 10 or 20 μm cell. The microscope should magnify up to 300 or 400
fold and be equipped with some kind of contrast enhancement (Nomarski, Phase or Hoff-
man). Nomarski (or Differential Interference Contrast) is best for critical placement of
the electrode because it gives a very crisp image with a narrow depth of field. Phase con-
trast is acceptable for less critical applications and provides better contrast for fine pro-
cesses. Hoffman presently ranks as a less expensive, slightly degraded version of Nomarski.
Regardless of the contrast method selected, an inverted microscope is preferable for two
reasons: 1) it usually allows easier top access for the electrode since the objective lens is
underneath the chamber, and 2) it usually provides a larger, more solid platform upon
which to bolt the micromanipulator. If a top-focusing microscope is the only option, one
should ensure that the focus mechanism moves the objective, not the stage.

The micromanipulator should permit fine, smooth movement down to a couple of

microns per second, at most. The vibration and stability requirements of the micromanip-
ulator depend upon whether one wishes to record from a cell-attached or a cell-free
(inside-out or outside-out) patch. In the latter case, the micromanipulator needs to be sta-
ble only as long as it takes to form a seal and pull away from the cell. This usually tales less
than a minute.

Finally, the electronic requirements for single-channel recording are more complex than
for extracellular recording. However, excellent patch clamp amplifiers, such as those of the
Axopatch™ amplifier series from MDS Analytical Technologies, are commercially avail-
able.

A recent extension of patch clamping, the patched slice technique, requires a setup that
borrows features from both in vitro extracellular and conventional patch clamping config-
urations. For example, this technique may require a chamber that continuously perfuses
and oxygenates the slice. In most other respects, the setup is similar to the conventional
patch clamping setup, except that the optical requirements depend upon whether one is
using the thick-slice or thin-slice approach (see Further Reading at the end of this).
Whereas a simple dissecting microscope suffices for the thick-slice method, the thin-slice
approach requires a microscope that provides 300- to 400-fold magnification, preferably
top-focusing with contrast enhancement.

2.3. VIBRATION ISOLATION METHODS

By careful design, it should be possible to avoid resorting to the traditional electrophysiol-
ogist’s refuge from vibration: the basement room at midnight. The important principle
here is that prevention is better than cure; better to spend money on stable, well-designed
micromanipulators than on a complicated air table that tries to compensate for a micro-
manipulator’s inadequacies. A good micromanipulator is solidly constructed and com-
pact, so that the moment arm from the tip of the electrode, through the body of the
manipulator, to the cell in the chamber, is as short as possible. Ideally, the micromanipula-

20 The Axon Guide — 2500-0102 Rev. C

 2.4. Electrical Isolation Methods

tor should be attached close to the chamber; preferably bolted directly to the microscope
stage. The headstage of the recording amplifier should, in turn, be bolted directly to the
manipulator (not suspended on a rod), and the electrode should be short.

For most fine work, such as patch clamping, it is preferable to use remote-controlled
micromanipulators to eliminate hand vibration (although a fine mechanical manipulator,
coupled with a steady hand, may sometimes be adequate). Currently, there are three main
types of remote-controlled micromanipulators available: motorized, hydraulic/pneumatic
and piezoelectric. Motorized manipulators tend to be solid and compact and have excel-
lent long-term stability. However, they are often slow and clumsy to move into position,
and may exhibit backlash when changing direction. Hydraulic drives are fast, convenient
and generally backlash-free, but some models may exhibit slow drift when used in certain
configurations. Piezoelectric manipulators have properties similar to motorized drives,
except for their stepwise advancement.

Anti-vibration tables usually comprise a heavy slab on pneumatic supports. Tables of vary-
ing cost and complexity are commercially available. However, a homemade table, consist-
ing of a slab resting on partially-inflated inner tubes, may be adequate, especially if high-
quality micromanipulators are used.

2.4. ELECTRICAL ISOLATION METHODS

Extraneous electrical interference (not intrinsic instrument noise) falls into three main
categories: radiative electrical pickup, magnetically-induced pickup and ground-loop
noise.

2.4.1. RADIATIVE ELECTRICAL PICKUP

Examples of radiative electrical pickup include line frequency noise from lights and power
sockets (hum), and high frequency noise from computers. This type of noise is usually
reduced by placing conductive shields around the chamber and electrode and by using
shielded BNC cables. The shields are connected to the signal ground of the microelec-
trode amplifier. Traditionally, a Faraday cage is used to shield the microscope and cham-
ber. Alternatively, the following options can usually reduce the noise: 1) find the source of
noise, using an open circuit oscilloscope probe, and shield it; 2) use local shielding around
the electrode and parts of the microscope; 3) physically move the offending source (e.g., a
computer monitor) away from the setup; or 4) replace the offending source (e.g., a mono-
chrome monitor is quieter than a color monitor). Note that shielding may bring its own
penalties, such as introducing other kinds of noise or degrading one’s bandwidth (see
Chapter 11). Do not assume that commercial specifications are accurate. For example, a
DC power supply for the microscope lamp might have considerable AC ripple, and the
“shielded” lead connecting the microelectrode preamplifier to the main amplifier might
need additional shielding. Solution-filled perfusion tubing entering the bath may act as an
antenna and pick up radiated noise. If this happens, shielding of the tubing may be
required. Alternatively, a drip-feed reservoir, such as is used in intravenous perfusion sets,
may be inserted in series with the tubing to break the electrical continuity of the perfusion

The Axon Guide — 2500-0102 Rev. C 21

 2. The Laboratory Setup

fluid. Never directly ground the solution other than at the ground wire in the chamber,
which is the reference ground for the amplifier and which may not be the same as the sig-
nal ground used for shielding purposes. Further suggestions are given in Section 6.6.,
“Line-Frequency Pick-Up (Hum)”.

2.4.2. MAGNETICALLY-INDUCED PICKUP

Magnetically-induced pickup arises whenever a changing magnetic flux passes through a
loop of wire, thereby inducing a current in the wire. It most often originates in the vicinity
of electromagnets in power supplies, and is usually identified by its non-sinusoidal shape
with a frequency that is a higher harmonic of the line frequency. This type of interference
is easily reduced by moving power supplies away from sensitive circuitry. If this is not pos-
sible, try twisting the signal wires together to reduce the area of the loop cut by the flux, or
try shielding the magnetic source with “mu-metal.”

2.4.3. GROUND-LOOP NOISE

Ground-loop noise arises when shielding is grounded at more than one place. Magnetic
fields may induce currents in this loop. Moreover, if the different grounds are at slightly
different potentials, a current may flow through the shielding and introduce noise. In
principle, ground loops are easy to eliminate: simply connect all the shields and then
ground them at one place only. For instance, ground all the connected shields at the signal
ground of the microelectrode amplifier. This signal ground is, in turn, connected at only
one place to the power ground that is provided by the wall socket. In practice, however,
one is usually frustrated by one’s ignorance of the grounding circuitry inside electronic
apparatuses. For example, the shielding on a BNC cable will generally be connected to the
signal ground of each piece of equipment to which it is attached. Furthermore, each signal
ground may be connected to a separate power ground (but not on Axon Cellular
Neuroscience amplifiers). The loop might be broken by lifting off the BNC shielding
and/or disconnecting some power grounds (although this creates hazards of electrocu-
tion!). One could also try different power sockets, because the mains earth line may have a
lower resistance to some sockets than others. The grounds of computers are notorious for
noise. Thus, a large reduction in ground-loop noise might be accomplished by using opti-
cal isolation (see Chapter 9) or by providing the computer with a special power line with
its own ground.

The logical approach to reducing noise in the setup is to start with all equipment switched
off and disconnected; only an oscilloscope should be connected to the microelectrode
amplifier. First, measure the noise when the headstage is wrapped in grounded metal foil;
microelectrode headstages should be grounded through a low resistance (for instance,
1 MΩ), whereas patch-clamp headstages should be left open circuit. This provides a refer-
ence value for the minimum attainable radiative noise. Next, connect additional pieces of
electronic apparatuses while watching for the appearance of ground loops. Last, install an
electrode and add shielding to minimize radiative pickup. Finally, it should be admitted
that one always begins noise reduction in a mood of optimistic rationalism, but invariably
descends into frustrating empiricism.

22 The Axon Guide — 2500-0102 Rev. C

 2.5. Equipment Placement

2.5. EQUIPMENT PLACEMENT

While the placement of equipment is directed by personal preferences, a brief tour of elec-
trophysiologists’ common preferences may be instructive. Electrophysiologists tend to
prefer working alone in the corners of small rooms. This is partly because their work often
involves bursts of intense, intricate activity when distracting social interactions are inad-
missible. Furthermore, small rooms are often physically quieter since vibrations and air
currents are reduced. Having decided upon a room, it is usually sensible to first set up the
microscope and its intimate attachments, such as the chamber, the manipulators and the
temperature control system (if installed). The rationale here is that one’s first priority is to
keep the cells happy in their quiescent state, and one’s second priority is to ensure that the
act of recording from them is not consistently fatal. The former is assisted by a good envi-
ronment, the latter by good optics and mechanics. Working outward from the micro-
scope, it is clearly prudent to keep such things as perfusion stopcocks and
micromanipulator controllers off the vibration isolation table. Ideally, these should be
placed on small shelves that extend over the table where they can be accessed without
causing damaging vibrations and are conveniently at hand while looking through the
microscope.

Choice and placement of electronics is again a matter of personal preference. There are
minimalists who make do with just an amplifier and a computer, and who look forward to
the day when even those two will coalesce. Others insist on a loaded instrument rack. An
oscilloscope is important, because the computer is often insufficiently flexible. Further-
more, an oscilloscope often reveals unexpected subtleties in the signal that were not appar-
ent on the computer screen because the sample interval happened not to have been set
exactly right. The oscilloscope should be at eye level. Directly above or below should be
the microelectrode amplifier so that adjustments are easily made and monitored. Last, the
computer should be placed as far as possible—but still within a long arm’s reach—from
the microscope. This is necessary both to reduce the radiative noise from the monitor and
to ensure that one’s elbows will not bump the microscope when hurriedly typing at the
keyboard while recording from the best cell all week.

A final, general piece of advice is perhaps the most difficult to heed: resist the temptation
to mess eternally with getting the setup just right. As soon as it is halfway possible, do an
experiment. Not only will this provide personal satisfaction, it may also highlight specific
problems with the setup that need to be corrected or, better, indicate that an anticipated
problem is not so pressing after all.

The Axon Guide — 2500-0102 Rev. C 23

 2. The Laboratory Setup

2.5.1. LIST OF EQUIPMENT

Traditional Patch-Clamp Setup
Item Suggested Manufacturers

Vibration isolation table Newport

Micro-g (Technical Manufacturing Corp.)

Microscope, inverted Carl Zeiss

Leica Microsystems
Nikon
Olympus

Micromanipulators
hydraulic Narishige International USA
motorized Newport
piezoelectric EXFO (Burleigh)
Sutter Instrument

Patch-clamp amplifiers MDS Analytical Technologies.

(Axon Cellular Neuroscience)

Oscilloscopes Tektronix

Pipette fabrication

glass Garner Glass

Friedrich & Dimmock
Sutter Instrument
pullers Sutter Instrument
microforges Narishige International USA
homemade - based on:
Carl Zeiss metallurgical microscope
Olympus CH microscope
coaters Narishige International USA
hydrophobic coating Dow Corning Sylgard 184
Q-dope

Microelectrode holders MDS Analytical Technologies

(Axon Cellular Neuroscience)
E. W. Wright

Chamber, temperature control Narashige International USA

Computers see Chapter 9

24 The Axon Guide — 2500-0102 Rev. C

 2.5. Equipment Placement

Patch-Slice Setup
Item Suggested Manufacturers

Microscope, low power Carl Zeiss

Vibratome Vibratome
(other requirements as for a traditional patch-clamp setup)
Extra/Intracellular Microelectrode Setup
Item Suggested Manufacturers

Vibration isolation table Newport

Micro-g (Technical Manufacturing Corp.)

Microscope Carl Zeiss

Leica Microsystems
Nikon
Olympus

Micromanipulators Narishige International USA

mechanical Stoelting
Sutter Instrument
hydraulic Narishige International USA
piezoelectric EXFO (Burleigh)
Sutter Instrument

Microelectrode amplifiers MDS Analytical Technologies

(Axon Cellular Neuroscience)

Oscilloscopes Tektronix

Electrode fabrication
glass Garner Glass
Friedrich & Dimmock
Sutter Instrument
pullers David Kopf
Sutter Instrument
GlasswoRx

Microelectrode holders MDS Analytical Technologies

(Axon Cellular Neuroscience)
E. W. Wright

Chamber, temperature control Narishige International USA

Computers See Chapter 9

Optical Recording Setup
Photomultipliers Hamamatsu

The Axon Guide — 2500-0102 Rev. C 25

 2. The Laboratory Setup

Imaging systems MDS Analytical Technologies

2.6. FURTHER READING

Conventional intra- and extracellular recording from brain slices
Dingledine, R. Ed. Brain Slices. Plenum Press, New York, NY, 1983.

Geddes, L. A. Electrodes and the Measurement of Bioelectric Events. Wiley Interscience,

1972.

Purves, R. D. Microelectrode Methods for Intracellular Recording and Ionophoresis. Aca-

demic Press, San Diego, CA, 1986.

Smith, T. G., Jr., Lecar, H., Redman, S. J., Gage, P. W. Ed. Voltage and Patch Clamping
with Microelectrodes. American Physiological Society, Bethesda, MD, 1985.

Standen, N. B., Gray, P. T. A., Whitaker, M. J. Ed. Microelectrode Techniques. The Com-
pany of Biologists Limited, Cambridge, UK, 1987.
General patch-clamp recording
Hamill, O. P., Marty, A., Neher, E., Sakmann, B., Sigworth, F. J. Improved patch-clamp
techniques for high-resolution current from cells and cell-free membrane patches. Pflügers
Arch. 391: 85–100, 1981.

Sakmann, B. and Neher, E. Ed. Single-Channel Recording. Plenum Press, New York, NY,
1983.

Smith, T. G., Jr. et al., op. cit.

Standen, N. B. et al., op. cit.

Patch-slice recording
Edwards, F. A., Konnerth, A., Sakmann, B., Takahashi, T. A thin slice preparation for
patch clamp recordings from neurons of the mammalian central nervous system. Pflügers
Arch. 414: 600–612, 1989.

Blanton, M. G., Lo Turco, J. J., Kriegstein, A. Whole cell recording from neurons in slices
of reptilian and mammalian cerebral cortex. J. Neurosci. Meth. 30: 203–210, 1989.
Vibration isolation methods
Newport Catalog. Newport Corporation, 2006.
Electrical isolation methods
Horowitz, P., Hill, W. The Art of Electronics. Cambridge, 1988.

Morrison, R. Grounding and Shielding Techniques in Instrumentation. John Wiley &

Sons, New York, NY, 1967.

26 The Axon Guide — 2500-0102 Rev. C

 3. Instrumentation for
Measuring Bioeletric Signals
from Cells1
There are several recording techniques that are used to measure bioelectric signals. These
techniques range from simple voltage amplification (extracellular recording) to sophisti-
cated closed-loop control using negative feedback (voltage clamping). The biggest chal-
lenges facing designers of recording instruments are to minimize the noise and to
maximize the speed of response. These tasks are made difficult by the high electrode resis-
tances and the presence of stray capacitances2. Today, most electrophysiological equip-
ment sport a bevy of complex controls to compensate electrode and preparation
capacitance and resistance, to eliminate offsets, to inject control currents and to modify
the circuit characteristics in order to produce low-noise, fast and accurate recordings.

3.1. EXTRACELLULAR RECORDING

The most straight-forward electrophysiological recording situation is extracellular record-
ing. In this mode, field potentials outside cells are amplified by an AC-coupled amplifier
to levels that are suitable for recording on a chart recorder or computer. The extracellular
signals are very small, arising from the flow of ionic current through extracellular fluid (see
Chapter 1, “Bioelectricity”). Since this saline fluid has low resistivity, and the currents are
small, the signals recorded in the vicinity of the recording electrode are themselves very
small, typically on the order of 10–500 μV.

The most important design criterion for extracellular amplifiers is low instrumentation
noise. Noise of less than 10 μV peak-to-peak (μVp-p) is desirable in the 10 kHz band-
width. At the same time, the input bias current3 of the amplifier should be low (< 1 nA)

1. In this chapter “Pipette” has been used for patch-clamp electrodes. “Micropipette” has been used for intracellular electrodes, except
where it is unconventional, such as single-electrode voltage clamp. “Electrode” has been used for bath electrodes. “Microelectrode” has
been used for extracellular electrodes.
2. Some level of capacitance exists between all conductive elements in a circuit. Where this capacitance is unintended in the circuit design,
it is referred to as stray capacitance. A typical example is the capacitance between the micropipette and its connecting wire to the
headstage enclosure and other proximal metal objects, such as the microscope objective. Stray capacitances are often of the order of a
few picofarads, but much larger or smaller values are possible. When the stray capacitances couple into high impedance points of the
circuit, such as the micropipette input, they can severely affect the circuit operation.
3. In the ideal operational amplifier (op amp), no current flows into the inputs. Similarly, in the ideal transistor, no current flows into
the gate. In practice, however, amplifying devices always have an input current. This current is commonly known as the input bias
current.

The Axon Guide — 2500-0102 Rev. C 27

 3. Instrumentation for Measuring Bioeletric Signals from Cells

so that electrodes do not polarize. Ideally, the amplifier will have built-in high-pass and
low-pass filters so that the experimenter can focus on the useful signal bandwidth.

3.1.1. SINGLE-CELL RECORDING

In single-cell extracellular recording, a fine-tipped microelectrode is advanced into the
preparation until a dominant extracellular signal is detected. This will generally be due to
the activity of one cell. The microelectrode may be made of metal, e.g., glass-insulated
platinum, or it may be a saline-filled glass micropipette.

While not specifically targeted at extracellular recording, the Axoclamp™ 900A and the
MultiClamp™ 700B amplifiers are particularly suitable for single-cell extracellular record-
ing if the extracellular electrode is a microelectrode of several megohms or more. The
input leakage current of these amplifiers is very low and headstages are designed to
directly accommodate the micropipette holder. If necessary, capacitance compensation
can be used to speed up the response. The x100 AC-coupled output signal of the
MultiClamp 700B amplifier, in particular, is useful for measuring small extracellular sig-
nals, because up to 2000x further amplification can be applied to the signal with the built-
in output gain. Thus the total amplification is 200,000x, allowing for easier measurement
of extracellular potentials of less than 1 mV.

The Axon CyberAmp® 380 programmable signal conditioner amplifier has special ultra
low-noise probes suitable for extracellular recording. These probes do not have the special
features of the Axoclamp 900A and the MultiClamp 700B amplifiers, but for low-resis-
tance electrodes, from tens of ohms up to a few hundred kilohms, these ultra low-noise
probes have superior noise characteristics. The AI 402 x50 probe for the CyberAmp 380
amplifier contributes less noise than the thermal noise of a 250 Ω resistor. Electrodes can
be connected directly to the CyberAmp 380 amplifier without using a separate low-noise
probe. In this case, the additional noise due to the amplifier will still be very low—less
than the thermal noise of a 5 kΩ resistor. If electrodes are directly connected to the main
instrument, there is always a risk of picking up line-frequency noise or noise from other
equipment. Using a probe located very close to the electrode greatly reduces the probabil-
ity that it will pick up extraneous noise.

3.1.2. MULTIPLE-CELL RECORDING

In multiple-cell extracellular recording, the goal is to record from many neurons simulta-
neously to study their concerted activity. Several microelectrodes are inserted into one
region of the preparation. Each electrode in the array must have its own amplifier and fil-
ters. If tens or hundreds of microelectrodes are used, special fabrication techniques are
required to produce integrated pre-amplifiers. If recording is required from up to 16 sites,
two CyberAmp 380 amplifiers can be alternately used with one 16-channel A/D system
such as the Digidata® 1440A digitizer.

28 The Axon Guide — 2500-0102 Rev. C

 3.2. Intracellular Recording Current Clamp

3.2. INTRACELLULAR RECORDING CURRENT CLAMP

3.2.1. VOLTAGE FOLLOWER
The traditional method for recording the cell interior potential is the current-clamp tech-
nique, also commonly known as “Bridge” recording, and occasionally as “voltage-fol-
lower” recording. The essence of the technique is the connection of a micropipette to a
unity gain buffer amplifier that has an input resistance many orders of magnitude greater
than that of the micropipette and the input resistance of the cell (Figure 3.1). The output
of the buffer amplifier follows the voltage at the tip of the electrode. The ideal buffer
amplifier draws no input bias current, therefore the current through the micropipette is
“clamped” at zero.

+1 A1

+1 A1
-

Figure 3.1: An ideal micropipette buffer amplifier.

In A the buffer (A1) is represented in block-diagram form as a unity-gain amplifier.

Note: The symbol +1 indicates a non-inverting input of gain x1. B shows how A1 is
built from an operational amplifier with unity feedback.

If a high-quality current injection circuit (current source) is connected to the input node,
all of the injected current flows down the micropipette and into the cell (see Figure 3.2).
The current source can be used to inject a pulse of current to stimulate the cell, a DC cur-
rent to depolarize or hyperpolarize the cell, or any variable waveform that the user intro-
duces into the control input of the current source.

The Axon Guide — 2500-0102 Rev. C 29

 3. Instrumentation for Measuring Bioeletric Signals from Cells

Vp + Vcmd +1 Vcmd
A2
+1
+
Ro Vcmd

I -

+1 A1 Vp
Vp

I Vm

Figure 3.2: A high-quality current source.

A high-quality current source can be made by adding a second amplifier to the buffer
amplifier circuit. The inputs to A2 are a command voltage, Vcmd, and the pipette voltage
(Vp) buffered by A1. The voltage across the output resistor, Ro, is equal to Vcmd regardless
of Vp. Thus the current through Ro is given exactly by I = Vcmd/Ro. If stray capacitances
are ignored, all of this current flows through the pipette into the cell, then out through the
cell membrane into the bath grounding electrode.

3.2.2. BRIDGE BALANCE

Can the intracellular potential be measured if the current injected down the micropipette
is a variable waveform? Without special compensation circuitry, the answer is no. The
variable current waveform causes a corresponding voltage drop across the micropipette. It
is too difficult to distinguish the intracellular potential from the changing voltage drop
across the micropipette. However, special compensation circuitry can be used to eliminate
the micropipette voltage drop from the recording. The essence of the technique is to gen-
erate a signal that is proportional to the product of the micropipette current and the
micropipette resistance. This signal is then subtracted from the buffer amplifier output
(Figure 3.3). This subtraction technique is commonly known as “Bridge Balance” because
in the early days of micropipette recording, a resistive circuit known as a “Wheatstone
Bridge” was used to achieve the subtraction. In all modern micropipette amplifiers, opera-
tional amplifier circuits are used to generate the subtraction, but the name has persisted.

30 The Axon Guide — 2500-0102 Rev. C

 3.2. Intracellular Recording Current Clamp

A
Vp +1
Vm
I ∝
( Vp = Vm+ IR p )
B

Vp

Vm

Figure 3.3: The “Bridge Balance” technique.

This technique is used to separate the membrane potential (Vm) from the total potential
(Vp) recorded by the micropipette. The technique is schematically represented in A. A dif-
ferential amplifier is used to subtract a scaled fraction of the current (I) from Vp. The scal-
ing factor is the micropipette resistance (Rp). The traces in B illustrate the operation of
the bridge circuit. When the current is stepped to a new value, there is a rapid voltage step
on Vp due to the ohmic voltage drop across the micropipette. Since the micropipette is
intracellular, changes in Vm are included in Vp. Thus the Vp trace shows an exponential
rise to a new potential followed by some membrane potential activity. The bridge ampli-
fier removes the instantaneous voltage step, leaving the Vm trace shown.

There are several ways to set the bridge balance. A commonly used technique is to apply
brief repetitive pulses of current to the micropipette while it is immersed in the prepara-
tion bath. The Bridge Balance control is advanced until the steady-state pulse response is
eliminated (Figure 3.4). At this point, the circuit is balanced and the micropipette resis-
tance can be read from the calibrated readout of the Bridge Balance control. The same
technique can even be used after the micropipette has penetrated the cell. Details of how
to achieve this are described in the Axoclamp 900A and MultiClamp 700B manuals.

The Axon Guide — 2500-0102 Rev. C 31

 3. Instrumentation for Measuring Bioeletric Signals from Cells

50 mV
1 ms

Figure 3.4: Illustration of bridge balancing while micropipette is extracellular.

All traces show the Vm output from the bridge-balance circuit. A 5 nA, 2 ms pulse is
applied to a 50 MΩ electrode. No bridge balance is used in the left trace, revealing the full
voltage drop across the micropipette. In the middle trace, optimum bridge balance is used
and the voltage drop across the micropipette is eliminated from the record. The transients
at the onset and finish of the step result from the finite bandwidth of the headstage ampli-
fier. In the right trace, the bridge-balance control is advanced too far; the voltage drop
across the micropipette is overcompensated and a reverse-polarity step appears.

It is important that the micropipette resistance be constant. Clearly, if the micropipette

resistance varies as the current flows there will be no unique setting of the Bridge Balance
control and the measurement will include an artifact due to the variable voltage drop
across the micropipette.

3.2.3. JUNCTION POTENTIALS

A second extraneous contributor to the measured voltage at the output of the micropi-
pette buffer amplifier is the sum of the junction potentials. Junction potentials occur
wherever dissimilar conductors are in contact. The largest junction potentials occur at the
liquid-metal junction formed where the wire from the amplifier input contacts the elec-
trolyte in the micropipette and at the liquid-liquid junction formed at the tip of the
micropipette. The sum of all of the junction potentials can be eliminated by introducing a
single DC potential of the opposite polarity. In micropipette experiments, the junction
potentials can often sum to a few hundred millivolts. The ±500 mV range of the offset
controls provided by all Axon Cellular Neuroscience microelectrode amplifiers is more
than sufficient to compensate for even the worst-case junction potentials.

3.2.4. TRACK
A special form of automatic offset compensation is available in many patch clamp ampli-
fiers, including the Axopatch™-1 and the Axopatch 200 amplifier series. In Axopatch
amplifiers this technique is called “Track.” In some instruments of other manufacturers it
is called “Search.” The Track circuit is used only during seal formation. During seal for-
mation the tip comes in and out of contact with the membrane and rapid shifts in the tip
potential of the headstage sometimes occur. These potential shifts lead to equivalent rapid
shifts in the current record, which may exceed the dynamic range of the headstage. Since

32 The Axon Guide — 2500-0102 Rev. C

 3.2. Intracellular Recording Current Clamp

the headstage itself saturates, AC coupling at the oscilloscope or any other form of exter-
nal DC offset removal will not be useful. Instead, the Track circuit dynamically adjusts the
offset potential of the micropipette itself, so that the current is forced to be zero on aver-
age. It is important that the Track circuit be switched off before recording of ionic cur-
rents commences since the Track circuit introduces the same type of distortion produced
by AC coupling.

3.2.5. CURRENT MONITOR

Current-clamp circuits include a monitor output proportional to the current waveform.
In some current-clamp amplifiers, the monitor output is merely a scaled version of the
control voltage driving the current-clamp circuit. As long as the micropipette resistance is
moderate and the current-clamp circuit output does not exceed its linear operating range,
this technique is accurate. However, if the micropipette resistance is very high and the cur-
rent-clamp circuitry saturates, the actual current through the micropipette will be less
than the expected current. This simple monitor output will not indicate the failure to pass
the expected current. A better technique is to actually measure the current by monitoring
the voltage drop across a resistor in series with the micropipette (Figure 3.5). This is a
superior technique used in all of Axon’s micropipette current-clamp amplifiers, including
the Axoclamp 900A and MultiClamp 700A and 700B.

+1 Vcmd
A2
+1

Ro +1
A3 VI
+1

+1 A1 Vp

C in
Cw

Im

Figure 3.5: Series current measurement.

The circuit in Figure 3.2 is extended by the addition of a differential amplifier (A3) to
measure the voltage drop across the current setting resistor, Ro. This voltage drop is pro-
portional to the total current (I) flowing through Ro. For high-frequency currents
(e.g., the transient current during a step change in the command potential), the current
through Ro is not quite identical to the membrane current (Im). There is an additional

The Axon Guide — 2500-0102 Rev. C 33

 3. Instrumentation for Measuring Bioeletric Signals from Cells

current that flows through the capacitance (Cw) of the wall of the micropipette and
another current that flows into the input capacitance (Cin) of the amplifier. Nevertheless,
if precautions are taken to minimize Cw (e.g., shallow bath immersion, Sylgard or mineral
oil coating) and Cin (e.g., bootstrapped buffer amplifier, capacitance neutralization), the
error current in these stray capacitances can be minimized and the output signal, VI, is a
good representation of the membrane current.

Another way to measure the true micropipette current is to use a separate circuit called a
“virtual ground” (Figure 3.6). Instead of connecting the bath ground electrode directly to
ground, it is connected to the virtual ground formed at the negative input of an inverting
operational amplifier. The negative input in this configuration is called a “virtual” ground
because the operational amplifier endeavors to keep the negative input equal to the
ground potential at the positive input. To achieve this goal, the operational amplifier out-
put must continuously pass a current through the feedback resistor (into the node at the
negative input) that is equal and opposite to the bath ground current. Thus the voltage at
the output of the operational amplifier is directly proportional to the bath current.

+1
Vcmd
A2
+1

+1 A1 Vp
I in

C in
Cw
Iw

Rf If

Im
-
A3 Vp
+

Figure 3.6: Virtual-ground current measurement.

The bath grounding electrode is connected to the negative input of operational amplifier
A3. The output (VI) continually adjusts to pass current through the feedback resistor (Rf )
equal and opposite to the total current flowing into the junction at the negative input of
A3. The potential at the negative input is “virtually” equal to ground at all times. Most of
the current measured by A3 is the membrane current, Im. However, during rapid potential
changes, transient current flows through the capacitance (Cw) of the wall of the micropi-

34 The Axon Guide — 2500-0102 Rev. C

 3.2. Intracellular Recording Current Clamp

pette and the input capacitance (Cin) of the amplifier. The current (Iw) through Cw flows
into the bath and is recorded by the virtual-ground circuit. The current (Iin) through Cin
flows into the system ground and is not recorded by the virtual-ground circuit.

Since the series current-measurement technique is much more convenient, it is routinely

used in preference to the virtual-ground current-measurement technique. The virtual-
ground technique is inconvenient for several reasons:

1 The input lead to the virtual ground is very sensitive and easily picks up line-
frequency and other interference.
2 It is extremely difficult to use two microelectrode amplifiers in the same preparation
bath if one or both of them uses a virtual-ground current-measurement circuit. If one
amplifier uses a virtual ground, the individual micropipette currents will be difficult
to identify because the virtual-ground current monitor measures the summed current
from all sources. If two amplifiers use a virtual ground, large DC-error currents will
flow between the virtual-ground circuits because of the small differences in offset
voltages between the two circuits. This is equivalent to the problem that would exist if
two voltage sources were connected in parallel.
3 The virtual-ground probe is just one more box of electronics that has to be mounted
in the crowded space around the preparation bath.
Nevertheless, the virtual ground technique is still used in some circumstances, most com-
monly with high-voltage two-electrode voltage-clamp amplifiers, because of the technical
difficulty of making a series current-measurement circuit operate at high voltages.

The Bath Error Potentials section below describes an important variation on the virtual-
ground design to eliminate the voltage error due to current flow through the grounding
electrode.

3.2.6. HEADSTAGE CURRENT GAIN

The HS-9A and HS-2A headstages used with the Axoclamp 900A and Axoclamp 2B
amplifiers all have unity voltage gain but come in a variety of current-passing capabilities.
The current-passing range and resolution are determined by the size of a single resistor in
the headstage. This resistor (Ro) has two functions. In addition to determining the size
and resolution of the current that can be passed, it also determines the sensitivity with
which current is measured. In the standard HS-9A headstage, Ro is 10 MΩ; this is arbi-
trarily assigned a current-passing gain of H = 1. To pass more current, a headstage with a
smaller value of Ro is required. The headstage with Ro = 1 MΩ can pass ten times as much
current as the standard headstage; it is assigned a current-passing gain of H = 10. The
headstage designed for ion-sensitive micropipettes can only pass very tiny currents. The
value of Ro in this headstage is 100 GΩ; the current-passing gain is H = 0.0001.

3.2.7. CAPACITANCE COMPENSATION

The high-frequency performance of the micropipette amplifier is compromised by the
presence of capacitance at the amplifier input. This capacitance comes from several

The Axon Guide — 2500-0102 Rev. C 35

 3. Instrumentation for Measuring Bioeletric Signals from Cells

sources: the capacitance across the glass wall (transmural capacitance) of the immersed
part of the micropipette; the stray capacitance from the rest of the micropipette to nearby
grounded surfaces; the stray capacitance from the micropipette holder; and the capaci-
tance to ground at the input of the buffer operational amplifier.

The input capacitance and the micropipette form a simple low-pass filter. That is, high-
frequency signals at the tip of the micropipette are shunted to ground by the input capac-
itance. There are several ways to improve the bandwidth of the recording system.

The best is to minimize the physical magnitude of the various elements contributing to
the input capacitance. The transmural capacitance of the immersed part of the micropi-
pette can be quite large. Typically, it is 1 pF or more per millimeter of immersion depth.
An effective way to reduce this capacitance is to thicken the wall of the micropipette. This
can be done by coating the micropipette with Sylgard or an equivalent material (see Chap-
ter 4). Other materials have been used, and in extreme cases, researchers have used a wide
micropipette to form a jacket over the narrower recording micropipette.

The other obvious way to reduce the transmural capacitance is to reduce the fluid level so
that the immersion depth is minimal. This is not always as effective as might be expected,
because surface tension causes the bath solution to creep up the surface of the micropi-
pette. This generates significant capacitance between the inside of the micropipette and
the film of solution on the outside. Forming this film can be prevented by making the sur-
face of the micropipette hydrophobic by dipping the filled micropipette into mineral oil
(or silane) immediately before using it. Note that it is essential to fill the micropipette first
with the aqueous electrolyte pipette solution so that the aqueous solution prevents the
mineral oil from entering the tip. Finally, one should try not to place the holder or the
microelectrode too close to grounded surfaces (such as the microscope objective).

Once the physical magnitude of the stray capacitance has been minimized, electrical tech-
niques can be used to reduce the effective magnitude. There are three such techniques:

1 Some researchers surround the micropipette with a metal shield that is connected to
the unity-gain output of the buffer amplifier. The principle here is to make sure that
the stray capacitance does not have a signal voltage across it. As long as there is no
signal voltage across the capacitance, no high-frequency current is drawn and the
bandwidth is increased. Unfortunately, this technique can significantly compromise
the noise performance of the amplifier because random noise generated in the buffer
amplifier is coupled back into its input via the shield capacitance. For this reason,
MDS Analytical Technologies does not recommend the use of a driven shield.
2 Unity-gain feedback can be used to reduce the component of stray capacitance that
exists between the amplifier input and its power supplies and case (Figure 3.7). In a
technique known as “bootstrapping,” sophisticated circuitry is used to superimpose
the unity-gain output of the buffer amplifier back onto its own power supplies and
the headstage case. This eliminates the high-frequency current loss through the power
supply capacitance and consequently increases the bandwidth. Since the power supply

36 The Axon Guide — 2500-0102 Rev. C

 3.2. Intracellular Recording Current Clamp

capacitance is present whether or not the power supply is bootstrapped, there is no

noise penalty due to implementing the technique. Bootstrapped power supplies are
implemented in the Axoclamp 900A and MultiClamp 700B amplifiers and
contribute to their excellent noise and bandwidth characteristics.

Vp + 10V
Vp +1 C+
A1 Vp
C-

Cext Vp - 10V
+1 A2 Vp

Figure 3.7: Bootstrapped power supplies.

Bootstrapped power supplies are used to increase the bandwidth by minimizing the
high-frequency current loss through the power-supply capacitance. The power-supply
capacitance is made up of C+ and C- from the positive input of A1 to the positive and
negative power supplies, respectively. All of the stray capacitances other than C+ and
C- are lumped together and called Cext in this figure. The power supplies of A1 are
batteries. The center point of the batteries is connected to a buffered version of Vp. If
the batteries are 10 V, the power supplies for A1 are (Vp + 10 V) and (Vp – 10 V). The
voltages across C+ and C- are constant at +10 V and -10 V, respectively. Hence, there
are no transient currents in neither C+ nor C-, except at very high frequencies where
these simple amplifier models are not appropriate. To ensure stability, the bandwidth
of A2 must be half or less of the bandwidth of A1. In practical implementations such
as the MultiClamp 700B and Axoclamp 900A microelectrode amplifiers, the batteries
are simulated by electronics.
3 Finally, but not least, the technique known as “capacitance compensation” or
“negative capacitance” can be used to reduce the effective value of whatever
capacitance remains. An amplifier at the output of the unity-gain buffer drives a
current-injection capacitor connected to the input (Figure 3.8). At the ideal setting of
the amplifier gain, the current injected by the injection capacitor is exactly equal to
the current that passes through the stray capacitance to ground. If the amplifier gain is
increased past the ideal setting, the current injected back into the input will cause the
input signal to overshoot. As the gain is increased past a certain setting, the circuit will
oscillate with potentially disastrous consequences for the cell.

The Axon Guide — 2500-0102 Rev. C 37

 3. Instrumentation for Measuring Bioeletric Signals from Cells

knVp
In A2 +k n

Cn
Ip
Vp
+1 A1 Vp
I in
C in
Rp

Vm

Figure 3.8: Capacitance neutralization circuit.

Amplifier A2 places a voltage proportional to the pipette potential (Vp) across the capaci-
tance-neutralization capacitor (Cn). At a particular setting of the gain (kn) of A2, the cur-
rent (In) through Cn is exactly equal to the current (Iin) through the stray capacitance
(Cin) at the input of A1. By conservation of current, the pipette current (Ip) is zero. In the
ideal case, the effective capacitance loading the pipette resistance (Rp) is zero and the
bandwidth is therefore very high. In practice, limitations due to the finite bandwidth of
the amplifiers, the distributed nature of Cin and other second-order effects limit the
recording bandwidth.

An important consideration in the design of a capacitance neutralization circuit is to avoid

introducing much additional noise from the circuitry itself. Low-noise amplifiers must be
used and the size of the injection capacitor must be small. Axon Cellular Neuroscience
HS-2 series headstages were produced in two varieties. The low-noise “L” series head-
stages used a 3 pF injection capacitor. The medium-noise “M” series headstages used a
10 pF injection capacitor. Generally, the M series headstages were only used for the cur-
rent-passing micropipette (ME2 on an Axoclamp 900A amplifier) in a two-micropipette
voltage-clamp setup. Noise in this micropipette is less important than noise in the voltage
recording micropipette (ME1), and it is sometimes useful to have a larger compensation
range on the current-passing micropipette.

With the Axoclamp 900A amplifier, the HS-9A headstage series was introduced. These
headstages have an increased capacitance neutralization range (-10 to 35.5 pF), indepen-
dent of the gain of the headstage. These headstages can be used equally effectively as volt-
age-recording or current-passing electrodes with the Axoclamp 900A amplifier. The
choice of HS-9A headstages for TEVC depends on the size of the typical current to be
measured.

A problem often encountered in TEVC is saturation of the amplifier output, because of

the limits imposed by the power supply ultimately supplying the current to the microelec-
trode. If the optimum gain of the capacitance neutralization amplifier is moderately high,

38 The Axon Guide — 2500-0102 Rev. C

 3.2. Intracellular Recording Current Clamp

(e.g., 2), the amplifier may saturate during large signals. In many experiments, large inputs
would be unusual, and even if they occurred the speed penalty incurred would not be
disastrous. However, in two-electrode voltage clamp experiments, large signal excursions
are common and speed is crucial. To eliminate the saturation of the capacitance compen-
sation circuit, the Axoclamp 900A amplifier uses ±180 V amplifiers in the capacitance
neutralization circuit in two-electrode voltage clamp mode.

If everything were ideal, the bandwidth of the recording system would approach the band-
width of the buffer amplifier alone (typically a few megahertz). However, in practice the
bandwidth is limited by the phase delays in the capacitance compensation pathway, and
by the false assumption that the stray capacitance can be represented by a lumped capaci-
tor at the input of the amplifier. The capacitance is distributed along the length of the
micropipette. Nevertheless, the techniques described above can be used to substantially
improve the bandwidth. It is not unusual to achieve bandwidths of 30 kHz with micropi-
pettes that would have a bandwidth of just one or two kilohertz if the above techniques
were not implemented.

3.2.8. LEAKAGE CURRENT

Ideally, a micropipette headstage should pass zero current into the micropipette when the
current command is at zero. In practice, there is a leakage current created from two
sources. The first is the inherent bias current of the operational amplifier’s input. This is
usually of the order of a few picoamps or less. The second is the current that flows
through the current injection resistor (Ro in Figure 3.2) because of voltage offsets in the
current-control circuitry. A trim potentiometer can be used to reduce this offset. In fact,
an optimal setting can be found where a small offset is left across Ro so that the resistor
current compensates the operational amplifier bias current and leaves a net zero current
through the micropipette. However, the leakage current will quickly grow again to unac-
ceptable levels if the offsets in the control circuitry change significantly with temperature
or time. To avoid this problem, MDS Analytical Technologies uses extremely high-quality,
low-drift operational amplifiers in its Axon Cellular Neuroscience micropipette amplifi-
ers.

3.2.9. HEADSTAGES FOR ION-SENSITIVE MICROELECTRODES

The most demanding application in terms of low-leakage current requirements is the
measurement of ion concentration using ion-sensitive microelectrodes (ISMs). In spite of
the current popularity about ion-sensitive intracellular dyes, ISMs are still the best means
available for determining extracellular and, in some cases, intracellular ion concentrations.
These electrodes can be very difficult to fabricate because of the efforts required to place
the highly hydrophobic ion-sensitive resin at the very tip of the electrode. Due to the
hydrophobic nature of the liquid resin, ISMs are also very sensitive to current passing into
the electrode. Even the tiniest amount of leakage current from the headstage can cause the
resin to move. If the resin leaks out of the electrode it can contaminate the preparation
and destroy cells in the vicinity of the electrode. If the resin moves up the barrel of the
pipette, the response of the electrode is slowed dramatically or destroyed.

The Axon Guide — 2500-0102 Rev. C 39

 3. Instrumentation for Measuring Bioeletric Signals from Cells

There are two ways to ensure low-leakage currents for these microelectrodes: First, use
special operational amplifiers that have bias currents of just a few tens of femtoamps. Sec-
ond, make the value of Ro very large (100 GΩ) so that the current induced by small offset
voltages across Ro is small.

3.2.10. BATH ERROR POTENTIALS

In most experiments, the bathing solution is grounded by a solid grounding electrode
(such as anAg/AgCl pellet in the bath, or which can be connected to the bath via an agar/
KCL bridge) and all measurements are made relative to ground, on the assumption that
the bath is also at ground. This assumption may not be true in experiments in which the
Cl- concentration or temperature of the bathing solution is significantly changed, or
where the membrane current is sufficiently large to cause a significant voltage drop across
the resistance of the grounding electrode. The latter circumstance would normally occur
only when voltage clamping very large cells such as frog oocytes, in which case the ionic
current may be of the order of several microamps or even several tens of microamps.

Depending upon the grounding method, the resistance of the bath grounding electrode
(Rb) could be as much as 10 kΩ, although with care it is not difficult to achieve Rb values
less than 1 kΩ.

In a simple two-electrode voltage clamp (TEVC) setup, the voltage drop across Rb is
indistinguishable from the membrane potential. That is, the potential (V1) recorded by
the voltage-recording micropipette is the sum of the transmembrane potential (Vm) and
the bath potential (Vb). Problems arise if the product of the clamp current (I2) and Rb is
significant. For example, for I2 = 5 μA and Rb = 2 kΩ, the error voltage is 10 mV. In some
experiments, a worst-case error of this magnitude might be tolerable; but if the error were
to be much greater, the position of the peak of I-V curves and other responses would be
seriously affected.

To faithfully record Vm, either Vb must be made equal to or nearly equal to zero, or the
value of Vb must be independently measured and subtracted from the potential recorded
by ME1. There are several methods to choose from:

1 The best method is to minimize Rb in the first place.

2 Electronically predict and subtract the error voltage. This technique is known as
“series-resistance compensation.”
3 Use an independent electrode to measure Vb and subtract this value from V1.
4 Set up an independent clamping circuit to clamp Vb, measured near the surface of the
cell, to zero.
5 Set up an independent virtual-ground circuit to clamp Vb, measured near the surface
of the cell, to zero, while simultaneously measuring the bath current.

40 The Axon Guide — 2500-0102 Rev. C

 3.2. Intracellular Recording Current Clamp

Minimizing Rb
There are three main contributors to Rb:

1 The cell access resistance from the membrane surface to the bath.
2 The resistance of an agar bridge, if used.
3 The resistance of the grounding wire.
Cell Access Resistance
The access resistance to a sphere is given by:

ρ
Ra = (1)
4πr

where ρ is the resistivity (typically 80 Ωcm for Ringer’s solution at 20 °C [Hille,

1984]) and r is the radius of the sphere in centimeters.

Modeling an oocyte as a 1 mm diameter sphere:

Ra = 127 Ω

Resistance of an Agar Bridge

The resistance of a cylindrical agar bridge is given by (Hille, 1984):

ρ × length (cm)
Ragar = (2)
area (cm 2 )

where length refers to the length of the agar bridge and area is the cross-sectional area
of the agar bridge.

For a 1 cm-long agar bridge with a 2 mm internal diameter (ID) filled with Ringer’s
solution:

Ragar = 2.6 kΩ

For an agar bridge 1 cm long x 1 mm ID, filled with Ringer’s solution:

Ragar = 10.2 kΩ

The Axon Guide — 2500-0102 Rev. C 41

 3. Instrumentation for Measuring Bioeletric Signals from Cells

The resistivity of 3 M KCl is approximately 20 times lower than for Ringer’s solution.
Thus, for an agar bridge filled with 3 M KCl, having the same length as above:

Ragar = 130 Ω if the ID is 2mm

Ragar = 510 Ω if the ID is 1mm

Resistance of Grounding Wire

When immersed in 0.9% saline, the impedance of a 1 mm diameter Ag/AgCl pellet
is in the range of 300–600 Ω, depending on how much of the surface is in contact
with the saline.4
Conclusion
The access resistance to the cell is quite small and beyond the control of the experi-
menter. The other major contributors to Rb are the resistance of the agar bridge and
the resistance of the grounding electrode.

To minimize Rb, it would be best to eliminate the agar bridge and ground the prepa-
ration directly with an Ag/AgCl pellet. The pellet should be as large as practical, and
the area of contact between the pellet and the solution should be maximized.

However, if the bathing solution is changed during the experiment, the DC offset of
the Ag/AgCl pellet will change with the chloride activity. Thus, in this case it is essen-
tial to either use an agar bridge to prevent the DC offset of the bath from changing,
or to use a bath-potential recording headstage with a 3 M KCl electrode to follow and
remove the shift. Another advantage of an agar bridge is that it prevents metal ions
from the grounding electrode from entering the bathing solution.

If an agar bridge is used, it is best to fill it with 3 M KCl instead of Ringer’s solution
in order to minimize Rb. When the agar bridge is filled with 3 M KCl, the sum of all
components of Rb will be approximately 1–2 kΩ. If leakage of KCl from the agar
bridge is a problem, it may be necessary to fill the agar bridge with Ringer’s solution.
In this case, Rb will be several kilohms or more.
Series-Resistance Compensation
This sophisticated technique is used to minimize the effects of the electrode resistance in a
whole-cell voltage clamp. It is described later in this chapter and is of paramount impor-
tance when there is no alternative available. However, series-resistance compensation is
never completely effective. The other methods discussed in this section are more reliable
and easier to apply when it comes to eliminating the voltage error across Rb.

4. Measurements made at Axon Instruments, Inc. using a 700 Hz signal source.

42 The Axon Guide — 2500-0102 Rev. C

 3.2. Intracellular Recording Current Clamp

Measure and Subtract Vb

Another technique used to eliminate the effects of the IR voltage drop across the bath
resistance is to actively measure the bath potential outside the cell.

In this method, the bath is grounded by any preferred method. The bath potential is
allowed to shift because of solution changes, temperature changes or current flow. The
changing values of the bath potential (Vb) are continuously monitored by a 3 M KCl-
filled electrode placed near the cell. The measured value of Vb is subtracted from the
potentials measured by the intracellular microelectrodes. For microelectrodes being used
for voltage recording, this subtraction yields the true transmembrane potential (Vm). The
signal provided to all subsequent circuits in the amplifier is Vm.

This method operates well. The minor inconveniences are the more complex electronics,
the noise of the bath electrode that is added to the noise of the intracellular recording
micropipette, and some (minor) potential for instability in a two-electrode voltage clamp
if the bandwidth of the bath potential headstage is greater than the bandwidth of the
intracellular recording micropipette.
Clamp Vb Using a Bath Clamp
Another means to eliminate the effect of the voltage drop across Rb is to actively control
the bath potential measured near the outside surface of the cell. This is achieved using a
two-electrode virtual-ground circuit (Figure 3.9). One electrode (SENSE) is a voltage-
sensing electrode. This electrode is placed in the preparation bath near the cell surface. It
is connected to the virtual-ground circuit by an agar bridge or other means, of resistance
Rb2. Since there is no current flow through this electrode, there is no voltage drop across
Rb2. The other electrode (IBATH) is placed in the preparation bath. This electrode carries
the ionic current. The feedback action of the operational amplifier ensures that the poten-
tial at the SENSE electrode is equal to the potential at the negative amplifier input (ignor-
ing the offset voltage at the inputs to the operational amplifier).

SOLUTION
CELL
R b1 I BATH Rf
I BATH

SENSE
R b2
SENSE
- V1
+

VG-2A x100

Figure 3.9: Two-electrode virtual-ground circuit.

In this implementation, separate electrodes are used to pass the bath current and control
the bath potential (unlike Figure 3.6 above). The basic operation is the same as for the vir-

The Axon Guide — 2500-0102 Rev. C 43

 3. Instrumentation for Measuring Bioeletric Signals from Cells

tual-ground circuit described earlier, except that instead of connecting the feedback resis-
tor directly to the negative input of the operational amplifier, it is connected indirectly via
the bath solution. The bath potential near the surface of the cell is recorded by the SENSE
electrode (resistance Rb2) and forced by the feedback action of the operational amplifier to
be near ground. The bath current required to achieve this is passed by the operational
amplifier through the feedback resistor (Rf ) and through the bath current electrode
(IBATH; resistance Rb1).
Clamp Vb Using a Virtual-Ground Current Monitor
This technique is identical to the bath clamp described above, except that in this case
instead of being ignored, the output of the circuit is used to monitor the current flowing
from the micropipettes into the bath. The output (VI) of this circuit is proportional to the
bath current.

V I = ( Rb1 + R f ) I BATH (3)

There are several problems with this technique. First, if the output is used to record the
bath current, it is usual to ignore the Rb1 term (because its value is unknown) and to
assume that

V I = R f I BATH (4)

Thus, there is a small error equal to Rb1/(Rb1 + Rf ). Second, if there are fluid-filled tubes
connected to the bath, and if some of these run outside of the Faraday cage, they will act
as antennas and conduct a lot of hum current into the bath that will be recorded by the
virtual-ground circuit. Third, if it is desired to use more than one amplifier with the same
preparation, it is essential that no more than one of them uses a virtual-ground circuit to
clamp the bath potential. This is not a serious problem, since it is easy to make one vir-
tual-ground serve the needs of both amplifiers. The problem is serious, however, if one of
the amplifiers introduces command signals via the virtual-ground bath clamp. This is a
practice employed by some manufacturers because in some circumstances it is easier to
design the electronics for this mode of operation. However, if command potentials are
introduced via the bath electrode, it is extremely difficult to perform experiments using
two amplifiers, e.g., whole-cell patch clamp of two fused cells.

When used with consideration for the measurement error, and if hum pick up is avoided,
a virtual-ground circuit to measure IBATH offers excellent performance.
Summary
As a first priority, we recommend that the value of Rb be minimized by the techniques
described above. If it is still considered to be important to compensate for the IR voltage
drop, a virtual-ground headstage (such as the VG-9A series) can be used to clamp the bath

44 The Axon Guide — 2500-0102 Rev. C

 3.2. Intracellular Recording Current Clamp

potential, or a unity-gain voltage follower headstage (such as the HS-9A and HS-2A
series) can be used to record and subtract the bath potential.

3.2.11. CELL PENETRATION: MECHANICAL VIBRATION, BUZZ AND CLEAR

Often when a micropipette is pressed against a cell, the membrane dimples but the
micropipette fails to penetrate. A variety of methods are used to overcome this problem.
The oldest and frequently effective trick is to tap the micromanipulator gently. Since this
“method” is difficult to calibrate and reproduce, two electronic alternatives are widely
used. One is to drive a brief, high-frequency oscillatory current through the micropipette.
This is often done by briefly increasing the capacitance compensation such that the com-
pensation circuit oscillates. On the Axoclamp amplifiers, a dedicated feature called "Buzz"
is provided for this purpose. On the Axoclamp 900A amplifier, this feature has a duration
control setting; on the Axoclamp 2B amplifier, buzz duration can also be controlled via its
remote buzz accessory unit. The duration used should be long enough for penetration,
but short enough to prevent damage to the cell. The mechanism of penetration is
unknown, but it may involve attraction between the charge at the tip of the micropipette
and bound charges on the inside of the membrane.

The second electronic alternative is to drive a large positive or negative current step into
the cell.

In the Axoclamp 900A amplifier, both the duration and height of the current pulse are
adjustable. Again, the mechanism of penetration is not known. One cannot even predict
a priori whether a positive or a negative pulse will be more effective.

None of the three methods described here can be described as “best.” For a particular cell
type, the researcher must experiment to find the method that works best with those cells.

3.2.12. COMMAND GENERATION

In general, current and voltage commands arise from a variety of sources; internal oscilla-
tors in the instrument, internal DC current or voltage controls, external input from a
computer, etc. These commands are usually summed in the instrument so that, for exam-
ple, an externally generated step command can be superimposed on the internally gener-
ated holding potential.

An important consideration when using externally generated commands is the attenua-

tion provided by the instrument’s command input. Manufacturers attenuate the external
command so that the effects of noise and offset in the external signal generator are mini-
mized. Noise is particularly common on the analog output of most computer interfaces.
To illustrate the problem, imagine a micropipette amplifier used for current injection into
cells that has a current passing capacity of ±100 nA maximum. Compare the situation for
two different values of the command input attenuation. In the first case, the amplifier
passes 1 nA for each millivolt from the computer. If there is a 2 mV of wide-band digital
noise on the output of the D/A converter, this will generate 2 nA of noise in the recording
a substantial fraction of the usable range. Further, if the D/A converter had a 3 mV offset,

The Axon Guide — 2500-0102 Rev. C 45

 3. Instrumentation for Measuring Bioeletric Signals from Cells

then instead of a command of zero generating 0 nA, it would generate 3 nA. In the second
case, the amplifier passes 1 nA for each 100 mV from the computer. This time, the 2 nA
of noise from the D/A converter only generates 20 pA of noise in the recording, and the
3 mV offset error only generates a 30 pA current error. The full ±100 nA range of the
amplifier can still be controlled, since most D/A converters produce up to ±10 V.

All of the Axon Cellular Neuroscience current-clamp and voltage-clamp instruments from
MDS Analytical Technologies use the maximum possible attenuation so that the
±10.24 V output from the D/A converter corresponds to the normal maximum operating
limits of the instrument.

Another important consideration when using externally generated commands is the pres-
ence or absence of polarity inversion. To minimize the risk of making mistakes during an
experiment, it is best that the instrument does not invert the command signal. Thus the
user should always know that a positive command from the computer will generate a pos-
itive current or voltage in the instrument.

3.3. INTRACELLULAR RECORDING—VOLTAGE CLAMP

In the voltage-clamp technique, the membrane potential is held constant (i.e., “clamped”)
while the current flowing through the membrane is measured. Usually, the investigator
has no inherent interest in the membrane current, but is interested in the membrane con-
ductance, since conductance is directly proportional to the ion-channel activity. Current
is measured because the investigator has no direct way of measuring conductance. By
holding the membrane potential constant (or at the very least, constant after a rapid step),
the investigator ensures that the current is linearly proportional to the conductance being
studied.

Micropipette voltage-clamp techniques for whole-cell current measurement traditionally

assign the role of potential measurement and current passing to two different intracellular
micropipettes. This is done using the conventional two-electrode voltage-clamp tech-
nique, or it is simulated by time-division multiplexing (time-sharing) using the discontin-
uous single-electrode voltage-clamp technique. Separation of the micropipette roles by
either of these methods avoids the introduction of errors in the measurement due to
unknown and unstable voltage drops across the series resistance of the current-passing
micropipette.

The so-called “whole-cell patch” voltage clamp is a technique wherein only one micropi-
pette is used full-time for both voltage recording and current passing. A method known as
series-resistance compensation (see Section 3.3.7., “Series-Resistance Compensation”)
attempts to eliminate the aberrations arising when one micropipette is made to perform
the work of two. For series-resistance compensation to work, the micropipette resistance
must be reasonably low compared to the impedance of the cell. Since the resistance of
intracellular micropipettes generally is too high, researchers use very blunt micropipettes
that can be tightly sealed to a patch of membrane. The patch of membrane is subse-

46 The Axon Guide — 2500-0102 Rev. C

 3.3. Intracellular Recording—Voltage Clamp

quently ruptured so that the micropipette provides a low-resistance access to the whole
cell, hence the term “whole-cell patch” clamp.

3.3.1. THE IDEAL VOLTAGE CLAMP

The ideal voltage clamp simply consists of a battery, a switch, a wire, the cell and an
ammeter (Figure 3.10). Since the wire has zero resistance when the switch is closed, the
membrane potential steps instantly to the battery voltage. This generates an impulse of
current that charges the membrane capacitance, followed by a steady-state current to sus-
tain the voltage across the membrane resistance.

Vm

Vcmd Rm Cm

Figure 3.10: The ideal voltage clamp.

The ideal voltage clamp simply consists of a battery, a switch, a wire, the cell and an
ammeter. When the switch closes, the membrane potential steps instantly to the battery
voltage. (For simplicity, it is assumed that Vm = 0 before the step.) There is an impulse of
current injecting a charge Q = CmVcmd; the steady-state current is Vcmd/Rm.

In this experiment, the voltage is the independent variable. Its value is controlled and
equal to the battery value. The current is the dependent variable; its value is measured by
an ammeter. For this reason, the voltage-clamp circuit is sometimes called a “current fol-
lower.”

3.3.2. REAL VOLTAGE CLAMPS

There are many reasons for real voltage clamps to be inferior to the ideal voltage clamp.
These include finite resistance of the current-passing micropipette, limited bandwidth of
the voltage-recording micropipette, capacitive coupling between the two micropipettes,
limited bandwidth of the clamp amplifier, and non-ideal phase shifts in the membrane.
These phenomena and the minimization of their impact are discussed in the following
sections.

The Axon Guide — 2500-0102 Rev. C 47

 3. Instrumentation for Measuring Bioeletric Signals from Cells

3.3.3. LARGE CELLS—TWO-ELECTRODE VOLTAGE CLAMP

The theory of two-electrode voltage clamping is discussed in detail by Finkel and Gage
(1985). Without attempting to duplicate that discussion, some of the most important
conclusions will be presented here.

A conventional two-electrode voltage clamp is shown in Figure 3.11. This figure has been
simplified by ignoring several frequency-dependent components: the input capacitance at
the input of micropipette ME1, the coupling capacitance between the micropipettes, and
the special phase lead and lag circuitry that is often introduced deliberately into the elec-
tronics. Furthermore, series resistance in the membrane and the bathing solution has been
ignored. In the figure, the sole frequency-dependent component is the membrane capaci-
tance.

A Vcmd
+μ
ε A2
+1 A1 -μ

R p1 R p2

ME1 ME2

B
Vm
Vm
Rm Cm R p2 Cm
τ= μ

Im

Figure 3.11: Conventional two-electrode voltage clamp.

A. The membrane potential (Vm) is recorded by a unity-gain buffer amplifier (A1) con-
nected to the voltage-recording microelectrode (ME1). Vm is compared to the
command potential (Vcmd) in a high-gain differential amplifier (A2; gain = μ). The
output of A2 is proportional to the difference (ε) between Vm and Vcmd. The volt-
age at the output of A2 forces current to flow through the current-passing micro-
electrode (ME2) into the cell. The polarity of the gain in A2 is such that the current
in ME2 reduces ε.
B. Unlike the ideal case, there is a finite time required to charge the cell capacitance.
The time constant for the current and potential transients is τ = Rp2Cm/μ, where
Rp2 is the resistance of ME2. If was infinite, or if Rp2 was zero, the response would
approach the ideal case.

48 The Axon Guide — 2500-0102 Rev. C

 3.3. Intracellular Recording—Voltage Clamp

Error
The steady-state membrane potential (Vm) after a step change in the command voltage
(Vcmd) is:

μK
Vm = Vcmd (5)
μK + 1

where μ is the gain of the clamp amplifier and K is the attenuation of the clamp amplifier
caused by the cell membrane resistance (Rm) and the resistance (Rp2) of the output
micropipette (ME2):

Rm
K= (6)
Rm + R p 2

As the product μK becomes very large, the difference between Vm and Vcmd becomes very
small. Ideally, the error will be very low: just a fraction of one percent. The gain of the
clamp μ is typically set by a front-panel gain control in the range of about 100 to 10,000.
If K were unity, the error would vary from 1 percent down to 0.01 percent. However, if K
is less than unity (as it always is), the error will be worse. If the output micropipette resis-
tance is 90 MΩ and the membrane resistance is 10 MΩ, K is 0.1 and the error will be ten
times worse than if K were unity. If the two resistances are equal, K will be 0.5. Thus, as a
rule of thumb it is desirable to use an output micropipette whose resistance is as low as
possible, ideally about the same size or smaller than the membrane resistance.
Step Response and Bandwidth
After a step command, the membrane potential relaxes exponentially towards its new
value. For μK >> 1, the time constant for the relaxation is:

R p 2C m (7)
τ=
μ

Increasing the clamp gain decreases the time constant for the step response. For example,
if Rp2 = 10 MΩ, Cm = 1000 pF and μ = 100, the time constant is 100 μs.

Stated differently, increasing the clamp gain also increases the bandwidth with which Vm
can follow changes in Vcmd. The -3 dB frequency of the bandwidth is:

μ
f −3 = (8)
2πR p 2C m

The Axon Guide — 2500-0102 Rev. C 49

 3. Instrumentation for Measuring Bioeletric Signals from Cells

Stability
The voltage clamp circuit shown in Figure 3.12 is unconditionally stable. The membrane
capacitance provides a 90° phase shift, which is required for stability in all negative feed-
back circuits. Unfortunately, other factors combine to make the circuit unstable at high
clamp gains.

The coupling capacitance (Cx) between the micropipettes is extremely destabilizing. Val-
ues as small as 0.01 pF can lead to oscillation if μ has a magnitude of several hundred or
more. There are several ways to minimize this coupling capacitance. The two best ways
are by:

1 Introducing the two micropipettes into the preparation at a wide angle, preferably
greater than 90°.
2 Placing a grounded metal shield between the two micropipettes. This shield should be
large enough to block all line-of-sight pathways between the two micropipettes and
their holders.
Another destabilizing factor is the non-ideal nature of the membrane. In Figure 3.11, the
membrane is simply modeled as a parallel resistor and capacitor. In practice, a distributed
model applies. The capacitance elements are themselves non-ideal; they should be mod-
eled by an ideal capacitor with a series-resistance component. For real membranes, the
phase shift at high frequencies is less than 90°. In the Axoclamp 900A amplifier, a phase-
shift control is included to allow the user to empirically introduce a phase lag to the cir-
cuit to build the total high-frequency phase shift up to 90°.

The input capacitance of the voltage-recording micropipette (ME1) adds another fre-
quency-dependent variable into the system that also tends to decrease the stability. The
effect of this input capacitance is usually minimized by carefully adjusting the capacitance
neutralization control to maximize the bandwidth of ME1.

Small instabilities in the voltage clamp usually show up as an overshoot in the current and
voltage responses. Full instability shows up as a continuous oscillation that can destroy the
cell. In some voltage clamps, special circuits are used to detect oscillations and take appro-
priate action to avoid damage to the cell. The Axoclamp 2 series and the GeneClamp®
500 amplifiers do not include an oscillation guard circuit. The Axoclamp 900A amplifier
includes an oscillation detection feature that automatically reduces the voltage clamp gain,
protecting the cells from harm.
Membrane Conductance Changes
It can be shown that the current response to a step change in the membrane conductance
is identical to the membrane potential response to a step change in the command voltage.
This is fortunate, because it means that if the voltage clamp is set up optimally by observ-
ing the response to a repetitive command step, it is also optimally set up for measuring
conductance changes. In order for this equivalence to be maintained, it is essential that
the experimenter not use tricks such as filtering the command voltage to eliminate an

50 The Axon Guide — 2500-0102 Rev. C

 3.3. Intracellular Recording—Voltage Clamp

overshoot in the response. This is a bad practice because it disguises the dynamic perfor-
mance of the clamp circuit.
Noise
There are several sources of noise in a two-electrode voltage-clamp circuit, but the most
important is the thermal and excess noise of the input micropipette, ME1. To the voltage-
clamp circuit, this noise appears as an extra command signal. It is imposed across the
membrane and leads to a noise current. Because of the membrane capacitance, this noise
current increases directly with frequency in the bandwidths of interest. The magnitude of
the noise is worst in cells with large Cm values.

Since the current noise increases with frequency, a single-pole filter is inadequate. An
active two-pole filter is required at minimum; while a four-pole filter is preferred. The
Axoclamp 900A amplifier includes a four-pole filter with a corner frequency up to 30 kHz
on each output. The user can choose between a Bessel and a Butterworth filter.
Micropipette Selection
Ideally, both micropipettes should have very small resistances. The resistance of ME1
should be small to minimize the micropipette noise, and the resistance of ME2 should be
small to maximize the product μK. However, low-resistance micropipettes are more blunt
than high-resistance micropipettes and do more damage to the cell. In general, it is more
important to use a low resistance micropipette for ME2. If its value is high, the amplifier’s
power supply will saturate at a lower current, causing a voltage-clamp error because the
ME2 headstage cannot inject enough current to hold the membrane voltage “clamped.”
As a result, the recorded data will be of dubious merit. A relatively high-resistance
micropipette for ME1 has its problems—increased noise and reduced bandwidth—but at
least it does not introduce an error.
Input vs. Output Offset Removal
Certain offset-removal circuits act only on the output signal delivered to the oscilloscope
or data acquisition system. Examples of these are the AC input of the oscilloscope itself,
the DC-offset removal in the CyberAmp 380 amplifier, and the output offset controls in
the Axoclamp 900A. Adjusting these controls does not affect the membrane currents
injected through the micropipette.

On the other hand, some offset-removal circuits act on the input signal seen by the volt-
age-clamp circuit. For example, the micropipette-offset controls alter the potential that is
clamped by the voltage-clamp circuit. Thus the actual membrane potential is affected
when these controls are altered and, consequently, the membrane current is affected. Care
should be taken not to alter the setting of an input-offset control once the experiment has
begun.

The Axon Guide — 2500-0102 Rev. C 51

 3. Instrumentation for Measuring Bioeletric Signals from Cells

3.3.4. SMALL CELLS—DISCONTINUOUS SINGLE-ELECTRODE VOLTAGE CLAMP

In discontinuous single-electrode voltage clamp (dSEVC) the tasks of voltage recording
and current passing are allocated to the same micropipette. Time-sharing techniques are
used to prevent interactions between the two tasks. The principles of operation are dis-
cussed in detail by Finkel and Redman (1985).

A block diagram and a timing diagram illustrating the technique are shown in
Figure 3.12. A single micropipette (ME1) penetrates the cell and the voltage recorded
(Vp) is buffered by a unity-gain headstage (A1). To begin the discussion, assume that Vp is
exactly equal to the instantaneous membrane potential (Vm). A sample-and-hold circuit
(SH1) samples Vm and holds the recorded value (Vms) for the rest of the cycle.

S1
voltage recording
0 volts
GT
CCS

Io current passing

Vms
+1 A1 SH1 - S2
Vp ε A2
ME1 + voltage current
clamp clamp
I cmd
Vcmd

Vm

T
current passing

S1 voltage recording

τp

Vp
sample
sample
Vm

Vcmd
Vms
Vms (average)

Figure 3.12: Block diagram and timing waveforms.

52 The Axon Guide — 2500-0102 Rev. C

 3.3. Intracellular Recording—Voltage Clamp

Vms is compared with a command voltage (Vcmd) in a differential amplifier (A2). The
output of this amplifier becomes the input of a controlled-current source (CCS) if the
switch S1 is in the current-passing position.

The CCS injects a current into the micropipette that is directly proportional to the volt-
age at the input of the CCS irrespective of the resistance of the micropipette. The gain of
this transconductance circuit is GT.

The period of current injection is illustrated at the start of the timing waveform. S1 is
shown in the current-passing position during which a square pulse of current is injected
into the micropipette, causing a rise in Vp.

The rate of rise is limited by the parasitic effects of the capacitance through the wall of the
glass micropipette to the solution and the capacitance at the input of the buffer amplifier.
The final value of Vp mostly consists of the IR voltage drop across the micropipette due to
the passage of current Io through the micropipette resistance Rp. Only a tiny fraction of
Vp consists of the membrane potential (Vm) recorded at the tip.

S1 then switches to the voltage-recording position. When the input of the CCS is 0 volts,
its output current is zero and Vp passively decays. During the voltage-recording period Vp
decays asymptotically towards Vm. Sufficient time must be allowed for Vp to reach within
a millivolt or less of Vm. This requires a period of up to nine micropipette time constants
(τp). At the end of the voltage-recording period, a new sample of Vm is taken and a new
cycle begins.

The actual voltage used for recording purposes is Vms. As illustrated in the bottom timing
waveform, Vms moves in small increments about the average value. The difference
between Vms,avg and Vcmd is the steady-state error (ε) of the clamp that arises because the
gain (GT) of the CCS is finite. The error becomes progressively smaller as GT is increased.
Minimum Sampling Rate and Maximum Gain
If the sampling rate (fs) is too slow, the dSEVC will become unstable. This is because the
resultant long current-passing period (Ti) allows the membrane potential to charge right
through and past the desired potential before the clamp has had an opportunity to take a
new sample of potential and adjust the current accordingly. The larger the cell membrane
capacitance (Cm) the larger the value of Ti (and hence the slower the sampling rate) that
can be used for a given average gain (GT). The stability criterion is:

GT Ti
0< <2 (9)
Cm

The Axon Guide — 2500-0102 Rev. C 53

 3. Instrumentation for Measuring Bioeletric Signals from Cells

For optimum response (critical damping) we require:

GT Ti
=1 (10)
Cm

Thus for a given GT, if Cm is small, Ti must be small (i.e., fs must be large).

For example, if GT = 1 nA/mV and Cm = 100 pF, then Ti must be 100 μs for optimum
response. If Ti is greater than 100 μs, the step response will overshoot and at 200 μs the
clamp will oscillate. (Note that in Axoclamp amplifiers, Ti = 100 μs corresponds to a sam-
pling rate of 3 kHz because the duty cycle (discussed later) is 30%.)

If Ti in this example cannot be as short as 100 μs because the micropipette response is too
slow, then a lower value of GT will have to be used to maintain optimum response.
Fastest Step Response
In a critically damped system, that is, where the gain has been turned up to achieve the
fastest possible response with no overshoot, the clamp can settle to its final value in as lit-
tle as one or two periods.
Error Number One—Clamping the Micropipette
If the micropipette voltage does not have sufficient time to decay to zero before a new
sample of Vm is taken, the sample will include a micropipette artifact voltage. The clamp
circuit will be incapable of distinguishing this micropipette artifact voltage from the mem-
brane potential and thus it will voltage clamp the sum of the two potentials. This leads to
inaccurate and bizarre data collection. For example, cells with known rectifiers will gener-
ate I-V curves that are almost linear.

This error is easily avoided by carefully observing the monitor output on the Axoclamp
900A amplifier. It is essential that the micropipette voltage trace on this output be given
sufficient time to decay to zero before the sample is taken. The user can achieve this by
adjusting the sampling frequency and by optimizing the micropipette capacitance com-
pensation. With the Axoclamp 900A amplifier, the micropipette voltage monitor is trans-
mitted to the computer through a USB 2.0 connection. When dSEVC is activated, a
window displaying the micropipette voltage is automatically displayed. This facilitates the
experimenter in observing the micropipette voltage and avoiding clamp errors.
Error Number Two—Steady-State Clamp Error
Even if the sampling is such that there is negligible voltage drop across the micropipette,
there will still be a steady-state error due to the finite gain, GT, of the clamp. To minimize
the error, it is necessary to use as high a gain setting as possible, consistent with stable
operation.

54 The Axon Guide — 2500-0102 Rev. C

 3.3. Intracellular Recording—Voltage Clamp

Ripple in the Membrane Potential

Due to the switching nature of the dSEVC circuit, the actual membrane potential changes
during each cycle. The size of the ripple is proportional to the command voltage; for a
clamp at the resting membrane potential, the ripple is zero. Typically, the ripple is no
more than one or two percent of the command voltage. The actual error between Vm and
Vcmd is worst at the end of each cycle, just when the sample is taken. Therefore, the pres-
ence of ripple means that the average value of Vm is closer to Vcmd than is suggested by
the sampled voltages.
Aliased Noise and the Anti-Aliasing Filter
The dSEVC is a sampling system, and thus it suffers from the noise amplification prop-
erty called “aliasing” (see Chapter 11). The only way to prevent aliasing is to sample two
or more times more rapidly than the bandwidth of the signal. For example, if the micropi-
pette has a 20 kHz bandwidth, the sample rate should be 40 kHz or more to prevent alias-
ing. Unfortunately, in the dSEVC there is a competing consideration: the sample period
(T) should be at least ten times longer than the micropipette time constant. For a 20 kHz
micropipette, the time constant is about 8 μs; therefore the sampling period should not be
less than 80 μs, equivalent to a 12.5 kHz sampling rate.

The anti-alias filter is a special filter used in the Axoclamp 2-series amplifiers to reduce the
bandwidth of the micropipette. Ordinarily, this is not desirable, because the sampling rate
has to be reduced to suit. However, some micropipettes have a biphasic decay after a cur-
rent pulse (Figure 3.13). There is an initial rapid decay as the micropipette capacitance is
discharged, followed by a smaller, slower decay resulting from ion redistribution in the tip.
For these micropipettes, the maximum sampling rate is severely restricted by the slow
phase of the micropipette decay. However, the wide-bandwidth noise of the micropipette
is related to the initial fast phase. For these micropipettes, it makes sense to slightly filter
the micropipette response to try to remove some of the wideband noise as long as this can
be achieved without increasing the ultimate settling time (which is limited by the slow
decay). This filtering is provided by a variable single-pole filter called the Anti-Alias Filter.
With the Axoclamp 900A amplifier, the anti-alias filter is applied along with the lag filter.
This implementation leads to increased clamp stability and speed.

Fast phase

Slow phase

Figure 3.13: Biphasic voltage response of a high-resistance micropipette.

The Axon Guide — 2500-0102 Rev. C 55

 3. Instrumentation for Measuring Bioeletric Signals from Cells

The voltage response of a high-resistance micropipette is often biphasic. The initial fast
phase reflects the time constant formed by the micropipette resistance and the stray capac-
itance. This time constant sets the cutoff frequency for thermal noise. The slow phase is
thought to be due to ion redistribution in the tip. The maximum sampling rate during
dSEVC is limited by the time taken for the slow phase to settle.
Capacitance Neutralization Noise
Another source of noise in discontinuous single-electrode voltage clamps arises from the
capacitance neutralization circuitry. A fundamental property of all capacitance neutraliza-
tion circuits is that they introduce noise in excess of what is contributed by the thermal
noise of the recording micropipette and the input noise of the buffer amplifier. The excess
noise becomes progressively larger as the amount of capacitance neutralization is increased
to reduce the micropipette time constant. In discontinuous systems the micropipette time
constant must be reduced more than in continuous systems because of the need for the
micropipette voltage to decay to Vm within the time allotted for passive recording. Com-
pared to a continuous two-electrode voltage clamp, the dSEVC will always be a factor of
three or more times noisier at similar gains.
Selecting the Duty Cycle
The duty cycle is the fraction of each period spent in current passing. As the duty cycle
approaches unity, the clamp fails to operate properly because insufficient time is left for
the micropipette artifact to decay, unless the sampling period is made unreasonably slow.
On the other hand, if the duty cycle is made very small, the magnitude of the current
pulse becomes very large (to maintain a given average current). Since there are conflicting
requirements on the selection of the duty cycle, a compromise must be found. It has been
shown that the best compromise is a duty cycle equal to 0.3. This is the value used in the
Axoclamp amplifiers.
Useful Signal Bandwidth
As a general rule, the useful signal bandwidth is about one tenth of the switching rate. If
the dSEVC is switching at 20 kHz, it is reasonable to expect to record membrane currents
to within a bandwidth of 2 kHz.
Current and Voltage Measurement
Since the system is injecting current through the micropipette during every cycle, there is
no place that a continuous record of the membrane current and voltage exists. Instead,
sample-and-hold amplifiers must be used to store the values measured at the appropriate
times.

For the Vm output, the appropriate time to measure the voltage is at the end of the period
of passive decay, just before the next current pulse begins. This sampled value is held for
the duration of the cycle.

For the Im output, the current is sampled during the middle of the current pulse. The
value is held until the next current pulse. Since the user is interested in the average current

56 The Axon Guide — 2500-0102 Rev. C

 3.3. Intracellular Recording—Voltage Clamp

entering the cell, the output of the sample and hold is multiplied by the duty cycle before
being presented on the output.
Conditions For Good dSEVC
1 The micropipette resistance should be as small as possible.
2 The micropipette capacitance should be minimized.
3 The membrane time constant must be at least ten times the micropipette time
constant.
Comparison with Two-Electrode Voltage Clamp
The TEVC is generally superior to the dSEVC. If the gains are adjusted for a similar
steady-state error, the TEVC will generally have just one third of the noise of the dSEVC.
If the gain of the TEVC is increased so that the noise of the two clamps is similar, the
TEVC will generally settle three times faster.

It is clear that a dSEVC should not be used in preference to a TEVC. It should only be
used in those cases where it is impractical to implement a TEVC, either because the cells
are not individually discernible or because they are too small.

3.3.5. DISCONTINUOUS CURRENT CLAMP (DCC)

Discontinuous current clamp mode (DCC) uses the same circuitry and principles as the
dSEVC mode. The main difference is that in dSEVC the amplitude of each current pulse
depends on the difference between the membrane potential and the command potential,
whereas in DCC the amplitude is determined by the experimenter. Thus, to inject a con-
stant current into the cell, a DC command is presented and successive current pulses
should be identical.

The main use of the discontinuous current clamp technique is to aid in the setup of the
instrument before selecting dSEVC. DCC-mode recording has been used in some experi-
ments in place of Bridge-mode recording to eliminate the possibility of a badly balanced
Bridge introducing an error in the voltage measurement. However, in general, the advan-
tage of DCC mode is small because in DCC mode the capacitance neutralization control
plays a similar role to the Bridge Balance control in continuous current clamping. If the
micropipette response is too slow, the transient after the current pulses will not have
decayed to baseline before the next sample is taken, and a current-dependent error voltage
will be measured. Thus, DCC mode just replaces one problem with another. At the same
time, DCC mode has more noise and a lower signal-recording bandwidth, so its use as a
substitute for Bridge mode should be undertaken with caution.

The Axon Guide — 2500-0102 Rev. C 57

 3. Instrumentation for Measuring Bioeletric Signals from Cells

3.3.6. CONTINUOUS SINGLE-ELECTRODE VOLTAGE CLAMP (WHOLE-CELL PATCH

CLAMP)
To implement a continuous single-electrode voltage clamp (cSEVC; also known as a
whole-cell patch clamp), a blunt, low-resistance pipette is sealed by suction to the surface
of the membrane. The patch of membrane enclosed within the tip of the pipette is rup-
tured by one of a variety of techniques. The electrolyte solution in the pipette then forms
an electrical continuity with the interior of the cell. It is equivalent to an extremely low-
resistance (e.g., 1–10 MΩ) intracellular micropipette.

The voltage at the top of the pipette is controlled by a voltage-clamp circuit. It is impor-
tant to realize that this is quite different from the situation in TEVC or dSEVC. In both
of the latter cases, the voltage at the tip of the voltage-recording micropipette is controlled
(remember, in dSEVC, time-division multiplexing effectively yields two micropipettes:
the voltage-recording micropipette and the current-passing micropipette). In cSEVC, the
same electrode is used simultaneously for voltage recording and for current passing. The
voltage recorded at the top of the pipette is the sum of the membrane potential Vm, which
the experimenter wishes to control, and the current-induced voltage drop across the
pipette.

The current through the series resistance of the pipette and the residual resistance of the
ruptured patch is often sufficiently large to introduce significant voltage errors. Tech-
niques exist for compensating these errors. To get a feeling for the magnitude of the errors,
assume that the maximum compensation is 90%, beyond which the system oscillates and
destroys the cell. Further assume that the access resistance (Ra; the sum of the pipette
resistance and the residual resistance of the ruptured patch) is 5 MΩ. After compensation,
the effective value of Ra (Ra,eff ) would be just 0.5–1 MΩ. In this case, a 10 nA current
would cause a 50 mV uncompensated voltage error, reduced to 5 mV by the compensa-
tion. Clearly, the cSEVC technique cannot be used to record large currents. Even for
modest whole-cell currents, care must be taken to compensate for the series resistance and
then correctly interpret the residual error. The dSEVC technique should be considered as
an alternative to the cSEVC technique when the access resistance is too large.

The temporal resolution of the whole-cell patch clamp is also affected by Ra,eff. The time
constant for resolving currents is the product of Ra,eff (assuming that the membrane resis-
tance is much greater) and the membrane capacitance (τ = Ra,effCm). Thus the technique
is also limited to small cells where this product is small enough for the desired time resolu-
tion to be achieved. Figure 3.14 illustrates the voltage and temporal errors caused by the
presence of Ra.

58 The Axon Guide — 2500-0102 Rev. C

 3.3. Intracellular Recording—Voltage Clamp

A R a,eff

Im
Vm

Vcmd = Vp Rm Cm

V1
B
Vcmd

V1
Vm1
τ τ
Vm

τ
I1
IR
f

τ
Vm1 = V1 - I 1 R a,eff
V1 R m
= R +R
m a,eff
Im= V1
R a,eff + R m

Figure 3.14: Voltage and temporal errors caused by the presence of Ra.

A. The cSEVC circuit is simply illustrated as a voltage source (Vcmd) in series with the
effective access resistance (Ra,eff) and the membrane (Rm, Cm). The cSEVC circuit
ensures that the pipette potential (Vp) is equal to Vcmd.
B. After Vcmd steps to V1, a steady-state current, Im, flows in the circuit. The mem-
brane potential is equal to Vcmd - ImRa,eff. After the step change in the command
potential, Im and Vm settle exponentially to their steady-state values with τ =
[Ra,effRm/(Ra,eff + Rm)]Cm, but since in general Rm >> Ra,eff, a good approximation
is τ ≈ Ra,effCm.

3.3.7. SERIES-RESISTANCE COMPENSATION

In the ideal experiment, the resistance of the patch micropipette in whole-cell experiments
would be zero. In this case, the time resolution for measuring membrane currents and
changing the membrane voltage would be limited only by the speed of the electronics
(typically just a few microseconds).
Positive Feedback (“correction”)
Series-resistance compensation using positive feedback is an attempt to achieve this ideal
electronically. This technique is also called “correction.” It is the same technique that has
been widely used in conventional two-electrode voltage clamps. Basically, a signal propor-

The Axon Guide — 2500-0102 Rev. C 59

 3. Instrumentation for Measuring Bioeletric Signals from Cells

tional to the measured current is used to increase the command potential. This increased
command potential compensates in part for the potential drop across the micropipette
(Figure 3.15). The amount of compensation achievable is limited by two considerations.
First, as the compensation level (α) approaches 100%, the increase in the command
potential hyperbolically approaches infinity. For example, at 90% compensation, the
command potential is transiently increased by a factor of ten (Vcmd/(1 – α)). Thus at
large compensation levels the electronic circuits approach saturation. Second, the current
feedback is positive; therefore, the stability of the circuit is degraded by the feedback and
at 100% compensation the circuit becomes an oscillator. In practice, the oscillation point
is much lower than 100% because of non-ideal phase shifts in the micropipette and the
cell membrane.

Rf

I
-
Vp
A1 +1
+ A2 VI
-1

Vm

Rm Cm
+1
A3
Vp +1
Vcmd

Figure 3.15: Series resistance correction.

In this figure, a single pipette is used to voltage-clamp the cell. Operational amplifier A1 is
configured as a current-to-voltage converter. Differential amplifier A2 subtracts the
pipette potential (Vp) to generate the current output (VI). A fraction (α) of VI is summed
with the command voltage (Vcmd) used to control Vp. This causes both a transient and a
steady-state increase in Vp compared with Vcmd. As a result, the membrane charges faster,
the voltage drop across the electrode resistance is compensated and the bandwidth is
increased.

The first problem, saturation of the electronics, could in principle be reduced by using
high-voltage (e.g., ±120 V) operational amplifiers. However, this approach has not been
pursued because these types of operational amplifiers have more noise and worse drift
than good conventional operational amplifiers. The second problem, stability, has been
partially reduced by adding a variable low-pass filter in the current-feedback loop (e.g., the
“lag” control of the Axopatch-1D and the Axopatch 200 series amplifiers, and “band-
width” in the MultiClamp 700 series amplifiers). By empirically setting the low-pass filter

60 The Axon Guide — 2500-0102 Rev. C

 3.3. Intracellular Recording—Voltage Clamp

cutoff frequency, large percentage compensations can be used, but these only apply to the
currents at bandwidths below that of the filter cutoff. Thus the DC, low- and mid-fre-
quency series resistance errors can be substantially reduced while the high-frequency errors
remain large.
Supercharging (“prediction”)
Another technique for speeding up the response to a command step is the “supercharging”
technique, also known as “prediction.” In contrast to the “correction” method of series-
resistance compensation, “prediction” is an open-loop method. That is, there is no feed-
back and there is little risk of oscillations.

Supercharging is accomplished by adding a brief “charging” pulse at the start and the end
of the command voltage pulse. This means that initially the membrane is charging
towards a larger final value than expected (Figure 3.16). In its crudest form, the charging-
pulse amplitude or duration are adjusted empirically by the investigator so that the mem-
brane potential does not overshoot. Since the membrane potential is not directly observ-
able by the user, this is accomplished by adjusting the controls until the current transient
is as brief as possible. Once the optimum setting for one step size has been found, the size
of the supercharging pulse for all other step sizes can be calculated by the computer. Since
the large, transient supercharging current has to be carried by the feedback resistor in the
headstage, the amount of supercharging that can be used is limited by saturation of the
current-to-voltage converter in the headstage.

The Axon Guide — 2500-0102 Rev. C 61

 3. Instrumentation for Measuring Bioeletric Signals from Cells

A Rf

-
Vp A1 +1 VI
+ A2
-1

Vcmd
Vm

B
Vpre

Vstep
Vcmd

Vpre
τm
Vstep
Vm
time

Figure 3.16: Prediction implemented empirically by the computer.

A. A brief charging pulse is added to the start and the end of Vcmd. The size and dura-
tion of the charging pulse is empirically determined to give the fastest charging
current. Vm charges rapidly to its final value. Vcmd and the supercharging pulses
are not part of a positive feedback circuit.
B. An expanded view of Vcmd and Vm. Initially, Vm increases towards the amplitude
(Vpre) of the pre-pulse. However, the pre-pulse is terminated early, just when Vm
reaches the desired command voltage, Vstep.

The Axopatch 200 amplifier pioneered an alternative technique for generating the super-
charging current. This technique takes advantage of having full knowledge of the Ra and
Cm values. Fortunately, this information is generally available on modern whole-cell
patch-clamp amplifiers as part of the technique used to unburden the feedback resistor
(Rf ) from the necessity of passing the charging current into the cell (i.e., whole-cell capac-
itance compensation). Once these parameters are determined, it is possible to automati-
cally boost changes in the command voltage to supercharge the cell (Figure 3.17). This
technique has the significant advantage that no empirical determination of the charging-
pulse amplitude or time is needed and that it works with any command-voltage shape
without requiring a computer to calculate the supercharging pulse.

62 The Axon Guide — 2500-0102 Rev. C

 3.3. Intracellular Recording—Voltage Clamp

Rf

-
Vp
A1 +1
+ A2
-1

Vcmd
+1
A3
Vm +1

Vcmd SHAPING Vcmd (s)

CIRCUIT

SUPERCHARGING
PROCESSOR

Figure 3.17: Implementation of prediction based on the knowledge of cell parameters.

The supercharging pulse added to Vcmd is a fraction (α) of a shaped version of Vcmd itself.
The shaped version, Vcmd(s), is generated in the whole-cell capacitance compensation cir-
cuit described later. Since the calculation of the supercharging pulse is determined by the
parameters of the cell and Vcmd itself, the system works with all commands (step, triangle,
sine, etc.) without the need for an empirical determination of the shape and amplitude for
each command shape and size.

Supercharging is appropriate for a specific type of experiment, but it is inappropriate in

others and its use should be carefully considered. The appropriate application is the mea-
surement of ionic currents that activate during the time that the membrane potential is
normally changing. If supercharging is applied, it might be possible to clamp the potential
to its final value before the ionic current substantially activates.

There are two significant shortcomings that the investigator should be aware of when
using the supercharging technique:

1 The technique does not correct for the voltage error that occurs when current flows
through the series resistance of the pipette.
2 The dynamic response of the circuit is not improved. That is, there is no
improvement in the speed with which changes in current (and hence, changes in
conductance) can be resolved. Membrane current changes are still filtered by the time
constant of the access resistance and the membrane capacitance. The investigator

The Axon Guide — 2500-0102 Rev. C 63

 3. Instrumentation for Measuring Bioeletric Signals from Cells

should not be misled by the rapid settling of the response to a command step into
thinking that this settling time represents the time resolution of the recording system.
Neither of these problems occurs with the correction method of series-resistance compen-
sation. The relative merits of correction versus prediction are shown in Table 3.1.
Table 3.1: Correction vs. Prediction.

Error Fast response Fast response

Stability
correction to Vcmd change to Rm change

Correction Yes Yes Yes Not guaranteed

Prediction No Yes No Guaranteed

Usually, the type of series-resistance compensation provided in a patch clamp amplifier is

a combination of correction and prediction. The Axopatch 200 and MultiClamp 700
amplifier series allow the prediction and correction components of the circuitry to be set
independently so that the researcher can optimize the technique that is most appropriate
for the experiment. Optimum performance is usually achieved by combining series-resis-
tance “correction” with series-resistance “prediction.”
Limitations of Series-Resistance Compensation
Series-resistance compensation is an attempt to electronically reduce the effect of the
pipette resistance. Because of practical limitations, it is never perfect. Even if 100% com-
pensation could be used with stability, this would only apply to DC and medium-speed
currents. Very fast currents cannot be fully corrected.

3.3.8. PIPETTE-CAPACITANCE COMPENSATION

When the command voltage is stepped, a large amount of current flows into the pipette
capacitance during the transition from one potential to the next. In an intracellular
micropipette amplifier, such as the Axoclamp 900A amplifier, this is reduced by setting
the capacitance neutralization controls, as discussed in the earlier section titled Capaci-
tance Compensation. In a patch-clamp amplifier, such as the Axopatch 200 and
MultiClamp 700 amplifier series, a variation of the technique is used for the same pur-
pose—to reduce or even eliminate the transient. This section discusses pipette capacitance
compensation in a patch clamp.

Compensating the pipette capacitance in a patch clamp has three purposes. First, many
researchers want to remove the transient from the records for “cosmetic” reasons. Second,
during the transient the potential at the top of the pipette is changing slowly while the
pipette capacitance charges. By rapidly charging the pipette capacitance through the com-
pensation circuitry, the potential at the top of the pipette is stepped more rapidly, reduc-
ing the likelihood that rapid-onset ionic currents will be distorted. Third, uncompensated
pipette capacitance has a detrimental effect on the stability of the series-resistance correc-
tion circuitry. The component of the current that flows into the pipette capacitance is not

64 The Axon Guide — 2500-0102 Rev. C

 3.3. Intracellular Recording—Voltage Clamp

in series with any resistance. Thus the series-resistance correction circuit exceeds 100%
compensation for this component of the current as soon as the circuit is switched in.

In the Axopatch and MultiClamp amplifiers, two pairs of pipette capacitance compensa-
tion controls are available. With these controls, it is often possible to reduce the pipette
transients to extremely low levels. The bulk of the transient is reduced by using the fast
magnitude and time constant (τ) controls. The magnitude control compensates the net
charge. The τ control adjusts the time constant of the charge compensation to match the
time constant of the command pathway and to compensate for small non-idealities in the
frequency response of the pipette and electronics.

The residual slow component seen in many pipettes is reduced by using the slow magni-
tude and τ controls. A simplified circuit of the fast and slow compensation circuitry is
shown in Figure 3.18.

I C1

C1 = 1 pF

Rf
Rp
OPEN
CIRCUIT -
Cp I
+
Ip
Vcmd

FAST τ

10 V cmd +
FAST MAG

SLOW τ +

+
+

SLOW MAG

Figure 3.18: Pipette capacitance compensation circuit

When the command potential (Vp) changes, current Ip flows into Cp to charge it to the
new potential. If no compensation is used, Ip is supplied by the feedback element (Rf ),
resulting in a large transient signal on the output (I).

By properly setting the fast and slow magnitude and τ controls, a current (IC1) can be
induced in capacitor C1 (connected to the headstage input) to exactly equal Ip. In this
case no current needs to be supplied by Rf, and there is no transient on the output.

The Axon Guide — 2500-0102 Rev. C 65

 3. Instrumentation for Measuring Bioeletric Signals from Cells

3.3.9. WHOLE-CELL CAPACITANCE COMPENSATION

When the membrane potential is stepped, there is a significant current transient required
to charge the membrane capacitance. This is true whether the cell is clamped using a two-
electrode clamp or a single-electrode clamp. Since it is impossible to record meaningful
currents during this transient (the membrane potential has not settled), this transient is
ignored in most two-electrode voltage-clamp or discontinuous single-electrode voltage-
clamp experiments. However, when whole-cell voltage clamping is performed using a
patch-clamp amplifier, it is common to suppress this transient from the recording.

It is important to understand that the transient is not suppressed in the belief that some-
how the data during this period is more meaningful in a patch-clamp amplifier than it is
in a two-electrode voltage clamp. The transient is suppressed for technological reasons. In
a patch-clamp amplifier the value of the feedback resistor for whole-cell clamping is typi-
cally 500 MΩ. In an instrument driven from ±15 volt power supplies, the maximum cur-
rent that can be passed through a 500 MΩ resistor is less than 30 nA. The transient
current required to apply a 100 mV step to a cell that is clamped through a 1 MΩ resistor
is typically 100 nA. If series-resistance compensation is used, much larger currents are
required. Since the 500 MΩ resistor cannot pass the required current, the system would
saturate and the time to effect a step would be prolonged significantly. In order to prevent
saturation of the system, the transient current is injected through a capacitor. Since the
current monitor output on patch clamps is only proportional to the current through the
feedback resistor, the transient current is not seen, even though in reality it is still being
passed through the pipette into the cell.

The settings of the whole-cell capacitance controls to eliminate the transient are unique to
the cell being clamped. The values of the cell membrane capacitance and the access resis-
tance can be directly read from the controls. Figure 3.19 is a simplified circuit illustrating
how the whole-cell capacitance controls operate.

66 The Axon Guide — 2500-0102 Rev. C

 3.3. Intracellular Recording—Voltage Clamp

C2 = 5 pF (50 pF)

Rf

Im Rp
-
I
+
Rm Cm
Vcmd

SERIES RESISTANCE

10 Vcmd +
WHOLE +
CELL
CAP
+

Vp

Figure 3.19: Whole-cell capacitance compensation circuit.

Assume that the fast and slow electrode compensation controls have already been set to
compensate for Cp. By appropriately adjusting the series-resistance and whole-cell capaci-
tance controls, the current injected through C2 will supply the transient membrane cur-
rent (Im). These adjustments do not alter the time constant for charging the membrane.
Their function is to offload the burden of this task from the feedback resistor, Rf. In many
cells, even a small command voltage of a few tens of millivolts can require such a large cur-
rent to charge the membrane that it cannot be supplied by Rf. The headstage output satu-
rates for a few hundred microseconds or a few milliseconds, thus extending the total time
necessary to charge the membrane. This saturation problem is eliminated by the appropri-
ate adjustment of the series-resistance and whole-cell capacitance controls. This adjust-
ment is particularly important during series-resistance correction since series-resistance
correction increases the current-passing demands on Rf. By moving the pathway for
charging the membrane capacitance from Rf to C2, the series-resistance circuitry can
operate without causing the headstage input to saturate. The effect of transferring the cur-
rent-passing burden from Rf to C2 is illustrated in Figure 3.20.

The Axon Guide — 2500-0102 Rev. C 67

 3. Instrumentation for Measuring Bioeletric Signals from Cells

Vcmd

Vm

I Rf

I C2

Before After
Whole-cell Compensation Whole-cell Compensation

Figure 3.20: Using the injection capacitor to charge the membrane capacitance.

Vcmd is the patch-clamp voltage command; Vm is the voltage across the cell membrane;
IRf is the current across the patch-clamp feedback resistor; IC2 is the current injected
across the patch-clamp compensation capacitor.

The traces in this figure were recorded using an Axopatch 200 amplifier and a model cell.
After perfect whole-cell compensation is applied, the current to charge the membrane
capacitor is removed from the IRf trace and only the steady-state current remains. All of
the transient current appears in the IC2 trace. (The IC2 trace in the figure was recorded
using an oscilloscope probe connected to the internal circuitry). The I and Vm outputs on
the Axopatch 200 amplifier show the IRf and Vcmd trace illustrated in Figure 3.20. It is
easy to mistakenly think that the time course for charging the membrane is very fast, but
this is clearly not the case. Use of an independent electrode in the cell would show that the
cell charging rate is not affected by these adjustments.
Absolute Value
The absolute value of the membrane capacitance is displayed on the whole-cell capaci-
tance dial after the whole-cell current transient has been eliminated. This value may be
used to estimate the surface area of the cell assuming that the membrane capacitance per
unit area is 1 μF/cm2.

3.3.10. RUPTURING THE PATCH

To go from the patch to the whole-cell mode, the usual method used to rupture the mem-
brane under the pipette is to apply a pulse of gentle suction. This can be done by mouth
or by syringe.

68 The Axon Guide — 2500-0102 Rev. C

 3.3. Intracellular Recording—Voltage Clamp

“Zap” is another alternative for going whole-cell. In this technique, a large, brief hyperpo-
larizing voltage pulse (about 1.5 V) is applied to the cell. This pulse initiates dielectric
breakdown of the membrane patch and allows access to the interior of the cell. It is impor-
tant to use the briefest pulse consistent with low-access resistance. If the pulse is too long,
the seal might be lost. Both the Axopatch 200B and MultiClamp 700B amplifier allow
the Zap pulse to be applied for 0.5 ms or less at the minimum.

With some cells it is difficult to go whole cell without losing the seal. An alternative is the
perforated patch technique, which is discussed in Chapter 5.

3.3.11. WHICH ONE SHOULD YOU USE: dSEVC OR cSEVC?

Two methods can be used to implement a whole-cell patch voltage clamp. The first is the
discontinuous single-electrode voltage clamp (dSEVC) method. The second is the contin-
uous single-electrode voltage clamp (cSEVC) method. The cSEVC method is more com-
monly known as the “whole-cell patch clamp”; but formally, the term cSEVC is more
descriptive of the method. Both methods have their pros and cons. The dSEVC is gener-
ally superior when the currents being clamped are modest or large in size ( > 5 nA), lead-
ing to a concern that the error due to the uncompensated series resistance might be
significant. In dSEVC mode, if the sampling rate is correctly chosen and the capacitance
compensation correctly set, there is no error due to series resistance. For small currents,
the dSEVC mode is less attractive because it is more difficult to set up and because it is
noisier. Moreover, the error due to the uncompensated series resistance can generally be
made negligible for small currents, making the cSEVC mode very attractive.

The cSEVC mode is available on patch-clamp amplifiers (Axopatch 200B and

MultiClamp 700B) as the “whole-cell patch” mode. In a patch-clamp amplifier, the volt-
age-clamp circuit is a current-to-voltage converter located in the headstage (described in
detail below; see Figure 3.21). The Axoclamp 900A amplifier offers the dSEVC mode,
and the voltage-clamp circuit is located in the main unit. There are fewer ancillary con-
trols in the Axoclamp 900A amplifiers pertaining to whole-cell patch clamp as compared
to the Axopatch 200B and MultiClamp 700B amplifiers. For example, the Axoclamp
900A amplifier does not have a whole-cell capacitance compensation control.

3.3.12. SPACE CLAMP CONSIDERATIONS

There is one limitation to the performance of the voltage clamp that cannot be electrically
compensated. This is the deviation of the cell from a sphere centered on the tip of the
voltage-recording micropipette. The voltage clamp is maintained at the tip of the voltage-
recording micropipette. If all portions of the cell membrane are separated from this tip by
equal access resistance, then the membrane will be uniformly voltage clamped. However,
many cells have processes such as axons, dendrites and filopodia attached to the cell body
(where the micropipettes are usually located). The membranes of these processes are sepa-
rated from the cell body by an axial access resistance whose value depends on the distance
to each portion of the membrane and the cross section in that region of the cell. Thus
there is a voltage drop across the access resistance that becomes substantial for distal com-
ponents of the membrane. Even though the somatic membrane potential may be well

The Axon Guide — 2500-0102 Rev. C 69

 3. Instrumentation for Measuring Bioeletric Signals from Cells

controlled, the axonal or dendritic membrane potential may be very poorly controlled. In
these cases, the time course of synaptic currents, regenerative currents and measurements
of reversal potentials may be grossly distorted.

As a general rule, the voltage clamp is considered to be acceptable if the length of the
attached axon or dendrites is no more than 1/10 of the length constants. Even this short
length will cause significant distortion of fast currents (see Figure 7 in Rall and Segev,
1985). Calculation of the length constant for a cell is complicated since it depends on the
geometry of the particular cell under investigation. Some of the common ways to avoid
the problems of poor space clamping are as follows:

1 Restrict investigations to spherical cells. Many cultured cells are convenient.

2 Ligate attached axons. For example, the axon of large molluscan neurons can be tied
off with nylon thread.
3 Use short segments. For example, short segments (100 μm) of arteriolar syncytia can
be separated from the arteriole by careful cutting with a razor blade.
4 Restrict the range of the clamp to a short segment of the cell. This is the essence of the
“sucrose gap” technique sometimes used on axons.
5 Restrict the measurement to currents that are generated close to the micropipettes. For
example, the end plate currents in muscle fibers can be well clamped, even though the
bulk membrane current is very poorly clamped.
6 Restrict the measurement to the current flowing through a large patch of membrane
instead of the whole cell. The “macro patch” technique is a special case of the single-
channel patch-clamp technique described in Chapter 5, in which there are sufficient
channels for an ensemble current to be recorded.
7 Apply the patch pipette not at the cell body, but out on the process. This technique
has been successfully applied to dendrites, for example, where electrical access to the
cell is achieved far from the cell body along a dendritic shaft. This should improve
voltage clamp in nearby dendritic spines.

3.4. SINGLE-CHANNEL PATCH CLAMP 5

The single-channel patch clamp is a special case of the voltage-clamp technique. This
technique permits the direct measurement of the dynamic activity of individual mem-
brane-bound channel proteins.

Like the whole-cell patch technique, a blunt pipette is sealed onto a patch of membrane.
If single-channel recording is intended, the membrane at the tip of the pipette is preserved
(i.e., not ruptured). The current recorded is then simply the current that flows through
the membrane at the tip of the pipette. Since this membrane area is very small, there is a

5. Parts of this section have been reprinted with permission from Finkel, A. S., Progress in Instrumentation Technology for Recording
from Single Channels and Small Cells, in Cellular and Molecular Neurobiology: A Practical Approach, Oxford University Press, 1991.

70 The Axon Guide — 2500-0102 Rev. C

 3.4. Single-Channel Patch Clamp

good chance that just one or a a few ion channels are located in the patched membrane.
Individual ion-channel currents can thus be recorded. Previously, the only way to estimate
kinetics or conductance of these channels was the technique of “noise” or “fluctuation”
analysis. Noise analysis yields an estimate of the mean lifetimes of the channels and the
mean amplitudes, but no information about actual opening and closing transitions nor
about the shape of the conductance changes.

In single-channel recording, the current through the series resistance of the pipette is neg-
ligible, perhaps only a few picoamps flowing through a series resistance of just 10 MΩ.
The resulting voltage error is just a few tens of microvolts and is always ignored.

3.4.1. THE IMPORTANCE OF A GOOD SEAL

When a heat-polished pipette is pressed against a cell membrane it may adhere tightly.
The interior of the pipette is then isolated from the extracellular solution by the seal that
is formed. If the resistance of this seal is infinite, no current can leak across it.

Current leakage though the seal resistance has a crucial bearing on the quality of the patch
current recording. Firstly, depending on the size of the seal resistance, a fraction of the
current passing through the membrane patch will leak out through the seal and will not
be measured. The lower the seal resistance the larger the fraction of undetected current.

Secondly, thermal movement of the charges in the conducting pathways of the seal consti-
tute the major source of noise in the recording unless the seal resistance is very high (sev-
eral gigohms or more). A high seal resistance is a prerequisite of low-noise recordings.

High-resistance seals, often called “gigaseals,” require that very clean pipettes be used and
that they only be used once. If this is the case, gigaseals will routinely form after the appli-
cation of gentle suction. When good gigaseals are formed, the noise due to the leakage
current is virtually eliminated and other sources of noise remain the dominant limitations
in the resolution of the recording technique. These are the noise of the electronics, the
pipette glass, the input capacitance and the feedback resistor. Commercial patch clamp
amplifiers such as the Axopatch 200B amplifier have minimized the noise of the electron-
ics and the feedback element in the headstage. Chapter 4 describes the attributes of desir-
able glasses and the ways to fabricate low-noise pipettes. It also discusses the use of Sylgard
coatings to minimize pipette capacitance.

3.4.2. RESISTOR FEEDBACK TECHNOLOGY

Current-to-Voltage Converter
The basic concept behind the design of patch-clamp electronics is very simple
(Figure 3.21). A sensitive current-to-voltage converter is fabricated using a high-megohm
resistor (Rf ) and an operational amplifier (A1). The pipette is connected to the negative
input and the command voltage is connected to the positive input. Since the operational
amplifier has extremely high gain, the potential at the negative input is forced to follow
the potential at the positive input. All current flowing in the micropipette also flows

The Axon Guide — 2500-0102 Rev. C 71

 3. Instrumentation for Measuring Bioeletric Signals from Cells

through Rf. This current is proportional to the voltage across Rf which is measured at the
output of the differential amplifier (A2).

C Rf

If Rf
I
-
A1 +1
+ A2 BOOST
-1 VI

Vcmd

Figure 3.21: Resistive headstage.

Operational amplifier A1 is configured as a current-to-voltage converter. Differential

amplifier A2 subtracts Vcmd from the output of A1 to generate a voltage that is purely
proportional to the voltage across Rf and hence the feedback current, If. The boost circuit
increases the high-frequency gain to compensate for the narrow bandwidth of the feed-
back resistor.

In principle, the patch clamp is equivalent to a conventional two-electrode voltage clamp

in which the output circuit is connected back to the input pipette. In practice, the patch
clamp is better behaved. The patch-clamp system is simpler since all of the gain is in a sin-
gle operational amplifier stage. The stray capacitance across the feedback resistor guaran-
tees stability. Furthermore, because the gain of this operational amplifier is so high, the
difference between the command potential and the potential of the micropipette is negli-
gible. Remember that the potential that is controlled is the potential at the top of the
micropipette, not the potential at its tip.
Problems
The special demands of patch clamping lead to complicating factors in the design of a
good headstage.

1 Integrated circuit operational amplifiers do not have the required combination of

ultra-low noise and sub-picoamp bias currents. Thus the operational amplifier has to
be made from a composite of low-noise FETs6, such as the U430 type, and
conventional operational amplifiers. The voltage noise of the input FETs leads to
noise current being injected into the input capacitance. If this input capacitance is

6. Field Effect Transistor.

72 The Axon Guide — 2500-0102 Rev. C

 3.4. Single-Channel Patch Clamp

large, this source of noise becomes dominant; it is proportional to the product of the
input capacitance and the noise of the input FETs.
2 The minimum theoretically achievable current noise is lower for larger Rf values;
therefore it is important to choose large Rf values. The largest value typically used is
50 GΩ, since it allows a reasonable maximum current of more than 200 pA to be
measured. Unfortunately, for reasons that are not well understood, high-value resistors
are several times noisier than predicted by thermal-noise theory. Since the noise of
these resistors cannot be predicted, various brands of resistors must be tested until the
best one is found.
3 The inherent bandwidth of a 50 GΩ resistor is limited by the stray capacitance across
the resistor (CRf in Figure 3.21). For example, a 50 GΩ resistor with 0.1 pF stray
capacitance has a 5 ms time constant, corresponding to a bandwidth of only 32 Hz.
This poor time resolution is unacceptable for measuring ionic currents. Therefore, the
high-frequency components of the headstage output signal must be boosted. This is
typically achieved by an analog frequency compensation circuit. This circuit is made
complicated by the fact that Rf cannot be considered as an ideal capacitor in parallel
with an ideal resistor; thus a simple high-pass filter cannot be used. Complex circuits
consisting of up to four poles and three zeros in the transfer function are commonly
used. The placement of the poles and zeros must be carefully set for the particular
resistor.
4 The current required to charge the input capacitance during a step voltage command
can easily exceed the maximum current that can be passed by Rf from the typical 13 V
swing of the operational amplifier. For example, to linearly charge 5 pF of input
capacitance to 100 mV in 10 μs would require a charging current of 50 nA. This is
well beyond the 260 pA that can be passed by a 50 GΩ resistor driven from ±13 V.
Thus, special circuits are typically added to inject the required charging current
through a capacitor.
5 The best high-value resistors seem to be available only in a miniature “surface mount”
form. This means that the headstage electronics have to be manufactured in a hybrid.
This manufacturing technique has some advantages. First, the U430 input FETs and
other FETs typically used to switch between different values of Rf can be used in an
unpackaged form. This is to be preferred, since the sealing glasses used in packaged
transistors typically have a leakage resistance that increases the noise. Second, it means
that the sensitive input components can be maintained in a hermetically sealed
environment that reduces the rate at which they age due to environmental
contamination.
If the Probe Output is Slow, How Can Voltage Clamping be Fast?
In resistive-feedback headstages (but not capacitive-feedback headstages, discussed below)
the current output of the current-to-voltage (I-V) converter in the probe is slow. The
high-frequency boost occurs afterwards and cannot influence the events at the pipette.
Thus, one might conclude that the voltage clamp of the pipette must also be slow.

The Axon Guide — 2500-0102 Rev. C 73

 3. Instrumentation for Measuring Bioeletric Signals from Cells

In fact, despite the slow current output of the I-V converter, the voltage clamp of the
pipette is rapid. The pipette is connected to the negative input (summing junction) of the
op amp. The command potential is connected to the positive input of the op amp. The
operation of the op amp in this configuration is to force the potential at the summing
junction to rapidly follow the potential at the positive input. If the command potential is
a step, the potential at the summing junction (and hence the pipette) is also a step. The
current required to achieve this step is passed through the feedback resistor (Rf ) and the
associated stray feedback capacitance (CRf ) of the I-V converter. The output of the I-V
converter settles to the final value with time constant RfCRf. This relatively slow settling
occurs despite the fact that the step at the summing junction is fast.

In this discussion, we have carefully referred to the fact that it is the pipette that is rapidly
voltage clamped. The membrane potential is voltage clamped to its final value much more
slowly. To a reasonable approximation, the time constant for voltage clamping the mem-
brane is RpCm, where Rp is the pipette resistance and Cm is the membrane capacitance.

3.4.3. CAPACITOR FEEDBACK TECHNOLOGY

Rationale
A newer technology in single-channel recording is the capacitor-feedback technique, also
known as the integrating headstage technique. While this technique was talked about for
many years, it was only first attempted and described in 1985 by Offner and Clark. Prac-
tical implementations did not become commercially available until 1989.

In this technique, the feedback resistor in the headstage is replaced by a capacitor (Cf in
Figure 3.22). When a capacitor is used instead of a resistor, the headstage acts as an inte-
grator. That is, for a constant input current, the output of the headstage is a ramp. The
slope of the ramp is proportional to the current. To recover a voltage that is simply pro-
portional to the input current, the integrator must be followed by a differentiator (A2).

Cf
Rd
I
- Cd
A1 - VI
+ A2
+

Vcmd

Figure 3.22: Capacitive feedback headstage.

74 The Axon Guide — 2500-0102 Rev. C

 3.4. Single-Channel Patch Clamp

The feedback element around operational amplifier A1 is capacitor (Cf ). Thus the output
of A1 is the integral of the pipette current. The actual current is recovered from the inte-
gral by differentiating it in the differentiator formed by Cd, Rd and A2.

Compared with resistor-feedback picoamp current-to-voltage converters (RIV), capacitor

feedback current-to-voltage converters (CIV) exhibit less noise, increased dynamic range,
wider bandwidth and improved linearity.
Noise, Dynamic Range, Linearity and Bandwidth
1 The noise of the CIV is lower for two reasons. First, the capacitors do not generate as
much thermal noise as resistors do. (There is an equivalent resistor noise source due to
the differentiator, but with careful design this can be made negligible.) Second,
capacitors are commercially available that are free of the excess noise sources that
plague high-gigohm resistors.
The noise benefit of the CIV is eliminated if Cf is large (10 pF or more). This is
because from the noise point of view, Cf sums with the other sources of input
capacitance. The voltage noise of the input FETs leads to a current injection into the
total input capacitance. If this input capacitance is large, this source of noise becomes
dominant.
The lower noise of the CIV can be realized only in situations where all other noise
sources are minimized. That is, the experimenter should use a low-noise glass and
pipette holder, Sylgard coating, and a high-resistance seal. By carefully removing other
sources of noise, the reduced noise of the CIV can be realized in real patches, not just
in theory.
2 The capacitor-feedback headstage has an equivalent transfer resistance (RT) given by

Cd (10)
RT = Rd
Cf

Because the noise is theoretically independent of the Cf value and the gain of the
differentiator, RT, may be kept quite low, e.g., 100 MΩ. At this gain, the maximum
current that can be recorded is 500 times greater than for an RIV using a 50 GΩ
feedback resistor. Thus the CIV potentially has vastly improved dynamic range.
3 In the section on Resistor Feedback Technology, it was pointed out that in order to
achieve an acceptable bandwidth of, e.g., 20 kHz for single-channel recording, a
complex boost circuit has to be used to correct the frequency response. Even if the
boost box has as many as four poles and three zeros, the frequency response of Rf is
not perfectly corrected. The imperfections of the correction are most easily seen by
observing the step response of the headstage. Frequently, there will be overshoots and
ripples that can amount to as much as 2% of the response. However, because excellent
capacitors are available, the step response is inherently square with a CIV.

The Axon Guide — 2500-0102 Rev. C 75

 3. Instrumentation for Measuring Bioeletric Signals from Cells

4 The bandwidth of a CIV can be very wide. It is maximized by using a small value of
Cf and a small value of Rd. For practical values (Cf = 1 pF and Rd = 150 kΩ), the
bandwidth is of the order of 70 kHz or more. This is considerably better than the
20 kHz of a good RIV using a 50 GΩ feedback resistor.
Whether it is practical to use this increased bandwidth or not is another matter. The
noise of both the RIV and the CIV headstages increases dramatically with frequency.
Problems
There are two major problems that make the capacitor feedback technology difficult to
implement. The first problem is that after a sustained net DC input current the integrator
voltage ramps towards one of the power-supply limits. When this happens, the integrator
must be reset by discharging the capacitor. The frequency of the resets depends on the size
of Cf and the average input current. For example, if Cf is 1 pF and the input current is
2 pA, the output of the integrator will ramp at 2 V/s. If the resetting circuitry is triggered
when the integrator voltage exceeds 10 V, resets will occur every five seconds.

The reset itself lasts approximately 50 μs. During the reset, sample-and-hold circuits are
used to maintain the current output at its value immediately prior to the start of the reset.
If it is acceptable to lose 0.1% of the data, resets as frequent as every 50 ms would be
acceptable. In the above example, this corresponds to an input current of 200 pA. For sin-
gle-channel recording, this is more than adequate.

Figure 3.23 shows the signal pathway for the capacitor-feedback headstage. The output
current (I) of the capacitor-feedback headstage is normally connected through a switch to
the output pre-filter amplifier, then to the low-pass filter, and finally to the post-filter
amplifier. The signal also goes through a low-pass filter to a sample-and-hold amplifier.
During reset, the switch shifts to the RESET position. Simultaneously, the sample-and-
hold amplifier is switched to the hold mode so that the signal immediately before the reset
transient occurs is presented to the output amplifiers.

x1 or x10 x1, 2, 5, 10, 20, 50, 100

SAMPLE RESET
I AND SCALED
HOLD
OUTPUT
LOW-PASS NORMAL
FILTER PRE-FILTER LOW-PASS POST-FILTER
HOLD
AMPLIFIER FILTER AMPLIFIER

50 μs RESET
DATA
NOT VALID

Figure 3.23: Signal handling during resets in the capacitor-feedback headstage.

A capacitor feedback is not practical for whole-cell recordings. Due to the large whole-cell
currents the resets would be too frequent unless the value of Cf was increased. In practice,
it is hardly worth it. It is simpler to switch to a modest-sized resistor (e.g., 500 MΩ) and

76 The Axon Guide — 2500-0102 Rev. C

 3.4. Single-Channel Patch Clamp

avoid the problem of resets altogether. Reverting to resistor-feedback methods for whole-
cell recordings does not represent a setback in terms of noise because the 500 MΩ resistor
is not the dominant source of noise.

During the measurement of voltage-activated single-channel currents, the resets can be

made irrelevant. This is done by using a control pulse to force a reset immediately prior to
the voltage step. This guarantees that Cf is in its discharged state at the beginning of the
voltage step and is therefore unlikely to need resetting during the step.

The second problem is that during resets transients are injected into the headstage input
by the reset circuitry. If left uncompensated, these would cause unwanted and possibly
damaging current pulses to be injected down the micropipette into the patch. Special
compensation circuitry needs to be included in the design to exactly balance out these
injected currents.

There are other transient problems that occur during and after the reset. These include
dielectric absorption in the differentiator and feedback capacitors, and other ill-defined
transients that work their way into the system. Compensation circuitry must also be
included to compensate for these transients.

Overall, the need to reset and to compensate for the many transients that occur during
reset make the design of an integrating patch clamp challenging. However, in a well-
designed system, the reset transients measured at the output will be less than 1 pA mea-
sured in a 1 kHz bandwidth.

3.4.4. SPECIAL CONSIDERATIONS FOR BILAYER EXPERIMENTS

The Advantage of Resets in Bilayer Experiments
While they are normally considered to be a nuisance, there is one application where resets
are quite an advantage. This is in voltage-step experiments in bilayers. During a reset, a
low resistance of about 10 kΩ is placed in parallel to the feedback capacitor. This resistor
allows currents up to 1 mA to pass out of the headstage to rapidly charge the bilayer
capacitance. Voltage steps as large as 100 mV are normally achieved with one or two resets
separated by less than a millisecond.
Noise vs. Access Resistance
In bilayer experiments, the access resistance is quite low, usually ranging from a few kilo-
hms to several tens of kilohms. The access resistance is in series with the membrane capac-
itance and produces voltage noise just as though the headstage had high intrinsic noise.

For bilayer recording done at a low bandwidth (<1 kHz), the enCin noise has not yet
become the major contributor to the overall noise (where en is the voltage noise of the
probe input FETs and Cin is the capacitance at the input of the headstage primarily due to
the bilayer membrane capacitance). If the access resistance is large enough, it becomes the
dominant noise contributor at low bandwidths.

The Axon Guide — 2500-0102 Rev. C 77

 3. Instrumentation for Measuring Bioeletric Signals from Cells

Figure 3.24 shows the dependence of peak-to-peak noise on access resistance in a 1 kHz
bandwidth for a given bilayer membrane capacitance. It can be seen that the lower the
access resistance, the lower the noise. It is therefore important that the patch-clamp
instrument be capable of operating without oscillation even when extremely small, or
zero, access resistances are used. The CV-203BU headstage for the Axopatch 200B ampli-
fier and the CV-7B/BL headstage for the MultiClamp 700B amplifier have been designed
for complete stability even with extremely low access resistances.

10 MEMBRANE
CAPACITANCE
BW =1 kHz
300 pF
(pA peak-to-peak)

200 pF
TOTAL NOISE

100 pF
1

0.1
100 1k 10 k
SERIES RESISTANCE (ohms)

Figure 3.24: Typical current noise in bilayer experiments.

3.4.5. HOW FAST IS “FAST”?

The speed of the observed transitions of single-channel currents is the same as the
response time of the electronics. Whenever recordings are made at wider bandwidths, the
observed transition rates become shorter. It is not clear that we will ever be able to resolve
the actual transition time using electrical measurements (perhaps optical techniques will
emerge to do the job); but the discovery of new lower limits to the transition rate will be
made possible by the low-noise wide-bandwidth patch clamps such as the Axopatch 200
series amplifiers.

3.4.6. MEASUREMENT OF CHANGES IN MEMBRANE CAPACITANCE

The measurement of minute changes in membrane capacitance, such as the changes that
occur during exocytosis, was made practical by the whole-cell patch-clamp technique.
Two methods can be used. The simplest and the most traditional way is to use a fitting
technique to find the time constant of the current response to a voltage step. By assuming
a simple cell model, the membrane capacitance can be easily deduced. This technique is
relatively easy to apply, but it has a resolution no better than 100–200 fF, which is insuffi-

78 The Axon Guide — 2500-0102 Rev. C

 3.4. Single-Channel Patch Clamp

cient to resolve the 10–20 fF capacitance increases occurring during fusion of single gran-
ules with the membrane.

A much more sensitive technique involves the use of a lock-in amplifier (also known as a
phase detector). In this technique, a sinusoidal command voltage is applied to the cell.
The magnitude of this voltage must be small enough so that the cell’s properties are essen-
tially linear. One measures two outputs that are proportional to the sinusoidal current
response at two orthogonal frequencies. It can be shown that the magnitudes of these two
responses are a function of the access resistance, the membrane resistance and the cell
capacitance. With a third measurement of the DC current, and assuming that the reversal
potential of the DC currents is known, there is sufficient information to calculate the
value of all three parameters. The resolution of this technique can be as good as 1 fF, with
measurements being made up to one hundred times per second.

Joshi and Fernandez (1988) have described a similar technique where all of the phase-sen-
sitive measurements and sinusoidal stimulations are performed by software. This has the
great advantage of not requiring any special equipment other than the patch clamp.

3.4.7. SEAL AND PIPETTE RESISTANCE MEASUREMENT

The micropipette resistance is easily measured with micropipette amplifiers by passing a
pulse of current through the micropipette. Generally, a convenient pulse size such as 0.1,
1 or 10 nA is used. If the pulse size is 1 nA, the micropipette potential without any Bridge
Balance is 1 mV/MΩ of micropipette resistance. For example, if the pulse response is
55 mV, the micropipette resistance is 55 MΩ. Alternatively, the Bridge Balance control
can be adjusted to null the pulse response and the micropipette resistance can be mea-
sured directly from the Bridge Balance dial.

In patch clamps, the measurement of the combined pipette and seal resistance is not quite
as straight forward because the patch clamp is normally used in voltage-clamp mode
rather than current-clamp mode. In this case, a voltage step is applied to the pipette. The
combination of the pipette and seal resistance is calculated from the current response
using the formula:

R = V/I

For example, if the test pulse is 1 mV and the current response is 67 pA, the resistance is
15 MΩ.

With the MultiClamp 700B and Axoclamp 900A amplifiers, the Commander software
controlling the amplifier can apply a pulse and automatically calculate the electrode resis-
tance, whether in current or voltage clamp mode.

3.4.8. MICROPIPETTE HOLDERS

High-quality micropipette holders are crucial to successful single-channel recording. Two
characteristics of the holder are important. First, the holder must be mechanically stable.
When working with a cell-attached patch, drift of the micropipette is unacceptable.

The Axon Guide — 2500-0102 Rev. C 79

 3. Instrumentation for Measuring Bioeletric Signals from Cells

Clearly, a low-drift micromanipulator is required (as discussed in Chapter 2). But if the
holder is loose in the headstage connector or if the pipette does not seat properly, drift will
occur despite the high quality of the micromanipulator. One common, exasperating prob-
lem associated with loose holders is movement of the pipette when suction is applied,
which either prevents seal formation or damages the cell. The HL-U holders from MDS
Analytical Technologies fit all recent Axon Cellular Neuroscience headstages and have
been designed to fit snugly into the headstage connector and to firmly grip the pipette to
prevent drift. Second, the holder must not introduce noise. To assure this:

1 Select the right holder material. When a holder is connected to the headstage, even
before a pipette is connected, the open-circuit noise goes up slightly. The reasons are
not understood. Empirical tests have shown that when used with glass micropipettes,
polycarbonate adds the least noise. The HL-U holders use polycarbonate for the
pipette cap and the body of the holder. The portion that plugs into the headstage is
made of Teflon. Teflon in this position does not add noise, and it facilitates smooth
insertion and removal.
2 Use a small pin for the electrical connection. Large pins have more capacitance
contributed by the surrounding grounded surfaces.
3 Do not use a shield. Surrounding the holder with a driven or grounded shield adds to
the input capacitance and thus increases the noise.

3.5. CURRENT CONVENTIONS AND VOLTAGE CONVENTIONS

The terminology used in this discussion applies to all Axon Cellular Neuroscience ampli-
fiers manufactured by MDS Analytical Technologies.

3.5.1. DEFINITIONS
Positive Current
The flow of positive ions out of the headstage into the microelectrode and out of the
microelectrode tip into the preparation is termed positive current.
Inward Current
Current that flows across the membrane, from the outside surface to the inside surface, is
termed inward current.
Outward Current
Current that flows across the membrane, from the inside surface to the outside surface, is
termed outward current.
Positive Potential
The term positive potential means a positive voltage at the headstage input with respect to
ground.

80 The Axon Guide — 2500-0102 Rev. C

 3.5. Current Conventions and Voltage Conventions

Transmembrane Potential
The transmembrane potential (Vm) is the potential at the inside of the cell minus the
potential at the outside. This term is applied equally to the whole-cell membrane and to
membrane patches.
Depolarizing / Hyperpolarizing
The resting Vm value of most cells is negative. If a positive current flows into the cell, Vm
initially becomes less negative. For example, Vm might shift from an initial resting value
of -70 mV to a new value of -20 mV. Since the absolute magnitude of Vm is smaller, the
current is said to depolarize the cell (i.e., it reduces the “polarizing” voltage across the
membrane). This convention is adhered to even if the current is so large that the absolute
magnitude of Vm becomes larger. For example, a current that causes Vm to shift from
-70 mV to +90 mV is still said to depolarize the cell. Stated simply, depolarization is a pos-
itive shift in Vm. Conversely, hyperpolarization is a negative shift in Vm.

3.5.2. WHOLE-CELL VOLTAGE AND CURRENT CLAMP

Depolarizing / Hyperpolarizing Commands
In whole-cell voltage clamping, whether it is performed by TEVC, dSEVC, cSEVC or
whole-cell patch clamp, a positive shift in the command voltage causes a positive shift in
Vm and is said to be depolarizing. A negative shift in the command voltage causes a nega-
tive shift in Vm and is said to be hyperpolarizing.
Transmembrane Potential vs. Command Potential
In whole-cell voltage clamping, the command potential controls the voltage at the tip of
the intracellular voltage-recording microelectrode. The transmembrane potential is thus
equal to the command potential.
Inward / Outward Current
In a cell generating an action potential, depolarization is caused by a flow of positive
sodium or calcium ions into the cell. That is, depolarization in this case is caused by an
inward current.

During intracellular current clamping, a depolarizing current is a positive current out of

the micropipette tip into the interior of the cell. This current then passes through the
membrane out of the cell into the bathing solution. Thus, in intracellular current clamp-
ing, a depolarizing (positive) current is an outward current.

An inward sodium current flows in some cells after a depolarizing voltage step. When the
cell is voltage clamped, the sodium current is canceled by an equal and opposite current
flowing into the headstage via the microelectrode. Thus it is a negative current. When
two-electrode voltage clamping was first used in the early 1950’s, the investigators chose
to call the negative current that they measured a depolarizing current because it corre-
sponded to the depolarizing sodium current. This choice, while based on sound logic, was
unfortunate because it means that from the recording instrument’s point of view, a nega-

The Axon Guide — 2500-0102 Rev. C 81

 3. Instrumentation for Measuring Bioeletric Signals from Cells

tive current is hyperpolarizing in intracellular current-clamp experiments but depolarizing

in voltage-clamp experiments.

To prevent confusion, Axon Cellular Neuroscience instruments always use current and
voltage conventions based on the instrument’s perspective. That is, the current is always
unambiguously defined with respect to the direction of flow into or out of the headstage.
Some instrument designers have put switches into the instruments to reverse the current
and even the command voltage polarities so that the researcher can switch the polarities
depending on the type of experiment. This approach has been rejected by MDS
Analytical Technologies because of the real danger that if the researcher forgets to move
the switch to the preferred position, the data recorded on the computer could be wrongly
interpreted. MDS Analytical Technologies believes that the data should be recorded
unambiguously.

3.5.3. PATCH CLAMP

By design, the patch-clamp command voltage is positive if it increases the potential inside
the micropipette. Whether it is hyperpolarizing or depolarizing depends upon whether
the patch is “cell attached,” “inside out” or “outside out.” The patch-clamp pipette current
is positive if it flows from the headstage through the tip of the micropipette into the patch
membrane.
Cell-Attached Patch
The membrane patch is attached to the cell. The pipette is connected to the outside sur-
face of the membrane. A positive command voltage causes the transmembrane potential
to become more negative, therefore it is hyperpolarizing. For example, if the intracellular
potential is -70 mV with respect to 0 mV outside, the potential across the patch is also
-70 mV. If the potential inside the pipette is then increased from 0 mV to +20 mV, the
transmembrane potential of the patch hyperpolarizes from -70 mV to -90 mV.

From the examples it can be seen that the transmembrane patch potential is inversely pro-
portional to the command potential, and shifted by the resting membrane potential
(RMP) of the cell. A positive pipette current flows through the pipette across the patch
membrane into the cell. Therefore a positive current is inward.
Inside-Out Patch
The membrane patch is detached from the cell. The surface that was originally the inside
surface is exposed to the bath solution. Now the potential on the inside surface is 0 mV
(bath potential). The pipette is still connected to the outside surface of the membrane. A
positive command voltage causes the transmembrane potential to become more negative,
therefore it is hyperpolarizing. For example, to approximate resting membrane conditions
of Vm = -70 mV, the potential inside the pipette must be adjusted to +70 mV. If the
potential inside the pipette is increased from +70 mV to +90 mV, the transmembrane
potential of the patch hyperpolarizes from -70 mV to -90 mV.

82 The Axon Guide — 2500-0102 Rev. C

 3.5. Current Conventions and Voltage Conventions

From the example it can be seen that the transmembrane patch potential is inversely pro-
portional to the command potential. A positive pipette current flows through the pipette
across the patch membrane from the outside surface to the inside surface. Therefore a pos-
itive current is inward.
Outside-Out Patch
The membrane patch is detached from the cell in such a way that the surface that was
originally the outside surface remains exposed to the bath solution. The potential on the
outside surface is 0 mV (bath potential). The pipette interior is connected to what was
originally the inside surface of the membrane. A positive command voltage causes the
transmembrane potential to become less negative, therefore it is depolarizing. For exam-
ple, to approximate resting membrane conditions, assuming that Vm = -70 mV, the
potential inside the pipette must be adjusted to -70 mV. If the potential inside the pipette
is then increased from -70 mV to -50 mV, the transmembrane potential of the patch
depolarizes from -70 mV to -50 mV.

The membrane potential is directly proportional to the command potential. A positive

pipette current flows through the pipette across the patch membrane from the inside sur-
face to the outside surface. Therefore a positive current is outward.

3.5.4. SUMMARY
1 Positive current corresponds to:
Cell-attached patch patch inward current
Inside-out patch patch inward current
Outside-out patch patch outward current
Whole-cell voltage clamp outward membrane current
Whole-cell current clamp outward membrane current

2 A positive shift in the command potential is:

Cell-attached patch hyperpolarizing
Inside-out patch hyperpolarizing
Outside-out patch depolarizing
Whole-cell voltage clamp depolarizing

3 The correspondence between the command potential (Vcmd) and the transmembrane
potential (Vm) is:
Cell-attached patch Vm = RMP – Vcmd
Inside-out patch Vm = – Vcmd
Outside-out patch Vm = Vcmd
Whole-cell voltage clamp Vm = Vcmd

The Axon Guide — 2500-0102 Rev. C 83

 3. Instrumentation for Measuring Bioeletric Signals from Cells

3.6. REFERENCES
3.6.1. PATCH CLAMP
Rae, J.L. and Levis, R.A. Patch voltage clamp of lens epitheleal cells: theory and practice.
Molec. Physiol, 6, 115–162, 1984.

Hamill, O. P., Marty, A., Neher, E., Sakmann, B., and Sigworth, F.J. Improved patch-
clamp techniques for high-resolution current recording from cells and cell-free membrane
patches. Pflügers Archiv., 391, 85–100, 1981.

Sakmann, B. and Neher, E., Eds. Single-Channel Recording. New York: Plenum Press,
1983.

Joshi, C. & Fernandez, J. M. Capacitance measurements: an analysis of the phase detector

technique used to study excocytosis and endocytosis. Biophys. J. 53, 885–892, 1988.

Finkel, A. S., Progress in instrumentation technology for recording from single channels
and small cells, in Cellular and Molecular Neurobiology: A Practical Approach. Oxford
University Press, 1991.

3.6.2. TWO-ELECTRODE VOLTAGE CLAMP

Finkel, A.S. & Gage, P.W. Conventional voltage clamping with two intracellular micro-
electrodes, in Voltage and Patch Clamping with Microelectrodes. Ed. T. Smith Jr. et al.,
Baltimore: Williams & Wilkins, 1985.

3.6.3. SINGLE-ELECTRODE VOLTAGE CLAMP

Finkel, A.S. & Redman, S.J. Theory and operation of a single microelectrode voltage
clamp. J. Neurosci. Meths. 11, 101–127, 1984.

3.6.4. SPACE-CLAMP CONSIDERATIONS

Rall, W. & Segev, I. Space-clamp problems when voltage clamping branched neurons
with intracellular microelectrodes. in Voltage and Patch Clamping with Microelectrodes.
Eds. T. Smith Jr. et al., Baltimore: Williams & Wilkins, 1985.

3.6.5. OTHER
Hille, B. Ionic Channels of Excitable Membranes. Sunderland, Massachusetts: Sinauer
Associates. 1984.

84 The Axon Guide — 2500-0102 Rev. C

 4. Microelectrodes and
Micropipettes
4.1. ELECTRODES, MICROELECTRODES, PIPETTES,
MICROPIPETTES AND PIPETTE SOLUTIONS1
Electrodes convert ionic current in solution into electron current in wires; they are made
of materials that can participate in a reversible reaction with one of the ions in the solu-
tion. The most frequently used electrode material in electrophysiology is a silver (Ag) wire
coated with a composite of Ag and silver-chloride (AgCl). For this electrode, chloride ions
(C1-) react with the Ag to produce AgCl plus an electron (e-), or an electron reacts with
AgCl to produce Ag plus Cl-. Thus, the current carried by chloride ions in the solution is
converted into electrons according to the following reversible reaction:

C1 - + Ag AgC1 + e−

The electrical potential at one of these electrodes is equal to the standard electrochemical
potential for Ag/AgCl plus RT/F ln(aCl-), where R is the gas constant (8.314 V C K-1
mol-1), T is the absolute temperature on the Kelvin scale, F is Faraday’s constant (9.648 x
104 C mol-1), and aCl- is the activity (i.e., the effective concentration) of Cl- in solution
near the electrode solution interface. The potential difference between a pair of these elec-
trodes should be zero if they are immersed in connected compartments with equal chlo-
ride concentrations. These concentrations must be kept constant during the course of an
experiment or an offset voltage will occur. An offset may be encountered when a Ag/AgCl
electrode is used as a bath electrode and the bath Cl- concentration is changed. To prevent
this type of voltage offset, many investigators connect their bath electrode through an agar
bridge that maintains a constant Cl- concentration in the immediate vicinity of the
Ag/AgCl electrode.

Intracellular and patch microelectrodes contain a AgCl-coated Ag wire placed in a solu-

tion-filled glass micropipette. The electrolyte solution provides the fluid connection from
the cell to the electrode. Thus, the glass “electrodes” one pulls are not electrodes at all but
simple conduits of proper geometry to establish this fluid bridge. In recognition of this,
microelectrodes are often referred to as “pipettes” or “micropipettes.”

1. Please see footnote 1 in Chapter 3.

The Axon Guide — 2500-0102 Rev. C 85

 4. Microelectrodes and Micropipettes

The composition of the pipette solution depends on the type of measurement being
made. Standard intracellular microelectrodes that are used to penetrate the cell membrane
must be filled with a highly concentrated electrolyte solution, usually 2–4 M salt. The
high electrolyte concentration reduces the electrode resistance. This minimizes the rectify-
ing current flow, lowers voltage noise and provides a wider recording bandwidth. A solu-
tion this concentrated dominates the liquid junction potential formed between the pipette
solution and the cell cytoplasm and results in a final junction potential that depends pri-
marily on the relative mobilities of the anions and cations in the filling solution. This is
important for the measurement of DC voltages because the cytoplasm anions are, pre-
dominantly, low mobility ions (mostly negative charges on large proteins), whereas the
cations are small, high-mobility charges. The difference in the mobilities of the cellular
anions and cations may result in a positive liquid junction potential. This positive liquid
junction potential would lead to the measurement of an artificially depolarized membrane
resting potential. The concentrated pipette solution is important because it negates the
development of the liquid junction potential, thus preventing an erroneous measurement
of the resting potential. However, there is a disadvantage in using the concentrated filling
solution since it can enter the cell and produce a hyperosmotic load that would cause the
cell to swell and alter its normal anion and cation content. While pipettes with very small
tip diameters can minimize or prevent concentrated solution from entering the cell, they
do so at the expense of noise, diminishing current passing ability, and limiting recording
bandwidth. These limitations result from the high resistance of small-tip pipettes. In prac-
tice, the choice of the pipette tip size reflects a compromise between its ability to prevent
the concentrated solution from entering the cell and the pipette’s resistance.

Patch micropipettes are usually filled with a solution that is iso-osmotic to cell cytoplasm
to allow current measurements in the absence of bulk water flow. Because of the much
larger diameter of the patch pipette tip (1–2 μm) as compared to the tip diameter of an
intracellular microelectrode (0.01–0.1 μm), the lower ionic-strength filling solution
results in an electrode whose resistance is as low as a few megohms. The specific choice of
the filling solution is dictated by the specific experiment to be done. Care must be taken
to minimize junction-potential artifacts. When the pipette tip is immersed in the bath, a
junction potential is generated. Its magnitude depends on the concentrations and mobili-
ties of the ions in both the pipette solution and the bath (see discussions of the Henderson
equation in electrochemistry textbooks). The junction potential is usually nulled while
the pipette tip is immersed in the bath. When the seal forms, the junction potential disap-
pears or changes its magnitude, and, consequently, the nulling voltage is no longer correct.
The junction potential also changes upon going whole cell. These effects must be taken
into account or offset errors will exist in either single-channel or whole-cell current
recordings.

The filling solution in a patch pipette can be changed during the course of an experiment
by inserting a fine tube into the back of the pipette through a special post in the electrode
holder. Different solutions can be injected into the tube. Since the tube can be placed
quite close to the pipette tip, the new solution can reach the vicinity of the pipette tip
after a short diffusion delay. The procedure of draining excess solution from the pipette

86 The Axon Guide — 2500-0102 Rev. C

 4.2. Fabrication of Patch Pipettes

depends on the specific setup. This technique cannot be applied to standard intracellular
microelectrodes; due to their small tip size, the tube cannot be placed close enough to the
tip to allow changing the tip solution within a reasonable time.

The pipette is mounted in a holder that contains a Ag/AgCl electrode and a proper con-
nector to plug into the headstage of the amplifier. The holder must be constructed of a
low-noise, inert material, which is often chosen empirically. Axon Cellular Neuroscience
holders are constructed from polycarbonate, which yields low noise with several patch-
clamp pipettes. HL-U headstages have an electrode well to accommodate pipettes with an
outside diameter 1.0–1.7 mm. The HL-1-17 (for the non-U-type CV-type headstages)
and HL-2-17 holders (for the HS-type and VG-type headstages) have an electrode well
that can accommodate pipettes with an outside diameter of 1.5–1.7 mm. Pipettes of this
size have an internal bore that allows insertion of a 1 mm diameter Ag/AgCl pellet elec-
trode. The HL-1-12 (for the non-U-type CV-type headstages) and HL-2-12 (for the HS-
type and VG-type headstages) holders have a pipette well that can accommodate pipette
sizes of 1.0–1.2 mm diameter. This pipette size requires an electrode made of small-gauge
Ag wire coated with AgCl. A simple yet effective coating procedure is to clean the silver
wire down to bare metal using fine sand paper and immerse the cleaned wire in bleach
(Clorox) for 20–30 minutes, until the wire is uniformly blackened by the Ag/AgCl coat-
ing. Coating of the silver wire must be repeated whenever the AgCl becomes scraped from
the wire. Disruption of the coat can be minimized either by heat polishing the back of
each pipette before insertion into the holder or by encapsulating the coated portion of the
wire in a perforated Teflon sleeve.

The selection of the glass used for making the electrodes is more important for patch
pipettes than for intracellular micropipettes. Most intracellular micropipettes are made
from Corning #7740 Pyrex glass (Corning Inc., Corning, NY). However, when used for
tiny current measurements as in patch recordings, this glass is too noisy. The best patch
pipettes are made from glasses with low loss factors (see Table 4.1). These glasses generate
lower noise, and often have lower capacitances and fewer time constants to compensate.

4.2. FABRICATION OF PATCH PIPETTES

Glass tubing selected for the fabrication of patch pipettes should have walls of substantial
thickness (0.2–0.3 mm). A thick wall generates lower electrical noise and increases blunt-
ness at the tip, thereby preventing penetration into the cell during seal formation. Most
investigators use glass tubing with a 1.5–2.0 mm outside diameter and a 1.15–1.2 mm
inside diameter.

If not pre-cleaned, the glass tubing should be cleaned prior to the preparation of the patch
pipette. Sonicating the glass in 100% ethanol in an ultrasonic cleaner is an effective clean-
ing procedure. After cleaning, the glass should be placed in an oven at 200 °C for
10–30 minutes to achieve complete drying. This heat treatment is necessary for low-noise
recording in environments with very high humidity.

The Axon Guide — 2500-0102 Rev. C 87

 4. Microelectrodes and Micropipettes

Patch pipettes require much blunter tips than standard intracellular micropipettes and it is
usually impossible to pull them adequately on single-stage electrode pullers. Many labora-
tories have modified standard vertical pipette pullers to operate as multiple-stage pipette
pullers. Pulling pipettes, however, has become remarkably easier with the advent of micro-
processor-driven microelectrode pullers like those available from the Sutter Instrument
Company. With these pullers, it is possible to implement very complicated multistage
pulls of glass pipettes and to store the parameters required for the desired operation in the
instrument’s memory.

For the lowest noise recordings, patch pipettes must be coated with a hydrophobic mate-
rial to prevent the bathing solution from creeping up the wall of the pipette and, thus,
limit a substantial noise source. A commonly used compound is Sylgard #184 from Dow
Corning (Midland, MI). Coating the glass with Sylgard also improves the electrical prop-
erties of the glass. After preparing the Sylgard solution per the manufacturer’s instructions,
it can be stored in the freezer in small, well-capped centrifuge tubes for several weeks.
When brought to room temperature for use in painting electrodes, this freezer-stored Syl-
gard will last for several hours before it begins to polymerize. Sylgard is painted on the
pipette tip using a small utensil, such as a piece of capillary tubing pulled over flame to a
reasonably fine tip. Sylgard painting can be done using the magnifications available with
standard dissecting microscopes. The pipette is coated to within 100 μm or less from its
tip. It is important that the Sylgard be directed away from the tip by gravity at all times
during the painting procedure. Otherwise, it will flow into the tip and make firepolishing
and/or sealing impossible. After painting, Sylgard can be cured by holding the tip for
5–10 seconds in the hot air stream coming from a standard heat gun of the type used in
electronics fabrication.

To promote gigaohm seals and to reduce the possibility of tip penetration into the cell
during seal formation, pipette tips should be firepolished. Firepolishing is accomplished
by connecting to a micromanipulator a rod of inert material to which a short loop of plat-
inum iridium wire has been fastened. The ends of this wire must be soldered to two other
pieces of wire that can be connected to a voltage or current source to allow current passing
through the platinum wire. The platinum loop is generally bent into a very fine hairpin so
that it can be brought to within a few microns of the electrode tip under direct observa-
tion using a compound microscope. Magnifications of 600x–1500x are required for ade-
quate visibility. The platinum wire is usually coated with a glass, such as Pyrex or Corning
#7056, to prevent sputtering. After positioning the electrode tip near the glass-coated
wire, current is passed through the wire to heat it. The amount of current required
depends on the softening temperature of the glass from which the pipette is constructed.
The hot platinum wire heats the electrode tip and causes sharp corners to round and
rough surfaces to smooth. This polishing is best done under direct observation at a magni-
fication of 600x–1500x.

88 The Axon Guide — 2500-0102 Rev. C

 4.3. Pipette Properties for Single-Channel vs. Whole-Cell Recording

4.3. PIPETTE PROPERTIES FOR SINGLE-CHANNEL VS.

WHOLE-CELL RECORDING
Whereas some properties of pipettes used for single-channel and whole-cell recordings are
similar, other properties are entirely different. Most importantly, reducing the noise of the
pipette is much more crucial in single-channel recording than in whole-cell recording.
The dominant noise source in whole-cell recording is the pipette resistance that is in series
with the capacitance of the entire cell. Hence, the noise contribution of the pipette is not
as important. However, in order to provide sufficient bandwidth for whole-cell current
recording and to limit command voltage errors, the resistance of a pipette used in whole-
cell recording should not exceed a few megohms. Limiting the pipette resistance is not
required in single-channel recording; nor does a higher pipette resistance, up to several
tens of megohms, result in a significant increase in the noise (see Chapter 11 for further
discussion of pipette noise). In either single-channel or whole-cell recording, capacitance
currents following voltage steps must be sufficiently small and simple in time course to
allow the researcher to compensate them by simple circuitry in the patch-clamp amplifier.
Additionally, whether used for single-channel or whole-cell recording, pipettes must be
made of glasses that do not leach compounds from their walls since these compounds
might alter the currents measured from a particular channel.

4.4. TYPES OF GLASSES AND THEIR PROPERTIES

Patch-clamp pipette glasses can be classified on the basis of the temperature at which they
soften, on the basis of their electrical properties, or on the basis of their major chemical
constituents. Many of these properties are itemized in specifications provided by the man-
ufacturers. Therefore, it is often possible to choose glasses that should be effective for
patch clamping just by considering their specifications. Table 4.1 lists the properties of a
number of glasses that have been used for patch clamping. Glasses are listed in increasing
order of loss factors.

Table 4.1: Electrical And Thermal Properties of Different Glasses.

Log10
Loss Dielectric Softening
Glass Volume Description
Factor Constant Temp. °C
Resistivity

7940 0.0038 11.8 3.8 1580 Quartz (fused silica)

1724 0.0066 13.8 6.6 926 Aluminosilicate

7070. 25 11.2 4.1 — Low loss borosilicate

8161 0.50 12.0 8.3 604 High lead

Sylgard 0.581 13.0 2.9 — #184 Coating compound

7059 0.584 13.1 5.8 844 Barium-borosilicate

The Axon Guide — 2500-0102 Rev. C 89

 4. Microelectrodes and Micropipettes

Log10
Loss Dielectric Softening
Glass Volume Description
Factor Constant Temp. °C
Resistivity

7760 0.79 9.4 4.5 780 Borosilicate

7040 1.00 9.6 4.8 700 Kovar seal borosilicate

KG-12 1.00 9.9 6.7 632 High lead

1723 1.00 13.5 6.3 910 Aluminosilicate

0010 1.07 8.9 6.7 625 High lead

EN-1 1.30 9.0 5.1 716 Kovar seal borosilicate

7720 1.30 8.8 4.7 755 Tungsten seal borosilicate

7056 1.50 10.2 5.7 720 Kovar seal borosilicate

3320 1.50 8.6 4.9 780 Tungsten seal borosilicate

7050 1.60 8.8 4.9 705 Series seal borosilicate

KG-33 2.20 7.9 4.6 827 Kimax borosilicate

7740 2.60 8.1 5.1 820 Pyrex borosilicate

1720 2.70 11.4 7.2 915 Aluminosilicate

N-51A 3.70 7.2 5.9 785 Borosilicate

R-6 5.10 6.6 7.3 700 Soda lime

0080 6.50 6.4 7.2 695 Soda lime

Note that the electrical properties and the thermal properties of the glasses are not neces-
sarily related to each other. Note also that Sylgard has better electrical properties than
most glasses shown in the table. It is, therefore, not surprising that heavy Sylgard coating
of pipettes fabricated from many types of glasses improves their electrical properties.

Table 4.2 shows the chemical constituents of many of the same glasses listed in Table 4.1.
This table may be useful in deciding which glasses are likely to contain leachable compo-
nents that might affect channel currents.

90 The Axon Guide — 2500-0102 Rev. C

 4.5. Thermal Properties of Glasses

Table 4.2: Chemical Compositions of Different Glasses.

Chemical Constituent

Glass SiO2 B2O3 Al2O3 Fe2O3 PbO BaO CaO MgO Na2O K2O Li2O As2O3 Sb2O3 SO3

1724 Not Available

7070 70.7 24.6 1.9 ___ ___ 0.2 0.8 0.8 ___ ___ 0.56 ___ ___ ___

8161 38.7 ___ 0.2 ___ 51.4 2.0 0.3 0.04 0.2 6.6 ___ 0.04 0.38 ___

7059 50.3 13.9 10.4 ___ ___ 25 ___ ___ 0.08 ___ ___ ___ ___ ___

7760 78.4 14.5 1.7 ___ ___ ___ 0.1 0.1 2.7 1.5 ___ 0.18 ___ ___

7040 66.1 23.8 2.9 ___ ___ ___ 0.1 0.1 4.1 2.7 ___ 0.1 ___ ___

KG-12 56.5 ___ 1.5 ___ 28.95 ___ 0.1 0.1 3.7 8.6 ___ 0.4 0.25 ___

1723 57.0 4.0 16.0 ___ ___ 6.0 10.0 7.0 ___ ___ ___ ___ ___ ___

0010 61.1 ___ ___ ___ 22.5 ___ 0.3 0.1 7.2 7.3 ___ ___ ___ ___

EN-1 65.0 18.0 7.6 ___ 0.01 2.7 0.1 0.1 2.3 3.2 0.6 ___ ___ ___

7720 71.4 15.2 2.0 ___ 6.1 0.3 0.2 0.1 3.7 0.3 ___ ___ 0.5 ___

7056 69.0 17.3 3.9 ___ ___ ___ 0.12 0.91 7.5 0.68 0.48 ___ ___

3320 75.3 14.3 ___ ___ ___ ___ 0.1 0.1 4.0 ___ ___ ___ 0.8 ___

7050 67.6 23.0 3.2 ___ ___ 0.1 0.1 0.1 5.1 0.2 ___ ___ ___ ___

KG-33 80.4 12.9 2.6 ___ 0.005 ___ 0.05 4.0 0.05 ___ ___ ___ ___

7740 80.4 13.0 2.1 ___ ___ ___ 0.1 0.1 4.1 ___ ___ ___ ___ ___

1720 62.0 5.3 17.0 ___ ___ ___ 8.0 7.0 1.0 ___ ___ ___ ___ ___

N51-A 72.3 9.9 7.3 ___ 0.02 ___ 0.9 0.05 6.5 0.7 ___ 0.02 ___ ___

R-6 67.7 1.5 2.8 ___ ___ 2.0 5.7 3.9 15.6 0.6 ___ ___ ___ 0.2

0080 73.0 0.04 ___ ___ ___ 0.1 4.8 3.2 16.8 0.4 ___ ___ ___ 0.22

4.5. THERMAL PROPERTIES OF GLASSES

Glasses that soften at lower temperatures offer several advantages over glasses that soften at
higher temperatures when fabricating patch-clamp pipettes. First, they do not impose a
burden on the heating filaments of microelectrode pullers. Since low-filament current is
required to pull these glasses, filaments rarely change their properties with extended use
and do not require replacement even after one or two years of continued operation. Sec-
ond, pipettes with extremely blunt tips can be more readily fabricated from these glasses
than from glasses with high softening temperatures. Blunt tips provide the lowest access

The Axon Guide — 2500-0102 Rev. C 91

 4. Microelectrodes and Micropipettes

resistance (Ra) for whole-cell recording. Furthermore, the blunt-tapered pipettes are less
likely to penetrate into the cell when pressed against the cell membrane during seal forma-
tion. High-lead glasses pull at the lowest temperatures and are remarkably amenable to
firepolishing. It is possible to pull pipettes at such a low temperature that their tips are
broken and jagged, forming tip openings with diameters in excess of 50 μm, and yet the
pipettes are easily firepolished into usable patch pipettes. Although the resulting tips are
exceedingly blunt, these pipettes are effective in forming a tight seal with the cell even
when the final pipette resistance is < 0.5 MΩ.

Blunt tips are very beneficial for perforated-patch recording because they are capable of
drawing in large omega-shaped pieces of membrane when suction is applied. The large
surface area of the omega-shaped membrane maximizes the number of parallel channels
created by amphotericin B or nystatin, thus minimizing the final Ra that is achievable (see
Chapter 5).

Most of the pipettes used for patch-clamp recording are fabricated from borosilicate
glasses. These glasses soften at temperatures in the 700–850 °C range (see Table 4.1).
Although those at the low end of the range are quite easily pulled and firepolished, they
are clearly not as good as the high-lead glasses in this regard. Compared to high-lead
glasses, the success of firepolishing borosilicate glasses depends more on the shape of the
tip obtained after pulling. However, with the advent of multi-stage, computerized pipette
pullers, one can routinely make both patch and whole-cell pipettes from almost any glass.

Aluminosilicate glasses are very hard glasses with high softening temperatures. They pro-
duce low-noise single-channel recordings. However, their low noise comes at a high price.
Glasses in this class soften at temperatures above 900 °C and, therefore, pulling them
wears out the coils and filaments of pipette pullers. Consequently, the coils change their
properties with time and must be replaced or readjusted frequently. Moreover, pipettes
fabricated from aluminosilicate glasses produce undesirably thin-walled tips after pulling.
This, along with their high softening temperature, makes them much more difficult to
firepolish than softer glasses.

4.6. NOISE PROPERTIES OF GLASSES

While there is no simple way to predict the noise that will be generated from a particular
glass when it is used for single-channel patch clamping, it has been observed that the noise
of a patch pipette is related to its loss factor. The loss factor is a parameter used by glass
manufacturers to describe the dielectric properties of a glass. Figure 4.1 illustrates the
dependence of the noise of a patch pipette on the loss factor of the glass. Patch pipettes
fabricated from various glasses, coated with Sylgard to within 100 μm of the tip and filled
with solution were sealed to Sylgard lining the bottom of a fluid-filled chamber. The rms
noise was measured using an rms meter, with the pipette in air just above the bath and
then again with the pipette sealed to Sylgard. The 10 kHz rms noise in this figure was cal-
culated by subtracting the noise measured in air from the noise measured when the
pipette was sealed to Sylgard.

92 The Axon Guide — 2500-0102 Rev. C

 4.6. Noise Properties of Glasses

8161

LOSS FACTOR / WALL THICKNESS 3

7760, EG-6, 0120

1723 EG-16
7040, KG-12
0010
5
7052
7720, EN-1
3320, 7056
7
7052TW

7740
10

1720

R-6

30 0080

50
0.0 0.1 0.2 0.3 0.4 0.5 0.6

RMS NOISE AT 10 KHz

Figure 4.1: Noise properties of glasses.

The dependence of the rms noise of a patch pipette on the loss factor of the glass.

As seen in Figure 4.1, glasses with the lowest loss factor exhibit the lowest noise. Further-
more, noise decreases as the wall thickness increases (compare 7052 with 7052TW, thin
walled). Even greater improvement in noise can be achieved using quartz (fused silica)
pipettes. The noise of pipettes fabricated from Corning 7070 (low loss electrical) or Corn-
ing 1724 (aluminosilicate) glasses could also be improved if proper techniques of fabricat-
ing patch pipettes from these glasses would be developed. The noise values shown in
Figure 4.1 are higher than would be obtained with the Axopatch™ 200B amplifier in sin-

The Axon Guide — 2500-0102 Rev. C 93

 4. Microelectrodes and Micropipettes

gle-channel recording mode since its capacitive headstage produces lower noise than is
produced by standard resistive headstages. It has been observed that as the noise of the
headstage electronics decreases, the noise of other components, such as holder and glass, is
reduced as well.

4.7. LEACHABLE COMPONENTS

Glasses are complicated substances made of many different compounds as shown in
Table 4.2. While glasses have major constituents that lead to their classification as soda
lime, aluminosilicate, borosilicate, etc., they have many trace compounds whose location
in the glass is unknown. Glasses may have components on their surfaces that can leach
into an aqueous environment with which they are in contact. Leachable components
could be particularly problematic in single-channel and whole-cell recording due to the
close proximity of the channels to the glass. The literature contains reports of undesirable
effects of the leaching of constituents from the glass on membrane currents.

4.8. FURTHER READING

Cota, G., Armstrong, C. M. Potassium channel “inactivation” induced by soft-glass patch
pipettes. Methods in Enzymology. Plenum Press. New York and London, 1983.

Sakmann, B., Neher, E. Geometric parameters of pipettes and membrane patches. Single-
Channel Recording. Sakmann, B., Neher, E., Eds. pp. 37–51. Plenum Press. New York
and London, 1983.

Cota, G., Armstrong, C. M. Potassium channel “inactivation” induced by soft-glass patch

pipettes. Biophys. J. 53, 107–109, 1988.

Corey, D. P., Stevens, C. F. Science and technology of patch-recording electrodes. Single-

Channel Recording. Sakmann, B., Neher, E. Eds. pp. 53–68. Plenum Press, New York
and London, 1983.

Hammill, O.P., Marty, A., Neher, E., Sakmann, B., Sigworth, F.J. Improved patch-clamp
techniques for high resolution current-recording from cells and cell-free membrane
patches. Pflügers Arch. 391, 85–100, 1981.

Rae, J. L., Levis, R. A. Patch-clamp recordings from the epithelium of the lens obtained
using glasses selected for low noise and improved sealing properties. Biophys. J. 45,
144–146, 1984.

Rae, J. L., Levis, R. A. Patch voltage clamp of lens epithelial cells: theory and practice.
Molec. Physiol. 6, 115–162, 1984.

Rae, J. L., Levis, R. A., Eisenberg, R. S. Ion Channels. Narahashi, T. Ed. pp. 283–327.
Plenum Press, New York and London, 1988.

94 The Axon Guide — 2500-0102 Rev. C

 4.8. Further Reading

Ogden, D. C., Stanfield, P. R. Microelectrode Techniques: The Plymouth Workshop

Handbook. Standen, N. B., Gray, P. T. A., Whitaker, M. J. Eds. pp. 63–81. The Com-
pany of Biologists Limited, Cambridge, 1987.

Furman, R. E., Tanaka, J. C. Patch electrode glass composition effects on ion channel cur-
rents. Biophys J. 53, 287–292, 1988.

Rojas, L., Zuazaga, C. Influence of the patch pipette glass on single acetylcholine channels
recorded from Xenopus myocytes. Neurosci. Lett. 88: 39–44, 1988.

The Axon Guide — 2500-0102 Rev. C 95

 4. Microelectrodes and Micropipettes

96 The Axon Guide — 2500-0102 Rev. C

 5. Advanced Methods in
Electrophysiology
5.1. RECORDING FROM XENOPUS OOCYTES
The Xenopus oocyte expression system was introduced by Gurdon in 1971 as a means to
study various aspects of the control of gene expression. Injection of exogenous DNA into
the oocyte nucleus or mRNA into the cytoplasm led to the expression of functional pro-
teins by the oocytes. Early studies dealt with the expression of proteins that are of little
interest to electrophysiologists, such as globin, interferon and various viral proteins. Con-
sequently, electrophysiologists and biophysicists paid scant attention to studies of protein
expression in oocytes. However, beginning in 1982, Miledi and co-workers demonstrated
that various types of ion channels and receptors could also be expressed in oocytes after
injection of mRNAs that had been isolated from the appropriate tissue. For example,
injection of Torpedo electric organ mRNA led to the expression of functional nicotinic
acetylcholine receptors, while injection of rat brain mRNA resulted in the expression of a
large number of different voltage- and ligand-gated ion channels, among them the volt-
age-gated Na+ channels, NMDA and non-NMDA subtypes of glutamate receptors, and
GABAA receptors.

Since these initial experiments, a number of workers have adopted this expression system
to study various aspects of the structure and function of ion channels and receptors. Some
of the more common types of studies that have employed expression of exogenous ion
channels and receptors in oocytes include:

1 Analyzing the properties of mutated channels as a means to understand structure-

function relationships in ion channels;
2 Studying the post-translational processing and the assembly of multisubunit channels
and receptors;
3 Comparing the properties of channels from various tissues expressed in a common
environment;
4 Examining the modulation of channel and receptor function by various second
messenger systems;
5 Analyzing various aspects of receptor-effector coupling; and
6 Functional screening of cloned genes that encode channels and receptors (expression
cloning).

The Axon Guide — 2500-0102 Rev. C 97

 5. Advanced Methods in Electrophysiology

Most of these experiments involve electrophysiological recording from oocytes. This

describes the mechanics of recording from oocytes and points out some oocyte features of
which one should be aware in order to obtain high-quality recordings.

5.1.1. WHAT IS A XENOPUS OOCYTE?

Xenopus oocytes are egg precursors that, upon proper hormonal stimulation, pass
through the frog’s oviduct and become eggs, which can then be fertilized to make more
frogs. Oocytes are stored in the abdominal cavity and can be removed surgically for exper-
imental purposes. They go through six developmental stages (termed stages I to VI). Most
researchers use only the large stage V and VI oocytes, which can be used interchangeably
in electrophysiological experiments. Oocytes are found in clumps called ovarian lobes,
which are made up of oocytes, connective tissue, blood vessels and follicle cells. Figure 5.1
shows an individual stage V or VI oocyte as found in an ovarian lobe.

Follicle Cell Animal

Layer Pole

Vitelline Vegetal
Membrane Pole

Figure 5.1: Stage V or VI oocytes as found in an ovarian lobe.

The oocyte is a large cell, with a diameter of approximately 1–1.2 mm. It is surrounded by
the vitelline membrane, which is a glycoprotein matrix that gives the oocyte some struc-
tural rigidity and helps it maintain a spherical shape. The mesh formed by this matrix is
rather large; consequently, small molecules and even small proteins such as α-bungaro-
toxin (molecular weight 8,000 dalton) can interact with proteins on the oocyte surface.
Since devitellinized oocytes are extremely fragile, the vitelline membrane is usually
removed only for single-channel recording, where the surface of the oocyte must be suffi-
ciently clean to allow the formation of a gigohm seal. A layer of follicle cells surrounds the
vitelline membrane. The individual cells in the follicle cell layer are electrically coupled to
each other and to the oocyte through gap junctions, so that electrical events taking place
in the follicle cells can be detected in the oocyte.

The follicle cell layer is a potential source for many problems. Therefore, most researchers
remove it prior to mRNA injection or recording by one of two ways: 1) an extensive colla-

98 The Axon Guide — 2500-0102 Rev. C

 5.1. Recording from Xenopus Oocytes

genase treatment to completely strip off the layer by immersing the oocyte for 1–3 hours
in a calcium-free saline containing 2 mg/mL collagenase Type IA (Sigma Chemical Com-
pany, St. Louis, MO) until about half of the oocytes have been released from the ovarian
lobe; or 2) a less extensive collagenase treatment followed by manual removal of the follic-
ular layer with watchmaker’s forceps. The latter is preferred, since oocytes that undergo
the extensive collagenase treatment sometimes do not survive as long as those treated less
extensively. Some of the problems introduced by the follicle cell layer are rather trivial. For
instance, it is more difficult to impale a follicle-encased oocyte with an injection needle or
a microelectrode. Other complications, however, can create major problems. The electri-
cal coupling between the oocyte proper and the follicle cells means that one records from
both. Numerous examples of oocyte “endogenous” receptors that turned out to be in the
follicle cell layer were published. In fact, the Xenopus oocyte is a nearly ideal expression
system for ion channels and receptors since it has very few types of endogenous channels
and receptors, an advantage that is defeated if the follicle cells are left on the oocyte.

One of the striking features of a Xenopus oocyte is its two-toned color scheme: the animal
pole hemisphere is dark colored, while the vegetal pole is light colored. This polarity is
maintained inside the oocyte: the nucleus is found in the animal pole, and different popu-
lations of mRNAs exist in the hemispheres. In addition, a standing Cl- current flows from
one pole to the other, indicating that the distribution of channels in the oocyte membrane
is not homogeneous. Published reports showed that ACh receptors expressed from exoge-
nous mRNAs are also unevenly distributed; although receptors were expressed all over the
oocyte, more receptors were expressed on the vegetal pole. It is not clear if there is any
relation between the site of RNA injection and the location of the expressed channels on
the oocyte membrane. This non-homogeneity is not important for whole-cell recording,
but could be important for single-channel recording.

5.1.2. TWO-ELECTRODE VOLTAGE CLAMPING OF OOCYTES

Two factors need to be taken into consideration when voltage clamping oocytes. First,
with an apparent surface area on the order of 106 μm2, there is an enormous amount of
membrane that must be charged in order to clamp the cell. In fact, the situation is even
worse than it may appear since there is considerable invagination of the surface mem-
brane, thereby doubling or even tripling the surface area compared to an “ideal” spherical
oocyte. Second, one can encounter currents up to 10 μA or larger after mRNA injection,
potentially causing appreciable series resistance errors. In the case of currents from cloned
K+ and Na+ channels, which activate rapidly and can give rise to very large currents, there
may be an inadequate control of the voltage during the initial phases of channel activa-
tion.

The response time (τ) of a voltage clamp to a step voltage change is:

RI C m (1)
τ=
A

The Axon Guide — 2500-0102 Rev. C 99

 5. Advanced Methods in Electrophysiology

where RI is the resistance of the current-passing electrode, Cm is the membrane capaci-

tance, and A is the gain of the command amplifier. Since not much can be done about Cm
(“Combining Whole-Cell Clamp with Focal Current Recording” on page 111), the only two
things one can do in order to achieve fast clamping of the cell is to use the lowest RI and
largest A possible. One “advantage” of the oocyte’s large size is that one can use low-resis-
tance electrodes for both the voltage-recording and current-passing electrodes. (This can
hardly be regarded as an advantage, since the large size of the oocyte is what caused the
problem in the first place). Electrodes that have resistances of 0.5–2 MΩ when filled with
3 M KCl are used. The particular type of glass is unimportant; both regular and fiber-
filled electrode glass have been used successfully. Since the electrodes have rather large tip
openings, they usually do not clog very often; the same set of electrodes can be used
repeatedly during the course of a day. The electrodes’ low resistance eliminates the need in
a negative-capacitance circuit to correct for the frequency response of the voltage-record-
ing electrode, thereby eliminating this potential source of clamp instability. To achieve the
fastest response, the electrodes must be shielded in order to reduce capacitive coupling
between them. Two rather simple ways of doing this are 1) Placing a grounded sheet of
metal between the two electrodes (making sure that it doesn’t make contact with the
bath), or 2) wrapping a wire around a piece of tubing that fits over the voltage electrode
like a sleeve, and then grounding the wire. In the latter case, the shield should extend as
close to the bath as possible.

The other parameter that one can control in order to maximize the speed of the voltage
clamp is the command amplifier gain. As with other types of cell-electrode combinations,
one must usually introduce a frequency response compensation network to ensure clamp
stability at high gains. Therefore, the absolute gain of the clamp and the command voltage
of the amplifier are less important than the maximum gain reached before the clamp
becomes unstable. The LAG and (AC) GAIN controls of the Axoclamp™ 900A micro-
electrode amplifier allow one to fine-tune the clamp. For oocytes, start with the LAG at
about 0.2 ms and the GAIN set to its lowest position, and gradually increase it as much as
possible until the clamp starts oscillating. The lower the LAG setting, the faster the clamp.
Axoclamp 900-series amplifiers include an Oscillation Detection circuit that automati-
cally reduces the voltage clamp gain when oscillation in the circuit is detected.

The goal is to achieve a fast voltage clamp. But, how fast is fast enough? The actual speed
required depends on the type of signal being measured. For example, when studying a
ligand-activated ion channel that is activated by bath application of an agonist, it makes
no difference if the clamp settles in 2 ms or 20 ms, since the exchange time of the cham-
ber is the limiting factor in the rate of activation. On the other hand, when studying volt-
age-gated channels, one wants the fastest clamp attainable. A well-tuned, two-
microelectrode voltage clamp can clamp an oocyte fast enough to study the gating kinetics
of most voltage-gated channels. However, it is probably impossible to clamp an oocyte fast
enough to study the activation kinetics of fast-gating voltage-gated Na+ channels in which
the rising phase of the current can be on the order of 1 ms or shorter. It is still possible to
study slower processes, such as inactivation or pharmacological properties, even if the
clamp is not fast enough to study the activation process. A faster clamp can be achieved

100 The Axon Guide — 2500-0102 Rev. C

 5.1. Recording from Xenopus Oocytes

either by recording from smaller oocytes (stage II or III) whose capacitances are approxi-
mately one-fifth those of the larger stage V and VI oocytes, or applying the more common
“macropatch” technique. Either of these two approaches provides a sufficiently fast clamp
for studying the kinetics of Na+ channel activation.

Electrode penetration is most easily achieved simply by advancing the electrode into the
oocyte until it dimples the membrane and then visibly pops into the cell. Due to the large
size of the oocyte, a coarse, inexpensive manipulator provides adequate control of move-
ment. Typical resting potentials in a physiological saline solution (100 mM NaCl,
2 mM KCl, 1.8 mM CaCl2, 1 mM MgCl2, 5 mM HEPES, pH 7.6) are approximately -
40 mV or more negative. The voltage clamp is tuned by applying a small (2 mV) square
wave at the resting potential and then gradually increasing the gain of the clamp. As the
gain increases, the voltage trace “squares up” and the capacitive transient in the current
trace gets faster and faster. Instability in the clamp will first appear as small oscillations in
the traces; the oscillations will increase as the gain increases. If oscillations do appear, the
TIME CONSTANT control should be adjusted until they disappear. This process should
be continued until one reaches the highest gain possible without any apparent oscillations.
In general, if the electrodes are shielded as described above, full gain can be achieved.

5.1.3. PATCH CLAMPING XENOPUS OOCYTES

While oocytes are obviously too large for the whole-cell mode of patch clamping, they are
amenable to all three configurations of single-channel recording, i.e., cell-attached, inside-
out and outside-out. One of the major requirements for achieving the high-resistance seals
necessary for single-channel recording is a clean membrane surface. With oocytes, one
must remove the vitelline membrane prior to patch clamping by placing the oocyte in a
hypertonic stripping solution (usually 200 mM K-aspartate, 20 mM KCl, 1 mM MgCl2,
10 mM EGTA, 10 mM HEPES, pH 7.4) and allowing it to shrink. As the oocyte shrinks,
the vitelline membrane detaches from the cell membrane and appears as a transparent
sphere around the oocyte. One can then manually remove the vitelline membrane using
fine watchmaker’s forceps. Since devitellinized oocytes will stick to glass or plastic (but not
to agarose) after a few minutes of contact, the shrinking and vitelline membrane removal
is usually carried out in small petri dishes with a layer of 2% agarose on the bottom. This
allows one to prepare a number of oocytes at once. At this point, the oocytes are extremely
fragile. In particular, stripped oocytes have a propensity for disintegrating when they are at
an air-water interface; therefore, extreme care must be taken when transferring them to
the recording chamber.

Once the vitelline membrane has been removed, gigohm seals can be obtained with high
success rates. Oocytes, like most cells, contain stretch-activated channels that can interfere
with recording “interesting” channels in the cell-attached mode. Therefore, most research-
ers prefer to use excised patches for single-channel work, where the incidence of stretch-
activated channel currents is much lower. The actual mechanics of obtaining excised
patches is the same as for any cell type, with the exception that some researchers prefer to
rupture the membrane by applying positive pressure rather than suction, since suction can
sometimes cause yolk platelets to clog the pipette tip.

The Axon Guide — 2500-0102 Rev. C 101

 5. Advanced Methods in Electrophysiology

5.2. FURTHER READING

Barnard, E.A., Miledi, R., Sumikawa, K. Translation of exogenous messenger RNA cod-
ing for nicotinic acetylcholine receptors induces functional receptor in Xenopus oocytes.
Proc. R. Soc. Lond. B 215:241–246, 1982.

Gurdon, J.B., Lane, C.D., Woodland, H.R., Marbaix, G. Use of frog eggs and oocytes for
the study of messenger RNA and its translation in living cells. Nature. 233:177–182,
1971.

Krafte D., Lester, H.A. Expression of functional sodium channels in stage II–III Xenopus
oocytes. J. Neurosci. Meth. 26:211–215, 1989.

Methfessel, C., Witzemann, V., Takahashi, T., Mishina, M., Numa, S., Sakmann, B.
Patch clamp measurements on Xenopus laevis oocytes: currents through endogenous
channels and implanted acetylcholine receptor and sodium channels. Pflügers Arch.
407:577–588, 1986.

Snutch, T.P. The use of Xenopus oocytes to probe synaptic communication. Trend Neu-
rosci. 11:250–256, 1988.

Stühmer, W., Methfessel, C., Sakmann, B., Noda, M. Y., Numa, S. Patch clamp charac-
terization of sodium channels expressed from rat brain cDNA. Eur. Biophys. J. 14:131–
138, 1987.

Stühmer, W., and Parekh, A.B., Electrophysiological Recordings from Xenopus Oocytes
in Single-Channel Recording, 2nd Ed., Sakmann, B., and Neher, E., eds., pp. 341–355,
1995.

5.3. PATCH-CLAMP RECORDING IN BRAIN SLICES

Although it is commonly accepted that patch-clamp techniques offer many technical
advantages over conventional intracellular microelectrode recording configurations, one
nagging limitation has been the belief that, in order to allow formation of gigohm seals
with patch electrodes, cells must be treated in some way that leaves their membrane
“clean.” Such treatment might entail dissociation of tissue by enzymatic digestion or
mechanical disruption, and might or might not be followed by growth in tissue culture.
These techniques invariably result in significant damage to the cells or severe alteration of
their environment, at least transiently, during the preparation of the cells for recording.
Alternatively, tissue explants or slices can be maintained in “organotypic” culture, allowing
many of the cell-cell interactions to be maintained and mitigating some of the problems
described above. However, with any of these approaches, the possibility remains that sig-
nificant alteration of cellular properties occurs prior to recording.

Therefore, it is of great interest to be able to apply patch-recording techniques to neurons

in acute tissue slices. The marriage of these two approaches offers many of the advantages
of both, with few of the limitations:

102 The Axon Guide — 2500-0102 Rev. C

 5.3. Patch-Clamp Recording in Brain Slices

1 Cells in tissue slices are likely to be much closer to their original state than cells
subjected to the above-mentioned treatments. No disruption of the normal cellular
environment need take place until the preparation of slices and disruption is limited
to the surface of the slice.
2 The increased signal-to-noise ratio and improved voltage-clamp quality, compared
with conventional intracellular recording using sharp microelectrodes, provide
substantial benefits for cellular electrophysiology. These benefits are particularly
helpful when measuring small or rapid events.
3 The ability to control the intracellular environment is greatly enhanced by whole-cell
recording, thus facilitating the study of interactions between intracellular
biochemistry and electrophysiology.
It should be noted however, that not all problems are solved by applying whole-cell tech-
niques. “Washout” of intracellular biochemical machinery may affect electrophysiological
properties, and control of membrane potential in branched or elongated neurons is often
inadequate.

Two strategies have been applied in obtaining patch-clamp recordings from tissue slices.
One approach, the “cleaning” technique, involves removing overlying neuropil or cellular
debris from a visually-identified cell using a relatively large pipette filled with the same
solution used to superfuse the slice. Recording is then possible in a manner similar to
standard patch recording. In the other approach, called the “blind” technique, the record-
ing pipette is lowered into a tissue slice without high-magnification visual guidance; seal
formation is monitored using electrical measurements.

5.3.1. THE CLEANING TECHNIQUE

Tissue preparation for this technique uses procedures slightly modified from those used
for standard slices. The slices must be thin enough to allow good cellular visibility (100–
300 μm, depending on the age of the animal), but thicker slices result in a greater number
of undamaged cells. Generally, vibrating microtomes are preferable to tissue choppers
because their use results in less dead or damaged tissue at the cut surface, and because they
facilitate preparation of thin slices. Special care should be taken when preparing thin slices
to ensure that the bathing medium remains ice cold in order to maintain the necessary
firm tissue consistency. Following their preparation and prior to recording, slices are com-
monly incubated at 37 °C for about one hour. This procedure softens the tissue and facil-
itates subsequent cleaning of cells. Various methods may be used for storing slices, but
care is needed to avoid damage to the cut surfaces resulting from procedures such as plac-
ing slices on filter paper. Slices may be stored at room temperature or warmed to
30–35 °C.

The only major piece of equipment required for this technique, in addition to a standard
recording setup, is an upright microscope equipped with a water-immersion objective,
most commonly 40x magnification. Use of an inverted microscope is not feasible since,
even if very thin slices are used, visibility through the tissue to a recording pipette is signif-

The Axon Guide — 2500-0102 Rev. C 103

 5. Advanced Methods in Electrophysiology

icantly impaired. Hoffman or Nomarski differential-interference-contrast optics can be

helpful, but are not necessary. Unless manipulators are mounted on the stage, it is
extremely helpful to use a fixed-stage microscope, so that changing focus does not cause
movement of the tissue with respect to cleaning or recording pipettes.

An objective with high numerical aperture is important. A commonly used apparatus is

the Zeiss 40x water-immersion objective, which has a numerical aperture of 0.75. This
objective offers a working distance of approximately 1.6 mm. The pipettes must, there-
fore, be placed at a very shallow angle (15° from the horizontal is common). One consid-
eration in mounting micromanipulators is that the movement in true horizontal and
vertical directions is preferable to a coordinate system with an axis oriented parallel to the
pipette. An advantage of this arrangement is the ability to lower the recording pipette
directly onto the target cell in a single motion.

The recording chamber must allow continuous superfusion of the slice; it must also per-
mit immobilization of the tissue despite the fluid flow, and do so using a device that does
not interfere with the objective or the pipettes. One solution to this problem is a net con-
sisting of a U-shaped piece of flattened platinum wire with a parallel array of very fine
nylon monofilaments, such as stocking fibers, glued across the arms of the U-shaped wire
(Edwards et al., 1989; Sakmann et al., 1989). The net can be placed on the slice and is
usually heavy enough to prevent movement. Although the filaments may cause some
damage to the surface of the tissue, they may be placed sufficiently far apart to leave con-
siderable working space (the precise distance depends on the cellular architecture in the
slice). Another solution is to attach the slice to the bottom of the recording chamber using
a plasma clot (Blanton et al., 1989).

The pipette used for cleaning should have a tip much larger in diameter than a recording
pipette, usually in the range of 5–20 μm. The optimal tip size depends on several factors;
in general, larger tips are preferable for larger cells and for deeper cleaning. However, the
consistency of the tissue and cell density may also affect the choice of tip diameter. For
example, with particularly “sticky” tissue, such as neocortex, small cleaning pipettes may
tend to become clogged with membranous debris. For a given preparation, it is best to
arrive empirically at an appropriate size. Additionally, since debris tends to accumulate on
the jagged edges of broken pipettes, it is usually best to pull unbroken pipettes to the
appropriate size. To minimize any damage that the process might inflict on the cell, the
cleaning pipette should be filled with the same solution used for the bathing medium.
Note that some researchers, instead of using a separate cleaning electrode, perform the
cleaning with the same recording electrode for the sake of expediency.

The first step in establishing a recording is to locate a healthy cell; it is usually possible to
distinguish cells in good condition by their appearance of solidity or opacity. In such neu-
rons, it is often possible to follow dendritic processes for some distance. Some healthy cells
may be observed whose somata are on the surface of the slice and which display a “clean”
membrane suitable for immediate patching with no further cleaning. Such neurons often
have dendrites that can be seen projecting into the depths of the slice. It seems that cells

104 The Axon Guide — 2500-0102 Rev. C

 5.3. Patch-Clamp Recording in Brain Slices

that have lost most of their dendrites due to the slicing procedure are often “sick,” as char-
acterized by a granular and/or transparent appearance of the soma.

Once a healthy cell has been located, the next step is to remove the overlying neuropil.
Mouth control of the pressure applied to the back of the pipette permits rapid, highly pre-
cise application of pressure and suction. It is often best to begin by applying positive pres-
sure gently with the pipette placed just at the surface of the slice and directly above the
targeted cell. Blowing accomplishes two goals. First, it provides information about the
consistency of the tissue, which may be rather variable from slice to slice and which differs
dramatically from one brain region to another. Second, application of positive pressure
begins the process of disrupting the tissue overlying the desired cell. When tissue has been
visibly disrupted, slight suction helps remove the loosened debris (Figure 5.2). Repeated
application of this blowing and sucking cycle, with appropriate movement of the pipette
to remove specific chunks of tissue, eventually results in a clean cell.

cleaning
pipette

brain tissue
slice

positive pressure
(blowing)

negative pressure
(sucking)

Figure 5.2: The Cleaning technique.

Use of a cleaning pipette to remove tissue overlying a cell in tissue slices. Repro-
duced with permission according to Edwards et al., (1989).

The Axon Guide — 2500-0102 Rev. C 105

 5. Advanced Methods in Electrophysiology

At first, this procedure may take a long time. It is almost inevitable that an inexperienced
experimenter, who has not yet developed a feel for the tissue, will kill cells by applying
pressure or suction that is too vigorous. With practice, however, one can clean even deep
cells quite rapidly. Careful and repeated adjustment of the focus aids greatly in determin-
ing the degree of cleaning required at each stage. Depending on the particular type of tis-
sue, a few tricks to facilitate and accelerate the process of producing clean cells may be
applied. For example, in regions of densely packed somata, such as the CA1 pyramidal cell
layer of the hippocampus, it is possible to remove tissue at the slice surface over a relatively
large length of the cell layer by applying continuous suction and moving the pipette along
the layer as if using a vacuum cleaner to remove dead tissue. Inspection of the underlying
neurons often reveals at least one clean pyramidal cell.

After completing the cleaning procedure, patch-clamp recordings of cleaned cells in slices
are carried out following essentially the same procedures that would be applied to cells in
culture. All four major patch-recording configurations—cell-attached, inside-out, out-
side-out and whole-cell—are possible, as well as the nystatin-perforated patch technique
(see Section 5.8., “Recording from Perforated Patches and Perforated Vesicles”).

5.3.2. THE BLIND TECHNIQUE

Apart from a microelectrode holder with the appropriate diameter for patch pipettes, and
with a side port for pressure application, no special equipment is necessary to adapt a stan-
dard slice setup equipped for intracellular recording to one suitable for blind-patch
recording. A dissecting microscope is sufficient. The recording pipette should have a resis-
tance in the bath of approximately 3–5 MΩ; the optimal resistance depends on the size
and type of cell being recorded.

The procedure for establishing gigohm seals is very simple (Figure 5.3). It is possible to
use either the current-clamp or voltage-clamp recording mode. A small repetitive current
or voltage pulse is applied to the electrode at relatively high frequency (e.g., 10 Hz) and
the voltage or current response is monitored with an oscilloscope. Continuous, slight pres-
sure is applied to the back of the pipette, for example, by blowing gently into the tubing
connected to the pipette and closing a stopcock to retain the pressure. The pipette is then
slowly lowered into the tissue in the region where the desired cells are found. It is
advanced slowly—1–5 steps per second, each step of 1–2 μm—into the tissue until a sud-
den increase in resistance is detected at the pipette tip, indicating contact with a cell.
Often a further advance of 1–2 μm causes a further increase in the resistance. Release of
the positive pressure should result in another approximate doubling of the resistance if the
electrode is indeed contacting a cell. Although gigohm seals sometimes form spontane-
ously at this stage, the next step is to apply light suction to cause the seal to form. This
may occur rapidly, in less than one second, or slowly, up to several seconds, requiring con-
stant, gentle suction. The patch is then ruptured by additional suction creating electrical
continuity between the pipette and cell interior.

106 The Axon Guide — 2500-0102 Rev. C

 5.3. Patch-Clamp Recording in Brain Slices

Approach neuron

Light suction

Gigohm seal

Rupture patch

Figure 5.3: The Blind technique.

Obtaining a recording using the blind technique by advancing the patch pipette
through a tissue slice. According to the method of Blanton et al., (1989).

Although this technique can be applied successfully, achieving a good recording on nearly
every attempt, it is almost inevitable that at some point one will encounter some difficulty
either in generating seals or establishing whole-cell recordings. Through experience, it is
possible to identify the problem-causing aspects of the procedure. Often, changing the
pipette size helps. A general rule is to use smaller tips to improve sealing, and larger ones
to facilitate the transition to the whole-cell mode. One should be prepared to use a num-
ber of pipettes, since the probability of obtaining a recording on the second attempt to
apply suction to the same pipette is very small. Another critical variable, which applies to
all patch-clamp recording, is the intracellular solution whose osmolarity should be verified
routinely, particularly when difficulties are encountered.

The Axon Guide — 2500-0102 Rev. C 107

 5. Advanced Methods in Electrophysiology

5.3.3. ADVANTAGES AND DISADVANTAGES OF THE TWO METHODS OF PATCH

CLAMPING BRAIN SLICES
It is impossible to give a complete discussion of the relative merits of the cleaning and
blind techniques here. However, some of the important considerations are outlined in the
following table.
Table 5.1: Brain Slice Techniques.

Cleaning Technique Blind Technique

• Need upright microscope and water- • Need no special equipment beyond that
immersion objective. for intracellular recording in slices.

• Vibrating microtome almost essential. • Vibrating microtome strongly

recommended, essential for some
tissues.

• Maximum slice thickness limited by • Maximum slice thickness limited by

visibility. viability.

• Once a cell is located, additional time is • Less preparation time for each
invested in cleaning. penetration, but no certainty of cell
health.

• Cell is visually identified. • Need post-hoc identification of cell type

unless electrophysiological
identification is possible or a
homogeneous cell population is used.

• Can see neighboring cells and apply • Visually-guided local stimulation is not
local stimuli. possible.

• All four standard patch-clamp • All four standard patch-clamp

configurations possible, plus the configurations possible, plus the
perforated patch technique. perforated patch technique.

• Recording quality similar to standard • High or unstable access resistance is

patch clamp techniques. often a problem with whole-cell mode
(possibly due to debris in the pipette
tip).

108 The Axon Guide — 2500-0102 Rev. C

 5.4. Further Reading

5.4. FURTHER READING

Blanton, M.G., Loturco, J.J., Kriegstein, A.R., Whole cell recording from neurons in
slices of reptilian and mammalian cerebral cortex. J. Neurosci. Methods. 30:203–210,
1989.

Coleman, P.A., Miller, R.F., Measurements of passive membrane parameters with whole-
cell recording from neurons in the intact amphibian retina. J. Neurophysiol. 61:218–230,
1989.

Edwards, F.A., Konnerth, A., Sakmann, B., Takahashi, T., A thin slice preparation for
patch clamp recordings from neurones of the mammalian central nervous system. Pflügers
Arch. 414:600–612, 1989.

Gähwiler, B.H., Organotypic cultures of neural tissue. Trends Neurosci. 11:484–489,

1988.

Hestrin, S., Nicoll, R.A., Perkel, D.J., Sah, P., Analysis of excitatory synaptic action in the
rat hippocampus using whole cell recording from thin slices. J. Physiol., 422, 203–225,
1990.

Rall, W., Segev, I., Space clamp problems when voltage clamping branched neurons with
intracellular microelectrodes. T.G.J. Smith, H. Lecar, S.J. Redman, P.W. Gage (Eds.),
Voltage and Patch Clamping With Microelectrodes. Bethesda: American Physiological
Society, pp. 191–215, 1985.

Sakmann, B., Edwards, F., Konnerth, A., Takahashi, T., Patch clamp techniques used for
studying synaptic transmission in slices of mammalian brain. Q J Exp Physiol 74: 1107–
18, 1989.

Sakmann, B., and Stuart, G., Patch-pipette Recordings from the Soma, Dendrites, and
Axon of Neurons in Brain Slices in Single-Channel Recording, 2nd Ed., Sakmann, B.,
and Neher, E., eds. pp. 199–211, 1995.

5.5. MACROPATCH AND LOOSE-PATCH RECORDING

Measuring the local density of a specific type of ion channel in different regions of a cell
provides valuable information. This kind of information is helpful for understanding the
specialized functions of axons, cell bodies and dendrites in neurons, as well as for studying
mechanisms of sensory transduction, the spatial distribution of ion channels near the
muscle end-plate, and localized changes in channel density or channel properties resulting
from single-transduction events. To efficiently address these areas of research, one needs a
method that is intermediate in scale between single-channel patch clamp and whole-cell
voltage clamp. Macropatch recording methods fulfill this need.

Two approaches are used to record ionic currents from patches of membrane containing
tens to hundreds of ion channels. The first applies standard gigohm-seal (gigaseal) patch-
clamp methods employing large-diameter patch pipettes (5–10 μm tip diameter) and is

The Axon Guide — 2500-0102 Rev. C 109

 5. Advanced Methods in Electrophysiology

used to sample currents from much larger areas than are commonly used in single-channel
recording (which is done with 0.5–2 μm tip diameter pipettes). The second method of
macropatch recording has been termed loose-patch recording. This method, which
enables one to gather data that could not be obtained with the gigaseal method, differs
from single-channel recording in three ways:

1 The tip of the loose-patch pipette is usually much larger than the conventional single-
channel patch pipette;
2 The loose-patch pipette does not form a gigaseal with the membrane, whereas gigohm
seals are required for single-channel recording; and
3 The loose-patch pipette can be reused and repositioned to map the spatial distribution
of channel subtypes on individual cells (Thompson and Coombs, 1988).

5.5.1. GIGASEAL-MACROPATCH VOLTAGE CLAMP

Macroscopic patch currents can be recorded using standard patch-clamp instrumentation
and a large patch pipette having a resistance of a few hundred kilohms. A gigaseal is
formed using suction. This method can be applied to cell-attached and excised patches.
Data analysis uses the same methods that are applied in whole-cell clamp experiments.
Consequently, data concerning the kinetic behavior and local amplitude of ionic currents
can be rapidly acquired. This method has been used to record macroscopic currents in
Xenopus oocytes (Hoshi et al., 1990), which are particularly amenable to this method of
recording because channel expression can be controlled, to a certain degree, by the experi-
mental conditions.

There are a few problems associated with this method that have to be considered in the
experimental design. The patch-clamp amplifier must be able to fully compensate for the
greater capacitance of larger patches. When voltage pulses are applied to macropatches in
the cell-attached configuration, current flowing in the patch will change the cell voltage;
this change can be significant. For example, with a macropatch current of 100 pA and a
cell input resistance of 100 MΩ, the cell voltage will change by 10 mV. Dynamic current-
dependent errors of this type are difficult to correct and are most severe in small cells. In
some preparations it may be impossible to use cell-attached macropatches to obtain accu-
rate I/V curves, especially for rapidly changing currents. One can eliminate this source of
error by studying excised patches.

Milton & Caldwell (1990) identified a more subtle error associated with macropatch
recording. They observed that the suction used to form the seal between the patch pipette
and the membrane can cause a change in ion channel density in the patch. With large
patch pipettes, a noticeable membrane bleb may form in the orifice of the pipette during
suction. This membrane bleb can be visualized under a microscope. It appears that the
ion-channel density decreases during bleb formation, possibly because of a flow of lipid
into the bleb causing membrane rearrangement and changes in channel distribution.
Therefore, the gigaseal-macropatch method may be unsuitable for mapping ion-channel

110 The Axon Guide — 2500-0102 Rev. C

 5.5. Macropatch and Loose-Patch Recording

densities because the procedure modifies the channel density before the measurement is
made.

5.5.2. LOOSE-PATCH VOLTAGE CLAMP

In loose-patch recording it is not necessary to form a tight seal between the patch pipette
and the plasma membrane; therefore, strong suction is not needed. This has at least two
advantages:

1 The pipette can be repositioned to sample currents from a number of patches on the
same cell; and
2 The loose-patch method can be used to measure local currents in voltage-clamped
preparations where establishing gigaseals is impossible due to the presence of adherent
connective tissue, basement membrane or glial cells. Note that loose-patch recording
is done only in the cell-attached configuration.
The low seal resistance between the pipette and the membrane presents the principal tech-
nical difficulty one must overcome in applying loose-patch methods. Low seal resistance
creates two types of problems. First, the current flowing through the seal resistance and
capacitance introduces noise in the current record, thus limiting the resolution (see Chap-
ter 3, “The Importance of a Good Seal”). A more significant problem arises from the fact
that membrane current flowing through the seal resistance is lost.

The following discussion will focus on the errors introduced by low seal resistance and the
procedures used to limit errors to acceptable levels. Three designs are available to combat
the problems associated with low seal resistance in loose-patch recording. The choice of
the most appropriate method depends on the nature of the preparation. The original liter-
ature should be consulted for further details.
Combining Whole-Cell Clamp with Focal Current Recording
The first method to be introduced combines whole-cell voltage clamping with loose-patch
current recording. This is the preferred method for many applications. The cell is voltage
clamped by a microelectrode or a whole-cell patch pipette using standard microelectrode
voltage-clamp or patch-clamp instrumentation, while current is recorded from a restricted
area of membrane with a loose-patch pipette using patch-clamp instrumentation. Voltage-
clamp pulses are applied via the whole-cell clamp while the loose-patch pipette is main-
tained at the bath potential to ensure that no current flows across the seal resistance. This
method has been used to record local currents in large molluscan neurons (Neher and
Lux, 1969; Johnson and Thompson, 1989; Thompson and Coombs 1989; Premack et
al., 1990) and in muscle cells (Almers et al., 1983). The method is particularly well suited
for measuring the kinetics and current/voltage relationships of macroscopic currents
expressed in Xenopus oocytes where a whole-cell voltage clamp is severely limited by the
large capacitance of the cell and high intracellular series resistance. The electrode arrange-
ment is shown in Figure 5.4.

The Axon Guide — 2500-0102 Rev. C 111

 5. Advanced Methods in Electrophysiology

Vc Vo
A

Patch Clamp
Im Amplifier

Ip
Cc
Rp
R ei R sl
R sr
Bath Clamp
C mp Amplifier

Figure 5.4: Combining whole-cell voltage clamping with loose-patch current recording.

Schematic diagram of a cell with voltage clamp electrodes and the large patch pipette in
place. The symbols used in the figure are as follows: A, voltage clamp open loop gain; Cc,
coupling capacitance between voltage clamp current electrode and patch electrode; Cmp,
capacitance of membrane patch; Ic, current error; Im, membrane current supplied by volt-
age clamp; Ip, patch membrane current supplied by patch clamp amplifier; Rei, resistance
of voltage clamp current electrode; Rp, axial resistance of patch electrode; Rsl, resistance of
seal between patch pipette and cell membrane; Rsr, series resistance between the outside of
the patch electrode and the bath clamp amplifier; Vc, voltage clamp command voltage;
Vo, voltage clamp output voltage. Reproduced with permission according to Johnson and
Thompson (1989).

A reusable loose-patch pipette (5–20 μm in diameter; axial resistance (Rp) of

100–500 kΩ) is used in cell-attached configuration. The resistance of the seal is moni-
tored by applying test voltage steps via the patch amplifier and measuring the amplitude
of the resulting current steps. After contacting the cell, gentle suction is applied until the
sum of the seal resistance (Rsl) and the axial resistance of the pipette (Rp) increases by at
least 10 fold; increases greater than 100-fold are common. The suction is released to avoid
pulling a “finger” of membrane and cytoplasm into the pipette since this would increase
the series resistance and lead to errors in voltage control and current measurement. Since
the lumen of the loose-patch pipette is held at the bath potential and Rp is low compared
to Rsl, there is no need to compensate for the finite value of Rsl.

This method allows one to simultaneously measure whole-cell and patch currents as well
as whole-cell and patch capacitances. The areas of the cell and of the patch can be esti-
mated by measuring the whole-cell and patch capacitances. A convenient way to do this is
to apply a triangle-wave voltage command to the whole-cell voltage clamp and record the

112 The Axon Guide — 2500-0102 Rev. C

 5.5. Macropatch and Loose-Patch Recording

resulting whole-cell and patch current waveforms. The patch capacitance (c) is calculated
from the change in the time derivative of the voltage (dV/dt) at a vertex of the triangle
wave and the corresponding jump in patch current (I), using the following equation
(Neher, 1971; Palti, 1971):

I
c= (2)
dV dt

Because the voltage of the patch pipette is equal to the bath voltage and only the voltage
across the membrane patch covered by the pipette is changing, the capacitance of the
patch pipette itself is rejected. The capacitance of the whole cell is measured in a similar
fashion. However, the interpretation is complicated by current contributed from poorly
clamped regions (see Johnson and Thompson (1989) for further discussion).

There are several issues to consider when applying this combined voltage-clamp method.
The most important requirement is that the patch current must be measured from a well-
clamped region of membrane, i.e., from a region that is “space clamped” by the whole-cell
voltage clamp amplifier. In order to accurately measure current-voltage relationships and
limit the capacitive current flowing across the patch during voltage steps, the loose-patch
pipette should be positioned close to the point where the whole-cell voltage is controlled.
Cell geometry and electrode positioning should be considered when designing the physi-
cal layout of the experiment. These considerations limit the kinds of preparations that can
be studied with the method. It is reasonable to use the loose-patch pipette to record local
current densities at different points on a well-clamped spherical cell body. However, it
would be inappropriate to record patch currents from a thin dendrite when the voltage
clamp is applied to the soma.

The most significant error results from current flowing in the seal resistance between the
loose-patch pipette and the bath. This error current (Ie), which is not measured by the
patch-clamp amplifier, is given by:

I p Rp
Ie = (3)
R p + Rsl

where Rp is the resistance of the loose-patch pipette and Rsl is the seal resistance. If
(Rp + Rsl) is ten times larger than Rp the error in the measured current is 10%. To mini-
mize the error, one can reduce Rp by using a patch pipette with a blunt taper and increase
Rsl by using suction to improve the seal with the membrane. Enzyme treatment with dis-
pase, pronase or trypsin can be used to clean the cell surface and improve the seal resis-
tance.

The Axon Guide — 2500-0102 Rev. C 113

 5. Advanced Methods in Electrophysiology

Two kinds of errors caused by series resistance should be considered:

1 The voltage drop that results from macroscopic patch currents flowing through the
axial resistance of the patch pipette (Rp) will produce an insignificant error in most
cases. For example, if the value of Rp is 300 kΩ and the maximal patch current is
500 pA, the resulting voltage error is 0.15 mV.
2 A more troublesome error occurs when the bath voltage outside the patch pipette
changes due to large whole-cell currents flowing in the series resistance (Rsr) between
the cell membrane and the bath ground or virtual ground.
The error current is given by:

I m Rsr
Ie = (4)
Rsl

where Rsr is the sum of the series resistances of the bath, the agar bridges and the liquid-
silver chloride junction at the bath electrode; it typically has a value of several kilohms.

Bath series resistance creates an insidious problem when membrane currents are large.
However, the error can be reduced by measuring the bath voltage with a separate electrode
and using it as the reference voltage at the patch-clamp amplifier. Good practice dictates
that the bath voltage be used to correct the membrane voltage signal applied to the whole-
cell clamp amplifier as well. Instability can result, however, if the bath voltage is measured
with a greater bandwidth than the membrane voltage. To remedy this, one can limit the
frequency response of the bath voltage follower, but this degrades performance and causes
other errors. A better solution is to eliminate the need to subtract the bath voltage at the
other circuits by voltage clamping the bath to ground potential. This can be accomplished
using an independent voltage-clamp amplifier operating at a high gain and bandwidth.
Two separate agar bridges and silver-chloride electrodes can be used to measure the bath
voltage and pass current to the bath. Because of the low resistance of these electrodes, a
simple voltage-clamp circuit employing a single operational amplifier is adequate for
clamping the bath to ground potential. This arrangement ensures that the bath voltage
does not change during clamp steps. The output of the circuit provides the record of
whole-cell current.

Other problems can be overcome by a careful design of the experimental layout. Since this
method uses several voltage-clamp amplifiers working simultaneously, it is necessary to
limit capacitive coupling between electrodes. This can be achieved by inserting a
grounded shield between the whole-cell current electrode and the macropatch pipette,
and by lowering the volume of saline to the minimum required for covering the cell. To
reduce high frequency noise in the patch recording, the submerged tip of the pipette
should be coated with Sylgard (Dow Corning, Midland, MI) and only the tip should be
filled with saline. Johnson and Thompson (1989) provide a thorough analysis of the prob-
lems and the approaches used to reduce errors to acceptable levels.

114 The Axon Guide — 2500-0102 Rev. C

 5.5. Macropatch and Loose-Patch Recording

Voltage Pulses Applied to the Loose-Patch Pipette

Ionic currents can also be studied by applying voltage pulses directly to the loose-patch
pipette, thus eliminating the need for an independent whole-cell clamp. Errors resulting
from the finite seal resistance and the axial resistance of the pipette are more severe in this
case and require more elaborate compensation. The advantage of this method is that
membrane currents can be recorded locally without the need to establish an overall space
clamp. This loose-patch method has been very successful in studies of acetylcholine (ACh)
receptor and Na+-channel distribution near muscle end-plates (Almers et al., 1983a,b).

Although extra effort is required with this method, it may be the best approach for some
preparations. It is important to remember that current flowing across a macropatch can
cause a significant change in cell voltage. In its simplest form, this method uses a large
diameter patch pipette in a cell-attached configuration and a conventional patch-clamp
amplifier (Almers et al., 1983a,b; Stühmer et al., 1983). In this configuration, the axial
resistance of the pipette (Rp) and the seal resistance (Rsl) act as a voltage divider. When a
voltage command pulse is applied to the pipette (Vp), it will be attenuated at the pipette
tip. The voltage at the pipette tip (Vt) is given by (Stühmer et al., 1983):

Rsl
Vt = V p (5)
R p + Rsl

When Rsl is large, as is the case in gigaseal recordings, the error is insignificant. In loose-
patch applications, however, Rp and Rsl are approximately equal and a large fraction of the
current delivered to the pipette will flow through the seal resistance. It is necessary to cor-
rect the amplitude of command-voltage steps by a factor equal to Rsl/(Rp+Rsl). To do so,
Rp is measured before the pipette contacts the cell and Rp + Rsl is monitored continually
throughout the experiment. Almers et al., (1983a,b) used a computer to calculate a cor-
rection factor A = Rsl/(Rp+Rsl) before each voltage step. The correction factor was used to
scale command-voltage steps applied by the computer by 1/A to ensure that the intended
voltage appeared at the pipette tip. Recorded currents were divided by A to correct for
current (often termed leakage current) flowing through the seal resistance. This digital
approach to seal resistance compensation is technically sound. To implement it, the user
has to develop a software program for continuously updating the parameters needed for
scaling the amplitude of the command pulses and correct for leakage current; an extra
effort that is well justified in some applications.

Seal resistance compensation can also be achieved with an analog circuit. Stühmer et al.,
(1983) pioneered a method that uses an active voltage divider and potentiometers to com-
pensate for measured values of Rp and Rsl. Their circuit allows one to null Rp and limit
the series resistance errors to acceptable values. When correctly adjusted, their design
ensures that the input impedance of the patch pipette is near zero so that there is minimal
loss of membrane current across the seal. A disadvantage of this method is that it requires

The Axon Guide — 2500-0102 Rev. C 115

 5. Advanced Methods in Electrophysiology

the user to build a specialized patch-clamp headstage and other circuitry following the
example provided by Stühmer et al., (1983).

With both digital and analog seal-resistance compensation, accuracy depends on one’s
ability to measure Rp and (Rsl + Rp) and on the assumption that these resistances remain
constant during voltage steps that activate conductances in the patch. The error increases
as the amplitude of the voltage step and the conductance of the membrane patch increase.
It is not certain that Rp and Rsl are independent of the current amplitude. In practice, one
often observes an unexpected flattening of the current-voltage relationship for inward or
outward current activation when the voltage step or the patch current become large.
Because most of the uncertainty in current measurement depends on Rsl, one should use
suction and/or enzyme treatment to maximize the seal. It also helps to limit the diameter
of the patch pipette to about 20 μm or less. Rp is minimized by using pipettes with a
blunt taper.
Loose-Patch Clamp with Concentric Patch Pipettes
Almers et al., (1984) introduced a modification of the loose-patch method that employs a
two-chambered concentric patch electrode. The electrode is fashioned from two glass cap-
illaries, one inside the other, with the inner capillary held in place by support rods during
the pull. The concentric electrode is pushed against the cell surface where it separates a
central membrane patch from an annular surround. The voltage-clamp amplifier is
designed to clamp both the center patch and the surround to the command potential.
Membrane current is measured in the central barrel of the electrode while the outer annu-
lus acts as a guard. Since the inner and outer barrels are at the same potential during the
command steps, current loss from the central membrane region to ground is minimized
and Rsl takes a value that is effectively infinite. This approach is historically related to
vaseline-gap and sucrose-gap voltage-clamp methods. It has been used successfully to volt-
age clamp mammalian skeletal muscle. However, the implementation is somewhat diffi-
cult since it requires the construction of an elaborate concentric electrode and some
specialized instrumentation. It may still be the best approach in some cases and the origi-
nal literature should be consulted for further details.
References
Almers, W., Stanfield, P.R., Stühmer, W., Lateral distribution of sodium and potassium
channels in frog skeletal muscle: measurements with a patch-clamp technique. J.Physiol.
336, 261–284, 1983a.

Almers, W., Stanfield, P.R., Stühmer, W., Slow changes in currents through sodium chan-
nels in frog muscle membrane. J. Physiol. 339, 253–271, 1983b.

Almers, W., Roberts, W.M., Ruff, R.L., Voltage clamp of rat and human skeletal muscle:
measurements with an improved loose-patch technique. J.Physiol. 347, 751–768, 1984.

Hoshi, T., Zagotta, W.N., Aldrich R.W., Biophysical and molecular mechanisms of
Shaker potassium channel inactivation. Science 250, 533–538, 1990.

116 The Axon Guide — 2500-0102 Rev. C

 5.6. The Giant Excised Membrane Patch Method

Johnson, J.W, Thompson, S., Measurement of non-uniform current density and current
kinetics in Aplysia neurons using a large patch method. Biophys. J. 55, 299–308, 1989.

Milton, R.L., Caldwell, J.H., Na current in membrane blebs: Implications for channel
mobility and patch clamp recording. J. Neurosci. 10, 885–893, 1990.

Neher, E., Two fast transient current components during voltage clamp on snail neurons.
J.Gen.Physiol. 58, 36–53, 1971.

Neher, E., Lux, H.D., Properties of somatic membrane patches of snail neurons under
voltage clamp. Pflügers Arch. 322, 35–38, 1971.

Palti, Y., Varying potential control voltage clamp on axons. in Biophysics and Physiology
of Excitable Membranes. W.F. Adelman, Jr., Ed. Von Nostrand Reinhold Co. pp. 194–
205, 1971.

Premack, B.A., Thompson, S., Coombs-Hahn, J., Clustered distribution and variability
in kinetics of transient K channels in molluscan neuron cell bodies. J. Neurosci. 9, 4089–
4099, 1990.

Stühmer, W., Roberts, W.M., Almers, W., The loose patch clamp. in Single-Channel
Recording. Sakmann, B., Neher, E., Eds. pp. 123–132, 1983.

Thompson, S. and Coombs, J., Spatial distribution of Ca currents in molluscan neuron

cell bodies and regional differences in the strength of inactivation. J. Neurosci. 8, 1929–
1939, 1988.

5.6. THE GIANT EXCISED MEMBRANE PATCH METHOD

Many physiologically important ion pumps, exchangers and co-transporters are electro-
genic. The giant excised membrane patch method has been developed to improve electro-
physiological studies of such mechanisms. Although whole-cell recording techniques are
often employed in transport studies, a method offering free access to the cytoplasmic
membrane surface and faster voltage clamping was strongly desirable. While the excised
membrane patch method was an attractive alternative, the conventional patch clamp tech-
niques were not useful. First, the single turnover rates of transporters (1–10,000 per sec-
ond) are far too small to expect resolution of the current generated by a single transporter.
Second, the magnitudes of “macroscopic” currents expected in excised patches are difficult
to measure because they are much smaller than the patch leak currents.

The giant excised membrane patch technique enables one to achieve 1–10 gigohm seals
routinely, using pipette tips with inner diameters of 12–40 μm. The giant membrane
patch produced by this method has 2–15 pF membrane capacitance, representing a 100-
fold increase of membrane area over the area of conventional, single-channel patch mem-
brane. Pipette diameters of > 50% of the cell diameter can be employed when recording
from small, spherical cells.

The Axon Guide — 2500-0102 Rev. C 117

 5. Advanced Methods in Electrophysiology

The giant patch method has proven to be advantageous for fast macroscopic current
recording whenever free access to a large surface area of the cytoplasmic membrane is
desired. To date, the giant patch method has been applied successfully to cardiac, skeletal
and tracheal myocytes, pancreas acinus cells, a Jurkat human T-cell line, Sf9 cells and
Xenopus oocytes. The method should be useful for a wide range of studies of electrogenic
mechanisms involving native proteins and proteins expressed from cloned genes. Promis-
ing applications include studies of macroscopic current modulation by intracellular com-
pound (e.g., enzymes or ions acting from the cytoplasmic side of the membrane), studies
of charge movement generated during channel gating, and studies of partial reactions of
transporters.

5.6.1. PRE-TREATMENT OF MUSCLE CELLS

Cell-surface invaginations of muscle cells limit the ability to form large-diameter seals. To
promote sealing, cells can be pretreated to induce separation of the surface membrane
from the underlying cell structure or produce large-scale surface membrane “blebbing.” In
some cases both separation and blebbing occur. To accomplish this, digested pieces of tis-
sue are placed in a “storage solution” consisting of 60–150 mM KCl, 10 mM EGTA,
1–5 mM MgCl2, 20 mM dextrose, 15 mM HEPES, pH 7, for 1–12 hours. This treat-
ment is similar to the treatment previously used to induce blebbing in skeletal muscle
(Standen et al., 1984). Disruption of volume regulation and the Donnan equilibrium is
clearly important for the blebbing process. Large-scale (20–50 μm) membrane bleb for-
mation or apparent lifting of sarcolemma away from myofilaments are routinely obtained
after 2 h treatment. The treatment has been found effective in inducing blebbing in many
different cell types. However, blebbing is not necessarily conducive to large-diameter seal
formation in all cell types, and other methods may be required. For instance, giant patches
have not been obtained on the blebbed membrane of oocytes, although they are routinely
obtained on hypertonically shrunken oocytes after mechanical removal of the vitelline
layer.

5.6.2. PIPETTE FABRICATION

Large-diameter (1.5–1.8 mm), thin-walled glass pipettes are used to fabricate pipettes
with large-diameter tips and relatively steep descents at their tips. The type of glass is
unimportant. A conventional double-pull technique using standard patch-pipette pullers
can be employed. For work with myocytes, light fire polishing with barely visible effects
on the tip favors the spontaneous formation of inside-out patches upon membrane exci-
sion, rather than vesicles that must be disrupted mechanically. Pipette A in Figure 5.5
shows a typical pipette tip with a 17 μm inner diameter.

118 The Axon Guide — 2500-0102 Rev. C

 5.6. The Giant Excised Membrane Patch Method

Figure 5.5: Pipette fabrication for giant excised patch method.

A. Oocyte electrode before fire polish.

B. Oocyte electrode after fire polish.
C. Myocyte electrode after cutting and fire polish.
D. Myocyte electrode after fire polish (no cutting).

Alternatively, pipettes can be pulled to smaller diameters and then scratched gently on the
edge of a soft glass bead to the desired diameter, leaving a rough edge (pipette B in
Figure 5.5, 17 μm diameter). When such electrodes are employed without fire polishing,
the membrane usually remains localized at the pipette opening during sealing, and vesicles
are obtained in most cases upon membrane excision. For very large cells, particularly for
oocytes, a different procedure is employed. Pipettes are first scratched to a 30–50 μm
diameter (pipette C in Figure 5.5) and then melted to a diameter of 18–35 μm (pipette D
Figure 5.5) close to a glass bead placed on a microforge. The resulting bullet-shaped
pipette tips are advantageous for fast voltage clamping, and the oocyte membrane remains
localized to the rim of the pipettes. We have had good success preparing electrodes by
melting the tip into a bead of soft glass, followed by cooling and “cutting” by a small pull.

A number of hydrocarbon mixtures consisting of oils, waxes and Parafilm (American

National Can Co.) allow large-diameter, high-resistance seal formation. Viscous oils can
be used by simply dipping the dry pipette tip into the oil of choice followed by washing
with the desired filling solution. The most consistent success has been obtained with
acetyltocopherol (Sigma Chemical Co.). Silicone oils, phospholipids and plasticizers are
usually not useful, and seals formed with oils alone are generally unstable. Seal stability
can be improved with a highly viscous mixture of Parafilm, acetyltocopherol (or light
mineral oil) and heavy mineral oil. The standard mixture is prepared by mixing approxi-
mately three parts by weight of finely shredded Parafilm with each of the two oils and vig-
orously stirring over heat for 30–60 minutes. When a thin mixture is prepared, pipette
tips can simply be dipped. With a thick mixture, a peanut-sized amount of the mixture is

The Axon Guide — 2500-0102 Rev. C 119

 5. Advanced Methods in Electrophysiology

smeared with a plastic rod on a clean surface, and a thin film of the mixture is drawn into
air by lifting the rod. The pipette tip is then passed quickly through the film, and the pro-
cedure is repeated once more. In order to reduce the pipette capacitance, a thick mixture is
applied as close as possible to the tip. The pipette is then backfilled and a brief, strong,
positive-pressure pulse is applied with a syringe to clear the tip. The pipette tip usually
appears clean and is visually indistinguishable from that of an untreated electrode.

5.6.3. SEAL FORMATION

Seals are formed in the usual manner by applying gentle negative pressure at the cell sur-
face. Maintenance of positive pressure up to the time of sealing is essential both to keep
the electrode tip clean and to prevent contamination of the pipette tip with bath solution
constituents. When working with myocytes, membrane vesicles rather than inside-out
patches are often obtained upon membrane excision, as indicated by the absence of the
characteristic currents. Although membrane position in the pipette tip can usually be
visualized, membrane patches cannot be visually distinguished from vesicles. The method
of choice to disrupt vesicles is to touch the tip of the pipette against an air bubble or a
bead of hydrocarbon placed on the wall of the patch chamber. Success rate is about 50%.

To date, the best success with oocytes has been obtained by shrinking oocytes with
200 mM K-aspartate rather than sucrose. Patches from oocytes are excised by gently mov-
ing the pipette tip from side to side in gradually increasing distances. Dark, granular cyto-
plasmic material is often excised with the oocyte patches, and vesicles are obtained only
rarely. The cytoplasmic material can be washed away by rapid pulses of solution directed
at the patch.

5.7. FURTHER READING

Collins, A., Somlyo, A. V., Hilgemann, D. W., The giant cardiac membrane patch
method: Simulation of outward Na/Ca exchange current by MgATP. J. Physiol. 454,
27–57, 1992.

Hilgemann, D. W., Giant excised cardiac sarcolemmal membrane patches: sodium and
sodium-calcium exchange currents. Pflügers Archiv. 415, 247–249, 1989.

Hilgemann, D.W., The Giant Membrane Patch in Single-Channel Recording, 2nd Ed.,
Sakmann, B., Neher, E., eds. pp. 307–326, 1995.

Hilgemann, D. W., Regulation and deregulation of cardiac Na-Ca exchange in giant

excised sarcolemmal membrane patches. Nature 344, 242–245, 1990.

Standen, N. B., Stanfield, P. R., Ward, T. A., Wilson, S. W. A new preparation for record-
ing single-channel currents from skeletal muscle. Proceedings of the Royal Society of Lon-
don (Biology) B221, 455–464, 1984.

120 The Axon Guide — 2500-0102 Rev. C

 5.8. Recording from Perforated Patches and Perforated Vesicles

5.8. RECORDING FROM PERFORATED PATCHES AND

PERFORATED VESICLES
The study of ion channel function and modulation by neurotransmitters and hormones
advanced dramatically with the advent of patch clamping in the mid-1970’s. Approxi-
mately a decade later, a variation of the single-channel patch-clamp recording method was
introduced and termed the perforated-patch recording method. Perforated-patch record-
ing techniques allow the measurement of whole-cell currents, single-channel currents and
transmembrane voltages much less invasively than do standard patch-clamp or microelec-
trode approaches.

5.8.1. PROPERTIES OF AMPHOTERICIN B AND NYSTATIN

Perforated-patch recording, as presently implemented, uses either nystatin or amphoteri-
cin B to gain electrical access to the cell’s interior. These polyene antibiotics form channels
in cholesterol- or ergosterol-containing membranes. The channels formed by both perfo-
rating compounds are permeable to monovalent cations and Cl- but exclude multivalent
ions such as Ca2+ or Mg2+. The monovalent cations are nine times more permeant than
Cl- through the channels formed by either perforating compound. When measured in
molar salt concentrations, the single-channel conductance of amphotericin channels is
twice that of nystatin channels.

5.8.2. STOCK SOLUTIONS AND PIPETTE FILLING

Because of their limited water solubility, amphotericin and nystatin stock solutions of
30 mg/mL are prepared in DMSO. These stock solutions lose activity upon prolonged
storage and freezing. Therefore, it is best to prepare them freshly before use and use within
1 hour. A convenient approach is to weigh 6 mg of the antibiotic powder into each of
40–50 microcentrifuge tubes and store them indefinitely in the freezer without loss of
activity. At the time of use, the powder is solubilized by pipetting 100 μL fresh DMSO
into the tube. Solubilization is enhanced by sonication or vortex mixing (but this is appar-
ently not required). Stock solution can be added directly to the desired pipette filling solu-
tion to a final concentration of 120–350 μg/mL. Since 120 μg/mL appears to be a
saturating concentration, it is not clear if higher concentrations offer any advantage. Vor-
tex mixing for a few seconds or sonication for a minute is recommended for maximum
solubilization of the antibiotic in the pipette filling solution. When the antibiotic is in
excess, some of it settles out of solution and adheres quite well to the wall of a plastic tube
or syringe. Therefore, the saturated solution can be used without adding much particulate
matter to the pipette interior. The filling solution can be filtered after adding the antibiot-
ics, but care must be taken since some filters might inactivate these compounds. With
either antibiotic, the final solution is yellow in color and slightly turbid. After filtration
with either a 0.22 or 0.45 micron filter, the final solution is perfectly clear. With nystatin,
it is possible to make up larger volumes of the stock solution and adjust the pH to neutral-
ity before use. Under these conditions, no precipitate forms when the final filling solution
is made. This approach has not been used with amphotericin B because of its higher cost.
The perforating activity of either compound begins to decrease about 3 hours after pre-
paring the stock solution. Since whole-cell recording experiments can easily last for 1–3

The Axon Guide — 2500-0102 Rev. C 121

 5. Advanced Methods in Electrophysiology

hours employing the perforated-patch approach, it is advisable to use stock and antibiotic-
containing filling solutions within one hour of preparation.

Both amphotericin and nystatin seem to interfere with seal formation. Although it is pos-
sible to seal some cells when either compound is present in the pipette filling solution, the
success rate is lower than when an antibiotic-free filling solution is used with the same
cells. Therefore, it is best to fill the tip of the pipette with an antibiotic-free solution and
then to backfill with the antibiotic-containing solution. The tip-filling process can be
accomplished by simply dipping the tip into the desired solution. With borosilicate
glasses, 1–5 seconds of tip immersion suffices, whereas with the more hydrophobic lead-
containing glasses, 30–60 seconds may be required. With high-lead glasses, gentle suction
for a second or so may be beneficial. With any glass, it is important that the antibiotic-free
solution not exceed 500 μm distance from the tip; otherwise, the time required for the
diffusion of the antibiotic from the backfilling solution to the tip will be excessive. For
example, if the pipette is filled up to 1.0 mm from the tip with an antibiotic-free solution,
more than an hour will be required for the antibiotic to reach a tip-concentration compa-
rable to that in the backfilling solution. Ideally, the tip should be filled to a distance that
allows the antibiotic to reach the tip within a few seconds following a gigohm seal forma-
tion. Since the time required for seal formation depends on the cells and the setup, tip fill-
ing distances should be optimized by each investigator. Consistently optimal tip filling is
routinely possible by first overfilling the tip and then forcing the excess solution out by
applying positive pressure to the back of the pipette while observing it under a dissecting
microscope.

5.8.3. PROPERTIES OF ANTIBIOTIC PARTITIONING

Following seal formation, the antibiotic slowly diffuses to the tip where it contacts the
membrane and begins to partition and form channels. While most of the time delay can
be attributed to diffusion, some investigators reported that incorporating a mild detergent
like Pluronic F-127 into the filling solution decreases the perforation time. Since it is
unlikely that this detergent could affect the diffusion time of the antibiotics, it seems that,
at least in some cells, the membrane partitioning (channel formation) time of the antibi-
otics is appreciable and that mild detergents decrease this time.

In a real experiment, one can observe the partitioning by applying a voltage pulse to the
pipette at frequent intervals. As perforation occurs, the size of the capacity transient
increases and the time constant of the transient decreases. The time constant to charge the
cell capacitance is approximately RaCm, where Ra is the access resistance and Cm is the cell
capacitance. When Ra is large, a brief voltage pulse will not charge the capacitance fully
before the pulse turns off. Therefore, as partitioning continues and Ra decreases, the
capacity transient gets larger since an increasing fraction of the cell capacity gets charged
during the brief pulse. The decaying phase of the transient gets shorter as Ra decreases. In
most cells, a properly filled pipette will produce minimized Ra by 20–30 minutes after seal
formation. Ra of less than 20 MΩ is often achieved in about 5 minutes. Some investiga-
tors begin recording at this time. However, Ra continues to fall and finally becomes stable
after 20–30 minutes. Once Ra stabilizes, it remains stable for 2.5–3 hours before begin-

122 The Axon Guide — 2500-0102 Rev. C

 5.8. Recording from Perforated Patches and Perforated Vesicles

ning to increase (presumably due to loss of activity of the antibiotic). The remarkable sta-
bility of Ra is one of the main virtues of the technique. For example, using amphotericin,
one investigator reported Ra of 3.4 MΩ that did not change by more than 100 kΩ for
three hours.

5.8.4. THE ADVANTAGES OF THE PERFORATED-PATCH TECHNIQUE

The perforated-patch technique has several advantages over conventional whole-cell
patch-clamp approaches:

1 The channels formed by either amphotericin or nystatin are impermeable to

molecules as large as or larger than glucose. Therefore, whole-cell recordings can be
done without dialyzing important substances from the cell’s cytoplasm. Currents run
down significantly slower; physiologically relevant second-messenger cascades and
mechanisms important to cell signaling and channel regulation remain operative.
2 Intracellular multivalent ions are not affected by the pipette-filling solution since the
channels formed by the antibiotics are not permeated by these ions. Therefore,
intracellular Ca2+, for example, can be measured by optical techniques simultaneously
with recording whole-cell currents.
3 With carefully fabricated pipettes and the use of appropriate composition of the filling
solution, the perforation technique is less damaging to cells than are intracellular
microelectrodes or standard whole-cell patch clamping. It is not uncommon to be able
to record from a single cell for 3 hours. In some instances, freshly dissociated cells
plate out on the bottom of the recording chamber while their whole-cell currents are
being measured. The ability to plate out is indicative of the cells’ viability.
4 Unlike the frequently experienced loss of seal when suction or voltage pulses are
applied to disrupt the membrane patch (to “go whole cell”), seals are rarely lost with
the perforating-patch approach.
5 The Ra achieved with the perforated-patch technique is as low or lower than that
obtained by standard approaches. Moreover, once achieved, the low Ra of the
perforated patch is usually considerably more stable than that produced with standard
techniques.
6 In many cells, whole-cell capacitance is easier to compensate when using perforated
patches. It appears that the equivalent circuit of the perforated patch is often better
approximated by a single RC than is the access resistance and capacitance in a
standard whole-cell recording.

5.8.5. THE LIMITATIONS OF THE PERFORATED-PATCH TECHNIQUE

The perforated-patch technique has some disadvantages when compared to the conven-
tional whole-cell patch-clamp approaches:

1 The perforated-patch approach does not allow one to dialyze the cell’s cytoplasm and
replace its content with compounds other than the small ions included in the filling

The Axon Guide — 2500-0102 Rev. C 123

 5. Advanced Methods in Electrophysiology

solution, as is possible with conventional methods. Therefore, the effects of such

compounds on intercellular mechanisms cannot be studied.
2 The pipette-filling solution and the cell cytoplasm exist in Donnan Equilibrium.
Thus, the pipette must contain an impermeant anion and a Cl- concentration that
match the cell interior. Failure to do so results in a Cl- flux either into or out of the cell
(depending on the direction of the mismatch). A charge balancing cation and water
will follow, resulting in changes in cell volume. Due to the low Cl- permeability of the
antibiotic channels as compared to their cation permeability, it may take
approximately 30 minutes to reach this equilibrium. It is not yet clear if there is a way
to alleviate this problem since it is difficult to find an impermeant anion that matches
the effective valence of the intracellular impermeant anions. Recently, it has been
reported that including 20–25 mM Cl- and 125–130 mM methanesulfonate in the
pipette solution appears to keep cell volume reasonably stable.
On the other hand, this disadvantage can be exploited for studies of volume-activated
currents by purposefully mismatching the pipette and cytoplasmic Cl- concentrations
causing swelling and shrinking of cells. To date, this approach has not been employed.
In addition to the influx and efflux of ions described above, the cells that are voltage-
clamped by the perforated-patch technique are very susceptible to volume changes
caused by external perturbations. In a standard whole-cell recording configuration,
cell volume is kept reasonably constant by the large volume of the pipette-filling
solution and the high hydraulic conductivity of the pipette tip. In contrast, when
applying the perforated-patch method, the access to a cell is provided by a million or
so tiny parallel channels. While these channels can provide electrical access
comparable to that obtained by a standard patch pipette, their composite hydraulic
conductivity is significantly lower than that of a single orifice with the same total
electrical resistance.
3 The perforation process requires considerably longer time for achieving access into the
cell interior, as determined by the low Ra, than do suction or voltage pulses. When
rapid measurements from a large number of cells is important, standard whole-cell
patch-clamp approaches would be more productive.
4 While the relative Na+, K+ and Cl- permeabilities of nystatin and amphotericin
channels are known, the permeation by other anions and cations, particularly
multivalent ions that are generally considered impermeant, needs further study.

5.8.6. SUGGESTED WAYS TO MINIMIZE THE ACCESS RESISTANCE

Clearly, the access resistance (Ra) cannot be lower than the resistance of the pipette itself.
The resistance of the patch of membrane to be perforated is in series with the electrode tip
and, therefore, the total resistance is the sum of the two. The final patch resistance
depends upon the total number of antibiotic channels formed (partitioning) and the sin-
gle-channel conductance of each channel. The total number of channels equals the num-
ber of channels formed in a unit surface area times the patch area. These simple

124 The Axon Guide — 2500-0102 Rev. C

 5.8. Recording from Perforated Patches and Perforated Vesicles

considerations suggest the following ways to minimize Ra:

1 The largest pipette tips compatible with seal formation should be used to minimize
the pipette resistance.
2 Pipette geometry should maximize the size of the omega-shaped membrane patch that
is drawn into the tip during seal formation. This maximizes the surface area available
for channel insertion.
3 The perforating compound that produces the highest total conductance should be
used. It may not be simple to predict the conductance-producing ability of a
compound since it requires both the partitioning (number of channels formed) and
single-channel conductance to be maximal. To date, amphotericin B has produced the
lowest Ra. This can be anticipated since the single-channel conductance of
amphotericin channels is twice that of nystatin, at least in measurements made in salts
at molar concentration. However, the relative conductance of the two types of
channels in lower salt concentrations (e.g., 150 mM) is unknown. In addition, there
are no relevant studies of the relative partitioning of the two compounds when
presented to the membrane in saturating concentrations. To date, the lowest patch
resistance obtained by perforating with amphotericin is about 2 MΩ. The lowest
resistance pipette that is capable of forming gigohm seals is approximately 0.5 MΩ.
Thus, at present, an Ra value of approximately 2.5 MΩ appears to be the lowest one
can achieve. Lower Ra values will require either perforating compounds with higher
single-channel conductance or finding ways to enhance the partitioning of the
existing antibiotics. It is unlikely that larger pipettes than those presently used would
be successful in forming seals.

5.8.7. OTHER USES FOR PERFORATED PATCHES

Cellular Voltage Measurements
In addition to whole-cell recording using patch-clamp amplifiers like the Axopatch™
amplifier in the voltage-clamp or current-clamp mode, the perforated-patch method is
useful for making cellular voltage measurements, such as for resting and action potentials.
These can be done either with patch clamp amplifiers operating in current-clamp mode or
with standard microelectrode amplifiers like the Axoclamp 900A or MultiClamp™ 200B
amplifiers. In any cellular voltage measurement, two main factors affect the accuracy of
the measurement:

1 The extent to which the cell’s voltage is altered by current flow through the shunt
resistance along the electrode’s outer surface at either the penetration or seal sites; and
2 The extent to which the pipette-filling solution alters the electrolyte concentrations in
the cell cytoplasm. High shunt resistances and minimal alteration of the cell content
are desirable.
Patch electrodes, with their tens to hundreds of gigohm seal resistances, clearly have much
higher shunt resistances than do intracellular microelectrodes. On the other hand, patch
electrodes sealed to cells cause rapid changes in the concentrations of the cellular electro-

The Axon Guide — 2500-0102 Rev. C 125

 5. Advanced Methods in Electrophysiology

lytes if the pipette-filling solution contains different electrolyte concentrations than the
cytoplasm. Intracellular microelectrodes, which are usually filled with molar concentra-
tions of potassium salts, also alter cellular electrolytes but not as rapidly as do patch elec-
trodes. Therefore, the perforated-patch technique provides both the advantage of the
higher shunt resistance than intracellular microelectrodes without the significant change
in cytoplasmic content caused by conventional patch electrodes.
Single-Channel Recording in Outside-Out Vesicles
The outside-out patch configuration has been particularly useful because it allows direct
application of agents to the extracellular side of the membrane while maintaining com-
plete voltage control during single channel recording. These patches have been required
for studying ligand-gated ion channels such as the nicotinic acetylcholine, GABAa, gly-
cine and glutamate receptors. Likewise, transmitters that regulate channels through non-
diffusible second messengers must be examined with outside-out patches. For example,
activation of atrial potassium channels is not seen with cell-attached patches when musca-
rinic agonists are applied to the bath, but is seen with direct application of transmitter to
an outside-out patch.

Unfortunately, conventional outside-out patches have a drawback. The formation of these

cell-free patches requires replacement of cytoplasm with artificial pipette solutions. In
many cases, this “washout” is accompanied by a loss of ion channel activity and/or modu-
lation. For example, in GH3 pituitary tumor cells, the neuropeptide thyrotropin releasing
hormone (TRH) triggers mobilization of intracellular calcium that in turn regulates cal-
cium and potassium channel activity. With outside-out patches, TRH fails to regulate
potassium channels. Furthermore, calcium channel activity runs down within 15 minutes.
These problems had prevented direct detection of transmitter-induced inhibition of single
calcium channels until the development of a new patch clamp configuration, termed the
perforated vesicle, that results in the formation of outside-out patches that retain cytoplas-
mic factors and organelles.

Developments in single-channel recording combine the outside-out patch recording

method with the perforated-patch technique. In this application, a perforating antibiotic
is used to produce an outside-out patch, termed perforated vesicle, that retains the cyto-
plasmic content of the cell forming the equivalent to tiny single-channel cells. Thus, the
perforated vesicle is an extension of the nystatin-perforated patch technique.

Similar to the protocol described above, nystatin stock solution (25 mg/mL) is prepared
by dissolving 2 mg nystatin in 80 μL DMSO. Since this pore-forming agent is susceptible
to oxidation, the stock solution can be frozen and used within two days. Simple pipette
solutions can be used (e.g., 150 mM KCl, 5 mM MgCl2, 10 mM KHepes, pH 7.1)
because nucleotides, proteins and calcium buffers cannot permeate through nystatin per-
forations. Four (4) μL of nystatin stock solution is pipetted into a glass tube and then
1 mL of filtered pipette solution is added. After covering the tube and vortexing, the
slightly cloudy solution is vigorously sonicated for 30 seconds in a cylindrical bath sonica-

126 The Axon Guide — 2500-0102 Rev. C

 5.8. Recording from Perforated Patches and Perforated Vesicles

tor to completely dissolve the nystatin, giving a final concentration of 100 μg/mL. This
aqueous solution can be used for only 3 hours.

Conventional patch pipettes are fabricated. When filled with pipette solution, their resis-
tance is 2–4 MΩ. Smaller pipettes minimize the surface area of the perforated patch and,
thus, give larger series resistances (Rs). On the other hand, larger pipette tips were found
to form fragile patches that could spontaneously break forming the conventional whole-
cell configuration. Pipettes are coated with Sylgard to reduce their capacitance. This not
only lowers noise for single-channel recording, but also permits accurate series resistance
measurements.

Pipettes are momentarily dipped in nystatin-free pipette solution and then backfilled with
nystatin-containing solution. A cell-attached seal is then formed. After compensating,
electrode capacitance (e.g., 3 pF), Rs is monitored by examining the unfiltered current
response to a 10 mV voltage step from -70 mV (holding potential). The height of the
capacitance transient is equal to the voltage step divided by Rs. Series resistance is moni-
tored for 5–20 minutes to ensure that it decreases smoothly to a value less than 50 MΩ.
Cells that show an abrupt decrease in Rs should be discarded.

The patch-clamp amplifier is switched from voltage-clamp (V-clamp) mode to current-

clamp (I-clamp) mode with zero resting current and the pipette is withdrawn from the cell
to form a perforated vesicle (Figure 5.6). Switching to I-clamp ensures that no large
changes in current could pass through the patch and disrupt the resealing of the mem-
brane that forms the vesicle. The patch-clamp amplifier is then switched to V-clamp
mode to record single-channel activity.

Nystatin Pipette
perforations

Receptor
withdraw
pipette

Ion channel
Cytoplasm
Perforated
vesicle
Cell-attached
perforated patch

Figure 5.6: Forming a cell-attached perforated patch and a perforated vesicle.

The formation of a perforated vesicle can be verified by many methods. In some cases,
vesicles can be seen with 400x magnification. The presence of cytoplasm can be con-

The Axon Guide — 2500-0102 Rev. C 127

 5. Advanced Methods in Electrophysiology

firmed by preloading the cells with a fluorescent dye. For example, cells may be treated
with 40 μM carboxyfluorescein diacetate, a membrane-permeant nonfluorescent com-
pound. After crossing the membrane, this agent is hydrolyzed to yield water-soluble car-
boxyfluorescein. The presence of organelles can also be verified with specific dyes.
Rhodamine 123 specifically partitions into functioning mitochondria and produces a
characteristic punctate staining that can be detected in GH3 cells and perforated vesicles.
Channel activity can also verify the presence of nystatin perforations. With normal vesi-
cles, channel currents appear attenuated and rounded. Furthermore, channels in the inner
face of the vesicle are not accessible to bath-applied hydrophilic inhibitors and have oppo-
site voltage dependence to channels in the vesicle surface that is in contact with the bath.
In contrast, perforated-vesicle channel activity can be recorded only from the portion of
the vesicle in contact with the bath. Through the use of fluorescent dyes and channel
recordings, the perforated vesicle can be differentiated from normal excised vesicles and
outside-out patches.

The nystatin-perforated vesicle method allows the study of channel activity and modula-
tion that is lost with conventional outside-out patches. Initial studies indicate that cyclic
AMP-dependent protein kinase, phosphodiesterase, phosphatase, G-proteins, phospholi-
pase C and intracellular calcium stores can modulate ion channels in perforated vesicles
excised from GH3 cells. Two channel types that quickly “wash out” of conventional
patches from these cells, L-type calcium channels and voltage-gated potassium channels,
survive for extended periods with perforated vesicles. Perforated vesicles were excised suc-
cessfully from a variety of cell lines and primary cultured pituitary cells. Thus, it appears
likely that this patch clamp configuration will aid the study of ion channel modulation in
many systems.

5.9. FURTHER READING

Falke, L.C., Gillis, K.D., Pressel, D.M., Misler, S., Perforated patch recording allows long-
term monitoring of metabolite-induced electrical activity and voltage-dependent Ca2+
currents in pancreatic islet B-cells. FEBS Lett., 251: 167–172, 1989.

Finkelstein, A., Water movement through lipid bilayers, pores, and plasma membranes:
theory and reality., Vol. 4, Wiley-Interscience, New York, p. 127, 1987.

Holz, R., Finkelstein, A., The water and nonelectrolyte permeability induced in thin lipid
membranes by the polyene antibiotics nystatin and amphotericin B. J. Gen. Physiol., 56:
125–145, 1970.

Horn, R., Diffusion of nystatin in plasma membrane is inhibited by a glass-membrane

seal. Biophys. J., 60: 329–333, 1991.

Horn, R., Marty, A., Muscarinic activation of ionic currents measured by a new whole-cell
recording method. J. Gen. Physiol., 92: 145–159, 1988.

128 The Axon Guide — 2500-0102 Rev. C

 5.10. Enhanced Planar Bilayer Techniques for Single-Channel Recording

Korn, S.J., Bolden, A., Horn, R., Control of action potential duration and calcium influx
by the calcium-dependent chloride current in AtT-20 pituitary cells. J. Physiol. (Lond),
439: 423–437, 1991.

Korn, S.J., Horn, R., Influence of sodium-calcium exchange on calcium current rundown
and the duration of calcium-dependent chloride currents in pituitary cells, studied with
whole cell and perforated patch recording. J. Gen. Physiol., 94: 789–812, 1989.

Korn, S.J., Marty, A., Connor, J.A., Horn, R., Perforated patch recording. Methods in
Neuroscience 4: 264–373, 1991.

Kurachi, Y., Asano, Y., Takikawa, R., Sugimoto, T., Cardiac Ca current does not run
down and is very sensitive to isoprenaline in the nystatin-method of whole cell recording.
Archiv. Pharm., 340: 219–222, 1990.

Levitan, E.S., Kramer, R.H., Neuropeptide modulation of single calcium and potassium
channels detected with a new patch clamp configuration. Nature, 348: 545–547, 1990.

Lindau, M., Fernandez, J.M., IgE-mediated degranulation of mast cells does not require
opening of ion channels. Nature, 319: 150–153, 1986.

Lucero, M.T., Pappone, P.A., Membrane responses to norepinephrine in cultured brown

fat cells. J. Gen. Physiol., 95: 523–544, 1990.

Rae, J., Cooper, K., Gates, G., Watsky, M., Low access resistance perforated patch record-
ings using amphotericin B. J. Neurosci. Methods., 37: 15–26, 1991.

5.10. ENHANCED PLANAR BILAYER TECHNIQUES FOR SINGLE-

CHANNEL RECORDING
The planar bilayer recording technique, in which ion channels are incorporated into an
artificial, planar lipid bilayer either by fusion of vesicles containing the channels with the
bilayer or by direct insertion of water-soluble channels, provides an unique approach for
studying single ion channels, enabling experimental designs that are impractical or impos-
sible using standard patch-clamp methods. For example, the effects of membrane compo-
sition on channel function can be rigorously studied, changes in solution composition on
either side of the membrane can be made easily, and channels can be incorporated from
membranes that are normally inaccessible to patch-clamp methods (e.g., cytoplasmic vesi-
cles, sarcoplasmic reticulum) (Wonderlin et al., 1990; Wonderlin et al., 1991). The utility
of the planar bilayer method for studying single ion channels, however, is often limited by
a low recording bandwidth, primarily due to the high background current noise, and poor
resolution of voltage step activation of ion channel activity due to the large, slow capaci-
tive current transient associated with a voltage step applied to a bilayer. These limitations
may be especially problematic in investigations of small-conductance, rapidly gating chan-
nels, the kinetics of channel-blocking drugs, or voltage-gated ion channels that must be
activated by voltage steps because of steady-state inactivation. The purpose of this brief

The Axon Guide — 2500-0102 Rev. C 129

 5. Advanced Methods in Electrophysiology

tutorial is to describe a combination of equipment and methods that partially overcome

these limitations and improve the quality of planar bilayer recordings. With this approach,
it is reasonable to attempt to record single-channel activity with low background noise
(< 0.35 pA rms at 5 kHz) and rapid resolution (<< 1 ms) of currents after a voltage step.
Several general considerations regarding the equipment and techniques necessary for mak-
ing high-resolution recordings will be presented first, followed by a more detailed descrip-
tion of the implementation of the approach. For more detailed discussion of the
equipment and methods, see Wonderlin et al., 1990.

5.10.1. SOLVING THE PROBLEMS OF HIGH RESOLUTION AND VOLTAGE STEPS

Minimizing the Background Current Noise
Bilayer recording systems are prone to high background current noise which can easily
obscure single-channel gating and often may require low-pass filtering of the current
record to a bandwidth of a few hundred hertz or less. This heavy filtering may obscure
rapid gating fluctuations, and it badly distorts the current response to a voltage step.
Unlike gigohm-seal (gigaseal) patch-clamp recording, thermal noise currents in either the
large value (e.g., 50 GΩ) feedback resistor of a resistive headstage amplifier or in the seal
resistance are generally insignificant for bilayer recordings. Rather, the majority of the
background noise in a planar bilayer recording system results from “voltage clamping”
across the bilayer capacitance (Cb) voltage noise present in the following sources:

1 The FET input of the patch amplifier,

2 The voltage command circuitry, and
3 The access resistance Ra, i.e., the resistance of the solutions and electrodes in series
with the bilayer.
The spectral density of the current noise S2i(f ) produced by voltage clamping a noise volt-
age in either the FET input or the voltage command across the bilayer capacitance is given
by Ohm’s law:

en2
S i2(f ) = (6)
X c2

where en is the rms noise voltage (V/√Hz) and Xc is the capacitive reactance of the bilayer.
The capacitive reactance is 1/(2αCbf ), where f is the frequency and Cb is the bilayer
capacitance. Therefore:

(7)
S i2( f ) = en2 (2πCb f ) 2

The spectral density of the current noise produced by voltage clamping the thermal volt-
age noise in the access resistance Ra across the bilayer capacitance is:

130 The Axon Guide — 2500-0102 Rev. C

 5.10. Enhanced Planar Bilayer Techniques for Single-Channel Recording

(8)
S i2(f ) = 4kT Re{Y (f )}

where Re{Y(f )} is the real part of the admittance of Ra in series with Cb, and is given by

α2
Re{Y (f )} = (9)
Ra (1 + α 2 )

with α = 2πfRaCb. For a constant bandwidth, at the lower limit of Ra (small α),
Re{Y(f )} = Ra(2πCbf )2, and equations (7) and (8) have a similar form. At the upper limit
of Ra (large α), Re{Y(f )} = 1/Ra. The transition from a linear to a reciprocal dependence
on Ra produces maxima in plots of this noise component against Ra (see Figure 2A of
Wonderlin et al., 1990).
Strategies for Minimizing the Background Current Noise
Reduce the Bilayer Capacitance. The single most effective approach to reducing the
background current noise is to reduce the bilayer diameter and hence its capacitance.
This assumes that the access resistance is not increased (see Wonderlin et al., 1990
and “Choosing a Recording Configuration” on page 137) for a discussion of open
chamber versus pipette/bilayer configurations). There is, however, a practical limit to
the extent of reduction as the bilayer diameter must be large enough both to allow
easy visualization and physical manipulation and to ensure an acceptable rate of
channel incorporation. Bilayer diameters in the 40–60 μm range (having 10–25 pF
Cb at 0.8 μF/cm2) represent a reasonable compromise between the competing needs
of reducing the capacitance versus ease of channel incorporation and manipulation. A
very good method for making these small bilayers is the shaved-aperture method,
which is described in detail below (Assembling a Bilayer Setup for High-Resolution
Recording: Making a Small Aperture).

Use an Amplifier with Low Internal Noise Voltage. Noise voltages are present in the
FET input of the headstage and in the circuitry with which voltage command inputs
are buffered and applied to the bilayer. Reduction of the noise voltage in these inter-
nal sources is difficult for the investigator to control without being involved in the
design and construction of the amplifier. However, when selecting an amplifier, the
magnitude of the internal noise voltage sources can be quickly but qualitatively deter-
mined by comparing the noise in the current output before and after connection of
the amplifier input through a bilayer-sized capacitor to ground. This test should be
done without the addition of any resistance in series with the test capacitor to avoid
confusing an internal noise voltage source with voltage noise in the access resistance.
It should be noted that amplifiers which exhibit low noise when used for patch
recording (where the input capacitance is small and the background noise is domi-
nated by thermal noise currents in the feedback resistor and seal resistance) can per-

The Axon Guide — 2500-0102 Rev. C 131

 5. Advanced Methods in Electrophysiology

form poorly when the input capacitance is increased from patch to bilayer values,
because of high internal voltage noise.

Use Low-Noise Voltage Command Sources. Although many commercially available

amplifiers are equipped with internal voltage command generators, voltage-step pro-
tocols will usually require the generation of the voltage command from an external
source, usually a computer-driven digital-to-analog (D/A) converter. This can present
a problem if the D/A source has a significant noise voltage. This is very likely if the
board containing the D/A is mounted internally in a computer, using the computer’s
power supply. The best configuration is an outboard D/A unit, with optically isolated
digital connections and its own, isolated power supply. Once again, a D/A source
that provides perfectly suitable voltage commands for patch clamping can fail misera-
bly when used to provide voltage commands for bilayer recordings, because of the
much greater sensitivity to voltage noise in the voltage command source. One should
also be careful about using digital function generators for producing command volt-
ages such as voltage ramps, as they can introduce considerably more noise than ana-
log function generators.

Minimize the Access Resistance. Even with perfect, noise-free electronics, considerable
background current noise can be generated when the thermal voltage noise in the
access resistance is voltage clamped across the bilayer capacitance. This source can
account for as much as one half the noise under many bilayer recording conditions
(Wonderlin et al., 1990). The access resistance responsible for generating the voltage
noise is composed primarily of the resistance of the solution within the aperture in
which the bilayer is formed and the convergence resistances between each end of the
aperture and the bulk solutions (see Wonderlin et al., 1990). Because the resistance of
these solutions is inversely related to the salt concentration, it may be possible to
reduce the intra-aperture and convergence resistances by increasing the salt concen-
tration, within the limits of the experimental design. Furthermore, the shaved-aper-
ture method can be used to minimize the length of the aperture and, therefore, its
resistance. The resistance of the bulk solution should not contribute appreciably to
the access resistance when an open chamber design is used; but it can be an impor-
tant factor if large bilayers are made on the tips of glass pipettes (Wonderlin et al.,
1990). Finally, if salt bridges are used to connect the Ag/AgCl pellets to each bath,
they should be designed with a low resistance in mind, such as by using relatively
large diameter (e.g., 2 mm) glass capillaries filled with a high-salt solution such as
3 M KCl.

Choice of Chamber Construction Materials. When selecting a type of plastic from

which to fabricate a recording chamber, one should consider, in addition to factors
such as machinability, durability and solvent resistance, the fact that some plastics
generate excess current noise when a hole in the plastic is filled with a salt solution
and connected to the input of a patch amplifier. For example, when an Axopatch-1B
amplifier was connected via an Ag/AgCl electrode to a 0.5 mL well filled with
3 M KCl, the following relative rms noise levels (30 Hz–5 kHz) were measured: high-

132 The Axon Guide — 2500-0102 Rev. C

 5.10. Enhanced Planar Bilayer Techniques for Single-Channel Recording

density polyethylene (1.0), polystyrene (1.0), polycarbonate (1.0), Teflon (1.0),

acrylic (1.4) and nylon (1.5) (Wonderlin et al., 1990). Although the absolute amount
of this excess noise is small relative to other sources listed above, this source of noise
can be eliminated easily by selecting an appropriate plastic.

Type of Solvent in a Painted Bilayer. We have found that the type of solvent present in
a painted bilayer can affect the background current noise. If a bilayer is painted from
phospholipids dissolved in decane, the background noise can often be reduced by
rewiping the bilayer with hexadecane alone. The reason for the improvement is not
known and appears somewhat paradoxical, since it is usually associated with a slightly
larger bilayer capacitance.
Maximizing the Bandwidth
Minimization of the background noise currents by one or more of the methods described
above may help increase the usable bandwidth by decreasing the amount of low-pass fil-
tering required to clearly discern single channel gating. If the low-pass filter cutoff (-3 dB)
frequency is increased to several kilohertz or more, a second, less obvious source of band-
width limitation may come into play. This limitation is the decreased efficiency of the
high-frequency boost circuit used in resistive-headstage amplifiers due to loading the
headstage input with large bilayer capacitance. The boost circuit is designed to restore the
bandwidth that is lost because of low-pass filtering (< 100 Hz) by the parallel combina-
tion of the feedback resistor and its stray capacitance. The high-frequency boost circuit
adds a scaled derivative of the headstage current to the raw headstage current, thereby
increasing the headstage bandwidth. For patch recording, a boost circuit can work well
because of the dominant role of fixed electrical elements, such as the feedback resistance
and stray feedback capacitance, whose properties can be optimally compensated in the
selection of circuit components. When a resistive headstage amplifier is used for bilayer
recordings, however, the loss of bandwidth due to the variable, and frequently large,
bilayer capacitance must also be restored. This is much more difficult because the optimal
amount of high-frequency boost that is required will change in parallel with any change in
bilayer capacitance. Therefore, it is difficult to select components in the design of the
amplifier that enable optimal boosting across a large range of bilayer capacitance values.
Furthermore, because the bilayer capacitance will often vary by 50% within a day’s exper-
iments, it may be very difficult to maintain optimal boost adjustment.
Strategies for Maximizing the Bandwidth
Use a Capacitive-Headstage Patch Amplifier. An integrating-headstage patch-clamp
amplifier (e.g., the Axopatch 200B) measures the patch current as the integral of the
current flow across a feedback capacitor, rather than as the voltage drop across a feed-
back resistor. A boost circuit is not required for an integrating-headstage amplifier
because the feedback capacitor does not produce any significant low-pass filtering of
the patch current. Integrating headstage amplifiers therefore provide a broader
recording bandwidth that is also independent of the size of the bilayer capacitance.

The Axon Guide — 2500-0102 Rev. C 133

 5. Advanced Methods in Electrophysiology

Resolving Voltage Steps Across Bilayers

A change of voltage across a bilayer of capacitance Cb by an amount Vstep requires the
movement across the bilayer of electrical charge Qb where:

Qb = V step C b (10)

For a resistive-headstage amplifier, the amount of time required to move Qb is:

Qb R f
Charging time = (11)
V max

where Vmax is the maximum voltage that can be applied across the feedback resistor Rf.
For example, a headstage with a 50 GΩ feedback resistor and a Vmax of 10 V can inject
charge at a maximum rate of 200 pA, requiring 25 ms to change the voltage across a
100 pF bilayer by 50 mV. During this 25 ms period, the output of the amplifier will be
saturated (at Vmax) and the bilayer will not be voltage clamped. This example illustrates
that the primary difficulty associated with applying voltage steps to a bilayer is the
requirement for rapid movement of a large amount of charge across the bilayer. For tradi-
tional resistive-headstage patch amplifiers this is a nearly insurmountable problem because
of the very slow rate at which charge can be delivered to the bilayer capacitance through
the large feedback resistance.

A secondary problem is that planar bilayers do not behave as ideal capacitors. The applica-
tion of a voltage step to a planar bilayer compresses the bilayer (i.e., causes electrostric-
tion). This electrostrictive force can change the relative areas of the bilayer and
surrounding annulus, producing a concomitant change in the capacitance of the bilayer.
During the period of time that the bilayer capacitance is changing, a capacitive current
will flow across the bilayer. Electrostrictive changes in bilayer capacitance are relatively
slow (milliseconds to seconds) and rarely large enough to produce capacitive currents that
obscure single-channel gating. However, the electrostrictive capacitive current tends to be
variable in magnitude and kinetics and it can change significantly as a bilayer “ages” dur-
ing the course of an experiment. This variability can make it difficult to subtract the elec-
trostrictive component of the capacitive current from the current record.
Strategies for Resolving Voltage Steps Across Bilayers
Reduce the Size of the Bilayer and Annulus. An essential first step is to make small
bilayers, because reducing the bilayer capacitance proportionally decreases the charg-
ing time. Unfortunately, reducing the bilayer diameter into the 40–60 μm range
(10–25 pF) may still result in an excessive saturation time (several milliseconds).
Because the magnitude of the electrostrictive change in capacitance depends on the
area of the annulus, decreasing the bilayer diameter also reduces this component. The
width of the annulus (i.e., in the plane of the bilayer) is also roughly proportional to

134 The Axon Guide — 2500-0102 Rev. C

 5.10. Enhanced Planar Bilayer Techniques for Single-Channel Recording

the thickness of the margin of the aperture. A substantial, further reduction in the
electrostrictive component can be achieved by using the shaved-aperture method,
because it produces a margin of the aperture that is very thin (a few micrometers).
Other techniques, such as drilling or punching apertures, produce a margin that is
much thicker. Finally, any manipulation of the bilayer that produces a “bulkier”
annulus, such as addition of excess solvent or repeated wiping, can greatly increase
the electrostrictive component and should be avoided.

Use Ccapacitance Compensation. The capacitance compensation circuits in most com-

mercially available patch clamps are only slightly effective in reducing the duration of
saturation because they are designed to compensate the relatively small capacitance of
a patch pipette rather than the much larger capacitance of a planar bilayer. Special
circuits capable of compensating the larger bilayer capacitance can be added to the
headstage, but they require the injection of compensating current into the amplifier
input through a large capacitor, which increases the total input capacitance and
degrades the noise performance. Sometimes, the electrostrictive capacitive current
can be partially subtracted by adjustment of a capacitance compensation circuit with
a slow time constant. The BL type of CV-7 headstages (such as the CV-7B/BL) in use
with MultiClamp 700 amplifiers has an increased fast pipette capacitance compensa-
tion range that allows compensation of up to 300 pF of pipette capacitance.

Decrease the Feedback Resistance. The charging time can be reduced by decreasing the
value of Rf, thereby increasing the rate at which charge can be delivered to the bilayer.
However, the price for this improvement is a proportional increase in the background
thermal noise currents in Rf and, therefore, a concomitant decrease in usable band-
width. Although this may work for large conductance channels, it is hardly a general
solution. A better solution is to include both a large (50 GΩ) and a small (50 MΩ)
feedback resistor in the headstage with logic-controlled circuitry that can switch
between the two. This approach was taken in the CV-4B headstage for the Axopatch
1 series of Axon Cellular Neuroscience amplifiers. During a voltage step, the bilayer
capacitance is rapidly charged through the small Rf. Immediately after the bilayer
capacitance is charged, the feedback pathway is switched to the large Rf, enabling
high resolution recording of single-channel activity. This approach was used success-
fully to study voltage-dependent activation of K+ channels from squid (Wonderlin et
al., 1990; Wonderlin et al., 1991).

Use a Capacitive-Headstage Patch Amplifier. Capacitive-headstage patch clamps can

very rapidly charge the bilayer capacitance with a maximum instantaneously injected
charge of:

Maximum charge = V max C f

where Cf is the feedback capacitance. With a Vmax of 10 V and Cf equal to 1 pF,

10 pC of charge can be delivered nearly instantly. This charge is sufficient to change

The Axon Guide — 2500-0102 Rev. C 135

 5. Advanced Methods in Electrophysiology

the voltage across a 100 pF bilayer by 100 mV without saturation of the headstage
output. Because integration of a constant patch current will eventually saturate Cf, an
integrating headstage requires a reset circuit that periodically discharges the feedback
capacitor; this is usually a logic-controlled switch that shunts the feedback capacitor.
In some designs, this same reset switch can be activated during a voltage step. Such a
“forced reset” allows rapid charging of the bilayer capacitance.

5.10.2. ASSEMBLING A BILAYER SETUP FOR HIGH-RESOLUTION RECORDINGS

Choosing an Amplifier
Traditionally, resistive headstage patch amplifiers have been used in bilayer recording sys-
tems. These amplifiers are widely commercially available or they can be built rather inex-
pensively because of the simple design of the circuitry. Integrating-headstage amplifiers
are commercially available.

How should one choose among these amplifiers? The ideal amplifier for high-resolution
bilayer recordings should have:

1 Low intrinsic current and voltage noise,

2 A wide bandwidth, and
3 The ability to rapidly charge the bilayer capacitance.
Although integrating headstages have less intrinsic current noise than resistive headstages,
there may be little difference in noise performance during actual bilayer recordings due to
the dominance in bilayer recordings of current noise produced by the interaction of the
bilayer capacitance with various noise voltage sources. With regard to bandwidth, the
integrating headstage is a clear favorite because it does not require a high-frequency boost
circuit. Adjustment of the high-frequency boost of a resistive headstage with a large, vari-
able bilayer capacitance at the input can be very frustrating. On the other hand, within
the 5–10 kHz bandwidth that one might hope to achieve with bilayer recordings, the
bandwidth of an integrating headstage should not be affected by the bilayer capacitance.
Finally, the integrating headstage is the better choice for rapid charging of the bilayer
capacitance during voltage steps, although a switching-feedback resistive headstage (such
as the CV-4B headstage for the old Axopatch 1 series of amplifiers) can perform nearly as
well (Wonderlin et al., 1990; Wonderlin et al., 1991), since it also provides a low-imped-
ance pathway for charging the bilayer capacitance. It should be emphasized that the rapid
charging by either an integrating headstage amplifier or a switching-resistive headstage
amplifier does not decrease the slow, electrostrictive component of the capacitive current,
whose time course is determined only by the dynamic properties of the bilayer/annulus
structure.

When all of the requirements for an amplifier are considered, it is clear that anyone
attempting high-resolution recording from bilayers with voltage steps should seriously
consider using a capacitive-headstage amplifier.

136 The Axon Guide — 2500-0102 Rev. C

 5.10. Enhanced Planar Bilayer Techniques for Single-Channel Recording

Choosing a Recording Configuration

There are two primary recording configurations for planar bilayers:

1 A bilayer formed over an aperture in a partition separating two open bath chambers
(open-chamber configuration); and
2 A bilayer formed on the tip of a polished glass pipette (pipette/bilayer configuration),
either by “punching” a preformed planar bilayer or by dipping the tip through a lipid
monolayer.
The relative merits of these configurations have been discussed in (Wonderlin et al.,
1990). Briefly, for bilayer capacitance values larger than a few picofarads, the open-cham-
ber configuration with a shaved aperture will generate less background current noise than
the bilayer/pipette configuration, due to the much higher access resistance of the interior
of the pipette compared to the open chamber. Therefore, if bilayers must be made large
enough to permit incorporation of channels by fusion of vesicles with the bilayer, the
open chamber configuration should be used. If channels can be incorporated during the
formation of the bilayer (e.g., “tip-dip”) so that bilayers with a very small area (less than a
few picofarads capacitance) can be used, then the pipette/bilayer configuration can be
used. Finally, the open chamber configuration offers greater flexibility with regard to
manipulation of the bilayer and bath solutions.
Making Small Apertures
It is fundamentally important to be able to work with small bilayers, usually with diame-
ters in the range of about 40–60 μm. This size range represents a trade-off between the
need to reduce the area (and capacitance) while maintaining a large enough area to ensure
adequate incorporation by fusion of vesicles. Among the many techniques available for
making apertures in plastic partitions, the shaved-aperture method is especially well suited
for making small apertures. This method has been previously described in detail for mak-
ing small apertures in plastic cups (Wonderlin et al., 1990).

Figure 5.7: Aperture formation in a plastic cup (not drawn to scale).

A. A metal stylus is warmed and pressed against the inner surface of the cup, pro-
ducing a cone-shaped depression.
B. After cooling, the stylus is retracted and the plastic shaved from the outer surface,
leaving a thin-edged aperture where the cut intersects the depression in the wall
of the cup, as in C. See text for additional details. Reproduced with permission
from Wonderlin et al., 1990.

The Axon Guide — 2500-0102 Rev. C 137

 5. Advanced Methods in Electrophysiology

Briefly, a conical metal stylus is warmed and pressed against the inner surface of a plastic
cup, forming a cone-shaped depression extending part way across the cup (Figure 5.7).
The stylus is then removed and the plastic shaved away from the outer surface, using a dis-
posable microtome blade, until the depression is intersected. The size of the aperture is
controlled by varying the depth of shaving. Apertures made using this method have a thin
margin beyond which the plastic rapidly increases in thickness, thereby decreasing the
stray capacitance across the partition and providing good mechanical strength. The thin
margin of the shaved aperture is very important in reducing the size of the annulus and,
therefore, the electrostrictive change in capacitance following voltage steps. Following are
the essential steps in implementing the method:

1 Making the stylus. The key to successfully using the shaved-aperture method is the
manufacture of a high-quality stylus. Using a lathe and dissecting microscope,
stainless steel darning needles can be ground to a very fine tip (< 5 μm diameter) and
polished to a very smooth finish using ultra-fine abrasive paper (Thomas Scientific,
Swedesboro, NJ). Softer metals can be substituted, but it is more difficult to produce
the highly polished tip. The stylus should be examined under high magnification
(400x) to ensure that the tip is very smooth.
2 Selecting the plastic. The original description of the shaved-aperture technique
(Wonderlin et al., 1990) recommended polystyrene as a good plastic for recording
cups. Since that time, repeated difficulties have been encountered with crazing of the
margin of apertures made in polystyrene cups, which sometimes appears to make the
cups electrically leaky. More recently, shaved apertures in cups were made from Ultra-
Clear centrifuge tubes (Beckman, Colombia, MD). Also, if a flat partition rather than
a cup is preferred, shaved apertures have been formed in plastic coverslips (Fisher
Scientific, Pittsburgh, PA). The margin of these apertures does not craze and bilayers
formed on apertures shaved in these plastics are very stable, lasting several hours.
Other plastics can probably be substituted, with the basic requirement being that they
cut cleanly so that the margin of the aperture is very smooth.
3 Melting the plastic. There are many choices of mechanical apparatus for manipulating
and heating the stylus. A rather simple method is to mount the stylus in a hole in the
tip of a variable temperature soldering iron, and then to mount the soldering iron on
a manipulator so that the tip can be carefully pushed into the plastic. Control of the
heat is important to ensure replicability and to avoid overheating, which may damage
the plastic and, perhaps, the stylus.
4 Shaving the aperture. It is easiest to shave the apertures while observing with a
dissecting microscope. The shaved apertures should be examined under high
magnification (400x) before use to ensure they are free from deformations that might
interfere with bilayer stability.

138 The Axon Guide — 2500-0102 Rev. C

 5.10. Enhanced Planar Bilayer Techniques for Single-Channel Recording

Viewing with a Microscope

It is not practical to attempt to work with small bilayers without observing them with a
microscope. Bilayers are usually observed under reflected light because the disappearance
of reflected light provides a good monitor of bilayer thinning. Quite often, the bilayer is
viewed through one eyepiece of a dissecting microscope, while the bilayer is illuminated
by light focused from a fiber bundle through the second eyepiece. This method is not use-
ful for small bilayers because of the lack of detail visible with reflected light and the rela-
tively low magnification (40–80x) of most dissecting microscopes. Using a horizontally
mounted compound microscope (Nikon Inc., Garden City, NY) with a long working-dis-
tance objective (Nikon, E-Plan 10x) and a bilayer recording chamber designed for transil-
lumination provides very good resolution of detail in the bilayer and annulus at a
magnification of 100x.
Testing the System
The performance of the recording system should be tested using a dummy bilayer, espe-
cially if voltage steps are to be used. The dummy bilayer should use a capacitor with a
value close to that expected for the bilayer capacitance. It may also be useful to add a resis-
tor in series with the capacitor to mimic the access resistance. In choosing a capacitor,
polystyrene capacitors are preferred. Some other types of capacitors, such as ceramic
capacitors, can exhibit very non-ideal properties, with odd relaxations similar to the elec-
trostrictive relaxation of planar bilayers. By simulating a bilayer recording using the
dummy bilayer, it is possible to determine the dependence of the background noise and
voltage-step response on electrical components with ideal electrical properties, as opposed
to the non-ideal electrical properties of components used during experiments, such as the
bilayer composition and salt solutions. It also permits optimization of the timing of volt-
age steps relative to logic-controlled switching between different feedback resistors in a
resistive headstage amplifier or the timing of a forced reset in a capacitive headstage ampli-
fier.
Making the Recordings
An example of the voltage-step activation of a delayed rectifier K+ channel from squid is
shown in Figure 5.8. The ultimate success of high-resolution recording in planar bilayers
depends not only on the various details described above, but also on the ease of fusion of
certain populations of membrane vesicles. Rates of incorporation are likely to be lower for
small bilayers than are typical for larger bilayers, but the difference may be reduced by
methods such as pressure application of vesicles from a pipette onto the bilayer. Although
the high-resolution recording method may increase the difficulty beyond that of tradi-
tional bilayer recording techniques, it opens exciting opportunities for new experimental
designs and the investigation of a broader range of ion channels.

The Axon Guide — 2500-0102 Rev. C 139

 5. Advanced Methods in Electrophysiology

Figure 5.8: Voltage-step activation of a K-channel from squid giant axon axoplasm.

A. A single, unsubtracted current record showing the response to a voltage step. In

A–C, the current is actually inward, but is shown inverted to be consistent with the
usual orientation of K-current records. The gain was switched to low 1 ms before
the voltage step, producing a small current transient (logic pulse is shown in trace
E). The gain was switched back to high 25 μs after the voltage step.
B. Current record from which the capacitive current and switching artifact have been
digitally subtracted. Blank traces (7 traces in which the channel was not active)
were averaged and subtracted from the active trace. Subtraction is very effective
in removing the step artifact, and high-resolution recording is established within
1ms after the voltage step.
C. An average current record simulating a macroscopic response. 240 current traces
were averaged, from which the average of the blank records was subtracted, as in
B.
D. Voltage command trace.
E. Logic pulse used to switch the headstage gain. The headstage was switched to
low gain during the rectangular pulse. Reproduced with permission from Wonder-
lin et al., 1990.

140 The Axon Guide — 2500-0102 Rev. C

 5.11. Further Reading

5.11. FURTHER READING

Wonderlin, W.F., Finkel, A., French, R.J., Optimizing planar lipid bilayer single-channel
recordings for high resolution with rapid voltage steps. Biophys. J., 58: 289–297, 1990.

Wonderlin, W.F., French, R.J., Ion channels in transit: voltage-gated Na and K channels
in axoplasmic organelles of the squid Loligo pealei. Proc. Natl. Acad. Sci. USA, 88: 4391–
4395, 1991.

Hamill, O.P., Marty, A., Neher, E., Sakmann, B., Sigworth, F.J. Improved patch-clamp
techniques for high-resolution current recording from cells and cell-free membrane
patches. Pflügers Arch. Eur. J. Physiol. 391:85–100, 1981.

Sigworth, F.J., Electronic design of the patch clamp. in Single-Channel Recording, 2nd
Edition, pp. 95–126, Sakmann, B., Neher, E. Eds. Plenum Press, New York, 1995.

The Axon Guide — 2500-0102 Rev. C 141

 5. Advanced Methods in Electrophysiology

142 The Axon Guide — 2500-0102 Rev. C

 6. Signal Conditioning and
Signal conditioners
It is rare for biological, physiological, chemical, electrical or physical signals to be mea-
sured in the appropriate format for recording and interpretation. Usually, a signal must be
“massaged” to optimize it for both of these functions. For example, storage of recorded
data is more accurate if the data are amplified before digitization so that they occupy the
whole dynamic range of the A/D converter, and interpretation is enhanced if extraneous
noise and signals above the bandwidth of interest are eliminated by a low-pass filter.

6.1. WHY SHOULD SIGNALS BE FILTERED?

A filter is a circuit that removes selected frequencies from the signal. Filtering is most
often performed to remove unwanted signals and noise from the data. The most common
form of filtering is low-pass filtering, which limits the bandwidth of the data by eliminat-
ing signals and noise above the corner frequency of the filter (Figure 6.1). The importance
of low-pass filtering is apparent when measuring currents from single-ion channels. For
example, channel openings that are obscured by noise in a 10 kHz bandwidth may
become easily distinguishable if the bandwidth is limited to 1 kHz.

High-pass filtering is required when the main source of noise is below the frequency range
of the signals of interest. This is most commonly encountered when making intracellular
recordings from nerve cells in the central nervous system. There are low-frequency fluctu-
ations in the membrane potential due to a variety of mechanisms, including the summa-
tion of synaptic inputs. The small, excitatory synaptic potentials that the user might be
interested in are often smaller than the low-frequency fluctuations. Since excitatory synap-
tic potentials are often quite brief, the low-frequency fluctuations can be safely eliminated
by using a high-pass filter. High-pass filtering is also referred to as AC coupling.

Another type of filter that is often used in biological recording is the notch filter. This is a
special filter designed to eliminate one fundamental frequency and very little else. Notch
filters are most commonly used at 50 or 60 Hz to eliminate line-frequency pickup.

The Axon Guide — 2500-0102 Rev. C 143

 6. Signal Conditioning and Signal conditioners

6.2. FUNDAMENTALS OF FILTERING

Filters are distinguished by a number of important features. These are:

6.2.1. -3 DB FREQUENCY
The -3 dB frequency (f-3) is the frequency at which the signal voltage at the output of the
filter falls to √1/2, i.e., 0.7071, of the amplitude of the input signal. Equivalently, f-3 is the
frequency at which the signal power at the output of the filter falls to half of the power of
the input signal. (See Section 6.3.9., “Decibels (dB)”.)

6.2.2. TYPE: HIGH-PASS, LOW-PASS, BAND-PASS OR BAND-REJECT (NOTCH)

A low-pass filter rejects signals in high frequencies and passes signals in frequencies below
the -3 dB frequency. A high-pass filter rejects signals in low frequencies and passes signals
in frequencies above the -3 dB frequency. A band-pass filter rejects signals outside a cer-
tain frequency range (bandwidth) and passes signals inside the bandwidth defined by the
high and the low -3 dB frequencies. A band-pass filter can simply be thought of as a series
cascade of high-pass and low-pass filters. A band-reject filter, often referred to as a notch
filter, rejects signals inside a certain range and passes signals outside the bandwidth
defined by the high and the low -3 dB frequencies. A band-reject filter can simply be
thought of as a parallel combination of a high-pass and a low-pass filter.

6.2.3. ORDER
A simple filter made from one resistor and one capacitor is known as a first-order filter.
Electrical engineers call it a single-pole filter. Each capacitor in an active filter is usually
associated with one pole. The higher the order of a filter, the more completely out-of-
band signals are rejected. In a first-order filter, the attenuation of signals above f-3
increases at 6 dB/octave, which is equal to 20 dB/decade. In linear terminology, this atten-
uation rate can be re-stated as a voltage attenuation increasing by 2 for each doubling of
frequency, or by 10 for each ten-fold increase in frequency.

6.2.4. IMPLEMENTATION: ACTIVE, PASSIVE OR DIGITAL

Active filters are usually made from resistors, capacitors and operational amplifiers. Passive
filters use resistors, capacitors and inductors. Passive filters are more difficult to make and
design because inductors are relatively expensive, bulky and available in fewer varieties and
values than capacitors. Active filters have the further virtue of not presenting a significant
load to the source. Digital filters are implemented in software. They consist of a series of
mathematical calculations that process digitized data.

6.2.5. FILTER FUNCTION

There are many possible transfer functions that can be implemented by active filters. The
most common filters are: Elliptic, Cauer, Chebyshev, Bessel and Butterworth. The fre-
quency responses of the latter two are illustrated in Figure 6.1. Any of these filter transfer
functions can be adapted to implement a high-pass, low-pass, band-pass or band-reject fil-
ter. All of these filter transfer functions and more can be implemented by digital filter
algorithms. Digital filters can even do the seemingly impossible: since all of the data may

144 The Axon Guide — 2500-0102 Rev. C

 6.3. Filter Terminology

be present when the filtering begins, some digital filters use data that arrive after the cur-
rent data. This is clearly impossible in an analog filter because the future cannot be pre-
dicted. Typical digital filters are the box-car smoothing filter and the Gaussian filter.

10 BESSEL
20
BUTTERWORTH
30

40

50

60

70
0.1 1 10
Normalized Frequency, f/f -3dB

Figure 6.1: Frequency response of 4th-order Bessel and Butterworth filters.

The spectra have been normalized so that the signal magnitude in the pass band is 0 dB.
The -3 dB frequency has been normalized to unity.

6.3. FILTER TERMINOLOGY

The terminology in this section is defined and illustrated in terms of a low-pass filter. The
definitions can easily be modified to describe high-pass and band-pass filters. Many of
these terms are illustrated in Figure 6.2.

6.3.1. - 3 DB FREQUENCY
f-3, defined previously, is sometimes called the cutoff frequency or the corner frequency.
While most engineers and physiologists specify a filter’s bandwidth in terms of the -3 dB
frequency, for obscure reasons some manufacturers label the filter frequency on the front
panel of their instruments based on a frequency calculated from the intersection fre-
quency of the pass-band and the stop-band asymptotes. This is very confusing. To make
sure that the filter is calibrated in terms of the -3 dB frequency, a sine wave generator can
be used to find the -3 dB frequency.

The Axon Guide — 2500-0102 Rev. C 145

 6. Signal Conditioning and Signal conditioners

6.3.2. ATTENUATION
Attenuation is the reciprocal of gain. For example, if the gain is 0.1 the attenuation is 10.
The term attenuation is often preferred to gain when describing the amplitude response of
filters since many filters have a maximum gain of unity. (For accurate measurements, note
that even filters with a stated gain of unity can differ from 1.00 by a few percent.)

6.3.3. PASS BAND

Pass band is the frequency region below f-3. In the ideal low-pass filter there would be no
attenuation in the pass band. In practice, the gain of the filter gradually reduces from
unity to 0.7071 (i.e., -3 dB) as the signal frequencies increase from DC to f-3.

6.3.4. STOP BAND

Stop band is the frequency region above f-3. In the ideal low-pass filter, the attenuation of
signals in the stop band would be infinite. In practice, the gain of the filter gradually
reduces from 0.7071 to a filter-function and implementation-dependent minimum as the
signal frequencies increase from f-3 to the maximum frequencies in the system.

6.3.5. PHASE SHIFT

The phase of sinusoidal components of the input signal is shifted by the filter. If the phase
shift in the pass band is linearly dependent on the frequency of the sinusoidal compo-
nents, the distortion of the signal waveform is minimal.

6.3.6. OVERSHOOT
When the phase shift in the pass band is not linearly dependent on the frequency of the
sinusoidal component, the filtered signal generally exhibits overshoot. That is, the
response to a step transiently exceeds the final value.

6.3.7. OCTAVE
An octave is a range of frequencies where the largest frequency is double the lowest fre-
quency.

6.3.8. DECADE
A decade is a range of frequencies where the largest frequency is ten times the lowest fre-
quency.

146 The Axon Guide — 2500-0102 Rev. C

 6.3. Filter Terminology

5
PASS BAND STOP BAND
0 Attenuation:
-3 -6 dB/octave
-20 dB/decade
-5
20 log Vout
Vin
-10
(dB)

-15

-20

-25
decade
octave
30
0.1 1 2 10 100
Frequency, f -3dB ( Hz )

Figure 6.2: Illustration of filter terminology.

A number of the terms used to describe a filter are illustrated in the context of a single-pole,
low-pass filter.

6.3.9. DECIBELS (DB)

Since filters span many orders of magnitude of frequency and amplitude, it is common to
describe filter characteristics using logarithmic terminology. Decibels provide the means
of stating ratios of two voltages or two powers.

Voltage:

Vout
dB = 20 log (1)
Vin

Thus, 20 dB corresponds to a tenfold increase in the voltage.

-3 dB corresponds to a √2 decrease in the voltage.

Power:

Pout
dB = 10log (2)
Pin

The Axon Guide — 2500-0102 Rev. C 147

 6. Signal Conditioning and Signal conditioners

Thus, 10 dB corresponds to a tenfold increase of the power.

-3 dB corresponds to a halving of the power.

Some useful values to remember:

Decibels Voltage Ratio Power Ratio

3 dB 1.414:1 2:1

6 dB 2:1 4:1

20 dB 10:1 100:1

40 dB 100:1 10,000:1

60 dB 1,000:1 1,000,000:1

66 dB 2,000:1 4,000,000:1

72 dB 4,000:1 16,000,000:1

80 dB 10,000:1 108:1

100 dB 100,000:1 1010:1

120 dB 1,000,000:1 1012:1

6.3.10. ORDER
As mentioned above, the order of a filter describes the number of poles. The order is often
described as the slope of the attenuation in the stop band, well above f-3, so that the slope
of the attenuation has approached its asymptotic value (see Figure 6.2). Each row in the
following table contains different descriptions of the same order filter.

Pole Order Slopes

1 pole 1st order 6 dB/octave 20 dB/decade

2 poles 2nd order 12 dB/octave 40 dB/decade

4 poles 4th order 24 dB/octave 80 dB/decade

8 poles 8th order 48 dB/octave 160 dB/decade

Typically, the higher the order of the filter, the less the attenuation in the pass band. That
is, the slope of the filter in the pass band is flatter for higher order filters (Figure 6.3).

148 The Axon Guide — 2500-0102 Rev. C

 6.3. Filter Terminology

Figure 6.3: Difference between a 4th- and 8th-order transfer function.

6.3.11. 10-90% RISE TIME

The 10-90% rise time (t10-90) is the time it takes for a signal to rise from 10% of its final
value to 90% of its final value. For a signal passing through a low-pass filter, t10-90
increases as the -3 dB frequency of the filter is lowered. Generally, when a signal contain-
ing a step change passes through a high-order filter, the rise time of the emerging signal is
given by:

(3)
t 10 -90 ≈ 0.3/f -3

For example, if f-3 is 1 kHz, t10-90 is approximately 300 μs.

As a general rule, when a signal with t10-90 = ts is passed through a filter with t10-90 = tf ,
the rise time of the filtered signal is approximately:

2 (4)
tr = t s2 + t f

6.3.12. FILTERING FOR TIME-DOMAIN ANALYSIS

Time-domain analysis refers to the analysis of signals that are represented the same way
they would appear on a conventional oscilloscope. That is, steps appear as steps and sine
waves appear as sine waves. For this type of analysis, it is important that the filter contrib-
utes minimal distortion to the time course of the signal. It would not be very helpful to

The Axon Guide — 2500-0102 Rev. C 149

 6. Signal Conditioning and Signal conditioners

have a filter that was very effective at eliminating high-frequency noise if it caused 15%
overshoot in the pulses. Yet this is what many kinds of filters do.

In general, the best filters to use for time-domain analysis are Bessel filters because they
add less than 1% overshoot to pulses. The Bessel filter is sometimes called a “linear-phase”
or “constant delay” filter. All filters alter the phase of sinusoidal components of the signal.
In a Bessel filter, the change of phase with respect to frequency is linear. Put differently,
the amount of signal delay is constant in the pass band. This means that pulses are mini-
mally distorted. Butterworth filters add considerable overshoot: 10.8% for a fourth-order
filter; 16.3% for an eighth-order filter. The step response of the Bessel and Butterworth
filters is compared in Figure 6.4.

Butterworth

100

mV
Bessel
50

Input
Step
0

0 20 40 ms

Figure 6.4: Step response comparison between bessel and butterworth filters.

The overshoots of fourth-order Bessel and Butterworth filters are compared.

In many experiments in biology and physiology (e.g., voltage- and patch-clamp experi-
ments), the signal noise increases rapidly with bandwidth. Therefore, a single-pole filter is
inadequate. A four-pole Bessel filter is usually sufficient, but eight-pole filters are not
uncommon. In experiments where the noise power spectral density is constant with band-
width (e.g., recording from a strain gauge), a single-pole filter is sometimes considered to
be adequate.

In the time domain, a notch filter must be used with caution. If the recording bandwidth
encompasses the notch filter frequency, signals that include a sinusoidal component at the
notch frequency will be grossly distorted, as shown in Figure 6.5. On the other hand, if
the notch filter is in series with a low-pass or high-pass filter that excludes the notch fre-
quency, distortions will be prevented. For example, notch filters are often used in elec-
tromyogram (EMG) recording in which the line-frequency pickup is sometimes much
larger than the signal. The 50 or 60 Hz notch filter is typically followed by a 300 Hz high-

150 The Axon Guide — 2500-0102 Rev. C

 6.3. Filter Terminology

pass filter. The notch filter is required because the high-pass filter does not adequately
reject the 50 or 60 Hz hum (see Figure 6.5).

Voltage

Time

Voltage

Time

Figure 6.5: Use of a notch filter: inappropriately and appropriately.

A. Shows an inappropriate use of the notch filter. The notch filter is tuned for 50 Hz. The input to the
notch filter is a 10 ms wide pulse. This pulse has a strong component at 50 Hz that is almost elimi-
nated by the notch filter. Thus, the output is grossly distorted.
B. Shows an appropriate use of the notch filter. An EKG signal is corrupted by a large 60 Hz compo-
nent that is completely eliminated by the notch filter.

6.3.13. FILTERING FOR FREQUENCY-DOMAIN ANALYSIS

Frequency-domain analysis refers to the analysis of signals that are viewed to make a
power spectrum after they are transformed into the frequency domain. This is typically
achieved using a fast Fourier transform (FFT). After transformation, sine waves appear as
thin peaks in the spectrum and square waves consist of their component sine waves. For
this type of analysis it does not matter if the filter distorts the time-domain signal. The
most important requirement is to have a sharp filter cutoff so that noise above the -3 dB
frequency does not get folded back into the frequency of interest by the aliasing phenom-
enon (see Chapter 11).

The simplest and most commonly used filter for frequency-domain analysis in biological
applications is the Butterworth filter. This filter type has “maximally flat amplitude” char-
acteristics in the pass band. All low-pass filters progressively attenuate sinusoidal compo-

The Axon Guide — 2500-0102 Rev. C 151

 6. Signal Conditioning and Signal conditioners

nents of the signal as the -3 dB frequency is approached from DC. In a Butterworth filter,
the attenuation in the pass band is as flat as possible without having pass-band ripple.
This means that the frequency spectrum is minimally distorted.

Usually, notch filters can be safely used in conjunction with frequency-domain analysis
since they simply remove a narrow section of the power spectrum. Nevertheless, they are
not universally used this way because many experimenters prefer to record the data “as is”
and then remove the offending frequency component from the power spectrum digitally.

6.3.14. SAMPLING RATE

If one intends to keep the data in the time domain, sufficient samples must be taken so
that transients and pulses are adequately sampled. The Nyquist Sampling Theorem states
that a bare minimum sampling rate is twice the signal bandwidth; that is, the -3 dB fre-
quency of the low-pass filter must be set at one half the sampling rate or lower. Therefore,
if the filter -3 dB frequency is 1 kHz, the sampling theorem requires a minimum sampling
rate of 2 kHz. In practice, a significantly higher sampling rate is often used because it is
not practical to implement the reconstruction filters that would be required to reconstruct
time-domain data acquired at the minimum sampling rate. A sampling rate of five or
more times the -3 dB frequency of the filter is common.

If you wish to make peak-to-peak measurements of your data for high-frequency signals,
you must consider the sampling rate closely. The largest errors occur when samples are
equally spaced around the true peak in the record. This gives the worst estimate of the
peak value. To illustrate the magnitude of this problem, assume that the signal is a sine
wave and that the following number of samples are taken per cycle of the sine wave. The
maximum errors that one could get are then as follows:

5 samples/cycle 19.0% error

10 samples/cycle 4.9% error

20 samples/cycle 1.2% error

The sampling of the peak values varies between no error and the maximum as stated
above.

If one intends to transform the data into the frequency domain, Butterworth or Elliptic
filters are more suitable than the Bessel filter. These filters have a sharper cutoff near the
-3 dB frequency than the Bessel filter, and thus better prevent the phenomenon known as
aliasing. With a fourth-order or higher Butterworth filter, it is usual to set f-3 to about
40% of the sampling rate. Frequently, researchers do not have a Butterworth filter handy.
If a Bessel filter is available, it can be used instead, but f-3 would normally be set to about
25–30% of the sampling rate. This is because the Bessel filter attenuation is not as sharp
near the -3 dB frequency as that of a Butterworth filter.

152 The Axon Guide — 2500-0102 Rev. C

 6.3. Filter Terminology

6.3.15. FILTERING PATCH-CLAMP DATA

The analog filters that are typically used with a patch clamp are the Bessel and Butter-
worth types, in either 4-pole or 8-pole versions. If simply regarded from the noise point of
view, the Butterworth filter is the best. Empirical measurements carried out MDS
Analytical Technologies reveal that the rms noise passed by the filters (relative to a 4-pole
Bessel filter) when used with an Axopatch™ 200 amplifier in single-channel patch mode is
as follows:

4-Pole 8-Pole 4-Pole 8-Pole

Noise
Bessel Bessel Butterworth Butterworth

1 kHz 1.0 0.97 0.95 0.93

5 kHz 1.0 0.97 0.85 0.82

10 kHz 1.0 0.94 0.83 0.80

The table shows that the main reduction in noise is gained by using a Butterworth filter
instead of a Bessel filter. The improvement achieved by going from a 4-pole filter to an
8-pole filter of the same kind is small.

However, for the reasons previously discussed above, the Butterworth filter cannot be used
for time-domain analysis. Since this is the most common kind of analysis performed on
patch-clamp data, Bessel filters are almost invariably used.

6.3.16. DIGITAL FILTERS

Some researchers prefer to record data at wide bandwidths to prevent the loss of poten-
tially important information. An analog filter is used to provide anti-alias filtering, fol-
lowed by a digital filter implemented at various lower -3 dB frequencies during the
analysis. There are many types of digital filters. Two types are described here:
Nonrecursive Filter
The output of a nonrecursive filter depends only on the input data. There is no depen-
dence on the history of previous outputs. An example is the smoothing-by-3’s filter:

x n −1 + xn + xn +1
yn = (5)
3

where yn and xn are the output and input samples at sample interval n.

Nonrecursive filters are also known as “finite impulse response” filters (FIR) because their
response to a single impulse endures only as long as the newest sample included in the for-
mula (i.e., xn+1 in the smoothing-by-3’s filter).

The Axon Guide — 2500-0102 Rev. C 153

 6. Signal Conditioning and Signal conditioners

Another example of a nonrecursive digital filter is the Gaussian filter. It has a similar form
to the smoothing-by-3’s filter described above, except that typically there are more terms
and the magnitudes of the coefficients lie on the bell-shaped Gaussian curve.

These types of filters have the advantage of not altering the phase of the signal. That is,
the mid-point for the rise time of a step occurs at the same time both before and after fil-
tering. In contrast, analog filters always introduce a delay into the filtered signal.

A problem with digital filters is that values near the beginning and end of the data cannot
be properly computed. For example, in the formula above, if the sample is the first point
in the data, xn-1 does not exist. This may not be a problem for a long sequence of data
points; however, the end effects can be serious for a short sequence. There is no good solu-
tion other than to use short filters (i.e., few terms). Adding values outside the sequence of
data is arbitrary and can lead to misleading results.
Recursive Filter
The output of a recursive filter depends not only on the inputs, but on the previous out-
puts as well. That is, the filter has some time-dependent “memory.” Recursive filters are
also known as “infinite impulse response” filters (IIR) because their response to a single
impulse extends indefinitely into the future (subject to computer processing limitations).

Digital-filter implementations of analog filters such as the Bessel, Butterworth and RC fil-
ters are recursive.
Correcting for Filter Delay
The delay introduced by analog filters necessarily makes recorded events appear to occur
later than they actually occurred. If it is not accounted for, this added delay can introduce
an error in subsequent data analysis. The effect of the delay can be illustrated by consider-
ing two common questions: How long after a stimulus did an event occur (latency mea-
surement)? And, what was the initial value of an exponentially decaying process
(extrapolation back to the zero time of a fitted curve)? If the computer program was
instructed to record 50 points of baseline and then apply a stimulus, it would be natural,
but incorrect, to assume in the subsequent analysis that zero time begins somewhere
between points 50 and 51 of the recorded data, since zero time may actually be at point
52 or later.

The delay can be seen in Figure 6.4. The times to the 50% rise or fall point of a step signal
(delay) are approximately 0.33/f-3 for a fourth-order low-pass Bessel filter, and 0.51/f-3 for
an eighth-order Bessel filter. In practice, when using a fourth-order Bessel filter with
f-3 = 1 kHz and sampling the trace at 6 kHz, the filter delay is 330 μs. So, the whole
record will have to be shifted by 2.0 points with respect to the stimulus events that were
programmed.

154 The Axon Guide — 2500-0102 Rev. C

 6.4. Preparing Signals for A/D Conversion

6.4. PREPARING SIGNALS FOR A/D CONVERSION

Analog-to-Digital (A/D) converters have a fixed resolution and measure signals in a grainy
manner. This means that all signals lying between certain levels are converted as if they are
the same value. To minimize the impact of this undesirable “quantization” effect, it is
important to amplify the signal prior to presenting it to the A/D converter. Ideally, the
gain should be chosen so that the biggest signals of interest occupy the full range of the
A/D converter but do not exceed it. In most laboratory systems, the full range of the A/D
converter is ±10 V, but other ranges such as ±5 V and ±1.25 V are often used in industrial
applications.

6.4.1. WHERE TO AMPLIFY

Amplification is possible at a number of different points along the signal pathway. Since
the amplification, filtering and offset circuits can themselves introduce noise, the location
of the amplification circuitry must be carefully considered. There are several options:
Inside the Recording Instrument
The ideal place to amplify the signal is inside the instrument that records the signal. For
example, the Axopatch 200B, MultiClamp 700B, and Axoclamp™ 900A amplifiers from
MDS Analytical Technologies contain a variable gain control in the output section that
can be used to provide low-noise amplification of the pipette current or membrane poten-
tial. The advantage of placing the amplification inside the recording instrument is that the
amount of circuitry between the low-level signal and the amplifying circuitry can be min-
imized, thereby reducing extraneous noise contributions.
Between the Recording Instrument and the A/D Converter
A good place to amplify the signal is after it emerges from the recording instrument,
before it is sent to the A/D converter. The CyberAmp® 380 amplifier from MDS
Analytical Technologies can be used for this purpose. For the best noise performance, a
small amount of initial amplification should be provided in the recording instrument if
the signal levels are low. The main advantage of using a multi-channel amplifier such as
the CyberAmp is that the gain of each recording pathway can be independently set by the
computer. An instrument such as the CyberAmp amplifier has more gain choices than are
usually available in a recording instrument.

In either of the examples discussed so far, anti-aliasing filtering is conveniently provided

on a per-channel basis after the gain amplification.
After the Channel Multiplexor on the A/D Converter Board
A common place to provide amplification is in a programmable gain amplifier (PGA)
located after the channel multiplexor on the A/D converter board. Briefly, in the A/D
converter board, many channels are typically digitized by a single A/D converter. The sig-
nals are sequentially presented to the A/D converter by a multiplexor circuit. A PGA
located after the multiplexor is very economical, since only one PGA is required regardless
of the number of channels.

The Axon Guide — 2500-0102 Rev. C 155

 6. Signal Conditioning and Signal conditioners

The main advantage of locating a PGA after the multiplexor on the A/D board is that it is
inexpensive. However, there are significant disadvantages:

1 The PGA has to have extremely wide bandwidth since it must be able to settle within
the resolution of the A/D converter at the multiplexing rate. Depending on the
number of channels being sampled, this could mean that the bandwidth of the PGA
has to be ten to several hundred times greater than the bandwidth of the analog
signals. Such fast amplifiers are difficult to design, but in many cases the required
speed can be achieved. The less obvious problem is that every amplifier introduces
intrinsic noise, and the amount of noise observed on the amplifier output increases
with bandwidth. Since the PGA may have several hundred times more bandwidth
than the analog signal, it is likely to contribute more noise to the recording than is
inherent in the signal. This problem cannot be eliminated or even reduced by filtering
because filtering would lengthen the settling time of the PGA. This serious problem
limits the usefulness of a shared PGA. This problem does not exist in the previously
discussed systems in which there is an amplifier for each channel.
2 If the PGA is located on the A/D board, the low-level signals must be brought by
cables to the A/D board. Typically, these cables are a couple of meters or more in
length and provide ample opportunity for pick up of hum, radio-frequency
interference and cross-talk between signals. With careful attention to shielding and
grounding, these undesirable effects can be minimized. In the alternative approaches,
in which the signals are amplified before they are sent to the A/D converter, the
relative impact of these undesirable interferences are reduced in proportion to the
amount of early amplification.
Systems that include a PGA on the A/D board are useful when the signals require only a
small amount of amplification (ten fold or less) or when the user cannot afford or has not
invested in external amplifiers.

6.4.2. PRE-FILTER VS. POST-FILTER GAIN

When low-level signals are recorded, it is essential that the first-stage amplification be suf-
ficient to make the noise contributed by succeeding stages irrelevant. For example, in a
microphone amplifier the tiny output from the microphone is first coupled into an
extremely low-noise pre-amplifier. After this first-stage amplification, circuits with more
modest noise characteristics are used for further amplification and to introduce treble and
bass filtering.

If the only rule was to maximize the early gain, all of the gain would be implemented
before any low-pass filter, notch filter or offset stages. However, with some signals, too
much gain in front of the low-pass filter can introduce a different problem. This occurs
when the signal is much smaller than the out-of-band noise.

This problem is illustrated by the signal in Figure 6.6. Panel A shows an input signal con-
sisting of a pair of 1 pA current pulses significantly corrupted by instrumentation noise.
When the noisy signal is first amplified by x100 and then low-pass filtered, the amplified

156 The Axon Guide — 2500-0102 Rev. C

 6.4. Preparing Signals for A/D Conversion

noisy signal saturates the electronics and the noise is clipped. (Note that for clarity, these
traces are not drawn to scale.) After low-pass filtering, the signal looks clean, but its ampli-
tude is less than x100 the original signal because the low-pass filter extracts a signal that is
the average of the non-symmetrical, clipped noise. When the noisy signal is first amplified
by x10, then low-pass filtered before a further amplification by x10, saturation is avoided
and the amplitude of the filtered signal is x100 the original input. Panel B shows the two
outputs superimposed to illustrate the loss of magnitude in the signal that was amplified
too much before it was low-pass filtered.

A.
Original signal

+ noise
1 kHz
x100 low-pass

x10 1 kHz
low-pass

x10

B.

Signal amplified x100 before filter

Signal amplified x10 before filter then x10 after filter

Figure 6.6: Distortion of signal caused by high amplification prior to filtering.

6.4.3. OFFSET CONTROL

In some cases, it is necessary to add an offset to the signal before it is amplified. This is
necessary if the gain required to amplify the signal of interest would amplify the DC offset
of the signal to the point that it would cause the gain amplifier to saturate.

An example is provided by an electronic thermometer. A typical sensitivity of an elec-

tronic thermometer is 10 mV/ °C , with zero volts corresponding to 0 °C. A 12-bit A/D
converter can measure with an approximate resolution of 5 mV, corresponding to 0.5 °C
in this example. If the data are to be analyzed at 0.01 °C resolution, amplification by a fac-
tor of at least x50, and probably x100, would be necessary. If the temperatures of interest
are between 30 °C and 40 °C and are amplified by x100, these temperatures will corre-
spond to voltages between 30 V and 40 V values well beyond the range of the A/D con-
verter. The solution is to introduce an offset of -350 mV before any amplification, so that
zero volts will correspond to 35 °C. Now when the signal is amplified by x100, the
30–40 °C temperature range will correspond to voltages between -5 V and +5 V.

The Axon Guide — 2500-0102 Rev. C 157

 6. Signal Conditioning and Signal conditioners

6.4.4. AC COUPLING AND AUTOZEROING

AC coupling is used to continuously remove DC offsets from the input signal. Signals
below the -3 dB frequency of the AC coupling circuit are rejected. Signals above this fre-
quency are passed. For this reason, AC-coupling circuits are more formally known as
high-pass filters. In most instruments with AC coupling, the AC-coupling circuit is a first-
order filter. That is, the attenuation below the -3 dB frequency increases at 20 dB/decade.

When a signal is AC-coupled, the DC component of the signal is eliminated and the low-
frequency content is filtered out. This causes significant distortion of the signal, as shown
in Figure 6.7.

0.1 Hz

1 Hz

10 Hz

Input

500 ms

Figure 6.7: Distortion of signal caused by AC-coupling at high frequencies.

A 1 Hz square wave is AC-coupled at three different frequencies. The distortion is pro-

gressively reduced as the AC-coupling frequency is reduced from 10 Hz to 0.1 Hz. In all
cases, the DC content of the signal is removed. The problem when using the lowest AC-
coupling frequencies is that slow shifts in the baseline may not be rejected and transient
shifts in the baseline might take a long time to recover.

Since there is less distortion when the AC-coupling frequency is lower, it is tempting to
suggest that the AC coupling should always be set to very low frequencies, such as 0.1 Hz.
However, this is often unacceptable because some shifts in the baseline are relatively rapid
and need to be eliminated quickly.

A significant difference between using an AC-coupling circuit and setting a fixed DC off-
set is the way the two circuits handle ongoing drift in the signal. With fixed DC offset
removal, the ongoing drift in the signal is recorded. With AC coupling, the drift is
removed continuously. Whether this is an advantage or a disadvantage depends on
whether the drift has meaning.

158 The Axon Guide — 2500-0102 Rev. C

 6.4. Preparing Signals for A/D Conversion

For very slow signals, even the lowest AC-coupling frequency causes significant distortion
of the signal. For these signals, an alternative technique known as autozeroing can be used.
This technique is available in some signal conditioners, such as the CyberAmp 380 ampli-
fier, as well as in some amplifiers, such as the Axopatch-1 amplifier. In this technique, the
signal is DC-coupled and a sample of the signal is taken during a baseline period. This
DC sample value is stored and continuously subtracted by the instrument from the signal
until the next sample is taken. In early instruments, the sample was taken using an analog
sample-and-hold circuit. These circuits exhibit the problem known as “droop.” That is,
the sampled value drifts with time. Later systems, such as the ones used in the CyberAmp
380 amplifier, use droop-free, digital sample-and-hold circuits.

The efficacy of the technique is illustrated in Figure 6.8.

DC

10 Hz

1 Hz

0.1 Hz

DC with
Autozero

Autozero

Figure 6.8: Comparison of autozeroing to AC-coupling.

The top trace shows an EKG signal that is sitting on a 5 mV offset resulting from elec-
trode junction potentials. At AC-coupling frequencies of 10 Hz and 1 Hz, the signal is
distorted. In the bottom trace, an Autozero command is issued to the signal conditioner
(the CyberAmp 380 amplifier) at the time indicated by the arrow. The DC component is
immediately removed, but the transients are unaffected.

Autozeroing should be restricted to cases where the time of occurrence of signals is

known, so that the Autozero command can be issued during the baseline recording period
preceding the signal.

The Axon Guide — 2500-0102 Rev. C 159

 6. Signal Conditioning and Signal conditioners

6.4.5. TIME CONSTANT

The AC-coupling frequency is related to the time constant of decay, τ (see Figure 6.7):

1
τ= (6)
2π f−3

The time constants for some common AC-coupling frequencies are:

f-3 (Hz) τ (ms)

100.0 1.6

30.0 5.3

10.0 16.0
3.0 53.0

1.0 160.0

0.1 1,600.0

In one time constant, the signal decays to approximately 37% of its initial value. It takes
approximately 2.3 time constants for the signal to decay to 10% of the initial value.

6.4.6. SATURATION
The AC-coupling circuit is the first circuit in most signal conditioning pathways. If a large
step is applied to the AC-coupled inputs, the AC coupling rejects the step voltage with a
time constant determined by the AC-coupling frequency. If the amplifiers are set for high
gain, the output might be saturated for a considerable time. For example, if the gain is
x100, the AC coupling is 1 Hz and the step amplitude is 1 V, the output will be saturated
until the voltage at the output of the AC-coupling circuit falls to 100 mV from its initial
peak of 1 V. This will take about 2.3 time constants. Since the time constant is 160 ms,
the output will be saturated for at least 370 ms. For the next several time constants, the
output will settle towards zero.

6.4.7. OVERLOAD DETECTION

An amplified signal may exceed the ±10 V acceptable operating range inside the instru-
ment in two places: at the input of the various filters and at the output of the final ampli-
fier stage. An example of an overload condition at the input of an internal low-pass filter is
shown in Figure 6.6.

It is common practice to place overload-detection circuitry inside the instrument at these

points. The overload-detection circuitry is activated whenever the signal exceeds an upper
or lower threshold such as ±11 V. The limit of ±11 V exceeds the usual ±10 V recom-
mended operating range in order to provide some headroom. There is no difficulty in

160 The Axon Guide — 2500-0102 Rev. C

 6.5. Averaging

providing this headroom since the internal amplifiers in most instruments operate linearly
for signal levels up to about ±12 V.

Normally, the overload circuitry simply indicates the overload condition by flashing a
light on the front panel. In more sophisticated instruments such as the CyberAmp ampli-
fier, the host computer can interrogate the CyberAmp amplifier to determine if an over-
load has occurred.

6.5. AVERAGING
Averaging is a way to increase the signal-to-noise ratio in those cases where the frequency
spectrum of the noise and the signal overlap. In these cases, conventional filtering does
not help because if the -3 dB frequency of the filter is set to reject the noise, it also rejects
the signal.

Averaging is applicable only to repetitive signals, when many sweeps of data are collected
along with precise timing information to keep track of the exact moment that the signal
commences or crosses a threshold. All of these sweeps are summed, then divided by the
total number of sweeps (N) to form the average. Before the final division, the amplitude
of the signal in the accumulated total will have increased by N. Because the noise in each
sweep is uncorrelated with the noise in any of the other sweeps, the amplitude of the noise
in the accumulated signal will only have increased by √N. After the division, the signal
will have a magnitude of unity, whereas the noise will have a magnitude of 1/√N. Thus,
the signal-to-noise ratio increases by √N.

6.6. LINE-FREQUENCY PICK-UP (HUM)

An important consideration when measuring small biological signals is to minimize the
amount of line-frequency pickup, often referred to as hum. Procedures to achieve this goal
by minimizing the hum at its source are discussed in Chapter 2.

Hum can be further minimized by using a notch filter, as discussed above, or by differen-
tial amplification. To implement the latter technique, the input amplifier of the data
acquisition system is configured as a differential amplifier. The signal from the measure-
ment instrument is connected to the positive input of the differential amplifier, while the
ground from the measurement instrument is connected to the negative input. If, as is
often the case, the hum signal has corrupted the ground and the signal equally, the hum
signal will be eliminated by the differential measurement.

6.7. PEAK-TO-PEAK AND RMS NOISE MEASUREMENTS

Noise is a crucially important parameter in instruments designed for measuring low-level
signals.

Invariably, engineers quote noise specifications as root-mean-square (rms) values, whereas

users measure noise as peak-to-peak (p-p) values. Users’ preference for peak-to-peak values

The Axon Guide — 2500-0102 Rev. C 161

 6. Signal Conditioning and Signal conditioners

arises from the fact that this corresponds directly to what they see on the oscilloscope
screen or data acquisition monitor.

Engineers prefer to quote rms values because these can be measured consistently. The rms
is a parameter that can be evaluated easily. In statistical terms, it is the standard deviation
of the noise. True rms meters and measurement software are commonly available and the
values measured are completely consistent.

On the other hand, peak-to-peak measurements are poorly defined and no instruments or
measurement software are available for their determination. Depending on the interpreter,
estimates of the peak-to-peak value of Gaussian noise range from four to eight times the
rms value. This is because some observers focus on the “extremes” of the noise excursions
(hence, the x8 factor), while others focus on the “reasonable” excursions (x6 factor) or the
“bulk” of the noise (x4 factor).

MDS Analytical Technologies has developed the pCLAMP® software to measure the rms
and the peak-to-peak noise simultaneously. The peak-to-peak noise is calculated as the
threshold level that would encompass a certain percentage of all of the acquired data. For
white noise, the corresponding values are:

Percentage of Data Peak-To-Peak

Encompassed Thresholds

95.0% 3.5–4 times rms value

99.0% 5–6 times rms value

99.9% 7–8 times rms value

These empirical measurements can be confirmed by analysis of the Gaussian probability

distribution function.

In this Guide and various Axon Cellular Neuroscience product specifications, noise mea-
surements are usually quoted in both rms and peak-to-peak values. Since there is no com-
monly accepted definition of peak-to-peak values, MDS Analytical Technologies usually
uses a factor of about 6 to calculate the peak-to-peak values from the measured rms values.

In an article from Bell Telephone Laboratories (1970), the authors define the peak factor
as “the ratio of the value exceeded by the noise a certain percentage of the time to the rms
noise value. This percentage of time is commonly chosen to be 0.01 per cent.” This ratio,
from tables listing the area under the Gaussian distribution, turns out to be 7.78 times the
rms value. Thus, according to this article, the appropriate factor to calculate the peak-to-
peak values from the measured rms values is closer to x8.

162 The Axon Guide — 2500-0102 Rev. C

 6.8. Blanking

6.8. BLANKING
In certain experiments, relatively huge transients are superimposed on the signal and cor-
rupt the recording. This problem commonly occurs during extracellular recording of
nerve impulses evoked by a high-voltage stimulator. If the isolation of the stimulator is
not perfect (it never is), there is some coupling of the stimulus into the micropipette
input. This relatively large artifact can sometimes cause the coupling capacitors in subse-
quent AC amplifiers to saturate. There might be some time lost while these capacitors
recover from saturation, and thus valuable data might be wasted.

If it is not possible to prevent the stimulus coupling, the next best thing to do is to sup-
press the artifact before it feeds into the AC-coupled amplifiers. This is made possible in
the Axoclamp 900A amplifier by providing sample-and-hold amplifiers early in the signal
pathway. The user provides a logic-level pulse that encompasses the period of the tran-
sient. This logic-level pulse forces the sample-and-hold amplifiers into the “hold” mode.
In this mode, signals at the input of the sample-and-hold amplifier are ignored. Instead,
the output of the sample-and-hold amplifier is kept equal to the signal that existed at the
moment the logic-level pulse was applied.

6.9. AUDIO MONITOR—FRIEND OR FOE?

When monitoring data, experimenters need not be limited to their sense of sight. The
data may also be monitored with great sensitivity using the sense of hearing. In such cases,
the data are input to an audio monitor and fed to a loudspeaker or a set of headphones.

Two types of audio monitor are used. The first is a power amplifier that applies the signal
directly to the speaker, thereby allowing signals in the audio bandwidth to be heard. This
type of audio monitor is frequently used in EMG monitoring or in central nervous system
recording. Each spike is heard as an audible click, and the rate and volume of clicking is a
good indicator of the muscle or nerve activity. This type of audio monitor is either called
an AM (amplitude modulated) audio monitor, or it is said to be operating in “click”
mode.

The second type of audio monitor is a tone generator whose frequency depends on the
amplitude of the input signal. Usually, the oscillator frequency increases with increasingly
positive signals. This type of audio monitor is often used for intracellular recording. It
provides an extremely sensitive measure of the DC level in the cell. This type of audio
monitor is either called an FM (frequency modulated) audio monitor, or it is said to be
operating in “tone” mode.

Most researchers have a very strong opinion about audio monitors; they either love them
or hate them. Those who love audio monitors appreciate the supplementary “view” of the
data. Those who hate audio monitors have probably suffered the annoyance of having to
listen for interminable lengths of time to the audio output from another rig in the same
room. A constant tone can be irritating to listen to, especially if it is high pitched, unless it
has an important message for you, such as “my cell is maintaining its potential admirably.”

The Axon Guide — 2500-0102 Rev. C 163

 6. Signal Conditioning and Signal conditioners

Audio monitors are standard features in Axon Cellular Neuroscience computer controlled
amplifiers. To minimize the potential for aggravation, we include a headphone jack so
that users can listen to the audio output without testing the patience of their colleagues.

6.10. ELECTRODE TEST

It is useful to be able to measure the electrode resistance for two reasons. The first reason
is to establish the basic continuity of the electrode circuit. Sometimes, electrode leads can
break, leaving an open-circuit input and, consequently, no incoming data. The second
reason is to verify that the electrode is acceptably attached. For example, it may be that to
achieve the low-noise recording levels needed in an EMG recording, the electrode resis-
tance must be less than 5 kΩ.

Some transducer amplifiers allow the electrode impedance to be easily measured. For
example, the CyberAmp amplifiers have an Electrode Test facility. When this is activated,
an approximate 1 μAp-p, 10 Hz square wave is connected to every input via individual
1 MΩ resistors. The electrode resistance can be directly determined from the amplitude of
the voltage response.

6.11. COMMON-MODE REJECTION RATIO

In general, the information that the researcher wants to record is the difference between
two signals connected to the positive and the negative amplifier inputs. Often, these sig-
nals contain an additional component that is common to both, but that does not contain
relevant information. For example, both outputs of a strain gauge might include a DC
potential of 2.5 V that arises from the excitation voltage. However, the strain on the gauge
generates a microvolt-size difference between the two outputs, but does not affect the
2.5 V “common-mode” voltage. Another example is often seen when recording EMG sig-
nals from an animal. Both the positive and the negative electrodes pick up line-frequency
hum that has coupled into the animal. The hum picked up by the electrodes may be as
large as 10 mV, but it is identical on both electrodes. The EMG signal of interest is the
small difference between the potentials on the two electrodes; it may be as small as a few
tens of microvolts.

To prevent the common-mode signal from swamping the much smaller differential signal,
the positive and negative gains of the amplifier must be nearly identical. In the above
EMG example, if the amplifier inputs are exactly unity, the 10 mV of hum that appears
equally on both electrodes does not show up at all on the amplifier output. The only sig-
nal to appear on the output is the small signal proportional to the EMG potential gener-
ated between the two electrodes.

In practice, the positive and negative inputs of the amplifier are never exactly equal. The
quality of their matching is measured by the common-mode rejection ratio (CMRR).
This is normally quoted in dB, where 20 dB corresponds to a factor of ten. Returning to
the EMG example, if the amplifier operates at unity gain with a CMRR of 60 dB (i.e.,
one part in a thousand), the 10 mV of common-mode hum results in 10 μV of hum

164 The Axon Guide — 2500-0102 Rev. C

 6.12. References

appearing on the amplifier output. This is small, but still significant compared with the
smallest EMG signals, so an amplifier with higher CMRR, e.g., 80 dB, may be desirable.

The CMRR of an amplifier varies with frequency. It is best at very low frequencies, while
above a certain frequency it diminishes steadily as the frequency of the common-mode
signal increases. It is therefore important to verify the CMRR of the amplifier at a fre-
quency that exceeds the expected frequency of the common-mode signal.

The CMRR of the recording system is adversely affected by imbalances in the source resis-
tances of the recording electrodes. This is because the source resistance of each electrode
forms a voltage divider with the input resistance of the amplifier. If the source resistances
of the two electrodes are not identical, the voltage dividers at the positive and negative
inputs of the amplifier are not equal. Returning to the EMG example, if the resistance of
one electrode is 9 kΩ, the resistance of the other is 10 kΩ, and the amplifier input resis-
tances total 1 MΩ, then the gain for one electrode is 0.9901 instead of unity, while the
gain for the other electrode is 0.9911. The difference is 0.001. Thus, even though the
amplifier may have a CMRR of 80 dB or more, the system CMRR is only 60 dB. In some
cases, 60 dB is acceptable, but in others it is not. The solution to this problem is to use an
amplifier that has very high input resistances of 100 MΩ or more.

If there is a large common-mode signal and a source imbalance of more than a few
kilohms, a high input resistance amplifier probe should be used. Several AI 400 series
probes are available from MDS Analytical Technologies that have input resistances of
10 gigohms (1010 Ω) or more. These probes are distinguished on the basis of noise, cost
and size.

6.12. REFERENCES
Bell Telephone Laboratories. Transmission Systems for Communications. By the mem-
bers of the technical staff at Bell Telephone Laboratories. Western Electric Company, Inc.,
Technical Publications. Winston-Salem, North Carolina, 1970.

6.13. FURTHER READING

Hamming, R. W., Digital Filters. Prentice-Hall, Inc., Englewood Cliffs, New Jersey,
1977.

Tietze, U., Schenk, Ch., Advanced Electronic Circuits. Springer-Verlag, Berlin, 1978.

The Axon Guide — 2500-0102 Rev. C 165

 6. Signal Conditioning and Signal conditioners

166 The Axon Guide — 2500-0102 Rev. C

 7. Transducers
Transducers are devices that convert a signal of one form to another. In physiological
applications this usually means converting variations of a physical signal, such as tempera-
ture or pressure, into variations in voltage that can then be recorded or manipulated as
necessary.

There are many different types of transducers available for physiological recording and
they range in their requirements for excitation voltages, amplification and recording tech-
niques as well as mechanical connections. In the past this has required a different ampli-
fier for each transducer type. The introduction of the CyberAmp® 380 amplifier
(MDS Analytical Technologies), a computer-controlled signal-conditioning amplifier,
heralded a new approach to interfacing transducers. A small 15-pin socket on each chan-
nel of the CyberAmp 380 amplifier provides the excitation voltage, the differential inputs,
the offset correction and the filtering required by a wide variety of transducers. This elim-
inates the need for a dedicated amplifier to suit the transducer type. The AI 400 series
probes and adapters can plug into any socket on the CyberAmp 380 amplifier. Within the
15-pin probe connector is a small memory device called an EEPROM (electrically eras-
able programmable read only memory) that stores such information as the transducer
type, its scale factor and the offset. When the transducer is plugged into the
CyberAmp 380 amplifier, this information may be automatically loaded by the acquisi-
tion software, thereby making the transducer ready for immediate use.

7.1. TEMPERATURE TRANSDUCERS FOR PHYSIOLOGICAL

TEMPERATURE MEASUREMENT
Three types of transducers are suitable for physiological temperature measurement: ther-
mistors, integrated circuit (IC) temperature transducers that produce an output current
proportional to absolute temperature, and IC temperature transducers that produce an
output voltage proportional to absolute temperature.

7.1.1. THERMISTORS
Thermistors are resistors whose resistance drops significantly as temperature increases.
They are composed of a compressed and sintered mixture of metallic oxides of manga-
nese, nickel, cobalt, copper, magnesium, titanium and other metals. Thermistors exhibit
temperature coefficients many times greater than those of pure metals. For most of them,
the resistance falls by 4–6% per °C with increasing temperature.

The Axon Guide — 2500-0102 Rev. C 167

 7. Transducers

The response is exponential with temperature, but can be considered to be linear over a
range of a few tens of degrees Celsius. By combining multiple thermistors and some resis-
tors, the useful linear range can be extended to span more than 100 °C. The Omega Engi-
neering, Inc. (Stamford, CT) thermistor composites 44018 and 44019 each contain a
matched pair of thermistors. When connected by precision resistors to an accurate DC
excitation voltage, these probes are interchangeable within ±0.15 °C over the recom-
mended range of -30 to +100 °C. These probes come in a wide range of shapes and sizes
for different applications such as esophageal, rectal and surface body temperature mea-
surements, or for immersion in hostile chemical environments.

Omega Engineering Inc. produces many other thermistor types as do other manufactur-
ers. These other types can also be interfaced to the CyberAmp 380 amplifier via the AI
490 Connector and AI 491 Cable Kits. Several different interface circuits are suggested in
the CyberAmp 380 manual.

7.1.2. IC TEMPERATURE TRANSDUCERS THAT PRODUCE AN OUTPUT CURRENT

PROPORTIONAL TO ABSOLUTE TEMPERATURE
The most notable of these temperature-dependent current sources are the Analog Devices,
Inc. (Norwood, MA) AD590 with a temperature range of -55 to +150 °C and the lower-
cost AD592 with a temperature range of -25 to +105 °C. These are not available in the
wide range of probe configurations provided by thermistor manufacturers. Most often
they are supplied in 0.23” (5.8 mm) diameter metal cans or in 0.25” x 0.093” (6.4 mm x
2.4 mm) ceramic flat packs. These need to be connected to wires and must be insulated
for special purposes. The only temperature probe available from Analog Devices is the
AC2626, which includes an AD590 enclosed in a 4” (102 mm) or 6” (153 mm) long
stainless steel sheath with a 3/16” (4.8 mm) outside diameter.

When connected to a DC voltage source (e.g., the +5 V excitation voltage in the

CyberAmp 380 amplifier) these transducers force the current that flows in the circuit to
be equal to 1 μA/°K. The external circuitry required to use the AD590 is very simple and
it is easy to configure for minimum and average temperature measurements by placing
transducers in series or in parallel, respectively. The absolute accuracy and interchange-
ability of these probes is inferior to the Omega Engineering probes described above. In
their best and most expensive grade, the interchangeability is ±0.5 °C at 25 °C. The main
advantages of the AD590 temperature probes are that the external support circuitry is very
simple and that they are suitable for remote sensing applications. Because the output is
insensitive to voltage drops, the devices can be used with twisted-pair cabling hundreds of
feet in length.

The Analog Devices semiconductor temperature probes can be interfaced to the

CyberAmp 380 amplifier via the AI 490 Connector and AI 491 Cable Kits, with the user
providing the interface design based on the range of application circuits provided by Ana-
log Devices and in the manual for the CyberAmp 380 amplifier.

168 The Axon Guide — 2500-0102 Rev. C

 7.2. Temperature Transducers for Extended Temperature Ranges

7.1.3. IC TEMPERATURE TRANSDUCERS THAT PRODUCE AN OUTPUT VOLTAGE

PROPORTIONAL TO ABSOLUTE TEMPERATURE
Two of the most suitable temperature-dependent voltage sources are the LM35A and the
LM135A sensors from National Semiconductor Corporation (Santa Clara, CA). These
have absolute accuracies (interchangeability) of ±0.5 °C and ±1.0 °C, respectively, at
25 °C. The LM35A has a zero voltage output at 0 °C and a sensitivity of 10 mV/°C. The
LM135A behaves as a zener diode with a zero voltage output at 0 °C and a sensitivity of
10 mV/°C. These devices are generally cheaper than the AD590 transducers. They are
supplied in approximately 0.2” (5 mm) metal cans or plastic packages, but they are not
available in ready-to-use probes.

The National Semiconductor temperature probes can be interfaced to the CyberAmp 380
amplifier via the AI 490 Connector and AI 491 Cable Kits with the user providing the
interface design using the range of application circuits provided by National Semiconduc-
tor and in the manual for the CyberAmp 380 amplifier.

7.2. TEMPERATURE TRANSDUCERS FOR EXTENDED

TEMPERATURE RANGES
Two types of transducers are commonly used for temperature measurements beyond the
physiological range: thermocouples and resistance temperature detectors.

7.2.1. THERMOCOUPLES
Thermocouples are economical and rugged transducers, having the advantage of small size
with a very fast response time and a wide temperature range. Thermocouples consist of
two dissimilar metals in contact. The offset potential between the metals is proportional
to temperature (see Table 7.1). The low sensitivity and broad operating range generally
make thermocouples more suitable for industrial applications than for physiological ones.
Because of their greater sensitivity, thermistors are more popular than thermocouples for
most physiological applications. On the other hand, thermocouples can be fabricated in
remarkably small sizes, creating the opportunity for some unusual biological applications.
For example, micron-dimension thermocouples that can be inserted into single living cells
have been described (Cain & Welch, 1974). Commercially available thermocouples can
be obtained with diameters as low as 25 μm and thermal time constants as short as 2 ms
(Omega Engineering, Inc., Stamford, CT).

The Axon Guide — 2500-0102 Rev. C 169

 7. Transducers

Table 7.1: Common Thermocouple Materials.

ANSI Sensitivity at
Min Value °C Max Value °C Material
Type 20 °C μV/°C
E -200 900 60.48 chromel/constantan
J -200 750 51.45 iron/constantan
K -200 1250 40.28 chromel/alumel
R 0 1450 5.80 platinum/Pt-13%
rhodium
S 0 1450 5.88 platinum/Pt-10%
rhodium
T -200 350 40.28 copper-constantan

Thermocouples require a reference temperature that traditionally was an ice bath but is
more commonly provided by a compensation circuit. Complete signal-conditioning mod-
ules, such as the Analog Devices AD594 and 3B47 signal conditioners, exist. They con-
tain the differential amplifiers and the temperature compensation circuitry. Alternatively,
voltage-output temperature transducers, such as the LM35A, can be used to derive a com-
pensation voltage to apply to the negative input of a CyberAmp 380 amplifier. To inter-
face thermocouples to the CyberAmp 380 amplifier, the user must provide an interface
circuit using a temperature transducer for compensation or using the Analog Devices or
equivalent conditioning modules. Circuit examples are given in the CyberAmp manuals.

7.2.2. RESISTANCE TEMPERATURE DETECTORS

Resistance thermometers can be made of metals or ceramic-like mixtures. By convention,
resistance thermometers made of metals are called Resistance Temperature Detectors
(RTD), while the ceramic-like resistance thermometers are called thermistors.

The resistance of most metals increases with increasing temperature. The sensitivity is
small, less than 0.4% per °C. The most commonly used metal is platinum because of its
wide linear resistance-to-temperature relationship.

Platinum RTD’s offer better stability and linearity than thermocouples, but are limited to
temperatures below 850 °C. Most platinum RTD’s have a resistance of 100 Ω at 0 °C and
a positive temperature coefficient of 0.385% per °C at 0 °C.

The simplest way to configure a platinum RTD is to excite it with a small, accurate DC
current. Currents of 1 mA or less are used so as to minimize self-heating. A differential
amplifier is used to measure the voltage across the RTD. If the excitation current is 1 mA,
the sensitivity is 385 μV per °C. To interface RTD’s to the CyberAmp 380 amplifier, the
user must provide an interface circuit. Circuit examples are given in the CyberAmp man-
uals.

170 The Axon Guide — 2500-0102 Rev. C

 7.3. Electrode Resistance and Cabling Affect Noise and Bandwidth

7.3. ELECTRODE RESISTANCE AND CABLING AFFECT NOISE

AND BANDWIDTH
The electrode impedance (resistance), the capacitance of the cable to the electrode, and
the amplifier input impedance combine to produce resistive dividers and filters
(Figure 7.1) that can substantially degrade the quality of a recorded signal. The easiest
solution to these potential problems is to use high-impedance probes (e.g., AI 401, AI
402) and place them as close as possible to the electrodes.

R E+
VS+ Cc +
RZ
VO

RE -
VS- Cc -
RZ

Figure 7.1: Electrode resistance and cabling can degrade a signal.

The electrode impedances (resistances) RE combine with the amplifier input impedances
RZ to produce resistive dividers. RZ also combines with the capacitance of the cable to the
electrodes CC to produce low-pass filters. VS and VO are the source and output voltages.

7.4. HIGH ELECTRODE IMPEDANCE CAN PRODUCE SIGNAL

ATTENUATION
The electrode resistance in series with the amplifier input impedance acts as a voltage
divider (Figure 7.2). The CyberAmp 380 amplifier has input impedances of 1 MΩ and if
the electrode resistance is 1 kΩ this reduces the signal by only 0.1%, since:

1MΩ
VI = Vs (1)
1kΩ + 1MΩ

However, if the electrode resistance is 100 kΩ, the signal would be reduced by about
9.1%, since:

1MΩ
VI = Vs = 0.909Vs (2)
100kΩ + 1MΩ

The Axon Guide — 2500-0102 Rev. C 171

 7. Transducers

The solution to high electrode resistance is to use a high-input impedance amplifier. If

using a CyberAmp 380 amplifier, then you should use a high-impedance probe between it
and the electrodes. The AI 401 x10 Low-Noise Differential amplifier probe and the AI
402 X50 Ultra Low-Noise Differential amplifier probe both have input impedances mea-
sured in the teraohms (1012 Ω).

VS

RE

VI

RZ
RZ
VI =
RE + RZ

Figure 7.2: High electrode resistance can attenuate a signal and increase crosstalk.

The resistance of each electrode in series with the amplifier input impedance acts as a pair
of voltage dividers. This can reduce the recorded amplitude of the signal. If the electrode
impedances are not matched, it will reduce the CMRR leading to amplification of com-
mon background signals. RE and RZ are the electrode and amplifier impedances. VS and
VI are the source voltage and input voltage to the amplifier.

7.5. UNMATCHED ELECTRODE IMPEDANCES INCREASE

BACKGROUND NOISE AND CROSSTALK
When making differential measurements such as EMG, EKG and EEG it is important
that both electrodes have similar properties. As we have seen above, high electrode imped-
ance can produce signal attenuation. If we apply Figure 7.2 to both inputs of Figure 7.1,
we see that we could easily get greater attenuation of the signal at one electrode than at the
other. This can substantially degrade the common mode rejection ratio (CMRR) and
result in amplification of common background signals that would otherwise be rejected.

For a differential amplifier, the CMRR is the ratio of the response for a signal at just one
input to the response for a common mode signal (applied to both inputs) of the same
amplitude. The CMRR of the CyberAmp 380 amplifier is 100 dB (at high gain). That is,
the output voltage with a signal applied to just one input is 100,000 times greater than
the output voltage with the same signal applied to both inputs. (The ratio of amplitudes is
measured in decibels where dB = 20 log10(Vout/Vin). A ten-fold increase in the ratio
equals a 20 dB increase (see Chapter 6).

172 The Axon Guide — 2500-0102 Rev. C

 7.6. High Electrode Impedance Contributes to the Thermal Noise of the System

If the electrode impedance of one electrode connected to a CyberAmp 380 amplifier is

1 kΩ and that of the other electrode 10 kΩ then the CMRR of the system as a whole is
reduced 1,000 fold to 41 dB.

Applied to a single channel of the CyberAmp 380 amplifier:

1MΩ
VI = V s = 0.999V s
1kΩ + 1MΩ

1MΩ
VI = V s = 0.990V s
10 kΩ + 1MΩ

The difference equals 0.009:

CMRR = 20 log 10 (0.999 / 0.009) = 41dB

As for the problem of the signal attenuation, the solution to a mismatch of electrode resis-
tances is to use an amplifier probe with very high input resistance such as the AI 401 and
the AI 402 probes. With each of these probes, the electrode impedances used in the exam-
ple are too small to have any influence on the CMRR of the probes.

If high-input impedance probes are not available, the CyberAmp 380 amplifier has a
built-in capacity to measure electrode impedances and this should be carried out prior to
each recording session to ensure a match. This is a good idea anyway as it can reveal the
development of faults such as breaks in the electrodes.

Crosstalk from other body signals can occur for additional reasons, although infrequently.
For example, if invasive EMG leads are routed subcutaneously near the heart, the EKG
can capacitively couple into the EMG leads. This problem is eliminated by recording dif-
ferentially using lead pairs that are twisted together, or by not routing the EMG leads near
the heart.

7.6. HIGH ELECTRODE IMPEDANCE CONTRIBUTES TO THE

THERMAL NOISE OF THE SYSTEM
With an active amplifier the only noise sources encountered are the thermal noise of the
electrode, the noise provided by the amplifier itself, and crosstalk from other body signals.

Thermal noise, also called Johnson noise, is a voltage produced across the terminals of all
resistive elements (including electrodes) due to the random motion of charge carriers
within the element. Thermal noise is proportional to the resistance (R) and absolute tem-
perature (T) of the resistive element (see Chapter 11). To minimize the thermal noise con-

The Axon Guide — 2500-0102 Rev. C 173

 7. Transducers

tribution, the electrode resistance must be minimized. This is usually accomplished by

maximizing the surface area of the electrode and ensuring good electrical contact.

The noise contribution to the signal from the amplifiers is negligible if the amplifier noise
is less than the thermal noise of the electrodes. All Axon Cellular Neuroscience amplifiers
have very low noise levels. The lowest noise amplifier, the AI 402 x50 differential ampli-
fier probe, has extremely low noise of just 1.1 μVp-p in the DC-10 kHz bandwidth
(0.18 μV rms). This is approximately equivalent to the thermal noise of a 250 Ω resistor.
This means that for electrodes whose resistance exceeds 250 Ω, the thermal noise of the
electrode exceeds the noise of the amplifier.

7.7. CABLE CAPACITANCE FILTERS OUT THE HIGH-

FREQUENCY COMPONENT OF THE SIGNAL
The source resistance, together with the capacitance of the cable to the amplifier, act as a
simple RC low-pass filter (Figure 7.1). Electrical cables can provide 30–100 pF capaci-
tance per foot (100–300 pF per meter). If a cable has a 1,000 pF capacitance and the elec-
trode resistance is 10 kΩ, then the cut-off frequency is about 16 kHz. A 100 kΩ electrode
resistance with the same cable would filter the signal at 1,600 Hz. The appropriate solu-
tion is to use an active probe as close to the signal source as possible.

7.8. EMG, EEG, EKG AND NEURAL RECORDING

7.8.1. EMG
Electromyograms (EMG) can be recorded in many ways, each with its own special
requirements. Surface EMG electrodes have the advantage of being non-invasive but suf-
fer from artifacts or even total loss of signal during movements. They are also not as selec-
tive as implanted electrodes.

Implanted electrodes must be capable of remaining in the same location and must be of
an appropriate size and separation. Bipolar electrodes should be placed in parallel with the
muscle fibers to record the maximum signal with an electrode spacing of 2–10 mm
appropriate for most mammalian muscles. This close spacing of electrodes reduces
crosstalk from other muscle sites and is therefore appropriate for selective recording from
local areas. Low-frequency components of electrical signals propagate for larger distances
in body tissues than do high-frequency components. The bipolar electrode configuration
acts as a high-pass filter whose cut-off frequency is determined by electrode spacing. Close
spacing results in a high cut-off frequency, thus filtering out some of the remaining low-
frequency components from distant muscle activity. However, close electrode spacing also
reduces the amplitude of the signal in a non-linear fashion. Larger electrodes reduce the
impedance (see “Noise Arising From Distributed Pipette Resistance and Capacitance” on
page 227) but are less selective regarding the site of muscle activity.

The frequency range of the EMG is 10–2,000 Hz, although the electrode configuration
and separation will have considerable influence on what frequencies are recorded. The sig-

174 The Axon Guide — 2500-0102 Rev. C

 7.9. Metal Microelectrodes

nal size ranges from 5 μV–20 mV for surface recording and from 50–1,000 μV for inva-
sive recording.

Several probes are available for recording EMG with the CyberAmp 380 amplifier. These
include the AI 401, 402, and 405 active amplifier probes, the AI 417 passive adapter and
direct user connection.

7.8.2. EKG
There are probably more books available on electrocardiogram (EKG) recording and anal-
ysis than any other electrophysiological topic. The signals are usually of large amplitudes
and readily recorded without the need for any amplifier probes. As electrode spacing is
reduced, there is an increasing possibility of recording unwanted EMG signals from
neighboring muscles. Consequently, the traditional electrode sites are worth considering.
The AI 417 passive 2 mm adapter provides a single differential EKG channel and plugs
directly into the CyberAmp 380 amplifier.

The normal frequency range of the mammalian EKG is 0.2–100 Hz; its amplitude size is
up to 2–3 mV.

7.8.3. EEG
The electroencephalogram (EEG) ranges from 10–300 μV in amplitude and has a fre-
quency range from 0.2–50 Hz. A single differential EEG channel is best recorded with the
assistance of one of the low-noise AI 400 series active probes.

7.8.4. NERVE CUFFS

Nerve-cuff recordings have a frequency response to 10 kHz and an amplitude in the low
microvolt range. The AI 402, x50 Ultra Low-Noise Differential amplifier probe is
designed for this application, with 10 kHz noise of less than 0.18 μVrms (1.1 μVp-p) in the
0.1–10 kHz bandwidth.

The nerve cuffs themselves are very simple to make by running bare stainless steel fine
wires through a small length of silicone tubing split longitudinally. Any bared wires that
are in contact with the outer surface of the cuff can be insulated with silicone adhesive.

7.9. METAL MICROELECTRODES

The impedance of metal microelectrodes may be several hundred kilohms or more. All of
the low-noise AI 400 series active probes are suitable for recording from high-resistance
microelectrodes. Since the input capacitance of the AI 401 differential amplifier probe is
approximately 5 pF, the largest electrode resistance consistent with maintaining a 10 kHz
bandwidth is about 3 MΩ.

The Axon Guide — 2500-0102 Rev. C 175

 7. Transducers

7.10. BRIDGE DESIGN FOR PRESSURE AND FORCE

MEASUREMENTS
Pressure and force transducers are often constructed using strain gauges connected in a
Wheatstone bridge circuit. The basic circuit is shown in Figure 7.3.

The Wheatstone bridge should be considered as two pairs of resistive dividers with the
output voltage equaling the difference between the voltages at (A) and (B). The bridge cir-
cuit is good for detecting small changes in resistance, and the output is linear in the region
of balance, i.e., when the voltage difference is small.

VDC

R1 R3
Vo
A B

Differential
R2 R4 Amplifier

Figure 7.3: Wheatstone Bridge circuit with amplifier.

The largest output comes from having strain gauges for all four resistive elements,
although physical constraints may require the user to have only two or even one “active”
element. For applications where only two gauges can be used, they should be placed in
positions R1 and R2 and the gauges should be located such that one increases in resistance
and the other decreases in resistance during the applied pressure or force. This is usually
achieved by placing the elements on opposites sides of the beam under strain. When
choosing strain gauges, keep in mind that semiconductor types have outputs ten times
higher than metal-film types.

The sensitivity of a bridge circuit decreases with increasing temperature and some applica-
tions may therefore require the addition of a temperature compensation circuit that
should be placed as close as possible to the bridge in order to experience the same temper-
ature changes.

7.11. PRESSURE MEASUREMENTS

In the past, high-quality pressure transducers were very expensive and suffered from sig-
nificant temperature sensitivity. The introduction of semiconductor pressure transducers

176 The Axon Guide — 2500-0102 Rev. C

 7.12. Force Measurements

has greatly reduced the cost, decreased the temperature sensitivity and enhanced the range
of transducer devices available.

An important point when measuring very low pressures is to set the height of the pressure
transducer to that of the sense location in order to avoid hydrostatic errors introduced by
fluid-filled catheters. Pressure transducers should also be kept out of the range of heat
sources such as heat lamps to avoid temperature-induced increases in the signal output.

Major producers of semiconductor pressure sensors include Honeywell, Inc. (Minneapo-

lis, MN). Blood pressure transducers from Cobe Laboratories, Inc. (Lakewood, CO) have
a usable range of -50 to + 300 mmHg. All pressure transducers can be interfaced to the
Axon CyberAmp 380 amplifier via the AI 490 Connector and AI 491 Cable Kits.

7.12. FORCE MEASUREMENTS

The requirements for force measurements are so diverse that many people need to make
their own force transducer, although some are commercially available (e.g., Grass Instru-
ments Company, Quincy, MA). Force transducers have some compliance and you must
consider this when choosing transducers. For example, with the Grass Instruments FT10,
a force of 100 N will stretch the device by 1 mm.

For user-designed force transducers, the two main choices are between resistive or semi-
conductor strain gauges. Resistive gauges are cheaper and generally more rugged while
semiconductor strain gauges are smaller and about 10 times more sensitive. Strain gauges
are available from Measurement Specialties, Inc. (Hampton, VA).

User-designed force transducers can be interfaced to the CyberAmp 380 amplifier via the
AI 490 Connector and AI 491 Cable Kits.

7.13. ACCELERATION MEASUREMENTS

Accelerometers can consist of piezoelectric pressure sensors attached to a test mass, small
enough for physiological experiments. Piezoelectric accelerometers can have a wide fre-
quency range (1 Hz to 25 kHz) and a wide dynamic range (0.01–2,500 g). Accelerome-
ters can also consist of strain gauges attached to a mass. Alternative systems use feedback
to prevent the test mass from being displaced; the amount of the applied feedback force
represents the output signal.

7.14. LENGTH MEASUREMENTS

7.14.1. IMPLANTABLE LENGTH GAUGES
Length can be measured inside an animal using mercury-filled or saline-filled silicone tub-
ing length gauges (Lemon and Prochazka, 1984). As the tubing is stretched, the imped-
ance of the transducer increases. To avoid the generation of bubbles in the tubing, a high-
frequency AC signal is used with an AC bridge circuit. In practice, these gauges have

The Axon Guide — 2500-0102 Rev. C 177

 7. Transducers

many difficulties and may not be much better than cinematography techniques using
markers on the limb joints for measuring muscle lengths.

7.14.2. LINEAR POTENTIOMETERS

Linear potentiometers are inexpensive and very simple to use but must be perfectly
aligned to avoid internal damage. Linear potentiometers can be interfaced to the Cyber-
Amp 380 amplifier via the AI 490 Connector and AI 491 Cable Kits.

7.14.3. LINEAR VARIABLE DIFFERENTIAL TRANSFORMERS (LVDT)

Linear variable differential transformers are used for accurate and stable measurement of
linear position and displacement. Rotary variable differential transformers (RVDT) are
similarly used to measure rotation. These devices are very rugged and since there is no
contact between the core and the body of the LVDT (or RVDT), there is no friction or
wear, providing essentially infinite life. Linearity of ±0.25% over the full range is typical
(ADS Transicoil, Inc., Collegeville, PA).

7.15. SELF-HEATING MEASUREMENTS

As a result of applying an excitation current or voltage to the temperature measurement
sensor, power is dissipated in the sensor. This power dissipation causes the temperature of
the sensor to rise above the ambient temperature that is being measured. This phenome-
non is referred to as “self heating.” For most of the temperature sensors used in physiology,
it is necessary to keep the power dissipation to a few milliwatts or less. If this condition is
met, temperature changes of less than 0.01 °C can be measured.

Other transducers, such as strain gauges, are temperature-sensitive; therefore it is impor-

tant not to allow self heating in these transducers to cause a temperature rise sufficient to
affect the measurement.

7.16. ISOLATION MEASUREMENTS

In industrial environments, it is common for signal leads from transducers to be run over
long distances past machinery that can induce large voltages in the leads. To protect the
measurement instrument, isolation amplifiers should be used for each transducer. Induced
voltages are not commonly a problem in animal physiology applications and, therefore,
isolation is not required. Note that if measurements are to be made from human subjects,
isolated probes that are approved for human use must be used.

7.17. INSULATION TECHNIQUES

Electrodes that are implanted for long-term recording eventually fail either due to the rup-
ture of a solder joint, breakage of wires at points of repeated flexion, or moisture penetra-
tion through the insulation. Although teflon-insulated wire is commonly used, when it is
connected to the leads of the transducer it is difficult to join the teflon insulation to the
insulation on the transducer leads. A solution is to etch the teflon so that it will adhere to

178 The Axon Guide — 2500-0102 Rev. C

 7.18. Further Reading

the adhesive used to join the two insulations. In practice it is often faster and better to use
PVC-insulated wires because it is much easier to make a strong adhesion, even though
PVC is more permeable to water over the long term. Also, Araldite AV138M (CIBA-
GEIGY) and Epoxylite #6001 (Epoxylite Corporation, Anaheim, CA) both provide excel-
lent water-resistant insulation for rigid applications.

Special attention should be paid to providing strain relief where wires, and particularly
connections, are repeatedly flexed. Sliding a length of silicon tubing over the stress region
often reduces this problem.

7.18. FURTHER READING

Cain, C., Welch, A. J., Thin-film temperature sensors for biological measurement. IEEE
Trans. Biomed. Eng. BME-21(5), 421–423, 1974.

Cooke, I.R., Brodecky, V., Becker, P.J., Easily implantable electrodes for chronic recording
of electromyogram activity in small fetuses. J. Neurosci. Meth. 33, 51–54, 1990.

Geddes, L. A., Baker, L. E., Principles of Applied Biomedical Instrumentation. John

Wiley & Sons, New York, 1989.

Horowitz, P., Hill, W., Measurements and signal processing. The Art of Electronics, 14.
Cambridge University Press, Cambridge, 1986.

Lemon, R., Prochazka, A. Eds. Methods for neuronal recording in conscious animals.
IBRO Handbook Series: Methods in the Neurosciences, Vol. 4. John Wiley & Sons,
Chichester, 1984.

Loeb, G. E., Gans, C., Electromyography for Experimentalists. University of Chicago

Press, Chicago, 1986.

The Axon Guide — 2500-0102 Rev. C 179

 7. Transducers

180 The Axon Guide — 2500-0102 Rev. C

 8. Laboratory Computer Issues
and Considerations
Chapter 8 provides an overview of personal computers. The earlier sections give more
detailed information on computer options and the issue of price vs. performance. The last
section provides a glossary of computer terms that you may come across when researching
computers.

8.1. SELECT THE SOFTWARE FIRST

The first consideration when setting up a personal computer is to determine which soft-
ware packages will be used, and what kind of computer that software will run on. For spe-
cialized applications, such as scientific analysis packages, it is not safe to assume that
comparable packages are available for both Macintosh and Windows computers. If one
needs a software package for the acquisition and analysis of microelectrode voltage-clamp
and current-clamp data, the Axon pCLAMP® software suite is available for Windows.

8.2. HOW MUCH COMPUTER DO YOU NEED?

8.2.1. THE MACHINE SPECTRUM: CAPABILITY VS. PRICE
Once you have determined the type of computer you are going to buy, you need to decide
how much you are willing to spend for performance and convenience. The main compo-
nents of the computer that you need to consider are discussed in the following sections.
Although very powerful computers have become relatively cheap in recent years, it is still
worth considering the motivations for upgrading the various system components.

8.2.2. MEMORY
RAM (Random Access Memory) is the physical device used to store programs and data
while the computer is running. When the computer power is off, RAM cannot hold any
information, so it is referred to as “volatile” memory. It is used because instructions and
data can be read from, and written to, RAM much faster than any other type of memory
device. How much RAM you need and can use depends on the type of computer you buy
as well as the capabilities of the programs you use.

The usefulness of extra memory depends on the type of applications that you will be run-
ning. Acquiring and analyzing very large datasets in general requires more memory than
acquiring and analyzing smaller amounts of data.

The Axon Guide — 2500-0102 Rev. C 181

 8. Laboratory Computer Issues and Considerations

8.2.3. HARD DRIVE

At the time of writing, 500 GB hard drives are not uncommon, and hard drives with
capacity of a terabyte or more are now widely available to consumers.

For traditional electrophysiology experiments, these hard drives provide more than
enough storage. For applications that include live cell video at high frame rates, a terabyte
of storage is not enough for long-term archiving of data. The good news is that hard
drives using USB, Firewire or SATA can be stacked externally, and are cheap enough to be
able to keep up with the increasing amounts of data.

For data acquisition, hard disk drive performance may be important. To record long seg-
ments of unbroken data, direct acquisition to disk (rather than memory) is necessary, and
the speed of acquisition is limited by the drive performance. High hard drive speeds of
5,400 RPM are standard, and for a small price premium 7,200 RPM drives are readily
available.

8.2.4. GRAPHICS CARD

Recent versions of the Windows and Macintosh operating systems have numerous extra-
neous “eye candy” features that can consume a large proportion of a system’s resources
without any tangible benefit. To free up the main system processor, dedicated high-pow-
ered graphics processors are now included in most modern computer systems.

Computations required to render high-speed or 3D graphics are performed on the graph-

ics card, freeing up the system processor for other tasks. The purchase of a high-end
graphics card can therefore increase overall system performance.

8.3. PERIPHERALS AND OPTIONS

One of the design principles of personal computers has been the modularity of design to
accommodate user-specific functionality. The descriptions below discuss the peripherals
most pertinent to laboratory setups.

8.3.1. DATA BACKUP

As common and everyday as personal computers have become, hard disks are complex
mechanical devices requiring precise tolerances in their operation, and are subject to
mechanical failure. The lifetime of hard disks is usually several years, but manufacturing
defects, physical shocks, or too many power-on cycles can make your hard disk ready to
fail at any moment. It is essential that a disk full of programs and data be backed up regu-
larly.

There are two common methods of backing up data: you can backup to another hard
drive, or back up to an optical medium such as CD or DVD. With the advent of rewrit-
able optical disk drives, it is possible to store 680 MB of data on a CD, and 4.4 GB on a
single-layer DVD.

182 The Axon Guide — 2500-0102 Rev. C

 8.4. Recommended Computer Configurations

Although outstripped by hard drives for storage capacity, optical storage has the advantage
of being compact for small amounts of data and electrically stable. Another clear advan-
tage of optical storage is that once a CD/DVD has been closed, there is a very low chance
of overwriting data. With a hard drive, there is a much higher chance of losing data
through accidental overwrite unless special precautions are taken.

On the other hand, hard drives allow data to be recovered and reorganized relatively easily.
Optical disks can be almost impossible to recover once they become scratched, and reor-
ganizing data requires data transfer and re-burning.

8.4. RECOMMENDED COMPUTER CONFIGURATIONS

Because computer state-of-the-art changes so quickly, please consult the MDS Analytical
Technologies web site for the latest recommended computer specifications.

8.5. GLOSSARY
A/D converter
An Analog-to-Digital converter (ADC) maps analog measurements to digital numbers
that the computer can store and use in calculations. See analog and digital.
Analog
Real signals in the physical world are described as analog, meaning that the values change
in a continuous manner. A mercury thermometer is an analog measurement device; the
mercury rises smoothly in response to changing temperature. However, the computer
requires discrete digital numbers to describe measurements of the data. See A/D converter
and digital.
ASCII
American Standard Code for Information Interchange. A standardized 7-bit code for
exchanging information between computer systems. The 128-character set consists of the
alpha-numeric and punctuation characters, as well as control codes such as Escape and
Carriage Return. Files on a computer are often stored in a simple ASCII format, so that
they can be easily transferred and read by different systems and programs.
Bit
Information is stored in computer memory as a series of bits, which are binary switches
(numbers) that can either be on (1) or off (0). A group of eight bits is called a byte, which
is a common unit for addressing and storing information.
Boot
To boot a computer refers to the sequence of events that occurs when a computer is
switched on. When the computer is turned on, its built-in software instructs the com-
puter to load the operating system from disk into memory, and turns over control of the
computer to the operating system.

The Axon Guide — 2500-0102 Rev. C 183

 8. Laboratory Computer Issues and Considerations

Byte
A byte is a series of eight bits of data. Since a bit can have two values (zero and one), a byte
can have 28 (256) values. A character on an IBM text screen is stored in a byte of memory,
so there are 256 distinct characters available. A byte is a typical unit for storing data and
computer instructions. See word.
Cache
A cache is a place to store often-accessed data or computer instructions, and is designed to
provide quicker access for increased performance. A memory cache is faster-than-usual
RAM for storing data that the microprocessor requires often from the normal RAM stor-
age area. A disk cache is a RAM storage area for data that is stored on disk but is being
read frequently.
CPU
Central Processing Unit. See Microprocessor.
D/A converter
A Digital-to-Analog converter (DAC) transforms digital information from the computer
into analog voltages to drive real world devices. See Analog and Digital.
Data acquisition
In the context of computers, data acquisition refers to the analog-to-digital conversion of
measurements for storage and analysis by the computer.
Device driver
Software that acts as an extension to the operating system, usually for supporting a hard-
ware device. Mouse pointing devices and network cards are common devices that require a
device driver to be loaded when booting the system up. On PC computers, device drivers
are files that traditionally have a .SYS filename extension.
Digital
Data represented in discrete, discontinuous form is digital, as opposed to the smooth rep-
resentation of data as measured by an analog device, such as a pen chart recorder. Com-
puters require discrete digital numbers for storage and processing. See A/D and analog.
DPI
Dots Per Inch is a measure of resolution on a printed page or computer screen. The more
dots per inch, the sharper that text or a graphic image will appear. A laser printer typically
has a resolution of 300 DPI.
EEPROM
Electrically Erasable Programmable Read-Only Memory. Allows ROM memory chips to
be reprogrammed with new information by users in the field. See ROM.

184 The Axon Guide — 2500-0102 Rev. C

 8.5. Glossary

I/O Interfaces
An I/O interface defines the physical interface and communication method for devices
attached to the computer, such as the display, printer, mouse and disk storage device. For
devices attached internally, such as the display adapter, the physical I/O interface is the
computer’s internal bus. For external devices, e.g., the mouse or printer, there are several
interfaces implemented on Windows and Macintosh computers.
PARALLEL PORT
The parallel port interface, which exists primarily in the Windows world, allows eight
bits of data to be transferred at a time (“in parallel”). It offers high speed, but its
design allows mostly one-way communication; the external device is very limited in
the signals it can send back to the computer. As a result the only devices designed to
use the parallel port are those that mostly receive data, the most common example
being the printer. While most consumer computers implement this port as a 25-pin
connector, commercial printers often use a 36-pin Centronics connector.

Parallel ports are becoming uncommon with the advent of USB printers.
SERIAL PORT
The RS-232 serial port is a general communication interface that allows full, two-
way communication, thereby making it much more versatile than the parallel port. It
exists on many types of computer systems. Common devices which attach to a serial
port include the mouse and the keyboard, but which used dedicated 6-pin PS/2 con-
nectors vs. the general-use 9-pin DB connector.

Serial ports have already been phased out for consumer devices, but because commu-
nication using the serial interface is relatively straightforward, serial ports remain use-
ful for low-level device communication. For device development, therefore, one or
two serial ports is essential.

When buying a new PC computer for use in a laboratory, you should ensure that it
has a serial port.
USB PORT
In just a few years, USB has lived up to its name and become the Universal Serial
Bus. Devices such as cameras, phones, printers, scanners, external hard drives, flash
drives, keyboards, mice and other consumer products now come with USB as the
standard interface. USB is much faster than older protocols such as serial, and so it is
suitable for applications that require large amounts of data to be transferred at high
speed. Modern scientific instruments such as the MultiClamp™ 700B and
Axoclamp™ 900A amplifiers are now being made with USB interfaces, and this
trend is likely to continue.

The original implentation of USB 1 has been supplanted by a high-speed version

known as USB 2. USB 2 allows high-speed devices, such as the Digidata 1440A
digitizer, to be easily connected to either desktop or laptop computers.

The Axon Guide — 2500-0102 Rev. C 185

 8. Laboratory Computer Issues and Considerations

KB
A kilobyte is approximately 1,000 bytes. As a computer’s natural system of numbers is in
base two, a kilobyte actually refers to 210 (1,024) bytes. It is abbreviated with a capital K
to distinguish it from the abbreviation for 1,000, as in kHz.
MB
A megabyte is 1,024 kilobytes, or 1,048,576 bytes. See KB.
Memory
See RAM and ROM.
Microprocessor
A microprocessor is a single integrated circuit chip that combines functions for doing log-
ical and mathematical operations, managing memory and communicating with external
devices. It is the main control center of a microcomputer.
Operating system
An operating system manages the hardware in a computer and provides a set of services
for an application program, such as writing data to a disk drive.
Parallel port
A type of communication interface that transfers data across several parallel lines. Com-
munication is mostly one-way, and the length of the connecting cable is limited. In the
PC, the parallel port is used mainly for connecting printers.
PC
Originally referred to the IBM PC, but now refers to any computer compatible with the
IBM PC designs based upon Intel CPUs.
RAM
Random Access Memory is composed of specialized integrated circuits designed for stor-
ing programs and data. RAM can be quickly read from and written to, so programs exe-
cute quickly.

However, when the computer power is off, RAM cannot hold any information.
Resolution
The “sharpness” of images and characters on the computer screen or on printer output.
Most accurately described in units of dots per inch.
ROM
Read Only Memory resides on integrated circuits that hold information which is not lost
when power is turned off. The contents of the ROM are usually loaded by the manufac-
turer and are not alterable by the user.

186 The Axon Guide — 2500-0102 Rev. C

 8.5. Glossary

RS-232
A common, standard protocol for communicating over serial ports. It is a general two-way
communication interface that transmits data on a single line and receives data on a single
line.
SCSI
Small Computer System Interface. This is a type of parallel interface that provides for
higher speeds than were possible in the older serial or parallel interfaces, while allowing
multiple devices to be connected to a single port.
Serial port
A legacy standard connection, which provides a general two-way communication interface
that transmits data on a single line and receives data on a single line. The serial interface
no longer exists in most types of computers. Also known as a
RS-232 serial port.
Throughput
A measure of performance describing how much data per unit of time can be transferred
from device to device in a computer system.
Video display card
A specialize circuit card to handle the complexity of graphics display. High-end graphics
cards can be added to computers to upgrade their standard display speeds, resolution and
number of colors.

The Axon Guide — 2500-0102 Rev. C 187

 8. Laboratory Computer Issues and Considerations

188 The Axon Guide — 2500-0102 Rev. C

 9. Acquisition Hardware
Scientists are always seeking better methods to record and store the growing amounts of
data generated in experiments. In electrophysiological experiments, the data are most
often in the form of voltage waveforms whose magnitudes vary with time. Data in this
form are appropriate for displaying on an oscilloscope or chart recorder, but entirely inap-
propriate for storing on a computer disk. Since computers can only store discrete num-
bers, a process of “analog-to-digital” conversion must be undertaken to convert the analog
data into a compatible format for the computer.

9.1. FUNDAMENTALS OF DATA CONVERSION

The analog-to-digital (A/D) conversion process can be illustrated by recording tempera-
ture during the course of a day, using pencil and paper. One way to do this is to look at
the mercury level in a glass thermometer every 30 minutes and write the temperature in a
column on a piece of paper. The result is a sequence of numbers, recorded at uniform
time intervals, which describes the variations in the temperature during the day. This pro-
cess illustrates the execution of an analog-to-digital conversion and the creation of an
array of data that could be typed into the computer and stored on disk.
Table 9.1: Example Temperature Data.

Time Digitized Temperature °C

6:00 AM 10
6:30 AM 11
7:00 AM 11
7:30 AM 12
8:00 AM 13
8:30 AM 13
9:00 AM 14
9:30 AM 14
10:00 AM 14
10:30 AM 15
11:00 AM 15
11:30 AM 15
12:00 PM 17
12:30 PM 18
1:00 PM 19
1:30 PM 19

The Axon Guide — 2500-0102 Rev. C 189

 9. Acquisition Hardware

Time Digitized Temperature °C

2:00 PM 19
2:30 PM 18
3:00 PM 16
3:30 PM 15

Analog Temperature Graph Digital Temperature Graph

20 20
18 18
16 16
14 14
12 12
10 10
8 8
6 6
4 4

6 7 8 9 10 11 12 1 2 3 6 7 8 9 10 11 12 1 2 3
Time Time

Figure 9.1: Analog-to-digital conversion.

This figure illustrates the analog and the digital representations of the temperature data
presented in Table 9.1.

A much easier way to accomplish the same task would be to use a data acquisition board
in the computer to directly record the temperature every 30 minutes. In this case the ther-
mometer must report the temperature in a form that the analog-to-digital converter
(ADC) can read. Nearly all ADCs require that the data be presented as a voltage. To do so,
one uses a different type of thermometer, such as a temperature-dependent resistor in an
appropriate circuit, that generates an analog voltage proportional to the analog tempera-
ture. The computer is instructed to initiate an analog-to-digital conversion every 30 min-
utes and store the results on disk.

Sometimes the computer is required to generate an analog waveform that will be propor-
tional to a list of numbers that are held in memory or on disk. In this case, a digital-to-
analog (D/A) converter (DAC) is used. The principles of operation of a DAC are similar
to those of an ADC, but the operation is the reverse.

9.2. QUANTIZATION ERROR

The ADC converter generates binary numbers that have finite resolution. That is, a small
range of analog values will all produce the same binary number after conversion. Return-
ing to the temperature-measurement example above, the hypothetical ADC rounded off
all of the temperature values to the nearest degree Celsius reading. That is, only numbers
between 0 and 99 °C in steps of 1 °C were allowed. Clearly, the temperature changes con-
tinuously between each 30-minutes reading, not in 1 °C jumps. The range of analog val-
ues that are recorded as the same digital number is the quantization error.

190 The Axon Guide — 2500-0102 Rev. C

 9.3. Choosing the Sampling Rate

In an ADC, the total measurement range (e.g., 0–100 °C) is divided into a fixed number
of possible values. The number of values is a power of two, often referred to as the number
of “bits.” Commonly, these values are:

8 bit = 28 = 256 values

12 bit = 212 = 4,096 values

16 bit = 216 = 65,536 values

To illustrate the impact on the resolution of using an 8-bit, 12-bit or 16-bit ADC, con-
sider the temperature-measurement example where the electronic thermometer circuit
generates an analog output from -10 V to +10 V for temperatures in the range -100 °C to
+100 °C. In this case, the resolutions are:

8 bit = 78.4 mV = 0.784 °C

12 bit = 4.88 mV = 0.0488 °C

16 bit = 0.305 mV = 0.00305 °C

If the ADC is preceded by a programmable-gain amplifier, signals that occupy only a

small range, e.g., ±1 V, can be amplified before conversion so that the full resolution of the
ADC can be utilized.

The resolution of a 12-bit ADC with an input range of ±10 V is 4.88 mV. Although this
resolution does not represent a difficult number for a computer-based system to use, some
researchers preferred a round number such as 5.00 mV. This could easily be achieved by
setting the span of the ADC to ±10.24 V instead of ±10 V. Now, many ADC systems are
now designed with the ±10.24 V span (e.g., the Axon Digidata® 1440A data acquisition
system).

9.3. CHOOSING THE SAMPLING RATE

There is a well-known theorem known as the sampling theorem or the Nyquist theorem
stating that data should be sampled at a frequency equal to twice the bandwidth of the sig-
nal or faster in order to prevent an artifactual increase in the noise, due to a phenomenon
known as aliasing (see Chapter 11), and guarantee that the analog signal can be recon-
structed unambiguously from the digital samples.

In practice, it is common to sample at a rate significantly faster than the minimum rate
specified by the sampling theorem. This is known as oversampling. Exactly how much
oversampling should be used depends upon the type of experiment.

For experiments where the data are analyzed in the frequency domain (e.g., noise analysis,
impedance analysis), it is common to oversample only modestly. The main concern is to
prevent aliasing. An anti-aliasing filter is introduced between the signal source and the
analog-to-digital converter to control the bandwidth of the data.

The Axon Guide — 2500-0102 Rev. C 191

 9. Acquisition Hardware

The factor of twice the analog bandwidth required by the sampling theorem is only appli-
cable if the anti-aliasing filter is ideal, i.e., the gain is unity in the pass band and it abruptly
changes to zero in the stop band. Ideal filters cannot be realized, although they can be
closely approximated. For frequency-domain analysis it is common to use sharp cutoff fil-
ters such as Butterworth or Chebyshev realizations. Sampling is typically performed at 2.5
times the filter bandwidth. For example, if the data are filtered at 10 kHz they should be
sampled at about 25 kHz. Slower sampling rates are unacceptable. Faster sampling rates
are acceptable, but offer little advantage and increase the storage and analysis require-
ments.

For experiments where the data are analyzed in the time domain (e.g., pulse analysis, I-V
curves), greater oversampling is required because reconstructing the analog signal requires
not only an ideal anti-aliasing filter but also an ideal reconstruction filter. The simplest
and most common reconstruction filter is to join each sample by a straight line. Other
techniques, such as cubic-spline interpolation, can be used, but due to their much more
demanding computational requirements they are used infrequently.

The reconstruction problem is discussed in the Sampling Rate section in Chapter 11.
There is no commonly accepted rule regarding the sampling rate of data for time-domain
analysis. In general, five times the data bandwidth is a common sampling rate, and ten
times is considered good. Sampling at 20 times is excessive and is rarely used.

9.4. CONVERTER BUZZWORDS

The terms in this section are defined for a D/A converter. With simple reorganization,
these definitions can be re-cast for an A/D converter. Only the more esoteric terms have
been defined in this section. It is assumed that the reader already understands conven-
tional terms such as “voltage offset.”

9.4.1. GAIN ACCURACY

Gain accuracy is the closeness with which the output of the D/A converter corresponds to
the specified full-scale value when the digital input to the D/A converter is the maximum
value.

9.4.2. LINEARITY ERROR

Linearity error is the maximum deviation of the D/A output from a straight line drawn
between the minimum and the maximum output.

9.4.3. DIFFERENTIAL NONLINEARITY

When the digital control word changes by a minimal step, i.e., one least-significant bit
(LSB), the output value should change by 2-n of the full-scale value, where n is the num-
ber of bits in the converter. Any deviation from this ideal change is called the differential
nonlinearity error. It is expressed in multiples of LSBs.

192 The Axon Guide — 2500-0102 Rev. C

 9.5. Deglitched DAC Outputs

9.4.4. LEAST SIGNIFICANT BIT (LSB)

LSB is the value of the smallest possible digital increment.

9.4.5. MONOTONICITY
Monotonic behavior means that for every increase in the digital input to the D/A con-
verter there will be an increase in the analog output. This requires that the differential
nonlinearity error be less than 1 LSB.

9.5. DEGLITCHED DAC OUTPUTS

When the input to a digital-to-analog converter changes value, the analog output ideally
moves rapidly and monotonically towards its new value. While usually this is the case, for
certain changes in value a transient is superimposed on the changing output. This tran-
sient is known as a “glitch.”

Glitches occur when several bits in the control word change. Each bit is connected to a set
of internal digital logic and analog switches. There is variability in the response time of
each part of the circuit, and therefore the response to the new control word does not hap-
pen at the same instant for each bit. In addition, charge is injected from the digital logic
circuitry into the analog switches. Glitches are worst when a large number of the bits of
the digital input to the D/A are changed. For example, when the analog output is required
to go from negative to positive, the digital input bits can change from 1111 1111 1111 to
0000 0000 0000. That is, every single bit changes. This change generates the worst glitch.

Glitches can be prevented by including a sample-and-hold circuit at the output of the

D/A converter. The sample-and-hold is normally in the “sample” mode so that the output
of the D/A converter feeds straight through. However, for one or two microseconds, per-
haps less, after the digital input to the D/A converter is updated, “hold” mode is selected
so that the glitch, if present, is blocked.

In modern integrated-circuit D/A converters, efforts are made in the design of the chip to
match the propagation and activation delays of each bit in the circuit and to minimize the
charge injection so that glitches are small in the first place. It is thus becoming increas-
ingly less common for a deglitching circuit to be included. On the other hand, the high-
level of integration in modern D/A converters sometimes introduces another problem,
called feedthrough noise. This problem occurs when the digital latches that contain the
word for the D/A converter are integrated into the same chip. It may happen that because
of the proximity of the digital circuits to the D/A converter, digital noise couples into the
D/A converter circuit. The coupled signal manifests itself as a pulse on the output of the
D/A converter circuit each time the digital word is updated. Strangely, this feedthrough
noise often appears even if the D/A value is being held constant. This is because, for sim-
plicity, most software that simultaneously performs A/D and D/A conversions updates the
D/A word once per A/D sample, even if the D/A value is not changing. The best solution
is to eliminate the feedthrough noise at its source by using separate integrated circuits for
the D/A converter latches. This is the approach used in the Digidata 1440A digitizer.

The Axon Guide — 2500-0102 Rev. C 193

 9. Acquisition Hardware

9.6. TIMERS
Most ADC systems have several timers available for a multitude of tasks. The most essen-
tial is the provision of a regular clock signal to initiate each A/D conversion. In most sys-
tems, one D/A conversion is performed for each A/D conversion, but this is not required
and some systems provide for running the D/A converter at a different clock rate to the
A/D converter.

Additional timers are used to implement “gear shifting.” This is a technique wherein the
acquisition rate is rapidly changed (gear shifted) during acquisition from a low to a high
rate or vice versa, without stopping the acquisition.

Uncommitted timers are generally provided for use in frequency measurement, event
counting or interval timing. For example, the frequency of nerve spikes can be counted if
they are first detected by an event detector that puts out a digital pulse for every spike. An
example of event counting is the measurement of the number of photons collected by a
photomultiplier tube in a fixed interval. An example of interval timing is measurement of
the period of a spinning optical filter wheel so that the optical data collected can be nor-
malized against fluctuations in the rotational speed of the wheel.

9.7. DIGITAL I/O

Digital outputs are used during experiments to control external equipment. For example,
an oscilloscope can be triggered before the acquisition commences, or a flash lamp or an
isolated stimulator could be activated during the acquisition. In other experiments, several
solenoids can be sequentially activated before, during and after the acquisition.

Digital inputs are used routinely in industrial control applications for monitoring the state
of solenoids and other apparatuses, but they are rarely used in electrophysiological experi-
ments. The main application for digital inputs in electrophysiology is for triggering acqui-
sitions and for indicating that a tag should be attached to the data.

9.8. OPTICAL ISOLATION

Computers provide quite hostile environments for plug-in data acquisition boards. The
rapid switching activity of the digital logic circuits causes electronic noise to radiate
directly into plug-in boards installed by the user and to be introduced into the ground ref-
erence used by the plug-in boards.

Nevertheless, it is usually possible to design data acquisition systems that do not suffer
from extraneous noise pickup. This is achieved by careful layout and good grounding.
Grounding and shielding, if applied with great skill and diligence, can sometimes be made
to work acceptably for the A/D converter, especially when combined with differential
recording. However, it is very difficult to achieve acceptable noise levels on the DAC out-
puts. The following approach is often invoked for 16-bit systems.

194 The Axon Guide — 2500-0102 Rev. C

 9.9. Operating Under Multi-Tasking Operating Systems

First, move the A/D converter off the plug-in board in the computer to a separate board
housed in an external box. This eliminates the problems due to direct pickup of noise gen-
erated inside the computer housing. However, it does not eliminate the noise introduced
through the system ground. The noise in the system ground can be eliminated by using
optical isolation. In this technique, the digital signals to and from the computer are not
connected directly. Instead, an optical coupler is interposed in each digital line. A separate
power supply is provided for the A/D and D/A side of the optical couplers. This tech-
nique allows a low-noise ground to be maintained for the A/D and D/A converters and
for the experimental setup.

9.9. OPERATING UNDER MULTI-TASKING OPERATING

SYSTEMS
Data acquisition systems are not naturally compatible with multi-tasking operating sys-
tems. As a general rule, data acquisition systems are designed to use all of the resources of
the computer (e.g., display and disk) and they use these resources on demand, rather than
when the computer makes them available.

Interrupt handling represents an acute problem for data acquisition in multi-tasking oper-
ating systems. Usually, when an interrupt occurs an immediate response is required. If the
multi-tasking operating system is busy updating a screen for a word processor, an immedi-
ate response will clearly not be forthcoming. Therefore, high performance clearly requires
that the data acquisition task runs as the foreground application. Another problem is that
modern computers generally have long interrupt latencies of several hundred microsec-
onds or more. One way to minimize the problem of excessive interrupt latency is to
ensure that the data acquisition routine has the highest priority and, in many cases, exclu-
sive control over the computer. Another way to minimize excessive interrupt latency is to
provide hardware support in the acquisition system. For example, a memory-buffered data
acquisition system can be configured by the acquisition software and left in a “primed”
state. When it “sees” an external trigger it can generate an interrupt, then commence the
acquisition before it even receives a response from the host.

9.10. SOFTWARE SUPPORT

Nowadays fewer researchers are writing their own data acquisition and analysis software
using low-level languages and device drivers. Most researchers are using one of two types
of commercially available software. The first type is a rich development environment such
as MATLAB or LabVIEW. This environment allows researchers to design their own soft-
ware without having to become experts in the low-level control of the data-acquisition
hardware. The second type of software is a turn-key package, such as pCLAMP® software.
These packages provide sophisticated acquisition and analysis, but only limited ability to
customize them. Nevertheless, they are extremely popular because they perform well and
are easy to learn.

The Axon Guide — 2500-0102 Rev. C 195

 9. Acquisition Hardware

When choosing hardware and software for a data acquisition system, we recommend that
you choose the software first and then purchase the hardware recommended by the soft-
ware applications that you have selected.

196 The Axon Guide — 2500-0102 Rev. C

 10. Data Analysis
Chapter 10 provides guidelines to selecting certain key data acquisition parameters, and
discusses current methods used in analysis of biological data, including fitting models to
data and analyzing single ion-channel recordings. The scope of this discussion is limited
to brief, practical introductions with references to more detailed discussion in this Guide
and in the literature. The manuals for the various software programs can be used for more
specific step-by-step instructions, and the text by Dempster (1993) is a useful collection of
standard techniques in this field.

10.1. CHOOSING APPROPRIATE ACQUISITION PARAMETERS

Successful analysis of a data set requires that the signals be recorded using appropriate data
acquisition parameters. Following are guidelines to selecting several critical recording
parameters.

10.1.1. GAIN AND OFFSET

Sufficient gain and offset should be maintained on all transducers and amplifiers during
data acquisition, because data review and analysis software can usually provide only lim-
ited amounts of additional (“software”) offset and gain. More than 10x of software gain
usually results in excessively jagged traces; this is because the digitized signal does not vary
smoothly but is quantized (see Chapter 9), and too much amplification allows the differ-
ence between one quantization level and the next to become visible. Software offset is lim-
ited to the full range of the analog-to-digital converter (ADC) input, which is usually
equivalent to about ±10 V referred to the input of the ADC. Signal resolution is best pre-
served when the signal fills this voltage range without exceeding it.

10.1.2. SAMPLING RATE

The sampling rate used should be selected considering the particular experiment. Use of
excessively high sampling rates wastes disk space and will increase the time required for
analysis. Furthermore, a higher sampling rate is usually associated with a higher filtering
frequency, which in turn allows a larger amount of noise to contaminate the signal. Subse-
quent analysis of the data may therefore require noise reduction using analysis-time filter-
ing, which can be time-consuming. Guidelines to choosing the correct sampling rate are
discussed in the following paragraphs (Colquhoun and Sigworth, 1983, and Ogden,
1987).
Biological signals are most commonly analyzed in the time domain. This means that the
time dependence of the signals is examined, e.g., to characterize the membrane response to

The Axon Guide — 2500-0102 Rev. C 197

 10. Data Analysis

a voltage-clamp pulse. The usual rule for time-domain signals is that each channel should
be sampled at a frequency between 5 and 10 times its data bandwidth. Knowing the value
of the data bandwidth is required in order to set the filter cut-off frequency during acqui-
sition and analysis.
For a sinusoidal waveform, the data bandwidth is the frequency of the sine itself. For most
biological signals, the data bandwidth is the highest frequency of biological information of
interest present in the recorded signal. This can be determined directly by examining a
power spectrum of rapidly sampled unfiltered data, though this is rarely done. Alterna-
tively, one can estimate the number of points per time interval required to give a data
record whose points can be easily “connected” by eye and calculate the sampling rate
directly. The data bandwidth and filter frequency can then be calculated from the sam-
pling rate. For example, if a fast action potential (1 ms to peak) is to be recorded, 25 sam-
ples on the rising phase would yield a reasonably good 40 μs resolution, requiring a
sampling rate of 25 kHz and an approximate data bandwidth of 5 kHz.
The rules are more straightforward in some special cases. Single-channel recording is dis-
cussed below. For signals with exponential relaxation phases, the sampling rate needed to
estimate a time constant depends on the amount of noise present; for moderately noise-
free data, at least 15 points should be taken per time constant over a period of 4 to 5 time
constants. Many fitting routines will fail if sampling is performed over only 3 time con-
stants, since the waveform does not relax sufficiently far towards the baseline. For a sum of
multiple exponentials, the sampling rate is determined in this way from the fastest phase;
sampling must extend to 4 time constants of the slowest phase. If this would result in too
many samples, a split clock (as in the Clampex program of the Axon pCLAMP® suite) or
other methods of slowing the acquisition rate during the acquisition, could be employed
as long as at least 15 points are taken over each time constant.
When a set of several channels is recorded (e.g., channels 0 through 3), most data acquisi-
tion systems sample the channels sequentially rather than simultaneously. This is because
the system usually has only one analog-to-digital converter circuit that must be shared
among the channels in the set. For example, if four channels are sampled at 10 kHz per
channel, one might expect that they would be sampled simultaneously at 0 μs, 100 μs,
200 μs, etc. Instead, channel 0 is sampled at 0 μs, channel 1 at 25 μs, channel 2 at 50 μs,
channel 3 at 75 μs, channel 0 again at 100 μs, channel 1 again at 125 μs, etc. There is
therefore a small time skew between the channels; if this causes difficulties in analysis or
interpretation, a higher sampling rate can be used to minimize the time skew (but this
may cause problems associated with high sampling rates, as mentioned above).
An additional consideration arises from the fact that on many data acquisition systems,
including the Digidata® 1440A digitizer from MDS Analytical Technologies, the digital-
to-analog converter (DAC) is updated whenever the ADC is read, even if there is no
change in the DAC output. This means that the DAC is updated only at the sample rate
over all channels. For example, if a stimulus is a 0 to 150 mV ramp and 50 samples are
acquired from one channel at a sampling interval of 25 μs, the DAC output will appear as
a series of steps each 25 μs long followed by an upward jump of 150 mV/50 = 3 mV,

198 The Axon Guide — 2500-0102 Rev. C

 10.2. Filtering at Analysis Time

which may be too large for some electrophysiological applications. Therefore, if a rapidly
changing continuous waveform is applied while acquiring slowly, the output waveform
should be checked with an oscilloscope and, if necessary, the sampling interval should be
increased. The computer preview of a waveform cannot be relied upon for this purpose
because it does not account for the effect of sampling. Note, however, that since most
users acquire significantly more samples per sweep than 50, this problem will not occur
except in very unusual situations.

10.1.3. FILTERING
The signal should be filtered using an analog filter device before it arrives at the ADC. As
discussed in Chapter 6 and in Colquhoun and Sigworth, 1983 and Ogden, 1987, this is
done to prevent aliasing (folding) of high-frequency signal and noise components to the
lower frequencies of biological relevance.
Acquisition-time filtering of time-domain signals is usually performed using a Bessel filter
with the cut-off frequency (-3 dB point; see Chapter 6) set to the desired value of the data
bandwidth. A 4-pole filter is usually sufficient unless excessive higher frequency noise
requires the 6- or 8-pole version. The Bessel filter minimizes both the overshoot (ringing)
and the dependence of response lag on frequency. The latter two effects are properties of
the Chebyshev and Butterworth filters (see Chapter 6 and Chapter 11, or Ogden, 1987),
which are less appropriate for time-domain analysis.

10.2. FILTERING AT ANALYSIS TIME

It is sometimes reasonable to sample data at higher rates than seems necessary, e.g., when a
greater bandwidth might be required during analysis. If excessive sample rates are used,
the filter frequency must be set to a higher value. Since there is more noise at higher fre-
quency, more noise is likely to contaminate the signal. Therefore, the data must be filtered
further during analysis in order to reduce the noise and avoid aliasing of high-frequency
signal content.
This analysis-time filtering is performed using filters implemented in software. The Gaus-
sian filter is most commonly used for this purpose because of its execution speed, though
the Bessel filter is employed as well. If one wants to write a computer program for a filter,
the Gaussian is easier to implement than the Bessel filter (see program example in Colqu-
houn and Sigworth, 1983). Note that all filters alter the waveform; for example, a Gauss-
ian-filtered step function deviates from the baseline before the time of transition of the
unfiltered signal. The user can examine data records filtered at different frequencies to
make sure that there is no significant distortion of the quantities of interest, such as time
of transition or time to peak.
Another common software filter is smoothing, the replacement of a data point by a simply
weighted average of neighboring points, used to improve the smoothness of the data in a
record or graph. In contrast to the smoothing filter, the Bessel and Gaussian types have
well-known filtering (transfer) functions, so that (i) it is easy to specify an effective cut-off

The Axon Guide — 2500-0102 Rev. C 199

 10. Data Analysis

frequency, and (ii) the effect of the filter may be compensated for by a mathematical pro-
cedure analogous to a high-frequency boost circuit in a voltage-clamp amplifier (Sachs,
1983). These advantages are important if the frequency properties must be known
throughout the analysis. If not, the smoothing filter is much faster to execute, especially
on long data records, and easier to implement if one writes one’s own software.

10.3. INTEGRALS AND DERIVATIVES

The integral function is used to calculate the area under a digitized data trace. Applica-
tions include measuring the total charge transfer from records of membrane current, and
measuring the area under a miniature endplate potential. The integral is generally calcu-
lated by direct summation of y(xi)Δxi between two cursors placed along the x axis. For a
data record, y(xi) is the amplitude at time point xi, and Δxi is the sampling interval at that
time. For a histogram, y(xi) is the amplitude of the bin at location xi, and Δxi is the bin
width of bin i (this allows for nonuniform bin width). The y values must be corrected for
any superfluous baseline before integration using a fixed or slanting baseline, as appropri-
ate. If the sample rate were low compared to the rate of change of the signal, so that y val-
ues show large changes from one xi to the next, the integral is considerably less accurate
than if significantly more samples were taken. Large inaccuracies can also occur if the sig-
nal does not return to baseline by the end of the data set or if part of the data is corrupted
by an extraneous signal. In some cases, errors in the integral can be reduced by first fitting
a smooth curve to the available data and then using the formula and parameters of the
best fit to calculate the integral.
The derivative function is used to determine the rates of change of a digitized signal. This
can be used to help in peak location (where the derivative will be near zero), transition
detection (large positive or negative derivative), sudden departure from baseline, etc. The
derivative will, however, amplify noise, making it difficult to determine trends. The signal
should therefore be filtered beforehand or fit to a function which can then be used to
obtain a noise-free form of the derivative. However, the worse the fit, the greater the error
in the derivative.

10.4. SINGLE-CHANNEL ANALYSIS

10.4.1. GOALS AND METHODS
The goal of single-channel current analysis is to reconstruct the idealized current wave-
forms from which information about the mechanisms of channel function is derived. Spe-
cific information can be deduced with respect to channel state diagrams, kinetics of
channel opening, closing and gating, channel barrier models, and the effects of channel
blocking agents and membrane constituent on channel function.
Articles that present approaches and methods used in single-channel current analysis
include Colquhoun and Hawkes, 1983; Colquhoun and Sigworth, 1983; McManus,
Blatz and Magleby, 1987; Sigworth and Sine, 1987; and French et al., 1990.

200 The Axon Guide — 2500-0102 Rev. C

 10.4. Single-Channel Analysis

The current from a single ion channel is idealized as a rectangular waveform, with a base-
line current of zero (closed-channel), an open-channel current dependent on the conduc-
tance and membrane potential, and rapid transitions between these current levels. In
practice, this idealized current is distorted by the limited bandwidth of the apparatus and
contaminated by noise. Shifts in the baseline may occur during the recordings, and capac-
itative current spikes are present in sweeps in which voltage changes are applied across the
membrane. The current signal is further altered by filtering and by periodic sampling.
These effects can impede making confident inferences from data regarding channel behav-
ior.

10.4.2. SAMPLING AT ACQUISITION TIME

In order to adequately reconstruct the transitions, the sampling rate should be no less than
5 times higher than the cut-off frequency of the filter (Colquhoun and Sigworth, 1983;
French et al., 1990).

10.4.3. FILTERING AT ACQUISITION TIME

As discussed above, the input signal should be filtered with a Bessel filter during acquisi-
tion. The cut-off frequency must usually be set sufficiently low to prevent unacceptably
frequent occurrences of noise, which cause false closings and false brief open-close events
during the construction of idealized channel currents. French et al., (1990) discuss how to
decide on the cut-off frequency for a particular situation: open events of longer duration
are more likely to be falsely closed by noise, so it is convenient to specify the longest dura-
tion dmax that can be recorded with less than 1% occurrence of these false closings. Once
dmax has been specified, the required cut-off frequency fc can be calculated using the fol-
lowing combination of equations (18) and (19) in French et al., (1990):

100 f c* (1)
fc = *
d max FTC

Here FTC* is the observed rate of false threshold crossings measured using recordings
made with an arbitrary cut-off frequency fc*. FTC* can be measured from idealized single-
channel records generated using a threshold set on the side of the baseline opposite to
where transitions are observed. This analysis can be achieved using pCLAMP software.

10.4.4. ANALYSIS-TIME FILTERING

The digital Gaussian filter can be used for additional analysis-time filtering of single-chan-
nel records. This introduces symmetrical time delays at both opening and closing and can
therefore be used for the unbiased estimation of latencies using the 50% criterion (Section
10.4.6., “Setting the Threshold for a Transition”).

The Axon Guide — 2500-0102 Rev. C 201

 10. Data Analysis

10.4.5. GENERATING THE EVENTS LIST

The first step during analysis of single-channel current records is to idealize the current
records to a series of noise-free open and closed states having infinitely short (or at least
shorter than the sample interval) transition times. This analysis, which can be performed
by pCLAMP software, results in a list of durations and amplitudes, called the events list.
Several effects that tend to complicate this reconstruction are briefly discussed below;
more thorough discussions are presented in the cited literature.

10.4.6. SETTING THE THRESHOLD FOR A TRANSITION

A transition between states of a channel occurs when the current amplitude passes
through a threshold between two levels. The most commonly used threshold is 50% of
the difference between the levels. The advantage of this threshold setting is that the event
durations are not biased because the values of the threshold are the same for both opening
and closing transitions.

10.4.7. BASELINE DEFINITION

The accuracy of the threshold method for transition detection depends on the stability of
the baseline (i.e., closed-channel) current or, if the baseline is unstable, on the ability of an
automated procedure to correctly identify the baseline as it changes. A number of ways
have been devised to track a moving baseline, including 1) averaging the baseline current
level to get the new baseline; 2) defining the new baseline as that level which maximizes
the number of times that the current signal crosses it during the closed state (“zero cross-
ings” method; Sachs, 1983); and 3) defining the new baseline at the peak of the histogram
of the baseline current amplitude (e.g., G. Yellen, quoted in Sachs, 1983). pCLAMP soft-
ware uses a hybrid approach in which the average of the most recent closed channel cur-
rent level is averaged with the old baseline level, weighted by a selectable factor. Regardless
of the method used, the user must carefully monitor the baseline to ensure that any auto-
matic procedure does not lose track of the baseline value.

10.4.8. MISSED EVENTS

Events will be missed if their true durations are shorter than the dead time of the system,
which is determined by the filter cut-off frequencies used during acquisition and analysis.
Events will also be missed if their superthreshold portions happen to miss the times when
the signal is sampled even if their durations are longer than this dead time. The resulting
error is minimal if the fastest time constant in the system is much longer than the sam-
pling interval because few events will be shorter than the sampling interval. If this is not
the case, the sampling rate must be increased with respect to the filter cut-off frequency
(see Section 10.4.2., “Sampling at Acquisition Time”).

10.4.9. FALSE EVENTS

The probability of detecting false events depends on the amount and spectrum of noise in
the system, the filter characteristics and the rate of sampling (French et al., 1990).

202 The Axon Guide — 2500-0102 Rev. C

 10.4. Single-Channel Analysis

10.4.10. MULTIPLE CHANNELS

The presence of multiple channels complicates the determination of the kinetic behavior
of a channel. If a record shows a transition from two open channels to one open, it cannot
be determined if the transition was due to the closing of the first or the second channel. A
number of methods have been proposed to deal with this ambiguity (French et al., 1990).
As a precaution, the amplitude histogram of the raw data can be inspected to determine if
multiple channels are present.

10.4.11. ANALYZING THE EVENTS LIST

The second step in the analysis procedure is the extraction of model-specific information
from the events list. Because the number of events is usually large, it is often convenient to
sort the event data into bins. The binned data may then be fit to a mathematical function
whose parameters are related to a state diagram of the channel.
The most common histograms include 1) the dwell time histogram, in which the dura-
tion of events is binned and which can be related to state models; 2) the first latency histo-
gram, in which the period of time from stimulus onset to first opening is binned and
which is used to extract kinetic information; and 3) the amplitude histogram, in which
the amplitudes of the levels or of all points in the records are binned and yield informa-
tion about the conductance, the system noise and the incidence of multiple channels.

10.4.12. HISTOGRAMS
The simplest histogram is one in which a series of bins are defined, each of which has an
associated upper and lower limit for the quantity of interest, e.g., dwell time. Each dwell
time will then fall into one of the bins, and the count in the appropriate bin is incre-
mented by one. In the cumulative histogram, a bin contains the number of observations
whose values are less than or equal to the upper limit for the bin.

10.4.13. HISTOGRAM ABSCISSA SCALING

Histograms may be scaled and displayed in a number of ways. The most common histo-
gram has constant bin width, a linearly scaled abscissa (x axis), and counts displayed on
the ordinate (y axis). A number of alternatives have been developed to improve the treat-
ment of exponential distributions, in which low-time bins have many counts and high-
time bins have few, yielding histograms that are difficult to visualize (e.g., McManus et al.,
1987; Sigworth and Sine, 1987). One solution is to increase the bin width logarithmi-
cally, so that for a single exponential, a constant number of events is expected per bin.
This has the disadvantage that the presence of multiple time constants is not evident from
the plot if the ordinate is not rescaled. These issues are discussed in the next section.

10.4.14. HISTOGRAM ORDINATE SCALING

If a constant bin width histogram is used, linear scaling of the ordinate is the most com-
mon (Figure 10.1 A). If a logarithmic bin width is used, several methods are common.
In one of the methods, the number of counts in each bin is divided by the bin width and
plotted on a logarithmic axis, in which case multiple time constants are evident

The Axon Guide — 2500-0102 Rev. C 203

 10. Data Analysis

(Figure 10.1 B). The main difficulty with this scaling is that the bins do not have the same
weight during the fit process (Section 10.4.17., “Fitting to Histograms”), because the vari-
ance of the bin depends on the number of counts it contains. This can be compensated
for using the transformation of Sigworth and Sine (1987). In this transformation the bin
width is constant and the square root of the number of counts per bin is plotted on the
ordinate, with time plotted logarithmically on the abscissa (Figure 10.1 C).

Figure 10.1: Histogram scaling.

Three representations of a dwell-time distribution with two exponential components.

5,120 random numbers were generated according to a distribution with time con-
stants of 10 ms (70% of the events) and 100 ms (30%) and binned for display as his-
tograms in the lower panel of each part of the figure. Superimposed are the
theoretical probability density functions for each component (dashed curves) and
their sum (continuous curve). In each part of the figure the upper panel plots the
absolute value of the deviation of the height of each bin from the theoretical curve,
with dashed curves showing the expectation value of the standard deviation for each
bin. The upper panels were plotted with vertical expansion factors of 2.1, 5.4, and 3.1
respectively.
A. Linear histogram. Events are collected into bins of 1 ms width and plotted on a lin-
ear scale. The 100-ms component has a very small amplitude in this plot.

204 The Axon Guide — 2500-0102 Rev. C

 10.4. Single-Channel Analysis

B. Log-log display with variable-width (logarithmic) binning. The number of entries in

each bin is divided by the bin width to obtain a probability density in events/s
which is plotted on the ordinate.
C. Square-root ordinate display of a logarithmic histogram. Note that the scatter
about the theoretical curve is constant throughout the display (reproduced with
permission from Sigworth and Sine, 1987).

10.4.15. ERRORS RESULTING FROM HISTOGRAMMING EVENTS DATA

Problems may appear as a result of the histogram process. The first problem may occur in
either amplitude or time histograms if the bin size is not an integral multiple of the resolu-
tion of the signal. For a dwell time, the resolution is the interval between successive sam-
plings of the ADC channel; for an amplitude, the resolution is determined by the gain
and the ADC. The symptom is that occasional or periodic peaks or valleys appear artifac-
tually in the histogram. This problem is less severe if the bin width corresponds to many
resolution intervals (e.g., 10). If bin widths are variable, the smallest bin width must like-
wise include many resolution intervals.
A second problem is called sampling promotion error. Sampling promotion error (Sine
and Steinbach, 1986) occurs because data are sampled periodically. Suppose data were
acquired with 1 ms sampling rate and dwell times binned into a histogram with bin width
equal to the sampling rate (1 ms) and centered around 3 ms, 4 ms, 5 ms, etc. The 4 ms bin
would therefore contain events whose true dwell times lie between 3 and 5 ms
(Figure 10.2). If the dwell times fall exponentially with increasing times, the 4 ms bin
would contain more events from 3 to 4 ms than from 4 to 5. The subsequent fit would
treat all these events as if they had occurred at 4 ms (the midpoint), thereby resulting in an
error. In a single exponential fit, this affects only the intercept and not the time constant;
but there may be more subtle effects in a multi-exponential fit. The error is small when
the bin width is much smaller (perhaps by a factor of 5) than the fastest time constant
present.

1 ms

sample
times

3 ms
event

5 ms
event

Figure 10.2: Sampling error.

The Axon Guide — 2500-0102 Rev. C 205

 10. Data Analysis

Figure 10.2 illustrates how a 1 ms sampling rate can result in an apparent 4 ms open time
for events of between 3 and 5 ms. The dots in the top part of the figure represent times
when the signal is sampled. The lower part of the figure shows two events whose wave-
forms are high at 4 sample times.
A third problem is termed binning promotion error. In an exponentially falling dwell time
distribution, a bin is likely to contain more events whose true dwell times are at the left
side of the bin than are at the right side. The average dwell time of the events in that bin is
therefore less than the midpoint time t. The error occurs when a fit procedure assumes
that all the bin events are concentrated at a single point located at the midpoint t, instead
of at the average, which is less than t. Binning promotion error can occur in addition to
sampling promotion error because binning takes place in a different step. Both of these
errors are due to the asymmetric distribution of true dwell times about the bin midpoint.
A correction procedure has been proposed by McManus et al., (1987).
The degree of these bin-width-related errors may be reduced independently of the correc-
tive procedures mentioned above, if the fit procedure explicitly uses the fact that a bin
actually represents an area instead of an amplitude. Some errors may be eliminated if the
individual data points are used without ever using a histogram, as in the maximum likeli-
hood fitting method.
Lastly, as discussed in the section on missed events, short events may not be detected by
the 50% threshold criterion. This can give rise to a number of errors in the extracted fit
parameters, which relate specifically to state models. For further details, consult the refer-
ences cited in French et al., 1990.

10.4.16. AMPLITUDE HISTOGRAM

Amplitude histograms can be used to define the conductances of states of single channels.
Two kinds of amplitude histograms are common: point histograms and level histograms.
The former use all the acquired data points; they are useful mainly for examining how
“well-behaved” is a data set. Abnormally wide distributions may result from high noise if
the signal were over-filtered, or if baseline drift had been significant. Similarly, the number
of peaks will indicate the presence of multiple channels or subconductance states. The all-
points histogram will probably not be useful for determining the conductances unless
baseline drift is small. Level histograms use only the mean baseline-corrected amplitudes
associated with each event in the events list. Such histograms can be fitted to the sum of
one or more Gaussian functions in order to estimate conductances.

10.4.17. FITTING TO HISTOGRAMS

Amplitude and dwell time histograms can be fitted by appropriate functions, usually sums
of several Gaussian functions for the former and sums of exponentials for the latter (see
“The Chi-Square Function (Least-Squares Method)” on page 209). The time constants and
amplitudes can be related to the parameters of several single-channel models, but this will
not be described here.

206 The Axon Guide — 2500-0102 Rev. C

 10.5. Fitting

Histogram bins containing zero counts should usually be excluded from a fit because the
chi-square function (see below) is not defined when a bin i contains Ni = 0 counts and
therefore has σi = 0. Alternatively, adjacent bins can be combined to yield a nonzero con-
tent.

10.5. FITTING
10.5.1. REASONS FOR FITTING
Fitting a function to a set of data points, such as a histogram or a time series, may be done
for any of the following reasons:
1 A function could be fitted to a data set in order to describe its shape or behavior,
without ascribing any “biophysical” meaning to the function or its parameters. This is
done when a smooth curve is useful to guide the eye through the data or if a function
is required to find the behavior of some data in the presence of noise.
2 A theoretical function may be known to describe the data, such as a probability
density function consisting of an exponential, and the fit is made only to extract the
parameters, (e.g., a time constant). Estimates of the confidence limits on the derived
time constant may be needed in order to compare data sets.
3 One or more hypothetical functions might be tested with respect to the data, e.g., to
decide how well the data were followed by the best fit function.
The fitting procedure begins by choosing a suitable function to describe the data. This
function has a number of free parameters whose values are chosen in order to optimize the
fit between the function and the data points. The set of parameters that gives the best fit is
said to describe the data, as long as the final fit function adequately describes the behavior
of the data. Fitting is best performed by software programs; the software follows an itera-
tive procedure to successively refine the parameter estimates until no further improvement
is found and the procedure is terminated. Feedback about the quality of the fit allows the
model or initial parameter estimates to be adjusted manually before restarting the iterative
procedure. Fitting by pure manual adjustment of the parameters (the so-called “chi by
eye”) may be effective in simple cases but is usually difficult and untrustworthy in more
complex situations.
The two following topics will be briefly discussed below: statistics, i.e., how good is the fit
and how confident is the knowledge of the parameters, and optimization, i.e., how to find
the best fit parameters. The statistical aspects are well discussed in Eadie et al., (1971);
Colquhoun and Sigworth (1983) provide examples relevant to the electrophysiologist. A
number of aspects of optimization are presented in Press et al., (1988).

The Axon Guide — 2500-0102 Rev. C 207

 10. Data Analysis

10.5.2. STATISTICAL ASPECTS OF FITTING

Statistics deals with the probability of occurrence of events. Probability is difficult to
define; there are two ways in which the word is used:
1 Direct probability: If we observe that N1 of N single-channel events have open
channel durations between 10 and 20 ms, we say that the probability p1 of this
occurring is N1/N, as long as N is very large. The probability density function (pdf ) is
an algebraic expression that when summed or integrated between 10 and 20 ms gives
the value of p1.
2 Inverse probability: If you are told by your physician that you have one of three
possible diseases D1, D2 or D3, and you ask what the probability is that you have D2,
the physician might say, “It’s either 0 or 1,” meaning that either you already have D2
or you do not have it, but the physician cannot yet determine which of these two
situations exists. To be helpful, your physician might give an inverse probability of
0.6, meaning that if you had D2, the probability that your particular set of symptoms
would have been observed is 0.6.
The Likelihood Function
Inverse probability is more appropriate for the scientist who may have N measurements of
open channel durations and wants to know which time constant best describes their expo-
nential distribution. For a particular time constant, τ1, one can calculate the direct proba-
bility of getting those N observed durations by first calculating the probability of
observing each duration and then multiplying these individual probabilities together. The
resulting number is called the likelihood of the time constant τ1. One could calculate the
likelihoods for many such values of τ and plot the logarithm of these likelihoods versus
their respective τ values (Figure 10.3). If N is sufficiently large, this curve will usually be
Gaussian. The value τ* at the peak of the function is called the maximum likelihood value
of τ. The root-mean-square spread of the function about τ is known as the standard devi-
ation of τ*, though this is not the same as a standard deviation of a set of numbers in the
direct probability case.

ln L
τ∗
τ

(2)

Figure 10.3: The likelihood function.

208 The Axon Guide — 2500-0102 Rev. C

 10.5. Fitting

The logarithm of the likelihood presented as a function of variable τ; the maximum likeli-
hood of the function occurs at τ*. For large numbers of samples, the shape of this curve
approaches a Gaussian.
It turns out that τ* reliably converges to the true time constant if the number of events N
is sufficiently large. For this reason it is important to either collect a large enough number
of data points or repeat the experiment several times so as to reduce the variation of the
parameters obtained over the data sets to an acceptable level. If the noise is large or there
are sources of significant variability in the signal, the data may be useless except in a qual-
itative way because of large variations of the best fit parameters between the runs. If one
long run is taken instead of several smaller ones, the run can be broken up into segments
and the analysis results of the segments compared with each other to assure that conver-
gence is near.
Although the maximum likelihood method is the most reliable, the time requested for the
calculations may be prohibitively long. The chi-square method, described below, is an
alternative that requires less time.
The Chi-Square Function (Least-Squares Method)
Suppose that a set of p measurements are made at the times x1, x2, ..., xp, and that the val-
ues measured are y1, ..., yp. If each yi is measured with a measurement error distributed as
a Gaussian with standard deviation σ1, ..., σp, the maximum likelihood method is equiv-
alent to minimizing the chi-square function:

p
( y i − y *i ) 2
χ2 = ∑
i =1 σ i2

where y*i is the fit value corresponding to yi. Minimizing chi-square is also called the
least-squares method. If the fit is made to a data sweep, each yi is the value measured at the
time xi, and each σi is the standard deviation or uncertainty of that value. In the typical
case when all the σi’s are equal (i.e., the uncertainty in the data does not depend on the
time), the σi’s can be ignored while performing the search for the best fit parameters, but
must be specified if the goodness of fit is to be calculated. If the fit is made to a histogram,
each yi is the numbers of events Ni in bin i, and each σi is √Ni.
It is much easier to maximize the chi-square function than to minimize the likelihood
function, whether for data or for many mathematical functions used as models.
Since the use of the chi-square function is equivalent to the use of the likelihood function
only if the uncertainties in the data (e.g., noise) are distributed as Gaussians, the correct-
ness of a least-squares fit can depend on the characteristics of this uncertainty.

The Axon Guide — 2500-0102 Rev. C 209

 10. Data Analysis

The Goodness of Fit

After a best fit has been obtained, the user might wish to know if the fit was good. If the
fit to p data points employed M free parameters, the probability of obtaining a chi-square
value greater than that of the best fit, given p–M degrees of freedom, can be read from a
table of probabilities of chi-square and compared to a chosen significance level.
Often several models are fitted to a single set of data so as to define the best-fit model.
Horn (1987) discussed choosing between certain types of models using their respective
chi-square values (with the F test) or the logarithm of the ratio of the likelihoods, which
follows the chi-square distribution. These tests can help decide whether the model con-
tains irrelevant parameters, e.g., if a three-exponential function was fitted to a set of data
containing only two exponentials.
Confidence Limits for the Parameters
An estimate of the confidence limits for each of the parameters is often useful. For a
model with one parameter a, some fit programs will derive the standard deviation σ from
the dependence of the likelihood or chi-square on the parameter. One then says that the
68.3% confidence limits are within a distance σa from a*, the chosen value for a. This
does not mean that 68.3% of the time the true value of a will fall between a* + σa and
a* – σa. It does mean that we will be right 68.3% of the time if we assert that these limits
include the true a. In the limit of a large number of data sets, a* will tend to converge to
the true value of a.
Suppose there are two parameters a and b to be fit. One can make a two-dimensional con-
tour plot of the likelihood or chi-square function, with each axis corresponding to one of
the parameters, showing the probability of the limits including the true parameter values
(Figure 10.4). The resultant ellipse is usually inclined at an angle to the parameter axes
due to the correlation of the two parameters. These correlations tend to increase the con-
fidence limits, which are indicated by the dotted lines which project to the respective axes.
The probability that the confidence limits for the best fit a includes the true a is 0.683,
but this does not specify anything about b, as indicated by the shaded region. If, in addi-
tion, limits ±σb are specified for the best fit b, the joint probability that both sets of confi-
dence limits include the true a and b has an upper limit of (0.683)2 or 0.393, i.e., the
probability content of the area inside the ellipse. If one has a five exponential fit with an
offset, the analogous cumulative joint probability will be (0.683)11, or 0.015, which is
quite small.

210 The Axon Guide — 2500-0102 Rev. C

 10.5. Fitting

b*+σb

b*-σb

a*- σa a*+σa

Figure 10.4: Confidence limits.

Two-dimensional contour plot of a likelihood or chi-square function vs. two parame-

ters a and b around the function minimum, with the two sets of dashed lines indicat-
ing the respective confidence limits.

Statistical Tests of Significance for Parameters

If one wishes to compare, for example, two time constants from two data sets obtained
under different conditions, one must first obtain the standard deviation of the time con-
stant, usually derived from the fitting procedure. Conventional statistical tests, such as the
chi-square table, the F test or Student’s t test, can then be applied to determine signifi-
cance.

10.5.3. METHODS OF OPTIMIZATION

Optimization methods are concerned with finding the minimum of a function (e.g., the
chi-square) by adjusting the parameters. A global minimum, i.e., the absolute minimum,
is clearly preferred. Since it is difficult to know whether one has the absolute minimum,
most methods settle for a local minimum, i.e., the minimum within a neighborhood of
parameter values. A number of algorithms have been developed to find such a minimum.
For example, to find time constants and coefficients in an exponential fit, pCLAMP soft-
ware allows the user to choose between the following:
> Minimizing the chi-square using the Levenberg-Marquardt method.
> Minimizing the chi-square using the Simplex method.
> Maximizing the likelihood using the variable metric or Simplex method.
Of the three methods, the Simplex method is relatively insensitive to shallow local min-
ima. Though it will reliably find the region of the global minimum or maximum, it may
not find the precise location of the minimum or maximum if the function is rather flat in

The Axon Guide — 2500-0102 Rev. C 211

 10. Data Analysis

that vicinity. The Levenberg-Marquardt method is more easily trapped in local minima of
the function, but it can provide better fits than the Simplex because it uses the mathemat-
ical characteristics of the function being minimized to find the precise location of the
minimum or maximum, within the numerical resolution of the computer. This method
also provides statistical information sufficient to find the confidence limits.
These methods are iterative, i.e., they continue refining parameter values until the func-
tion stops changing within a certain convergence criterion. They also require reasonable
starting estimates for the parameters, so that the function to be minimized or maximized
is not too far away from its optimum value; a poor starting set can lead some fit programs
to a dead end in a shallow local minimum.
For Boltzmann and exponential fits, the Axon Cellular Neuroscience analysis program
(Clampfit in the pCLAMP software suite) provides a non-iterative method, in which the
data points and function to be fit are transformed using a set of orthogonal Chebyshev
polynomials, and the fit function coefficients are quickly calculated using these trans-
formed numbers in a linear regression. This method is very fast and requires no initial
guesses, though the parameters may be slightly different than those found by the methods
listed above because the underlying algorithm minimizes a quantity other than the sum of
squared differences between fit and data.

10.6. REFERENCES
Colquhoun, D. and Sigworth, F.J. Fitting and Statistical Analysis of Single-Channel
Records. in Single-Channel Recording. Sakmann, B. and Neher, E., Eds. Plenum Press,
New York, 1983.
Colquhoun, D. and Hawkes, A.G. The Principles of the Stochastic Interpretation of Ion-
Channel Mechanisms. in Single-Channel Recording. Sakmann, B. and Neher, E., Eds.
Plenum Press, New York, 1983.
Dempster, J. Computer Analysis of Electrophysiological Signals. Academic Press, London,
1993.
Eadie, W.T., Drijard, D., James, F.E., Roos, M, Sadoulet, B. Statistical Methods in Exper-
imental Physics. North-Holland Publishing Co., Amsterdam, 1971.
Horn, R. Statistical methods for model discrimination. Biophysical Journal. 51:255–263,
1987.
McManus, O.B., Blatz, A.L. and Magleby, K.L. Sampling, log binning, fitting, and plot-
ting distributions of open and shut intervals from single channels and the effects of noise.
Pflügers Arch, 410:530–553, 1987.
Ogden, D.C. Microelectrode electronics. in Microelectrode Techniques. The Plymouth
Workshop Handbook. Standen, N.B., Gray, P.T.A. and Whitaker, M.J., eds.
The Company of Biologists Ltd., Cambridge, 1987.

212 The Axon Guide — 2500-0102 Rev. C

 10.6. References

Press, W.H., Flannery, B.P., Teukolsky, S.A. and Vetterling, W.T. Numerical Recipes in C.
Cambridge University Press, Cambridge, 1988.
Sachs, F. Automated Analysis of Single-Channel Records. in Single-Channel Recording.
Sakmann, B. and Neher, E., Eds. Plenum Press, New York, 1983.
Sigworth, F.J. and Sine, S.M. Data transformations for improved display and fitting of
single-channel dwell time histograms. Biophysical Journal, 52:1047–1054, 1987.
Sine, S.M. and Steinbach, J.H. Activation of acetylcholine receptors on clonal mamma-
lian BC3H-1 cells by low concentrations of agonist. Journal of Physiology (London),
373:129–162, 1986.
Wonderlin, W.F., French, R.J. and Arispe, N.J. Recording and analysis of currents from
single ion channels. in Neurophysiological Techniques. Basic Methods and Concepts.
Boulton, A.A., Baker, G.B. and Vanderwolf, C.H., Eds. Humana Press, Clifton, N.J.,
1990.

The Axon Guide — 2500-0102 Rev. C 213

 10. Data Analysis

214 The Axon Guide — 2500-0102 Rev. C

 11. Noise in
Electrophysiological
Measurements
In the most general sense, noise can be defined as any disturbance that interferes with the
measurement of the desired signal. In electrophysiological measurements, such distur-
bances can arise from the preparation itself, the electrodes that couple the preparation to
the measurement instrument, the electronic instrumentation (e.g., voltage-clamp ampli-
fier, patch-clamp amplifier), interference from external sources (e.g., electrostatic and elec-
tromagnetic coupling between the circuitry and 50 or 60 Hz power lines, fluorescent
lights, video monitors, acoustic noise, and noise associated with mechanical vibrations),
and, if the data is digitized (as is usually the case), from the digitization process itself (e.g.,
quantizing noise, aliasing).

We will begin by discussing the basic noise mechanisms that arise from the physics of the
materials and devices that comprise the electrical system. Interference from external
sources and noise associated with digitization will be considered later.

The main fundamental types of noise are: thermal noise, shot noise, dielectric noise, and
“excess” noise (e.g., 1/f noise). These types of noise form the basis for a discussion of
amplifier noise and electrode noise.

All the fundamental types of noise are completely random in nature. Their average prop-
erties can be measured, but their actual values at any particular instant in time cannot be
predicted. The most convenient measure of the amplitude of noise is its root-mean-square
(rms) value. Many noise processes have a Gaussian distribution of instantaneous ampli-
tudes versus time. The area under the Gaussian distribution represents the probability that
a noise event of a particular amplitude will occur (the total area is unity). The probability
that a noise peak will exceed plus or minus one times its rms value is 0.32; the probability
of a particular noise peak exceeding plus or minus three times its rms value is 0.003. It is
common engineering practice to define peak-to-peak noise as 6 times the rms value; if the
noise process has Gaussian distribution, the noise will remain within this bound 99.7% of
the time. However, peak-to-peak noise is not as clearly defined as is rms noise, and it is
not uncommon to find peak-to-peak noise operationally defined as anything from 5 times
to 8 times rms.

When considering noise in the time domain, it is important to know the bandwidth over
which the noise process is observed. Noise is made up of many frequency components,

The Axon Guide — 2500-0102 Rev. C 215

 11. Noise in Electrophysiological Measurements

frequently extending from DC to many megahertz or gigahertz. Some noise processes are
naturally restricted in bandwidth, but most require the appropriate use of filtering to
restrict the bandwidth of the noise while allowing adequate resolution of the signal. When
a noise amplitude (rms or peak-to-peak) is discussed, it is appropriate to also note the
bandwidth over which the noise is observed and the type of filter that has been used to
restrict the bandwidth.

Due to the random nature of noise, uncorrelated noise sources will add in an rms fashion.
Thus if E1, E2, and E3 are the rms values of three uncorrelated noise sources, the total rms
noise, ET, of these sources together is given by:

(1)
ET = E12 + E 22 + E32

Because of this relationship, the largest individual source of noise will tend to dominate
the total noise. For example, if two noise sources with rms values of 100 μV and 20 μV are
added together, the resulting noise will have an amplitude of {(100 μV)2 + (20 μV)2}1/2 =
102 μV rms.

11.1. THERMAL NOISE

Thermal noise results from random motion of thermally excited charge carriers in a con-
ductor. It is often referred to as Johnson noise or Nyquist noise. For a resistor, thermal
noise can be represented as a series voltage noise source or a parallel current noise source as
illustrated in Figure 11.1. These two representations are equivalent.

R (noiseless)
1/2
R (noiseless) Ith= (4kTB/R)

Eth

1/2
Eth = (4kTRB)

Figure 11.1: Noise equivalent circuits of a resistor.

The power spectral density (PSD) of thermal noise is white, i.e., it does not vary with fre-
quency. Its value, S2thV, is given as a voltage noise by:

2
S thV = 4 kTR ( units: Volt 2/Hz ) (2)

216 The Axon Guide — 2500-0102 Rev. C

 11.1. Thermal Noise

or, equivalently, as a current noise PSD, S2thI, by:

2 4kT
SthI = (units: Amp2/Hz ) (3)
R

where k is Boltzmann’s Constant (1.38 x 10-23 J/ °K), T is the absolute temperature in

degrees Kelvin (°K = °C + 273), and R is the resistance in ohms. It should be noted that
spectral densities are also often expressed in units of Volt/√Hz and Amp/√Hz.

The variance (also called the noise power) over a bandwidth B (units: Hz) is then:

E th2 = 4 kTRB (units : Volt 2 ) (4)

or

4kTB
I th2 = (units: Amp2) (5)
R

The rms noise over a bandwidth B is given by:

Eth = 4kTRB (units:Volt rms) (6)

or:

4kTB
Ith = (units: Amp rms) (7)
R

Spectral densities, rms current and voltage noise for bandwidths of 1 and 10 kHz are listed
in Table 11.1 for resistances of 100 Ω to 100 GΩ. Note that while thermal voltage noise
increases with increasing resistance, thermal current noise decreases as the resistor’s value
increases.

The Axon Guide — 2500-0102 Rev. C 217

 11. Noise in Electrophysiological Measurements

Table 11.1: Thermal Noise of Resistors.

1 kHz 10 kHz 1 kHz 10 kHz Voltage Current

Resistance
V noise V noise I noise I noise Density Density
Value
(μV rms) (μV rms) (pA rms) (pA rms) (nV/√Hz) (pA/√Hz)
100 Ω 0.040 0.126 400 1260 1.26 12.6
1 kΩ 0.126 0.40 126 400 4.0 4.0
10 kΩ 0.4 1.26 40 126 12.6 1.26
100 kΩ 1.26 4.0 12.6 40 40 0.4
1 MΩ 4.0 12.6 4.0 12.6 126 0.126
10 MΩ 12.6 40 1.26 4.0 400 0.040
100 MΩ 40 126 0.40 1.26 1260 0.0126
1 GΩ 126 400 0.126 0.40 4000 0.004
10 GΩ 400 1260 0.040 0.126 12600 0.0013
100 GΩ 1260 4000 0.013 0.040 40000 0.0004

11.2. SHOT NOISE

Shot noise arises when current flows across a potential barrier, e.g., a p-n junction in a
semiconductor device. Since potential barriers are not present in simple resistive elements,
resistors do not display shot noise. Over a bandwidth B, the rms value of the shot noise
current is given by:

I sh = 2qIB (units: Amp rms) (8)

where q is the charge of the elementary charge carrier (for electrons q = 1.6 x 10-19 cou-
lomb) and I is the DC current in Amps.

An important example of shot noise is the equivalent input current noise of an operational
amplifier that arises from its input bias current Ib (also referred to as gate current for FET1
input devices). For bipolar transistor input devices, Ib is typically in the range of 1 nA to
1 μA; for FET input operational amplifiers, Ib (at room temperature) is typically in the
range of 1 to 10 pA, and can be less than 0.5 pA for special devices. The rms value of shot
noise for operational amplifiers with Ib ranging from 0.1 pA to 1 μA in a 1 and 10 kHz
bandwidth are listed in Table 11.2. In capacitive-feedback patch voltage-clamp amplifiers
(e.g., the Axopatch™ 200 series amplifiers), shot noise current from the headstage ampli-
fier sets the noise floor at low-to-moderate frequencies. By design these devices display
very low levels of shot noise; selected units can have gate currents as low as 0.2 pA, result-
ing in low-frequency current noise less than that of a 250 GΩ resistor.

1. FET is Field Effect Transistor.

218 The Axon Guide — 2500-0102 Rev. C

 11.3. Dielectric Noise

Table 11.2: Shot Noise.

1 kHz Shot Noise 10 kHz Shot Noise

Op Amp Bias Current
(pA rms) (pA rms)
1 μA 18 57
100 nA 5.7 18
10 nA 1.8 5.7
1 nA 0.57 1.8
100 pA 0.18 0.57
10 pA 0.057 0.18
1 pA 0.018 0.057
0.1 pA 0.0057 0.018

11.3. DIELECTRIC NOISE

An ideal lossless capacitor does not produce thermal noise. However, all real dielectric
materials display some loss that results in the generation of thermal noise. For dielectrics
with relatively low losses, the spectral density of this noise SD2 can be described in terms
of the dissipation factor D and the capacitance CD of the dielectric:

SD 2 = 4kTDCD (2π f ) (units: Amp2/Hz ) (9)

Where f is the frequency in Hz. The PSD of dielectric noise characteristically rises linearly
with increasing frequency. The dissipation factor is often listed as tan δe where δe is the
loss angle (note that for small δe, tan δe ≈ δe); the “quality factor” Q is the inverse of
D (Q = 1/D). For a bandwidth extending from DC to an uppermost cut-off frequency B,
the rms noise current ID, resulting from a lossy dielectric, is given by:

I D = 4kTDC Dπ B 2 (units: Amp rms) (10)

For the best solid dielectrics (e.g., quartz, sapphire and some ceramics), D is on the order
of 10-5 to 10-4. For poorer (lossier) dielectrics D can be much higher; e.g., some thermo-
setting plastics have D in the range of 0.01 to 0.1. For some glasses used to fabricate patch
pipettes, D is at least 0.01. The dissipation factor has some frequency dependence,
although in the important range from about 1 kHz to 100 kHz it is usually reasonable to
approximate it as a constant. D also shows some temperature dependence (typically
decreasing with lower temperatures), although again in the usual range of temperatures it
may be considered constant.

Dielectric noise is one consideration in the selection of feedback and compensation capac-
itors in capacitive-feedback headstages for patch clamping. It must also be considered in
the packaging materials, electrode holders, etc. for any high-sensitivity current or charge

The Axon Guide — 2500-0102 Rev. C 219

 11. Noise in Electrophysiological Measurements

amplification device. By using high-quality capacitors (D ≤ 0.0001) the dielectric noise of

the feedback and compensation capacitor can be made a relatively insignificant compo-
nent of the total noise of present day capacitive-feedback headstages. For example, using
equation (10) it can be estimated that for D = 0.0001 and CD = 2 pF (1 pF feedback
capacitor and 1 pF compensation capacitor), the dielectric noise of these components in a
10 kHz bandwidth is only about 0.032 pA rms. Dielectric noise associated with packaging
(i.e., with the input gate lead of the patch-clamp headstage) can be somewhat higher. The
most important source of dielectric noise in common electrophysiological measurements
is the dielectric noise of the glass used to fabricate pipettes for patch voltage clamping.
This subject will be considered in greater detail below (see Section 11.6., “Electrode
Noise”). It is worth noting that a dielectric does not actually have to be in contact with the
input of a sensitive current amplifier, such as a patch-clamp headstage, in order to produce
noise. For example, a 1 pF capacitor formed by a high-loss dielectric with D = 0.1 that is
coupled through air to the input by a capacitance of 1 pF can result in approximately
0.35 pA rms in a 10 kHz bandwidth. Thus it is a good idea both in the design of head-
stage electronics and of an experimental patch clamp set-up not to have exposed, high-loss
dielectrics in the immediate vicinity of the input.

11.4. EXCESS NOISE

Excess noise can broadly be defined as any noise that is present in a circuit element in
addition to the fundamental noise mechanisms already described. For example, in the
presence of direct current all resistors exhibit low-frequency noise whose power spectral
density varies inversely with frequency. This noise is present in addition to the thermal
noise of the resistor and is usually referred to as “1/f noise.” Different resistor types display
different amounts of 1/f noise, with metal film and wirewound types showing the least
excess low-frequency noise and carbon composition resistors showing the most. The PSD
(in V2/Hz) of this excess noise rises with decreasing frequency essentially as 1/f. Its magni-
tude is directly proportional to the DC current flowing through the resistor. Semiconduc-
tor devices also exhibit 1/f noise.

High-value (gigohm range) resistors, such as those used as the feedback element in resis-
tive-feedback patch-clamp headstages, also display a form of excess noise that rises with
increasing frequency. Generally, such resistors only achieve their thermal noise level
(which is the minimum possible) in the frequency range from about 10 Hz to 1 kHz. At
low frequencies, 1/f noise is observed, and at high frequencies the noise PSD typically
rises as f α, where α is usually in the range of 1 to 2. This noise can result in rms noise
from a resistor having several times the predicted thermal noise value in a 10 kHz band-
width. The elimination of this noise source is one of the motivations behind using capaci-
tive-feedback in patch clamps such as the Axopatch 200 series amplifiers.

220 The Axon Guide — 2500-0102 Rev. C

 11.5. Amplifier Noise

11.5. AMPLIFIER NOISE

The intrinsic noise of an operational amplifier can be described by an equivalent input
voltage noise En in series with the negative (-) input and an equivalent input current noise
In between the positive (+) and negative (-) inputs (see Figure 11.2). These noise sources
have been referred to the input for convenience of analysis. It should be noted, however,
that they are measured at the output. For example, in an open-loop situation (such as
shown in (Figure 11.2), if the inputs are grounded, the voltage at the negative (-) input
will be zero (ground). En must be inferred by measuring the output noise voltage and
dividing it by the open-loop gain (both are frequency dependent) of the amplifier. On the
other hand, in a closed-loop situation (which is always the case when using operational
amplifiers, as illustrated in Figure 11.3), En will actually appear at the negative (-) input
due to the action of the feedback loop.

En

In

Figure 11.2: Operational amplifier noise model.

ef
Rf

CURRENT en
INPUT
Cg in

Figure 11.3: Noise model of a simplifier current-to-voltage converter.

The Axon Guide — 2500-0102 Rev. C 221

 11. Noise in Electrophysiological Measurements

When analyzing the noise of an operational circuit it is convenient to consider the noise
sources in terms of their power spectral densities. In this case the noise sources will be
denoted by lower case symbols, e.g., en(f ) (units: V/√Hz) or en2(f ) (units: V2/Hz). As a
useful example of noise analysis in an operational circuit, consider the simplified current-
to-voltage converter illustrated in Figure 11.3. In this circuit en and in are the input volt-
age and current noise PSDs, respectively, and ef is the PSD of the noise (thermal and
excess) of the feedback resistor Rf. The current to be measured is introduced at the termi-
nal labeled “input.” For simplicity, the positive (+) input is shown to be grounded. Cg is
any capacitance between the negative (-) input and ground; this includes the input capac-
itance of the amplifier plus strays and is usually about 15 pF. The noise PSD at the output
of the IV converter, Sout2(f ), can be shown to be:

2
S out ( f ) = in2 R f2 + e 2f + e n2 ( 1 + 4π 2 f 2 R f2 C g2 ) (units : V 2/Hz) (11)

It is useful to present this result in terms of current so it can be compared directly with the
current signal being measured. This is accomplished by dividing the above expression by
Rf2, thus referring the noise to the input. The input referred PSD, Sin2(f ), is then given
by:

e 2f 1
Sin2 ( f ) = in2 + + en2 ( + 4π 2 f 2C g2 ) (units : Amp2/Hz) (12)
R f2 R f2

In general, in, en and ef are all functions of frequency: in is the shot noise of the input gate
current of the amplifier but usually displays some 1/f behavior at low frequencies; ef is the
thermal and excess noise of Rf; and en also displays 1/f behavior at low-to-moderate fre-
quencies. Nevertheless, it is convenient to assume that each of these terms is independent
of frequency so that equation (12) can be easily integrated over a frequency band (for
instance, DC to some uppermost bandwidth denoted by B). The square root of this result
is then the rms noise for the bandwidth B:

e 2f B 4
I in = in2 B + B + en2 + π 2C g2 B 3 (13)
R 2f R 2f 3

It should be noted that if it is assumed that ef is only the thermal noise of the feedback
resistor, then the term ef2/Rf2 is simply 4kT/Rf (see equation (3) above).

It is instructive to consider the relative magnitudes of the various terms in equation (12)
or (13). In the case of a patch voltage clamp, the value of Rf is selected to be very high,

222 The Axon Guide — 2500-0102 Rev. C

 11.6. Electrode Noise

both to provide adequate gain to resolve tiny (pA) currents and to minimize the noise
contribution of this resistor; 50 GΩ is common for single-channel recording. The PSD of
the thermal noise of an ideal 50 GΩ resistor is 3.2 x 10-31 Amp2/Hz and the rms noise
associated with such a component over a bandwidth of 10 kHz is 0.057 pA rms (although
as already noted, excess high-frequency noise can increase this value several fold). The shot
noise, which accounts for in, associated with an input gate current of 0.2 pA would, by
itself, produce noise of about 0.025 pA rms in a 10 kHz bandwidth. For a typical high-
quality patch-clamp headstage, en is about 2–3 nV/√Hz at frequencies above about
1 kHz; Cg is usually about 15–20 pF. With these values, the term ( 43 π 2en2C g2 B3 )1 / 2
amounts to roughly 0.15 pA rms in a 10 kHz bandwidth. These three terms should be
uncorrelated; therefore, they sum in an rms fashion so that the total predicted noise is
(0.0572 + 0.0252 + 0.152)1/2 = 0.162 pA rms. For a resistive-feedback headstage, actual
noise in a 10 kHz bandwidth is normally substantially higher (about 0.25–0.30 pA rms)
both because of excess noise from the feedback resistor and because of the characteristics
of the low-pass filters that do not abruptly cut off noise at the -3dB bandwidth (Section
11.12., “Filtering”). Nevertheless, it is important to note that for bandwidths above a few
kilohertz, the term arising from en and Cg dominates total noise. To intuitively under-
stand the origin of this term, it is only required to remember that the current through a
capacitor is directly proportional to the rate of change of voltage across the capacitor
(I = C(dV/dt) ). Therefore, the voltage noise at the negative (-) input of the amplifier pro-
duces current noise through the capacitor Cg. This noise increases with increasing fre-
quency simply because higher frequency components of the voltage noise involve more
rapidly changing voltages. The noise current through the capacitor is supplied by the feed-
back resistor and therefore appears as noise at the amplifier output.

11.6. ELECTRODE NOISE

Noise associated with recording electrodes is significant and often dominant in most elec-
trophysiological measurements. In microelectrode voltage clamps and the whole-cell vari-
ant of the patch voltage-clamp technique, the thermal voltage noise of the electrode is of
greatest importance since this noise will exist across the cell membrane and produce large
levels of current noise in conjunction with the membrane capacitance. In patch single-
channel measurements using the patch clamp, the voltage noise of the electrode is also
important; but here the situation is further complicated by such factors as the dielectric
noise of the pipette that produces a major source to overall noise. We will consider noise
associated with the electrode in single-channel recording and whole-cell patch recording
separately, beginning with the single-channel recording.

11.6.1. ELECTRODE NOISE IN SINGLE-CHANNEL PATCH VOLTAGE CLAMPING

There are a variety of mechanisms by which the patch pipette contributes noise to the
measured current in the patch voltage-clamp technique. The first mechanism we will con-
sider here is the fact that the holder plus pipette add a significant amount of capacitance
to the headstage input. This capacitance reacts with the input voltage noise, en, of the

The Axon Guide — 2500-0102 Rev. C 223

 11. Noise in Electrophysiological Measurements

headstage amplifier to produce current noise with a PSD that rises with frequency in
essentially the same fashion as the PSD of the open-circuit headstage. In fact, this noise is
perfectly correlated with the noise resulting from the intrinsic capacitance, Cin, associated
with the headstage input. Note that Cin consists of the input capacitance of the
JFET (Ciss = Cgs + Cgd, where Cgs is the gate-source capacitance and Cgd is the gate-drain
capacitance), plus any compensation capacitors connected to the input, plus, for a capaci-
tive-feedback headstage, the feedback capacitor (1 pF for the Axopatch 200 amplifier) and
capacitance associated with the reset switch, plus about 1–2 pF of stray capacitance: in
total, Cin is usually about 15 pF. Denoting the holder plus electrode capacitance as Che,
the PSD of the current noise arising from en is given by 4π2en2(Cin+Che)f2. The capaci-
tance of the holder can vary from about 1–5 pF; larger capacitances are associated with
holders with metallic shielding. The pipette capacitance depends on the depth of immer-
sion of the pipette, the type of glass used, the thickness of the pipette wall, and the use of
Sylgard coating. The total capacitance of the pipette will generally be in the range of
1–5 pF. Thus, Che can range from 2–10 pF. With Che = 2 pF, the increment in wideband
noise from this mechanism above that of the open-circuited headstage should not be
much more than 10%. However, with Che = 10 pF the noise increment could be more
than 50% from this mechanism alone. Obviously, from the point of view of noise, it is
important to minimize the capacitance of the holder and pipette.
Dielectric Noise
The basic characteristics of dielectric noise have already been described above (see equa-
tions (9) and (10)). Dielectric noise of the pipette can be a major contributor to total
noise in patch voltage clamping; in some situations it can be the dominant noise source.
The dielectric noise arising from the pipette depends on the dissipation factor D of the
glass used to fabricate the pipette, on the pipette capacitance (CD in equations (9) and
(10)), and on the presence of Sylgard coating. The dissipation factor D of glasses other
than quartz that have been successfully used for patch pipettes generally falls in the ranges
of 0.001–0.01. The dissipation factor for quartz is variously reported to be 10-5 to as
much as 4 x 10-4. For Amersil T-08 quartz, which has been used in all of the preliminary
tests of quartz pipettes described below, the reported dissipation factor is 10-4. For
uncoated pipettes the value of CD is determined by the dielectric constant ε of the glass,
the ratio of the outer diameter (OD) to the inner diameter (ID) of the pipette, the depth
of immersion of the pipette into the bath, and to some extent by the pulling characteris-
tics of the glass and the geometry near its tip. The dielectric constant ranges from as little
as 3.8 for quartz to more than 9 for some high-lead glasses. Typical glass capillaries used in
the fabrication of patch pipettes have an OD of 1.65 mm and an ID of 1.15 mm; thus
OD/ID ≈ 1.43. If it is assumed that these proportions (OD/ID = 1.43) are maintained as
the glass is drawn out in pulling, then for uncoated pipettes the value of CD will be
approximately 0.15 ε pF per mm of immersion into the bath. For thick-walled glass with
OD/ID = 2, this value would fall to 0.08 pF per mm of immersion; while for thin-walled
glass with OD/ID = 1.2 the capacitance would increase to about 0.30 ε pF per mm.

Actual measurements with uncoated or lightly Sylgard-coated pipettes fabricated from

glasses with known ε and immersed to a depth of 2–3 mm indicate that these values often

224 The Axon Guide — 2500-0102 Rev. C

 11.6. Electrode Noise

underestimate pipette capacitance; therefore, the dielectric noise is also underestimated.

This is probably due to non-uniform thinning near the tip and to some uncertainty as to
the true depth of immersion (due to a meniscus of solution around the pipette). For
example, using the popular Corning 7052 glass, which has ε = 4.9 and with
OD/ID ≈ 1.43, it is not uncommon to measure a pipette capacitance as high as ≈ 3 pF,
i.e., about twice the predicted value, when an uncoated pipette or a pipette very lightly
coated with Sylgard is immersed at a depth of 2 mm.

Despite this precautionary note, it is clear that, all else being equal, the value of CD varies
linearly with the dielectric constant ε of the glass. Equation (10) indicates that if the depth
of immersion and the OD/ID ratio are constant, the rms noise for a given bandwidth aris-
ing from the lossy dielectric characteristics of pipettes fabricated from various glasses will
be proportional to (Dε)1/2. Thus the lowest noise glasses are expected to be those that
minimize the product of the dissipation factor and the dielectric constant. This relation-
ship has been born out experimentally in tests of some 20 different types of glass used to
fabricate patch pipettes (Rae and Levis, Personal Communication).

To illustrate the range of results expected for different glass types, we will consider three
pipettes with identical geometry fabricated from three different glasses. It will be assumed
that OD/ID ≈ 1.4, CD = 0.2 ε pF per mm of immersion, the immersion depth is 2 mm,
the pipettes are uncoated or are coated very lightly with Sylgard (so that the coating does
not reduce the dielectric noise significantly), and the rms noise is measured at a 10 kHz
bandwidth (-3 dB, 8-pole Bessel filter). Corning 7052 glass (D ≈ 0.003 and ε ≈ 5) repre-
sents reasonably low-loss glasses used to fabricate patch pipettes. Under the above stated
conditions, the CD value of this pipette is 2 pF and the dielectric noise contribution pre-
dicted by equation 10 is about 0.17 pA rms, or about 0.2 pA rms if the characteristics of
an 8-pole Bessel filter are taken into account. On the other hand, 0080 soda lime glass
(D ≈ 0.01, ε ≈ 7) represents high-loss glasses, which were commonly used in the early
days of patch clamping. Its CD = 2.8 pF and dielectric noise of about 0.44 pA rms is pre-
dicted. Finally, Corning 7760 is a very low-loss glass with D ≈ 0.0017 and ε ≈ 4.5. With
CD = 1.8 pF, a dielectric noise of slightly less than 0.15 pA rms is predicted. These figures
are in reasonable agreement with experimental findings that have attempted to separate
the components of total noise arising from the dielectric noise of pipettes fabricated from
various glasses.

Since the capacitance CD increases approximately linearly with increasing depth of

immersion, the dielectric noise for any particular type of glass and OD/ID ratio should
vary approximately as the square root of the depth immersion. From the point of view of
noise reduction, it is clearly useful with excised patches to withdraw the pipette tip as close
to the surface of the bath as possible. If the patch cannot be excised, then the bath should
be as shallow as possible. For example, for a pipette fabricated from Corning 7760, as in
the previous example, dielectric noise is expected to decrease from about 0.15 pA rms in a
10 kHz bandwidth for a 2 mm immersion depth to roughly 0.05 pA rms for a 0.2 mm
immersion depth. The effect of immersion depth on pipette noise has, at least qualita-
tively, been verified experimentally. A method employing Silicone fluid to minimize the

The Axon Guide — 2500-0102 Rev. C 225

 11. Noise in Electrophysiological Measurements

effective pipette immersion depth was introduced by Rae and Levis (1992) and is
described below.

The above discussion dealt with expected behavior for uncoated pipettes. However, it is
common practice (and highly recommended for low-noise measurements) to apply a coat-
ing of Sylgard #184 covering the entire tapered region of the pipettes (i.e., approx.
5–10 mm) and extending to within roughly 100 μm of the tip (see Chapter 4). Coating
with a hydrophobic substance such as Sylgard is necessary to prevent the formation of a
thin film of solution that will creep up the outer wall of an uncoated pipette. Such a film
can produce very large amounts of noise in uncoated pipettes. Sylgard coating not only
virtually eliminates this noise source but also thickens the wall of the pipette thereby
reducing its capacitance. The dielectric constant of Sylgard #184 is 2.9 and its dissipation
factor is ≈ 0.002, which is lower than that of most glasses that have been used for patch
pipette fabrication. Thus, coating with Sylgard will reduce dielectric noise of patch
pipettes. It is expected that the improvement in noise associated with Sylgard coating will
be greatest for glasses with a high D product (e.g., soda lime glass); this has been con-
firmed experimentally. Improvement of noise should be less for glasses with very low val-
ues of D, but coating with Sylgard will reduce the dielectric noise of all glasses. The effects
of Sylgard coating on noise are, however, difficult to quantify theoretically primarily
because the thickness of the coating is usually not uniform. In particular, it is difficult to
achieve a very thick coating very near the tip.

Experimental tests of the noise properties of patch pipettes fabricated from 19 different
kinds of glass (see Chapter 4) have confirmed the general conclusions described above.
With few exceptions, the noise attributable to the pipette is inversely correlated with the
D product of the glass. In addition, thicker-walled pipettes and shallow depths of immer-
sion reduce noise for any particular glass type. Sylgard coating has its greatest effect on the
glasses with the poorest inherent electrical properties, but it is important to remember
that such coating is necessary for all types of glass.

It has been obvious for some time that pipettes fabricated from quartz should produce
only very small amounts of dielectric noise due to the low dielectric constant of quartz
(ε = 3.8) and, more importantly, its extremely low dissipation factor (D ≈ 10-5 – 10-4).
However, due to the high melting point of quartz (≈ 1600 °C), only with the advent of
automated laser pullers has it become practical to pull patch pipettes from quartz tubing.
A quartz pipette with D = 10-4 that is immersed to a depth of 2 mm (again assuming
0.2 ε pF per mm of immersion) would be predicted to produce only about 0.03 pA rms of
dielectric noise in a bandwidth of 10 kHz (-3 dB, 8-pole Bessel filter); for D = 10-5 this
value would fall to 0.01 pA rms. Preliminary measurements using Amersil T-08 quartz
suggest that the amount of dielectric noise in this situation is closer to 0.04–0.05 pA rms.
A more detailed discussion of preliminary estimates of the noise properties of quartz
pipettes is provided below.

Dielectric noise is probably the largest source of noise for pipettes fabricated from all
glasses other than quartz. For pipettes fabricated from quartz, due to its very low dissipa-

226 The Axon Guide — 2500-0102 Rev. C

 11.6. Electrode Noise

tion factor, sources of noise other than dielectric noise are expected to dominate total
pipette noise (“Noise Properties of Quartz Patch Pipettes” on page 230).

To summarize, dielectric noise can be minimized by using thick-walled glasses with low
values of D and coating the pipette with Sylgard #184. The effects of Sylgard coating are
greatest for glasses with relatively poor electrical properties. For excised patches, dielectric
noise can be minimized by withdrawing the tip of the pipette as close as possible to the
surface of the bath.

It should be noted that dielectric noise will also contribute to the noise associated with the
holder. For an Axopatch 200A amplifier with an open circuit noise of 0.06 pA rms in a
5 kHz bandwidth, total noise should not increase to more than about 0.07 pA rms in this
bandwidth when the Axon-supplied polycarbonate holder is attached. Part of this noise
increment is due to the fact that the holder adds about 1–1.5 pF of capacitance to the
headstage input. The rest of the increment in noise associated with the holder is presum-
ably dielectric noise, which can be estimated to account for roughly 0.03 pA rms in a
5 kHz bandwidth.
Noise Arising From Distributed Pipette Resistance and Capacitance
Most of the resistance of a patch pipette resides at or very near its tip. On the other hand,
the capacitance of an uncoated pipette can be expected to vary linearly with its depth of
immersion into the bath. Therefore, it has sometimes been assumed that current noise
arising from the thermal voltage noise of the pipette resistance in conjunction with the
pipette capacitance can be assumed to be negligible in comparison with other noise
sources. However, significant resistance resides in regions of the pipette that are further
removed from the tip. The thermal voltage noise of this resistance will greatly exceed the
input voltage noise of the headstage itself, with which it is in series. The actual situation is
complicated because both the pipette resistance and capacitance are distributed. In this
section we will consider the pipette capacitance to be lossless, and initially the effects of
Sylgard coating will not be considered. In order to estimate the noise arising from the dis-
tributed pipette resistance and capacitance, we will consider rather idealized pipette geom-
etry.

The pipette has been modeled as a shank and a conical region approaching the tip; the
angle of the cone is 11.4° and the tip opening is 1 μm in diameter. With this cone angle
the ID of the pipette increases to 100 μm at a distance of 0.5 mm from the tip, 200 μm at
a distance of 1 mm from the tip, 300 μm at a distance 1.5 mm from the tip, etc. When
filled with a solution with a resistivity of 50 Ω cm the pipette will have a total resistance of
about 3.2 MΩ. About 2.1 MΩ of this total resistance resides in the first 10 μm from the
tip; slightly more than 3 MΩ occurs in the first 100 μm from the tip. However, about
80 kΩ resides in the region from 100–200 μm from the tip, an additional 27 kΩ resides
in the region from 200–300 μm; and about 37 kΩ occurs in the region from 300 μm to
1 mm from the tip. The region from 1–4 mm from the tip adds another 12 kΩ. It has
been assumed that the capacitance is uniformly distributed along the pipette with a value
of 1 pF/mm of immersion. An Ag/AgCl wire extends into the pipette coming to within

The Axon Guide — 2500-0102 Rev. C 227

 11. Noise in Electrophysiological Measurements

4 mm of the tip and it is assumed that the resistance of the pipette is negligible beyond the
tip of the wire (due to the shunting effect of the wire). A noise equivalent circuit of the
pipette can be created by lumping the resistance and capacitance of each small segment
(e.g., 20–50 μm) of pipette. The approximate circuit is then a ladder network formed of a
series of resistors and their equivalent thermal voltage noise sources with a capacitor to
ground at each node representing a portion of the pipette immersed in the bath. Rough
calculations with such an equivalent circuit indicate that for a depth of immersion of
2 mm the noise arising from this mechanism will be about 0.13 pA rms in a 10 kHz band-
width. For a 1 mm depth of immersion this value would fall to about 0.10 pA rms. For an
idealized pipette identical to the one described here but with a cone angle of 22.6° (total
resistance is about 1.6 MΩ) these values fall to about 0.07 pA rms for a 2 mm depth of
immersion and about 0.05 pA rms for a 1 mm depth of immersion, both for a 10 kHz
bandwidth. Over the frequency range of interest to patch voltage clamping, the PSD
(Amp2/Hz) of this noise should rise with increasing frequency as f 2.

These calculated values are highly approximate and the assumed geometry is obviously an
over-simplification. A variety of factors could increase the noise arising from this mecha-
nism. For example, the noise should increase if there is an extended region behind the tip
where the angle of taper is quite shallow, resulting in increased resistance in this region.
Noise will also increase if the wire does not protrude as close to the tip as assumed above.
In addition, some glasses tend to thin near the pipette tip, resulting in increased capaci-
tance in this region. It should also be noted that shallow depths of immersion would not
decrease this noise quite as much as might be expected since this would decrease the
pipette capacitance, but not its resistance.

The noise from this mechanism can be reduced in several ways. From the above estimates
it seems that noise arising from the distributed resistance and capacitance of the pipette
should be smaller than the dielectric noise of pipettes fabricated from even the best known
glasses. Nevertheless, with better (lower loss) glasses, particularly quartz, this mechanism
could be the dominant source of noise from the pipette itself, and its reduction may
become important. First, it is obvious that the geometry of the first few millimeters of the
pipette will be an important determinant of the magnitude of this noise; therefore, when
possible, pipettes should be pulled to minimize the resistance distal to the tip. Anything
that reduces the pipette capacitance per unit length will also reduce this noise. Thus,
thick-walled pipettes and glasses with low dielectric constants should provide the best
results in terms of noise. Perhaps of more practical importance, coating the pipette with
Sylgard #184 can significantly reduce the pipette capacitance, and consequently reduce
noise of the type considered here. However, as already noted, it is more difficult to build
up a thick coat of Sylgard in the region within the first few hundred micrometers from the
tip than in more distant regions. As was the case with dielectric noise, shallow depths of
immersion will also reduce the noise arising from pipette resistance and capacitance.
Finally, it should also be possible to reduce this noise regardless of the immersion depth by
using a fine wire (Ag/AgCl or platinized Ag/AgCl possibly sharpened at the tip) that pro-
trudes as far as possible toward the tip of the pipette. Such a wire will, in effect, short out
(in a frequency-dependent fashion) the resistance of the pipette-filling solution in the

228 The Axon Guide — 2500-0102 Rev. C

 11.6. Electrode Noise

region into which it extends; thus, while the pipette capacitance will be unaltered for any
given depth of immersion, its resistance up to the end of the wire would be significantly
reduced.

In 1992, Rae and Levis have introduced a technique that has been shown to minimize
pipette noise arising from all of the mechanisms discussed above. In this technique, the
surface of the bathing solution was covered with a layer of Silicone oil (#200 Fluid from
Dow Corning, Midland, MI). With excised patches it was then possible to raise the tip of
the electrode to within a few micrometers of the saline-oil interface; a sharp line of demar-
cation could be observed along the pipette at this interface. The electrical characteristics
of this oil are apparently quite good2, so that only the tip of the pipette that remains in
saline appears to contribute to total noise. If 10 μm of the pipette tip remain in saline, it
would be expected that the capacitance associated with this length should be roughly
0.01 pF. Dielectric noise associated with the immersed portion of the pipette should be
about one tenth of the values predicted for a 1 mm depth of immersion. It should be rea-
sonable to approximate distributed R-C noise in this situation by a lumped resistance
equal to the resistance of the pipette other than that arising from the tip itself is series with
approximately 0.01 pF. If this resistance is taken to be 2.5 MΩ, the expected noise is only
about 7 fA rms in a 10 kHz bandwidth; even for 10 MΩ it is less that 15 fA rms in this
bandwidth. The noise increment associated with the pipette capacitance and the head-
stage input voltage noise is also minimized in this arrangement since the capacitance
remains essentially the same as that with the electrode in air. It should also be noted that
even when patches cannot be excised, this approach can still be effective if the bath is
designed so that the solution level can be lowered, thereby bringing the layer of silicone
fluid very close to the cell surface.

Using quartz pipettes with this technique has resulted in noise levels in actual recording
situations as low as 0.08 pA rms in a 5 kHz bandwidth (4-pole Butterworth filter). For
these experiments, the Axopatch 200A amplifier was used with an open-circuit headstage
noise of 0.057 pA rms in a 5 kHz bandwidth; with the holder and electrode attached, but
with the electrode tip in air, noise typically increased to about 0.07 pA rms in this band-
width. In several experiments it was shown that the total noise closely approximated the
rms sum of the noise of the headstage plus holder/electrode (electrode in air) and the ther-
mal current noise of the measured DC seal resistance. This implies that dielectric noise
and “distributed R-C noise” of the pipette contributed only a negligible amount to total
noise in this situation. Rae and Levis (1993) showed that quartz pipettes can be fabricated
with the Sutter Instrument P-200 laser-based puller to yield lower noise than they previ-
ously achieved, obviating the need for Silicone oil.

Preliminary tests have also been made using this technique with pipettes fabricated from
two other types of glass (Corning 7760, Kimble KG-12). Although the noise reduction

2. Although the dielectric characteristics of the specific Silicone fluid used in these experiments is not available, Table 2.5 of Electronics
Designers' Handbook (1977, L. Giacoleto ed., McGraw-Hill) lists a dissipation factor (D) of 8.5x10-5 at 1 kHz and 2x10-5 at 100 kHz
for Silicone fluid (methyl- or ethyl-siloxane polymer); the dielectric constant ε is 2.68.

The Axon Guide — 2500-0102 Rev. C 229

 11. Noise in Electrophysiological Measurements

was significant, results were never as good as those achieved with quartz. For example, the
best results achieved with KG-12 were slightly more that 0.10 pA rms in a 5 kHz band-
width, even when seal resistances in the range of 50–100 GΩ were obtained. For a total
noise of 0.105 pA rms, and assuming that the headstage plus holder/electrode in air pro-
duced 0.075 pA rms and that a 50 GΩ seal produces 0.04 pA rms, all in a 5 kHz band-
width, the pipette would be estimated to contribute about 0.06 pA rms in this bandwidth.
On the basis of the above discussion, this is several times the amount of noise expected for
a pipette fabricated from this glass with 10 μm of its tip in saline. One possible explana-
tion of at least part of this discrepancy would be to assume that the glass had thinned
excessively near the tip. It is also possible that the noise attributed to the seal in arriving at
this estimate is too low. More work will be needed to clarify this issue.
Noise Properties of Quartz Patch Pipettes
Since 1991, it is now possible to routinely fabricate patch pipettes from quartz using the
Model P-2000 puller from Sutter Instrument Company (Novato, CA). This puller uses a
laser to overcome the extremely high melting temperatures needed to draw pipettes from
quartz tubing. Preliminary results were made using Amersil T-08 quartz pipettes
although, it was difficult to pull small-tipped blunt pipettes with 1.65 mm OD 1.15 mm
ID tubing. Pipettes fabricated from this tubing typically had a rather long and relatively
narrow shank. This geometry is not ideal for the lowest possible noise. Even so, these
pipettes produced significantly less noise than pipettes fabricated from the best glasses
used previously (e.g., 7760, 8161). In actual recording situations with pipettes immersed
about 3 mm into the bath and with seal resistances ≥ 50 GΩ, rms noise in a 5 kHz band-
width was typically about 0.12–0.13 pA rms using an Axopatch 200A amplifier. Estimates
of the contribution of the pipette to this noise were in the range of 0.075–0.09 pA rms in
this bandwidth. With the new technique described above (Rae & Levis, 1992), which
allows the pipette tip to be positioned within a few micrometers of a layer of Silicone oil
covering the surface of the bath, background noise levels as low as 0.08 pA rms in a 5 kHz
bandwidth have been achieved in actual single-channel recording situations with quartz
pipettes. In such cases the contribution of the dielectric noise and distributed R-C noise
of the pipette appears to be negligible in comparison to other noise sources.

More accurate estimates of the noise attributable to the pipette could be made by sealing
quartz pipettes to Sylgard. A typical result for approximately a 3 mm depth of immersion
is 0.115 pA rms in a 5 kHz bandwidth for pipettes coated with a thin layer of Sylgard to
within about 100–200 μm of the tip. The power spectral density (PSD) of this noise was
also measured. An estimate of the PSD attributable to the dielectric noise and distributed
R-C noise of the pipette was obtained by subtracting from the total PSD the PSD of the
headstage with the electrode in air with a correction made for the effects of the additional
capacitance of the immersed pipette (about 2.6 pF) in conjunction with the headstage
input voltage noise. The thermal current noise level associated with the DC seal was also
subtracted. The resulting PSD was taken to be the best estimate of the noise attributable
to the pipette per se. This PSD (Amp2/Hz) increased with frequency with a slope of
approximately f 1.85 in the frequency range from 2–20 kHz. From the PSD it could be
estimated that the pipette accounted for about 0.07 pA rms of noise in a 5 kHz band-

230 The Axon Guide — 2500-0102 Rev. C

 11.6. Electrode Noise

width and about 0.17 pA rms in a 10 kHz bandwidth. If it is assumed that the slope of
f 1.85 was composed of a component with a slope of f attributable to the dielectric noise of
the pipette and a component with a slope of f 2, attributed to distributed resistance-capac-
itance noise of the pipette, it can be estimated that dielectric noise would account for
slightly less than 0.06 pA rms in a 10 kHz bandwidth. Distributed R-C noise would then
account for about 0.16 pA rms at this bandwidth. The estimated value of dielectric noise
is somewhat higher than would be expected from the value of D listed for Amersil
TO-8 quartz (0.0001 at 1 MHz) and the measured value of CD (2.6 pF); with these val-
ues, equation 10 predicts ≈ 0.04 pA rms of dielectric noise in a 10 kHz bandwidth. If the
above parsing of the noise components is correct, then it can be estimated that the actual
value of D for the quartz used was closer to 0.00025. Even if this value is correct, the Dε
product for this quartz is still about an order of magnitude lower than that of any other
glass used to date for the fabrication of patch pipettes. At the time of this writing, neither
different grades (i.e., higher purity) of quartz from the same supplier nor quartz from
other manufacturers have been investigated. It should also be noted that the noise compo-
nent attributed to distributed R-C noise probably could have been reduced if the pipettes
tested had a more favorable geometry.

When the same procedure was used to estimate the pipette noise when the depth of
immersion was decreased such that the increment in capacitance was about 0.7 pF above
that measured with the pipette tip just above the solution, it was found that the noise
attributable to the pipette was roughly a factor of two lower than the figures reported
above. In this case the estimated pipette noise PSD increased with increasing frequency
approximately as f 1.9 in the range from 2–20 kHz. This is consistent with the prediction
that the relative decrease in dielectric noise will be somewhat greater than that of distrib-
uted R-C noise as the depth of immersion decreases.

It should be noted that the results for quartz pipettes are qualitatively as well as quantita-
tively different from those that have been obtained previously from pipettes fabricated
with other types of glass. Estimates of pipette noise based on procedures like those
described above have been performed for many types of glass (i.e., Figure 11.4). Even for
the lowest Dε product glasses (e.g., 7760, 8161, 7740), the estimated pipette noise PSD
rises with increasing frequency as f 1.1– f 1.3 (frequency range 2–20 kHz, immersion depth
≈ 2 mm). This indicates that dielectric noise dominates total pipette noise for these
glasses. However, for quartz not only is the pipette noise significantly less for a given
depth of immersion, but the estimated pipette noise PSD rises much more steeply with
increasing frequency (f 1.8 – f 1.9 in the 2–20 kHz frequency range)3. This indicates that
dielectric noise is no longer the dominant source of noise for patch pipettes fabricated

3. Note, however, that the PSD for quartz pipettes is lower than that of pipettes made from other glasses at all frequencies measured. At
sufficiently high frequencies distributed R-C noise should dominate the total noise for pipettes fabricated from any type of glass.
However, for quartz pipettes the PSD of distributed R-C noise should exceed that of dielectric noise by 1 kHz; for pipettes made from
7760, distributed R-C noise should not exceed dielectric noise up to 10–20 kHz, and for soda lime (e.g., 0080) pipettes this frequency
might be closer to 100 kHz or more. Distributed R-C noise depends on the pipette geometry and a variety of other factors already
described; the only influence of the glass type itself—accept in so far as it effects the geometry which can be pulled—is its dielectric
constant.

The Axon Guide — 2500-0102 Rev. C 231

 11. Noise in Electrophysiological Measurements

from quartz. The data described above are consistent with the interpretation that for
quartz pipettes “distributed R-C noise” has become the dominant noise mechanism due
to the greatly reduced contribution of dielectric noise. As more data from quartz pipettes
becomes available these estimates and conclusions will doubtlessly be refined.
Noise Arising From Pipette Resistance and Patch Capacitance
The capacitance, Cp, of the patch is in series with the entire pipette resistance, Re. The
thermal voltage noise of Re will produce current noise in conjunction with the patch
capacitance. The PSD of this noise, S2p (f ), is given by:

S p2 ( f ) = 4π 2ee2C p2 f 2 (14)

where e2e = 4kTRe is the thermal voltage noise of the pipette. The rms noise arising from
this mechanism from DC to a bandwidth B is then ( 4 π 2ee2C p2B3 )1 / 2 . In general, this
3
noise should be quite small, but it can become significant when the patch area is large;
e.g., when a large “bleb” of membrane is sucked into the pipette tip. Sakmann and Neher
(1983, in Single-Channel Recording, Sakmann, B. and Neher E. eds. pp. 37–51) mea-
sured patch capacitance for a large number of pipettes with resistances ranging from about
1–10 MΩ. They found that the value of Cp varied from as little as 0.01 pF to as much as
0.25 pF. Despite a very large amount of scatter, they found that Cp and Re were corre-
lated, with Cp increasing as Re decreases. The best fit to their data was Cp = 0.126 pF(1/R
+ 0.018), where R is the pipette resistance in MΩ. Using this average relationship it can be
predicted that for a “typical” 10 MΩ pipette (Cp = 0.015 pF) the noise in a 10 kHz band-
width (8-pole Bessel filter) arising from this mechanism will be about 0.03 pA rms, while
for a 1 MΩ pipette (typical Cp = 0.128 pF) this value will increase to 0.08 pA rms. Sak-
mann and Neher’s results indicate that in the most favorable situations in terms of
“Re-Cp” noise, the rms noise from this mechanism can be as low as 0.01 pA rms in a
10 kHz bandwidth. However, in the least favorable situation (Re ≈ 2 MΩ, Cp ≈ 0.25 pF)
the noise from this mechanism can be as large as 0.23 pA rms in a 10 kHz bandwidth
(8-pole Bessel filter).

11.7. SEAL NOISE

The membrane-to-glass seal that is essential to the patch voltage-clamp technique is the
final source of noise associated with the pipette that will be considered here. Seal noise is
perhaps the least understood source of noise in the patch clamp technique. The power
2
spectral, Ssh (f ), of the noise arising from the seal for zero applied voltage is given by:

2 (15)
S sh ( f ) = 4kT Re{Ysh }

232 The Axon Guide — 2500-0102 Rev. C

 11.7. Seal Noise

where Re{Ysh} is the real part of the seal admittance Ysh. The minimum estimate of seal
noise results from the assumption that Re{Ysh} = 1/Rsh, where Rsh is the DC seal resis-
tance. If this assumption is correct, then for a 10 kHz bandwidth, seal resistances of 1, 10
and 100 GΩ would produce noise of 0.4, 0.13 and 0.04 pA rms, respectively. Since values
of Rsh in the range of 100–200 GΩ are not uncommon, this would imply that the noise
associated with a very tight seal would be small in comparison with other sources of
patch-clamp noise. However, it is possible that the PSD of seal noise may rise above the
minimum level given by 4kT/Rsh as frequency increases (i.e., the real part of the seal
admittance may increase with frequency) due to the capacitance of the glass and the mem-
brane which make up the wall of the seal. Unfortunately, we have no good theoretical
basis upon which to estimate Ysh since the precise nature of the membrane-glass seal is not
known.

It is also very difficult to empirically dissect out the noise associated with the seal from
total patch clamp noise. We believe that previous attempts to do this have overestimated
this noise. For example, as shown in Figure 1-11 of Sigworth (1983), data from F. Sachs
and E. Neher indicate a frequency-dependent seal noise PSD (Rsh ≈ 20 GΩ) that would
amount to at least 0.13 pA rms in a bandwidth from DC to 5 kHz. However, with an
integrating headstage (e.g., the CV 203BU of the Axopatch 200B amplifier), total noise
levels (i.e., including the noise of the headstage, holder, pipette, seal, etc.) lower than this
value have often been achieved in the same 5 kHz bandwidth in actual recording situa-
tions.

Rae and Levis have reported in 1992 measurements using an Axopatch 200A amplifier
with quartz pipettes and a novel technique involving placing a layer of Silicone oil on the
surface of the bathing solution containing the cells to be patched. This technique allows
excised patches to be brought within a few micrometers of the oil-water interface, thereby
minimizing the noise contribution of the pipette. With this approach, total noise levels as
low as ≈ 0.08 pA rms in a 5 kHz bandwidth were achieved. In several experiments with
excised patches, the DC seal resistance was measured (range 40–60 GΩ) and the noise was
measured with the tip just beneath the layer of oil and again with the tip withdrawn into
the oil. The rms difference of these two noise measurements should be dominated by the
seal noise plus the small amount of noise arising from the pipette resistance in series with
the patch capacitance. In all cases, it was found that this rms difference for a bandwidth of
5 kHz was close to the thermal current noise predicted from a measured seal resistance
(i.e., (4kTB/Rsh)1/2). Nevertheless, it is well known to anyone who has struggled to
achieve the lowest possible noise in patch-clamp measurements that there is considerable
variability in the total noise even among patches with very high seal resistances and with
all other conditions seemingly identical. Despite the conclusions drawn above, it seems
reasonable to guess that some of this variability arises from the seal. Noise associated with
the membrane-glass seal represents one of the fundamental limitations of the patch clamp
technique, but it now seems clear that under favorable conditions such noise can be as low
as 0.03 pA rms in a 5 kHz bandwidth. Rae and Levis (1993) achieved ultra low-noise
recording without silicone oil by improving the fabrication of the quartz pipettes.

The Axon Guide — 2500-0102 Rev. C 233

 11. Noise in Electrophysiological Measurements

11.8. NOISE IN WHOLE-CELL VOLTAGE CLAMPING

The noise associated with the whole-cell variant of the patch voltage-clamp technique at
moderate to high frequencies will almost always be dominated by current noise arising
from the series (pipette) resistance Rs, in conjunction with the membrane capacitance,
Cm, of the cell. It should be noted that the measured value of Rs is often 2 or even 3 times
higher than the resistance of the pipette prior to achieving a whole-cell recording. Due to
the filtering effect of the access resistance and cell capacitance, series-resistance compensa-
tion is required to increase the actual bandwidth of current measurement beyond
1/2πRsCm. The level of series-resistance compensation is an important determinant of
whole-cell voltage clamp noise as well as of the bandwidth and fidelity of recording (see
Section 3.3.7., “Series-Resistance Compensation”). The voltage noise PSD of Rs is approxi-
mated by 4kTRs (V2/Hz). It should be noted, however, that in addition to thermal noise a
1/f component is expected, particularly when relatively large currents are passed through
the pipette. The PSD, S2wc (f ), of the current noise arising from Rs and Cm is given by:

2 4π 2 f 2 e s2 C m2
S wc ( f )= (units: Amp 2/Hz) (16)
1 + 4π 2 f 2τ sr2

where e2s = 4kTRs and τsr = RsrCm and Rsr is the residual (i.e., uncompensated) series
resistance, e.g., if Rs = 10 MΩ then for series-resistance compensation levels of 50%, 70%
and 90%, Rsr will be 5 MΩ, 3 MΩ and 1 MΩ, respectively. It should be noted that for
100% series-resistance compensation, equation (16) reduces to 4π 2 f 2es2Cm 2 . This
emphasizes that series-resistance compensation (if it is properly designed) only restores the
noise to the level that would have resulted from the thermal voltage noise of Rs in series
with Cm in the absence of the filtering effect of Rs mentioned above. The rms noise aris-
ing from Rs and Cm over a bandwidth from DC to B Hz can be obtained by integrating
equation (16) over f from 0 to B.

Patch clamps commonly use (switch in) a 500 MΩ feedback resistor for whole-cell voltage
clamping. The noise of this resistor dominates open-circuit headstage noise up to band-
widths of about 10 kHz. However, for typical cells the noise arising from Rs and Cm will
be larger than that of the headstage for all bandwidths above a few hundred Hz. Even for
relatively ideal situations in terms of noise (e.g., a small cell with Cm = 5 pF voltage
clamped through Rs = 5 MΩ), the noise of a 500 MΩ feedback resistor will not dominate
total noise for bandwidths above about 1 kHz. Figure 11.4 shows the noise PSD and rms
noise for a typical whole-cell voltage clamp situation with Rs = 10 MΩ and Cm = 33 pF
(these are the values used in the Model Cell provided with the Axopatch-1 and the
Axopatch 200 series of patch clamp amplifiers). The top panel shows the noise PSD (as
computed from equation 16 plus headstage noise) for series-resistance compensation lev-
els of 0%, 50%, 70% and 90%; the noise PSD of an open-circuit headstage alone is
shown for comparison. It is instructive to derive an expression for the high frequency pla-
teau of these PSDs (from equation 16). If the fraction of series-resistance compensation is
defined as α and β = 1 – α (e.g., for 80% compensation β = 0.2; also note that Rsr = βRs),

234 The Axon Guide — 2500-0102 Rev. C

 11.8. Noise in Whole-Cell Voltage Clamping

then as f → ∞, S2wc(f ) → 4kT/(β2R). Even without series-resistance compensation (β =

1) the high frequency plateau is 4kT/Rs; i.e., 50 times larger than that of a 500 MΩ resis-
tor. With 90% compensation (β = 0.1) the plateau level is 4kT/(0.01Rs), which, in this
case, is equivalent to the thermal current noise of a 100 kΩ resistor.

A 10
-24 Series Resistance
% Correction
-25 90
10

70
-26
10 50
Spectral Density
0
( A2 / Hz ) -27
10

-28
10 system noise
Whole Cell mode
-29
10
10 100 1k 10 k
Frequency ( Hz )
Series Resistance
B 100 % Correction
90
70
10 50
0

Total Noise system noise

( pA rms ) 1
Whole Cell mode

0.1

0.01
10 100 1k 10 k
Bandwidth ( Hz )

Figure 11.4: Noise in whole-cell voltage clamping.

A. Power spectral density as a function of frequency for a whole-cell voltage clamp.

Cell membrane capacitance is 33 pF and series resistance is 10 MΩ; cell input
resistance has been assumed to be much larger than 500 MΩ. Series-resistance
compensation levels of 50%, 70% and 90% are shown. The lower curve (system
noise) approximates the open circuit headstage noise in whole-cell mode with a
feedback resistor of 500 MΩ.
B. RMS noise as a function of bandwidth for the same whole-cell voltage clamp situ-
ation illustrated in A. Note that the headstage noise dominates total noise in this
situation only at bandwidths less than ≈ 100 Hz.

The Axon Guide — 2500-0102 Rev. C 235

 11. Noise in Electrophysiological Measurements

The lower panel of Figure 11.4 shows the total rms noise as a function of bandwidth for
the same cell parameters with Rs compensation levels of 0%, 50%, 70% and 90%; the
rms noise of the open circuit headstage is also shown. Note that by a bandwidth of only
about 200 Hz the total noise is twice that of the headstage alone; therefore, even if a head-
stage with negligible noise had been used, at this bandwidth total noise would only
decline to about 70% of the value shown. As the bandwidth of current measurement
increases, the noise of the headstage itself becomes progressively more negligible; with
90% compensation at a bandwidth of 1 kHz, a headstage with no noise would have
reduced total noise by less than 1% (recall that the noise of the headstage and the Rs-Cm
noise considered here are not correlated and, therefore, add in an rms fashion). It should
also be pointed out that the bandwidths in this figure refer to the setting of an external fil-
ter (a perfect “brick-wall” filter has been assumed, see Section 11.12., “Filtering”). But it is
important to realize that the actual bandwidth of current measurement is limited to
1/2πRsrCm (1 pole RC filter). Without series-resistance compensation, in this example
the bandwidth is only about 480 Hz and the use of external filter bandwidths much above
this will only add noise, not new signal information. The effective bandwidth increases
with increasing series-resistance compensation, reaching nearly 5 kHz with 90% compen-
sation.

11.9. EXTERNAL NOISE SOURCES

Interference from external sources can be almost completely eliminated in a well-designed
system, but can become the dominant source of noise if proper precautions are not taken.
The most familiar form of interference is line-frequency pickup (50 or 60 Hz and har-
monics) from power supplies, fluorescent lights, etc. Well-designed instruments will not
introduce significant amounts of interference from their internal power supplies. How-
ever, a typical laboratory environment is full of potential sources of interference from
sources external to the electronic instrumentation involved in a particular measurement.
In addition to line-frequency pickup, other potential sources of interference include
nearby motors, elevators, radio and television stations, and the video monitor of your
computer (which produces an annoying timing signal at 16 kHz or higher frequency).
High impedance measurements, such as patch clamping or work with intracellular micro-
electrodes, are particularly sensitive to such external interference.

In most cases such noise sources can be controlled by careful grounding, shielding and fil-
tering. (For a detailed discussion of these techniques see Chapter 2). In some situations,
however, shielding can actually increase noise. An example is metal shielding of the
pipette holder used in the patch clamp. Such shielding inevitably increases the capacitance
at the input of the amplifier (Cg in equations (11) – (13) above) by several picofarads.
Even if the mean voltage on the shield is precisely the same as that of the negative input of
the amplifier, the noise voltages will differ and lead to increased high frequency noise. For
this reason MDS Analytical Technologies does not offer or recommend such shielded
holders; it is our experience that grounding of nearby metal objects, such as the micro-
scope, usually provides adequate shielding. Vibration, either transmitted through the floor

236 The Axon Guide — 2500-0102 Rev. C

 11.10. Digitization Noise

or through the air, is another source of external interference and adequate vibration isola-
tion is almost always required for sensitive electrophysiological measurements.

11.10. DIGITIZATION NOISE

Noise arising from digitization is often ignored. Sometimes this is justified since such
noise can be negligible with respect to other sources of noise. However, in some situations
this potential source of noise can become significant. In order to ensure that this noise
remains negligible one needs to understand the types of noise that can arise from digitiza-
tion and to use analog-to-digital converters, preamplifiers, filters, etc. that are appropriate
for the measurement of the signal.

Quantization is the approximation of each value of a signal by an integer multiple of an

elementary quantity δ, which is the quantizing step. For a 12-bit analog-to-digital con-
verter (ADC) with full-scale range (FSR) of ±10 V, δ = 20 V/212 = 4.88 mV; for a 16-bit
converter with the same FSR, δ = 305 μV. This approximation leads to the addition of a
noise signal, called quantizing noise, to the original signal. When the signal being digi-
tized is reasonably large relative to the quantizing step δ, the power of the quantizing noise
can usually be approximated by:

δ2
12

and the rms value of the quantizing noise is therefore:

δ2
12

For a 12-bit ADC with a 20 V FSR this noise value is 1.41 mV rms or about 8.5 mV
peak-to-peak.

In the process of analog-to-digital conversion, the signal is sampled as well as quantized.

The sampling frequency is denoted by fs; e.g., when converting at 1 point every 10 μs, fs is
100 kHz. In this case all of the quantizing noise power in the ADC output will appear in
the frequency band from DC to fs/2. The PSD is usually white (i.e., constant over the fre-
quency band) and has a value of δ2/6fs.

It is obvious that the quantizing step δ should be small relative to the signal being mea-
sured. This is easily accomplished in most situations by the use of appropriate preamplifi-
cation to scale the desired signal such that it fills a reasonable portion of the dynamic
range of the ADC. Difficulties can arise, however, if you need to measure small changes
embedded in large signals. Analog instruments can often have a dynamic range that con-
siderably exceeds that of ADCs. Again, using the capacitive-feedback patch clamp as an

The Axon Guide — 2500-0102 Rev. C 237

 11. Noise in Electrophysiological Measurements

example, noise levels as low as 0.02 pA rms can be achieved at a bandwidth of 1 kHz;
moreover, such an instrument can achieve noise this low even with a gain as low as
100 μV/pA. This would amount to an rms noise of only 2 μV. In order to utilize the full
dynamic range of such an instrument at this bandwidth, an ADC with 22-bit resolution
(and capable of sampling at 2–5 kHz) would be required. To the best of our knowledge,
this much dynamic range is not required for electrophysiological measurements and 12-
to 16-bit resolution in conjunction with a variable amount of preamplification is quite
adequate (see Chapter 9 for further discussion).

11.11. ALIASING
A signal can be determined completely by a set of regularly spaced samples at intervals
T = 1/fs only if it does not contain components with a frequency equal to or greater than
fs/2. This is basically a statement of the sampling theorem; fs is just the sampling fre-
quency mentioned above. The term fs/2 will be denoted by fn and is often called the
Nyquist frequency or folding frequency for reasons that will be described below. Another
way of putting this is that for sampled data, frequency is only defined over the range from
0 to fn; components with frequencies higher than fn cannot be described (at least 2 points
per cycle are needed to uniquely define a sine wave). Obviously there is nothing (other
than good sense) that will stop one from digitizing a signal with frequency components
extending many times beyond fn. However, digitizing frequency components of the signal
that lie above fn will result in “folding back” of these higher frequency components into
the frequency range from 0 to fn, consequently producing aliases.

The term fn is referred to as the folding frequency because the frequency axis of the power
spectral density will fold around fn in a manner similar to folding a road map or a carpen-
ter’s scale. This folding effect is illustrated in Figure 11.5; frequency components above fn
are shifted to lower frequencies (in the range 0 to fn). If fx is the frequency of a signal com-
ponent (desired or noise) above fn, then the frequency of its alias, fa, is given by:

f a = f x − kf s (17)

where the brackets, | |, indicate absolute value, fs is the sampling frequency, and k is a pos-
itive integer taking on whatever value is required so that fa falls in the range 0 ≤ fa ≤ fn.
For example, with fs = 20 kHz (fn = 10 kHz), a frequency component at 18 kHz will alias
to a component at 2 kHz (fa = | 18 kHz – 1 x 20 kHz | = 2 kHz) in the digitized wave-
form. Similarly, as shown in Figure 11.5, frequency components at 22 kHz, 38 kHz,
58 kHz, 62 kHz, etc., will all alias to 2 kHz in the sampled data.

238 The Axon Guide — 2500-0102 Rev. C

 11.11. Aliasing

0 2 kHz 10 kHz (fn )

18 kHz
20 kHz 10 kHz

22 kHz 30 kHz
20 kHz

38 kHz
40 kHz 30 kHz

42 kHz 50 kHz
40 kHz

58 kHz
60 kHz 50 kHz

62 kHz 70 kHz
60 kHz

Figure 11.5: Folding of the frequency axis.

In this folding of the frequency axis for fs = 20 kHz (fn = 10 kHz), note that all of the
round points shown (18 kHz, 22 kHz, ...) alias to 2 kHz.

As an example of aliasing and the problems it can produce, consider white noise extending
from DC to 10 MHz. To be specific we will assume that the PSD of this noise is
10-14 V2/Hz (100 nV/Hz); the total noise in the 10 MHz bandwidth is then 316 μV rms
(about 2 mV peak-to-peak). If this noise is sampled at fs = 20 kHz without the use of an
anti-aliasing filter, it should be obvious that the rms value of the sampled points will be
the same as that of the original data, i.e., 316 μV. However, the sampled data cannot
describe any frequency component greater than fn, here 10 kHz. If a smooth curve is fit-
ted through the sampled points (e.g., using a cubic spline), you will find that the noise
appears to be bandlimited from DC to 10 kHz and that its amplitude is the same as the
original data; its power spectral density will be 10-11 V2/Hz, i.e., 1,000 times higher than
that of the original data because the sampling has folded over the original spectrum 1,000
times. The frequency components above 10 kHz have all been aliased into the frequency
band extending from DC to fn. Clearly, aliasing has not increased the total amount of
noise, but it has shifted it from higher to lower frequencies. It is worth considering what
will happen if the sampled data is subsequently digitally filtered with a cutoff frequency of
1 kHz. This will result in reducing the noise from 316 μV rms to 100 μV rms. However,
if the original noise signal had been passed through an analog filter with the same cutoff
frequency (1 kHz), the noise amplitude would have been reduced to only 3.16 μV rms.
Once aliasing has occurred it cannot be undone by any digital operation. The solution
here is either to sample much faster (> 20 MHz in this example) or, if a 20 kHz sample

The Axon Guide — 2500-0102 Rev. C 239

 11. Noise in Electrophysiological Measurements

rate is required, to use an analog anti-aliasing filter to adequately reduce the amplitude of
all frequency components above 10 kHz.

If the PSD of the noise is not white but instead rises as f 2 with increasing frequency—as is
the case for high frequency noise from a patch voltage clamp—the consequences of alias-
ing can be even more severe. As an extreme example, consider a voltage noise of
100 nV/√Hz in series with a 10 pF capacitor (see equation (13) and related discussion); in
a 10 MHz bandwidth this would produce a current noise of 115 nA rms. Again, assuming
a digitization rate of 20 kHz with no anti-aliasing filter, all of this noise would be aliased
into the frequency band from DC to 10 kHz, even though the noise in a bandwidth of
10 kHz would have only been about 3.6 pA rms.

Moreover, subsequent digital filtering of the digitized noise with a cut-off frequency of
1 kHz, as in the previous example, would only reduce the noise amplitude to about 35 nA
rms, whereas an analog filter set to a 1 kHz bandwidth would have reduced the noise to
only 0.115 pA rms—300,000 times less than achieved by digitally filtering the aliased
noise. Of course patch clamps do not have a bandwidth of 10 MHz; even with more real-
istic bandwidths, failure to use proper anti-aliasing filters can greatly increase noise
beyond that existing in the bandwidth resolved by the digitization process (i.e., fn) and
can reduce the effectiveness of subsequent digital filtering of the data.

It should be noted that in the above examples it has been assumed that the filter used had
an abrupt cut-off at its -3 dB bandwidth. This will normally not be the case when measur-
ing signals in the time domain. Filters with Gaussian or Bessel characteristics (as are used
most frequently for electrophysiological measurements) roll off quite gradually beyond
their -3 dB bandwidth (fc) and it is therefore not appropriate when using such filters to
eliminate aliasing to set fc = fn. Recall that the requirement to avoid significant aliasing is
that all frequency components above fn must be adequately attenuated. When using a
Bessel filter (typically 4- or 8-pole) for anti-aliasing, the choice of the cut-off frequency
relative to fn depends on the characteristics of the noise and how much aliasing can be tol-
erated. We advise the use of fc ≤ 0.4–0.5fn (0.2–0.25fs) as a useful and generally reason-
able practice.

11.12. FILTERING
The above discussion naturally leads to a brief discussion of filtering. In general, the band-
width of a filter is selected to reduce noise to acceptable levels so that the desired signal
can be adequately observed. As described above, filtering prior to digitization is also neces-
sary to prevent aliasing. If the signal to be measured is large in comparison with the back-
ground noise, then the filter bandwidth—and the appropriate digitization rate—can be
chosen on the basis of the desired time resolution; wider bandwidths allow more rapid
events to be resolved. However, in many electrophysiological measurements very wide
bandwidths will result in background noise levels that will obscure the signals of interest.
In such situations it is necessary to make compromises between the amount of noise that
can be tolerated vs. the time resolution that is achieved after filtering.

240 The Axon Guide — 2500-0102 Rev. C

 11.12. Filtering

There is an interesting—although perhaps unfortunate—relationship between a function

and its Fourier transform: as the function gets narrower its transform becomes wider. The
impulse response of a filter (time domain; note that the integral of the impulse response is
the step response) and its transfer function (frequency domain) are a Fourier transform
pair. There are limits on the degree to which signals can be simultaneously “concentrated”
in both the time and the frequency domain (in fact, if stated formally, this is the uncer-
tainty principle in the units we are using). In practical terms this means that a filter with a
narrow impulse response, and therefore a rapid rise time, will have a rather gradual roll-off
in the frequency domain. Conversely, a filter with a sharp cut-off in the frequency domain
will have a more spread-out impulse response and slower rise time.

Among commonly used filters, those that provide the best resolution with little or no
overshoot and ringing of the step response are the Gaussian and Bessel filters (as the order
of a Bessel filter increases it more closely approximates a Gaussian filter). A true Gaussian
filter is easy to implement digitally, but is not often produced as an analog filter (although
passive filters that are a good approximation of this type can be constructed). Bessel filters
are more commonly used in analog applications. The basic characteristics of both filter
types are quite similar. A Gaussian filter has an impulse response with the shape of the
Gaussian distribution; its step response has no overshoot. The 10–90% rise time of the
step response of a Gaussian filter is approximately 0.34/fc, where fc is the -3 dB band-
width in hertz. However, in the frequency domain the roll-off of this filter is quite grad-
ual. Denoting the transfer function by H(f ), H(fc) = 0.707 (-3 dB), H(2fc) = 0.25
(-12 dB), and H(3fc) = 0.044 (-27 dB). An 8th-order Bessel filter closely approximates
this response in both the time domain and the frequency domain. Clearly, in terms of
noise reduction, a filter whose transfer function rolls off much more quickly after it
reaches fc would appear to be desirable. Analog filters with such characteristics are readily
available (e.g., Elliptical and Chebyshev types); in fact, you can buy sharp cut-off filters
with H(f ) = 0.01 (-40 dB) when f = 1.06fc. Digital filters can achieve even sharper cut-
offs. Unfortunately, however, sharp cut-off filters are characterized by severe overshoot and
prolonged ringing in their step responses. Additionally, for a given value of fc, their rise
times can be nearly twice that of a Gaussian or Bessel filter. Because of this, very sharp cut-
off filters are desirable for frequency domain measurements but quite undesirable for mea-
surements in the time domain. In fact, in order to achieve the same time resolution with a
sharp cut-off filter that is achieved with a Gaussian or Bessel filter it is necessary to use a
higher value of fc. In this case, the sharp cut-off filter (with its higher fc) will pass as much
or even more noise (depending on the spectral characteristics of the noise) as the more
gradual roll-off Gaussian or Bessel filter when the two have been set to provide essentially
equivalent time resolution (as judged by step response rise time; see next paragraph).
Some “exotic” filter types can produce the same rise time as the Gaussian filter with mini-
mal overshoot and reduce the noise by a small amount; however, unless the noise PSD
rises very steeply with increasing frequency, the improvement is only a few percent.

It is instructive to quantitatively compare the performance of a Gaussian or Bessel filter

with a very sharp cut-off filter for time-domain measurements. When the underlying sig-
nal to be resolved is a square pulse, as is the case with single-channel currents, it is reason-

The Axon Guide — 2500-0102 Rev. C 241

 11. Noise in Electrophysiological Measurements

able to relate the time resolutions of the filter to its 10–90% rise time. For single-channel
measurements time resolution is often thought of in terms of the minimum detectable
event duration. To some extent, of course, such a minimum duration is dependent on the
detection algorithm used. Even so, it is reasonable to approximate the minimum detect-
able duration in terms of the filter bandwidth as 1/Tr, where Tr is the 10–90% rise time of
the filter. With such an operational definition of time resolution—or minimum detect-
able duration—it is possible to compare the performances of different filter types. As
already noted, for a Gaussian or Bessel filter (8th order) Tr ≈ 0.34/fc, where fc is the -3 dB
bandwidth in hertz. A 10th-order Chebyshev filter (0.1 dB pass-band ripple) is a reason-
able selection to approximate the “brick wall” characteristic mentioned above; for this fil-
ter H(1.09fc) = 0.1 (-20 dB), H(1.22fc) = 0.01 (-40 dB), and H(2fc) = -95 dB. However,
for this filter Tr ≈ 0.58/fc, i.e., approximately 1.7 times the rise time of a Gaussian or
Bessel filter with the same fc. Moreover the step response is characterized by a peak over-
shoot of about 20% and sustained ringing which is noticeable up to about 10/fc. In order
to achieve the same 10–90% rise time with the 10th-order Chebyshev filter that is
achieved with a Gaussian or 8th-order Bessel filter, it is necessary to select the -3 dB band-
width of the Chebyshev filter to be about 1.7 times higher than the -3 dB bandwidth of
the Gaussian or Bessel filter. For example, the rise time of a Chebyshev filter with fc = 17
kHz will be approximately the same as that of a Gaussian or Bessel filter with fc = 10 kHz;
it should be noted that the step response of the Chebyshev filter will ring severely. Even if
the ringing of its step and impulse response is ignored so that it is assumed that the 10th-
order Chebyshev filter and the Gaussian or Bessel filter have the same time resolution pro-
vided that fc (Chebyshev) ≈ 1.7fc (Gaussian), it is found that the Gaussian or Bessel filter
significantly out-performs the Chebyshev filter in terms of noise in the filtered signal.

For white noise it would be found that the Chebyshev filter would pass approximately
1.3x the noise passed by the Gaussian or Bessel filter, provided, of course, that both filters
had the same 10–90% rise time. For noise with a PSD that rises as f 2, the Chebyshev filter
would actually pass about 1.5x the rms noise passed by the Gaussian or Bessel filter, again
provided that the filters’ corner frequencies were set to produce the same rise time. Obvi-
ously, when time resolution is considered, extremely sharp cut-off filters are not the best
selection. Sharp cut-off filters should only be used when frequency domain data (e.g.,
power spectral density) is being collected.

Finally, as an illustration of the distinction between the -3 dB bandwidth and the upper-
most frequency component of noise that can pass through a Gaussian or Bessel filter, con-
sider a noise process (e2(f )) whose PSD (V2/Hz) rises as f 2 with increasing frequency. On
linear scales such noise is shown in Figure 11.6 along with the transfer function of a Gaus-
sian filter (squared, H2(f )) and the noise PSD that will be observed at the filter output
(i.e., H2(f )e2(f )). Note that the square root of the integral of the filtered noise PSD (i.e.,
the square root of the area under the curve) will be the rms noise. The PSD of the filtered
noise does not fall sharply at fc. In fact, the filtered PSD reaches its peak at about 1.2 fc
and does not fall to negligible levels until more than 3 fc. This must be remembered when
selecting a digitizing rate for any particular -3 dB filter setting if aliasing is to be avoided.
In comparison with a filter with a “brick wall” roll-off at the same fc, the Gaussian filter

242 The Axon Guide — 2500-0102 Rev. C

 11.12. Filtering

will have somewhat more than 40% more noise at its output. But recall that the sharp cut-
off filter will have a much poorer time-domain response. In fact, as described above, if the
-3 dB bandwidth of the sharp cut-off filter is selected to produce essentially the same time
resolution as the Gaussian or Bessel filter, it will pass even more noise in this situation.

4.0

Power Spectral Density (V /Hz e2 (f) and e2 (f)H (f))

3.5

2
3.0

2.5
2

2.0
e2( f )

1.5

Transfer Function
(Squared H (f))
2
1.0 H ( f ) 1.0

2
2
0.5 e2 ( f )H ( f ) 0.5

0 1 2 3 4
Normalized Frequency (f/fc )

Figure 11.6: Squared transfer function of a Gaussian filter.

Illustrated are squared transfer function of a Gaussian filter (H2(f )), noise process with
PSD that arises as f 2 with increasing frequency (e2(f )), and the same noise process PSD
after passing through the Gaussian filter (e2(f )H2(f )). The scales are linear; frequency is
normalized to the -3 dB bandwidth (fc) of the filter; PSD scale is arbitrary. Note that due
to the gradual roll-off of the Gaussian filter, the filtered noise still has significant power
well beyond fc.

The Axon Guide — 2500-0102 Rev. C 243

 11. Noise in Electrophysiological Measurements

A useful alternative to adjusting the analog filter bandwidth to attempt to achieve an opti-
mum signal-to-noise ratio is to digitize the data at a rapid rate through a Bessel filter with
its -3 dB bandwidth set to prevent aliasing. The wideband digitized data may then be fil-
tered digitally to achieve an acceptable signal to noise level. The digital filter should once
again generally be of the Gaussian or Bessel type. In such situations more than one filter
are used in series; if these are all of the Gaussian or Bessel type, then the final -3 dB band-
width, fcF, is approximately given by:

1
f cF =
1 1 1
2
+ 2 + 2 + ...
f1 f2 f3

where f1, f2, f3 ... are the -3 dB bandwidths of the individual filters in series. Of course,
the same expression holds for a series combination of analog filters. The disadvantage to
this general approach is that initial data storage will be increased due to the high digitiza-
tion rate; once the data has been digitally filtered it can also be decimated if desired.

11.13. SUMMARY OF PATCH-CLAMP NOISE

11.13.1. HEADSTAGE
At the time of publishing this Guide, the best capacitive-feedback headstage (the CV
203BU of the Axopatch 200B amplifier) has an open-circuit noise of about 0.13 pA rms
in a 10 kHz bandwidth (-3 dB, 8-pole Bessel filter). Resistive-feedback headstages have
noise of about 0.25–0.30 pA rms in this bandwidth. It is possible that developing capaci-
tive-feedback technology could yield further improvements in noise. In fact, in theory, it
should be possible to produce a capacitive-feedback headstage with noise that is only
about half of that achieved presently.

11.13.2. HOLDER
The pipette holder contributes noise by increasing the capacitance at the headstage input;
some dielectric noise must also be expected. The unshielded polycarbonate holders pro-
vided by MDS Analytical Technologies by themselves will only increase the headstage
noise by about 10% above its open circuit value even for the lowest noise capacitive-feed-
back devices. For example, for the Axopatch 200B amplifier with an open-circuit noise of
0.060 pA rms in a 5 kHz bandwidth (as shown on the panel meter), the total noise with
the holder attached should not increase beyond 0.070 pA rms (note that larger noise
increments probably mean that the holder needs to be cleaned), and will often be as low as
0.065 pA rms. Inserting a saline-filled electrode into the holder will further increase noise
even while the electrode tip is in air. With a saline-filled electrode, the noise of the
Axopatch 200B with an open-circuit noise of 0.060 pA in a 5 kHz bandwidth will gener-
ally increase to about 0.075 pA rms.

244 The Axon Guide — 2500-0102 Rev. C

 11.13. Summary of Patch-Clamp Noise

11.13.3. ELECTRODE
The noise associated with the patch pipette is determined by the type of glass used, the
wall thickness of the glass, the pipette geometry, the use of Sylgard coating, and the depth
of immersion into the bath. Dielectric noise is probably the dominant source of noise
associated with the electrode for all glasses except quartz. For glasses with lower Dε prod-
ucts—most notably quartz—it can be expected to become a much smaller component of
overall noise. Additional noise will arise from the distributed resistance and capacitance of
the pipette. For the best glasses presently in common use, this source of noise is probably
somewhat less than the dielectric noise of the glass; for very low-loss glasses, such as
quartz, it could be the dominant source of pipette noise. A small amount of noise will also
result from the thermal noise of the pipette resistance in series with the patch capacitance;
only under certain situations (a large patch area) will this “Re-Cp” noise become signifi-
cant.

Theoretical and experimental results indicate that a pipette fabricated from the lowest-
noise glasses (other than quartz) used to date (Corning #7760, #8161) with a moderate
coating of Sylgard will produce noise of about 0.2 pA rms in a 10 kHz bandwidth (8-pole
Bessel filter) for an immersion depth of about 2 mm. Under favorable circumstances, this
value can be cut at least in half when the tip of the pipette is withdrawn to within 100–
200 μm of the surface of the bath. For pipettes fabricated from quartz, preliminary results
indicate that for a 1 mm depth of immersion, noise of somewhat less than 0.1 pA rms can
be expected in a bandwidth of 10 kHz; with the very small immersion depth possible
(10 μm or less) with the Silicone-fluid technique described above, it can be estimated that
the noise contributed by a quartz pipette falls to less than half of this value. Results
obtained for quartz have involved pipettes with relatively long narrow shanks and resis-
tances of roughly 10 MΩ. Such a geometry is not ideal for achieving the lowest possible
noise. As techniques for pulling quartz pipettes improve, and as other grades and suppliers
of quartz are investigated, further improvements are likely.

11.13.4. SEAL
The noise associated with the seal is less easily predicted. Certainly a minimum estimate
of this noise in a bandwidth B is given by (4kTB/Rsh)1/2, i.e., the thermal current noise of
the DC seal resistance, Rsh. For excellent seals in the range of 50–200 GΩ this would
mean that the minimum noise attributable to the seal is in the range of 0.03–0.06 pA rms
in a 10 kHz bandwidth. Recent data suggests that under favorable circumstances seal
noise may be as low as these predicted values. However, experience indicates that there is a
good deal of variability in the noise of patches even when the seal resistances are very high
and all other precautions necessary to minimize noise have been strictly observed. Some of
this variability may well arise from the seal.

The Axon Guide — 2500-0102 Rev. C 245

 11. Noise in Electrophysiological Measurements

11.14. LIMITS OF PATCH-CLAMP NOISE PERFORMANCE

A capacitive-feedback headstage, such as the CV 203BU of the Axopatch 200B amplifier,
has an open circuit noise of less than 0.13 pA rms in a 10 kHz bandwidth, (bandwidths
are the -3 dB frequency of an 8-pole Bessel filter). The total noise is due to contributions
from the headstage, holder, pipette and seal. However, peltier cooling has removed the
thermal noise from this headstage, reducing total noise from 0.16 to about 0.13–0.14 pA
rms in a 10 kHz bandwidth. Of course, the noise associated with the seal is a fundamental
limitation of the patch clamp technique. An absolute minimum estimate for seal noise in
a 10 kHz bandwidth is about 0.03–0.04 pA rms; this assumes a 100–200 GΩ seal pro-
ducing only thermal noise. An estimate of the minimum noise associated with an
improved holder, quartz pipette and seal in a 10 kHz bandwidth is then about 0.06 pA
rms. These figures suggest that as holder and pipette technology continues to improve, it
will be worthwhile to continue to seek methods to reduce the noise of the headstage itself.
It is reasonable to expect that total noise levels roughly two thirds of the best values
achieved to date will become possible. However, to achieve such noise levels, careful atten-
tion to every aspect of the patch-clamp technique will be important.

11.15. FURTHER READING

Rae, J.L., Levis, R.A., A method for exceptionally low noise single-channel recording.
Pflügers Arch. 420, 618–620, 1992.

Rae, J.L., Levis, R.A., Quartz pipettes for single-channel recording. Biophys. 64, A201
(P394), 1993.

Sakmann, B. and Neher, E., Eds. Single-Channel Recording. New York: Plenum Press,
1983.

246 The Axon Guide — 2500-0102 Rev. C

 A. Appendix: Guide to
Interpreting Specifications
A.1. GENERAL
Following are terms commonly used in electrophysiological measurements.

A.1.1. THERMAL NOISE

All resistors generate thermal noise. Thermal noise can be represented by a mean square
voltage generator (e n2 ) in series with a noiseless resistor:

en2 = 4kTRB

where k = Boltzmann’s constant (1.138 x 10-23 VC/°K)

T = temperature in °K

R = resistor value in ohms

B = bandwidth in Hertz

The noise specified by this equation has a constant power spectral density within the
bandwidth and is zero outside the bandwidth. When noise is measured through a low-
pass filter having -3 dB bandwidth f-3, the noise will be greater than the amount predicted
because real filters continue to pass signals at frequencies > f-3.

The noise of a 10 MΩ resistor measured through various filters is shown in Table A.1.
Table A.1: Noise of a 10 MΩ Resistor in Microvolts rms.

2-Pole 4-Pole
Bandwidth 1-Pole RC Infinitely Sharp
Butterworth Butterworth
DC-100 Hz 5.07 4.27 4.11 4.06
DC-1 kHz 16.0 13.5 13.0 12.8
DC-10 kHz 50.7 42.8 41.1 40.6
DC-100 kHz 160 135 130 128

The Axon Guide — 2500-0102 Rev. C 247

 A. Appendix: Guide to Interpreting Specifications

A.1.2. RMS VERSUS PEAK-TO-PEAK NOISE

It is equally valid to specify noise as rms (root mean square) or peak-to-peak. The ratio of
the zero-to-peak value to the rms value is called the crest factor. The value of the crest fac-
tor depends on the type of noise. For white noise the crest factor is about 4. Thus the
peak-to-peak noise is about 8 times the rms noise.

A.1.3. BANDWIDTH—TIME CONSTANT—RISE TIME

For an exponential response, there are simple relationships between the -3 dB bandwidth
(f-3, Hertz), the time constant (τ, seconds) and the 10–90% rise time (t10–90; seconds):

1 1.1
f −3 = ≈
2πτ π t10 −90
t 10 -90 ≈ 2.2τ

A.1.4. FILTERS
Low-pass filters are commonly used in electrophysiology to reduce noise. The important
parameters of a low-pass filter are the -3 dB frequency, the type and the order.

The -3 dB frequency (f-3) is the frequency where the voltage response falls to √2 (0.707).
Some manufacturers specify the “cutoff ” or “bandwidth” based on the phase response or
an asymptotic approximation to the high and low frequency amplitude response. There-
fore when using a new filter you should check that the specified settings refer to the -3 dB
frequency.

The three types of low-pass filters commonly used in electrophysiology are the Bessel,
Butterworth and multiple coincident pole (RC) types. The most important differences
are:

1 When driven by a step voltage, the RC filter has no overshoot, the Bessel filter has
< 1% overshoot and the Butterworth filter has around 10% overshoot.
2 When driven by a noisy source, RC filters show the least rejection of the noise at
frequencies above f-3, Bessel filters show moderate rejection and Butterworth filters
show the most.
The order of a filter refers to the number of poles. Each resistor-capacitor section contrib-
utes one pole (inductors are rarely used for bandwidths < 100 kHz). Thus 4th-order,
4-pole and 4 RC sections all refer to the same thing.

The attenuation of high-frequency noise increases with order. At frequencies well above
f-3, the attenuation increases at 20 dB/decade (6 dB/octave) for each pole. Thus a 4-pole
filter rolls off at 80 dB/decade.
Note: 20 dB attenuation corresponds to a drop in voltage amplitude to one tenth.

248 The Axon Guide — 2500-0102 Rev. C

 A.2. Microelectrode Amplifiers

A.2. MICROELECTRODE AMPLIFIERS

Noise magnitude is very sensitive to the measurement technique. Specifications should
state:

1 -3 dB bandwidth and order of the filter used in the measurement circuit. Typically the
noise will look 20–30% better if a fourth-order low-pass filter is used instead of a first-
order low-pass filter.
2 Microelectrode bandwidth. The capacitance neutralization should be adjusted so that
the -3 dB bandwidth of the microelectrode is the same as the -3 dB bandwidth of the
measurement circuit.

A.2.1. VOLTAGE-CLAMP NOISE

Almost impossible to specify. Depends on cell mode, electrode model, capacitance neu-
tralization setting, electrode interactions, electrode resistances, current measurement tech-
niques, bandwidth of measurement circuit and clamp gain.

A.2.2. 10–90% RISE TIME (t10–90)

The time for the response to go from 10% to 90% of the final value.

If a voltage step is applied to a microelectrode headstage via a resistor, the resistor should
have a very low stray capacitance. Otherwise the stray capacitance couples the step directly
into the headstage input and artificially fast rise times are measured. A better estimate of
the rise time can be made by passing a current step out of the headstage into a load resis-
tor.

For consistent comparisons, rise times should be measured with the capacitance neutral-
ization adjusted for zero overshoot.

A.2.3. 1% SETTLING TIME (t1)

When microelectrodes are to be used in a discontinuous (switching) single-electrode volt-
age clamp, the more relevant specification is t1, the time taken for the response to go from
0% to 99% of the final value.

A.2.4. INPUT CAPACITANCE

In a circuit using capacitance neutralization and other compensation techniques, it is
meaningless to specify a value for the input capacitance. The efficacy of the neutralization
circuit depends on the magnitude of the electrode resistance and the measurement tech-
nique.

The Axon Guide — 2500-0102 Rev. C 249

 A. Appendix: Guide to Interpreting Specifications

For an electrode with resistance R and an exponential response to a step input, the effec-
tive input capacitance can be estimated from:

0.45t10 −90 0.22t1

Cin = =
R R

For example, if t10-90 = 10 μs and R = 10 MΩ, then Cin = 0.45 pF.

A.2.5. INPUT LEAKAGE CURRENT

Both well designed and badly designed headstages are normally adjustable to zero leakage
current. This ability in itself is not an important test of quality. The correct measure of
quality is the sensitivity of the input leakage current to temperature changes which occur
after the adjustment is made. Changes of ±10 °C can be expected in a typical laboratory
environment.

A.3. PATCH-CLAMP AMPLIFIERS

A.3.1. BANDWIDTH
The bandwidth of a patch-clamp headstage is usually measured by injecting current into
the headstage input through a small capacitor. No holder or electrode is attached to the
input.

The bandwidth during an experiment is usually much lower. It is typically limited by the
electrode resistance and membrane capacitance, as well as by stray capacitances. For exam-
ple, if a 5 MΩ electrode is used to clamp a whole cell having 20 pF of membrane capaci-
tance, the time constant for measuring membrane currents is 100 μs (5 MΩ x 20 pF).
This corresponds to a -3 dB bandwidth of 1.6 kHz, far below the 100 kHz that may be
specified. 80% series-resistance compensation would improve this to 8 kHz. Nevertheless,
it is still better to have a headstage with very wide bandwidth because in some cases series-
resistance compensation may work better.

A.3.2. NOISE
The ideal RMS current noise in selected bandwidths is shown in Table A.2 for various
feedback resistors. These figures assume ideal resistors (which only have thermal noise)
and noiseless electronics.
Table A.2: rms Current in Picoamps.

BANDWIDTH 500 MΩ 1 GΩ 5 GΩ 10 GΩ 50 GΩ
DC-1 kHz 0.182 0.128 0.057 0.041 0.018
DC-3 kHz 0.331 0.234 0.099 0.074 0.033
DC-10 kHz 0.574 0.406 0.181 0.128 0.057

250 The Axon Guide — 2500-0102 Rev. C

 A.3. Patch-Clamp Amplifiers

No patch-clamp amplifier can meet these ideal specifications. The difference between a
value in the table and the actual value is the excess noise. The excess noise is always pro-
portionately worse in the wider bandwidths.

ã=WARNING: The specified noise in a 10 kHz bandwidth will be misleadingly low if the
headstage does not have a 10 kHz bandwidth. If the manufacturer only specifies the
10–90% rise time or the time constant, use the conversion factors given in the General
section to verify that the bandwidth is sufficient to make the noise specification meaning-
ful.

The Axon Guide — 2500-0102 Rev. C 251

 A. Appendix: Guide to Interpreting Specifications

252 The Axon Guide — 2500-0102 Rev. C

 Index
Numerics use in stock solutions 121
Amplifier noise 221
1% settling time 249
Amplifiers
1/f noise 220
amplification 155
10-90% rise time 249
microelectrode 249
1% settling time 249
A 10-90% rise time 249
AC coupling 158 input capacitance 249
input leakage current 250
overload detection 160
noise 249
saturation 160 patch clamp 13
time constant 160 patch-clamp 250
Acceleration measurements 177 bandwidth 250
Accelerometers 177 noise 250
Acquisition hardware 189 voltage clamp 13
analog-to-digital conversion 189 Analog-to-digital (A/D) conversion 189
analog-to-digital converter 190 optical isolation 194
differential nonlinearity 192 quantization error 190
digital I/O 194 sampling rate 191
digital-to-analog converter 190 Analog-to-Digital conversion 155
gain accuracy 192 preparing signals 155
glitch 193 Analog-to-digital converter (ADC) 190, 237
least significant bit 193 Anti-aliasing 55
linearity error 192 Antibiotic partitioning 122
monotonicity 193 Anti-vibration tables 21
multi-tasking operating systems 195 Audio monitor 163
optical isolation 194 amplitude modulated 163
software support 195 frequency modulated 163
timers 194 Autozeroing 158
Agar bridge 85 Averaging 161
Aliasing 238
Bessel filter 240
B
folding frequency 238
Backup 182
Gaussian filter 240
Bandwidth 248
Nyquist frequency 238
Bath error potentials 40
Amphotericin B 121

The Axon Guide — 2500-0102 Rev. C 253

 Clamp Vb using a bath clamp 43 Capacitance compensation 35
clamp Vb using a virtual ground current monitor pipette-capacitance compensation 64
44 whole-cell 66
measure and substract Vb 43 Capacitance neutralization 56
minimizing Rb 41 Capacitor feedback headstage 74
series resistance compensation 42 Capacitors
Bessel filter 240, 241 currents through 12
aliasing 240 electrical fields 11
Bilayer experiments 77 Cell penetration 45
Bilayer techniques 129 Chebyshev filter 242
maximizing the bandwidth 133 Chebyshev polynomial fitting 212
minimizing noise 130 Chloriding silver wire 87
resolving voltage steps 134 Clear 45
setup 136 for cell penetration 45
Bioelectricity 1 Command generation 45
Blanking 163 Common-mode rejection ratio (CMRR) 164
sample-and-hold 163 Computer issues 181
Bootstrapping 36 software selection 181
Brain slices 19 Conductance 3, 7, 9, 15, 16
blind technique 106 Conductors 3
cleaning technique 103 Continuous single-electrode voltage clamp (cSEVC)
patch-clamp recording 102 58
thick slice 20 Controlled current source 53
thin slice 20 Crosstalk 172
Bridge balance 30 Current
Bridge design through capacitors 12
pressure and force measurements 176 Current clamp 14, 29
Buzz 45 Current conventions 80
for cell penetration 45 Current monitor 33

C D
Cable capacitance Data acquisition parameters 197
filtering capabilities 174 D/A update rate 198
Cabling data bandwidth 198
affect on noise and bandwidth 171 filtering 199, 201
Capacitance gain 197
input 249 offset 197

254 The Axon Guide — 2500-0102 Rev. C

 sampling rate 197, 201 contributes to thermal noise 173
multiple channels 198 unmatched
multiple exponentials 198 increases noise and crosstalk 172
single-channel recording 198 Electrode noise 223
time skew 198 Electrode resistance
Data backup 182 affect on noise and bandwidth 171
Dielectric noise 219, 224 see pipette resistance 79
quartz pipettes 224 Electrode test 164
Sylgard 224 Electrodes 85
Differential nonlinearity 192 chloriding silver wire 87
Digital I/O 194 glasses 87, 89
Digital-to-analog converter (DAC) 190 holders 87
feedthrough noise 193 junctional potential 86
Digitization noise 237 liquid junction potential 86
Discontinuous current clamp 57 metal microelectrodes 175
Discontinuous single-electrode voltage clamp microelectrode pullers 88
comparison with TEVC 57 microelectrodes 85
Dissipation factor 224 micropipettes 85
Driven shield 36 noise 89
Duty cycle selecting 56 patch micropipettes 86
patch pipettes 87
E silver, silver-chloride 85
silver/silver chloride 9, 10
Electrical currents 1
Electroencephalogram (EEG) 172, 175
Electrical instruments
Electromyogram (EMG) 172, 174
perfect and real 7
Equipment placement 23
Electrical interference 21
Excess noise
ground-loop noise 22
1/f noise 220
magnetically-induced pickup 22
Excess noise (1/f noise) 220
radiative electrical pickup 21
Extracellular recording 27
Electrical isolation methods 21
multiple-cell 28
Electrical potentials 1
setup 19
Electrical properties 89
single-cell 28
Electrocardiogram (EKG) 172, 175
Electrode
noise 223 F
Electrode impedance Filter
affect on noise and bandwidth 171 anti-aliasing 55
high 171

The Axon Guide — 2500-0102 Rev. C 255

 Filter terminology patch-clamp data 153
10-90% rise time 149 pole vs. order vs. slopes 148
-3db frequency 145 RC 248
attenuation 146 related terminology 144, 145
decade 146 sampling rate 152
decibels 147 time domain analysis 149
octave 146 Fitting 207
order 148 Chebyshev polynomials 212
overshoot 146 chi-square 209
pass band 146 confidence limits 210
phase shift 146 goodness of fit 210
stop band 146 likelihood 208
Filtering 240 motivation 207
acquisition time 199, 201 optimization methods 211
analysis time 199, 201 statistical aspects 208
Filters 248 statistical tests 211
-3db frequency 144 Folding frequency 238
band-pass 144 Force measurements 176, 177
band-reject 144
Bessel 240, 248
G
Bessel filter 150
Gain 156
Butterworth 248
pre-filter vs. post-filter gain 156
Butterworth filter 150, 151, 152
Gain accuracy 192
Chebyshev 242
Gaussian filter 240, 241
correcting for filter delay 154
aliasing 240
digital 153
Giant excised membrane patch method 117
finite impulse response 153
Gaussian 154 pipette fabrication 118
nonrecursive 153 seal formation 118
recursive 154 Gigaseal-Macropatch voltage clamp 110
frequency domain analysis 151 Gigaseals 71
function 144 Glass
high-pass 143 dissipation factor 224
low-pass 143 Glass compositions 91
noise 247 Glasses 89
notch filter 143, 150 electrical properties 89
Nyquist Sampling Theorem 152 see electrodes 87
order 144 thermal properties 91

256 The Axon Guide — 2500-0102 Rev. C

 types current clamp 29
noise properties 92 current monitor 33
Glitch 193 headstage current gain 35
Ground-loop noise 22 junction potentials 32
leakage current 39
H track 32
virtual ground 34
Headstage
voltage clamp 46
bandwidth 75
voltage follower 29
capacitor feedback 74
Wheatstone bridge 30
current gain 35
Ions
Dynamic range 75
in electrodes 9
for ion-sensitive microelectrodes 39
in solutions 9
Linearity 75
Isolation measurements 178
noise 75
resistor feedback 71
Headstages J
for intracellular recording 38 Johnson noise 216
for ion-sensitive microelectrodes 39 Junction potentials 32
High electrode impedance Junctional potential 86
can produce attenuation 171
Holders 87
for micropipettes 79
L
Hum 21, 161 Laboratory setup 19
differential amplification 161 list of equipment 24
notch filter 161 Large cells
voltage clamp 48
Leachable components 94
I Leakage current 39
I/O interfaces 185 Least significant bit 193
Implantable length gauges 177 Least squares 209
Input leakage current 250 Length measurements
Insulation measurements 178 implantable length gauges 177
Intracellular recording linear potentiometers 178
bath error potentials 40 linear variable differential transformers 178
bootstrapping 36 Linear potentiometers 178
bridge balance 30 Linear variable differential transformers 178
capacitance compensation 35 Linearity error 192
command generation 45 Line-frequency pick-up 161

The Axon Guide — 2500-0102 Rev. C 257

 Liquid junction potential 86 N
Loose-patch recording 109
Negative capacitance
Loose-Patch voltage clamp 111
see capacitance compensation 37
Nerve cuffs 175
M Noise 172, 215, 250
Macropatch recording 109 aliased, in voltage clamp 55
Magnetically-induced pickup 22 aliasing 238
Maximum likelihood 208 amplifier 221
Mechanical requirements capacitance 227
extracellular recording 19 capacitance neutralization 56
Mechanical vibration 45 dielectric 215, 219, 224
for cell penetration 45 digitization 237
Membrance capacitance electrode 223, 245
per unit area 68 excess (1/f noise) 215, 220
Membrane capacitance external sources 236
measurements 78 headstage 244
Membrane potential 55 high-impedance probes 171
ripple 55 holder 244
Metal microelectrodes 175 in single-channel patch voltage clamping 223
Microelectrode pullers 88 in two-electrode voltage clamp recording 51
Microelectrodes Johnson 173
glass patch capacitance 232
tight seals 16 patch clamp 246
Micromanipulators 19, 20 peak-to-peak 161, 215, 248
remote-controlled 20 pipette resistance 227, 232
Micropipette holders 79 quantizing noise 237
Micropipette selection 51 quartz patch pipettes 230
Microscope recording from bilayers 130
brain slice 20 rms 248
extracellular recording 19 rms noise 161
single-channel recording 20 root-mean-square (rms) 215
Monotonicity 193 seal 232, 245
Multiple-cell recording 28 shot 215, 218
Multi-tasking operating systems 195 silicone oil 229
Muscle sources 215
measuring thermal 173, 215, 216, 247
implantable length gauges 177 power spectral density (PSD) 216

258 The Axon Guide — 2500-0102 Rev. C

 thermal (Johnson, Nyquist) 216 holder 244
voltage-clamp 249 seal 245
whole-cell voltage clamping 234 Patch micropipettes 86
Nyquist frequency 238 Patch pipettes 87
Nyquist noise 216 coating 88
Nyquist theorem concentric, with loose-patch clamp 116
see Sampling theorem 191 electrical properties 89
Nystatin 121 fabrication 87
use in stock solutions 121 firepolishing 88
from quartz tubing 230
glasses 89
O noise 89
Offset control 157 noise properties 92
Offset removal 51 pulling 88
Ohm’s law 5 thermal properties 91
Oocytes Perforated patches
Xenopus 97 cellular voltage measurements 125
Optical isolation 194 other uses 125
Optical requirements recording from 121
extracellular recording 19 perforated vesicles
single-channel recording 19 recording from 121
Overload detection 160 Perforated-patch technique
advantages 123
P disadvantages 123
minimizing access resistance 124
Patch
Peripherals 182
rupturing 68
Pipette fabrication 118
zap 68
Pipette resistance measurement 79
Patch clamp 15, 153
Pipette solutions 85
amplifier noise 221
Pipette-capacitance compensation 64
capacitor feedback headstage 74
Pipettes 85
data, filtering 153
glasses 89
resistor feedback headstage 71
micropipettes 85
single channel 70
properties
Patch clamp noise 244
single-channel vs. whole cell recording 89
limits 246
Positive feedback (correction) 59
summary 244
Pressure
electrode 245
measuring very low 176
headstage 244

The Axon Guide — 2500-0102 Rev. C 259

 Pressure measurements 176 AC coupling 158
amplification 155
Q audio monitor 163
autozeroing 158
Quantization error 190
averaging 161
Quantizing noise 237
blanking 163
Quartz patch pipettes 230
common-mode rejection ratio 164
Quartz pipettes 224
differential amplification 161
electrode test 164
R gain 156
Radiative electrical pickup 21 line frequency pick-up (hum) 161
Resistor feedback headstage 71 noise measurements 161
Resistors 3 offset control 157
seal 16 overload detection 160
single-channel recording 3 peak-to-peak 161
Rise time 248 programmable gain amplifier (PGA) 155
saturation 160
stimulus coupling 163
S time constant 160
Sample-and-hold 163 Signal conditioning 143
Sampling rate Silicone oil 229
choosing 191 Single-cell recording 28
oversampling 191 Single-channel
sampling theorem 191 patch clamp 70
Sampling theorem 191 Single-channel analysis
Saturation 160 baseline estimation 202
Seal 71, 79 events
Seal formation 118 false 202
Seals missed 202
pretreatment of muscle cells for best 118 multiple channels 203
Self-heating measurements 178 events list 202, 203
Series resistance compensation 59 false closings 201
limitations 64 filtering 201
positive feedback (correction) 59 goals 200
supercharging (prediction) 61 histograms 203
Shielding 21, 22 errors 205
Shot noise 218 sampling rate 201
Signal conditioners 143 threshold 202

260 The Axon Guide — 2500-0102 Rev. C

Single-channel patch clamp pressure measurements 176
seal 71 resistance temp. detectors 170
Single-channel recording self-heating measurements 178
bilayer techniques 129 temperature transducers 167
setup 19 thermistors 167
Small cells thermocouples 169
voltage clamp 52
Software 181
V
Software support 195
Vibration isolation methods 20
Space clamp
Virtual ground 34
considerations 69
Voltage clamp 14, 46
Specifications 247
anti-aliasing filter 55
Stimulus coupling 163
capacitance neutralization 56
Supercharging (prediction) 61
common errors 54
Sylgard 224, 228
controlled-current source 53
current and voltage measurement 56
T error 49
Temperature transducers 167, 169 gigaseal-macropatch 110
resistance thermometers 170 ideal 47
thermistors 167 large cells 48
thermocouples 169 loose-patch 111
Thermal noise 173, 216 membrane conductance changes 50
Time constant 160, 248 micropipette selection 51
Time-domain analysis 149 noise 51
Timers 194 real 47, 48
Track 32 signal bandwith 56
Transducers 167 small cells 52
acceleration measurements 177 stability 50
accelerometers 177 step response and bandwith 49
for extended temp. ranges 169 Voltage conventions 80
force measurements 176, 177 Voltage divider 7
IC temperature, current 168 Voltage follower 29
IC temperature, voltage 169 Voltage pulses
insulation techniques 178 applied to loose-patch pipette 115
isolation measurements 178
length measurements 177
W
metal microelectrodes 175
Wheatstone bridge 30, 176

The Axon Guide — 2500-0102 Rev. C 261

 Whole-cell capacitance compensation 66
Whole-Cell clamp
combining with focal current recording 111
Whole-cell mode
dSEVC or cSEVC 69
rupturing the patch 68
Whole-cell voltage clamping 234
noise 234

X
Xenopus oocytes
definition 98
patch clamping of 101
recording from 97
two-electrode voltage clamping of 99

Z
Zap 69

262 The Axon Guide — 2500-0102 Rev. C