ME WLETT-PACKARD JOURNAL

lowiitt ttii MM mi
Copr. 1949-1998 Hewlett-Packard Co.

HEWLETT-PACKARD JOURNAL
Technical Information from the Laboratories of Hewlett-Packard Company

JULY 1983 Volume 34 Number 7

Contents:
A High-Speed System for AC Parametric Digital Hardware Analysis, by Andreas Wilbs
and Klaus-Peter Behrens Here's a 50-MHz multichannel stimulus/response system for test
ing digital circuits at realistic speeds.
A High-Speed Data Generator for Digital Testing, by Ulrich Hbner, Werner Berkel, Heinz
Nssle, and Josef Becker It offers high timing accuracy, precise pulse-level definition,
ease of operation, and versatility.
High-Speed Data Analyzer Tests Threshold and Timing Parameters, by Dieter Kible,
Bernhard Roth, Martin Dietze, and Ulrich Schottmer Two innovative features are pro
grammable sampling point delay and real-time compare mode.
Data Eckhard Software/Firmware Design, by Roberto Mottola and Eckhard Paul A skilled
task the makes full use of the interrupt structure of the CPU.
Power Supplies for the Stimulus/Response System, by Ulrich Otto and Horst Link The
objectives were high load current and serviceability within a restricted space.
New Multi-Frequency LCZ Meters Offer Higher-Speed Impedance Measurements, by
Tomio Wakasugi, Takeshi Kyo, and Toshio Tamamura These new instruments can be easily
combined with a component handler for more rapid production-line testing.

Authors
In this Issue:
Quality is high on everyone's priority list these days, and one of the principal paths to
[ quality is testing, thorough and realistic, at all phases of the process of designing and
producing a product. This month's issue deals with two families of test instruments that are
aimed resistors, assuring the quality of the myriad diodes, transistors, resistors, capacitors, induc
tors, integrated circuits, circuit boards, and modules that go into electronic equipment. Each
family capabilities a target range of testing applications and a repertoire of special capabilities
needed in those applications.
On this timing cover are three views, at progressively finer resolution, of the timing
diagram along a RAM (an integrated circuit random-access memory chip). Memory circuits like this one, along
with microprocessors, gate arrays, and other digital integrated circuits (ICs) are becoming faster and more
complex at a startling rate sometimes the next generation is announced before its predecessor is in produc
tion. The of that produced this timing diagram is designed to keep up with the rapidly changing state of
the art detail. instruments, test fast, complex digital ICs at realistic speeds and in minute detail. This family of instruments,
consisting of the 8180A Data Generator, the 81 81 A Extender, and the 8182A Data Analyzer, is distinguished
by a time resolution of 100 picoseconds (there are a million picoseconds in a millionth of a second) and the
analyzer's programmable sampling point delay. These features make it possible to zero in on the briefest of
1C events. You can read all about the design of this system on pages 3 to 31 .
While the high-performance 80A/81 A/82A System is designed for relatively low-volume testing of complex, high-performance
digital inspection in engineering, quality assurance, depot maintenance, and certain kinds of incoming inspection
and production testing, the 4276A and 4277 A LCZ Meters are aimed mainly at high-volume incoming inspection
and production testing of discrete components (LCZ stands for inductance-capacitance-impedance). They
also have components uses in the engineering laboratory and in semiconductor evaluation. By testing components
faster instruments and realistically than previously available equipment, these instruments improve the quality and
lower handler several of the products we purchase. Fed by an automatic component handler (available from several
manufacturers), they'll check up to ten components per second and sort them into as many as ten classes,
and keep it up hour after hour without attention. For reporting test data, they're compatible with computers
and plotters, and for realistic testing, they offer a variety of test frequencies and voltages. The designers tell
their story on pages 32 to 38.
-R.P. Dotan
Editor, Richard P. Doian Associate Editor, Kenneth A. Shaw Art Director, Photographer, Arvid A. Danieison Illustrators. Nancy S- Vanderbloom,
Susan E. European Administrative Services, Typography, Anne S. LoPresti, Susan E. Wright European Production Supervisor, Henk Van Lammeren
2

H E W L E T T - P A C K A R D

J O U R N A L

J U L Y

1 9 8 3

H e w l e t t - P a c k a r d

Copr. 1949-1998 Hewlett-Packard Co.

C o m p a n y

1 9 8 3

P r i n t e d

i n

U . S . A .

A High-Speed System for AC Parametric

Digital Hardware Analysis
This new 50-MHz stimulus-response system is a
state-of-the-art tool for comprehensive and
rapid characterization of all types of digital circuits.
by Andreas Wilbs and Klaus-Peter Behrens
RAPIDLY ADVANCING SEMICONDUCTOR technol
ogy now makes it possible to integrate thousands of
functions on a single chip, making digital circuits
increasingly more complex. They are also getting faster.
Subnanosecond propagation delays per gate are already
feasible, allowing clock rates of 50 MHz and higher even for
complex devices. Furthermore, new circuit designs, both
custom and semicustom (gate arrays), are receiving broad
acceptance in the marketplace.
This evolution has changed the requirements for digital
test equipment. The most significant new requirements are:
Higher data rates and more channels
More accurate timing and stimulus/comparator levels
A capability for longer test patterns.
It is also important for instrumentation to be compact,
quick to set up, and adaptable to different devices and
measurement problems.
Traditional measurement solutions do not fulfill these
new requirements. Big expensive test systems, for example,

are well suited to dc parametric tests and provide high

production throughput, but for low-volume engineering
applications, they are seldom readily available, and most
are too slow for testing fast devices at speed. The lack of a
live keyboard and difficult operating and programming
procedures make such systems inflexible and prevent the
quick access necessary in a laboratory environment.
Another traditional solution is the self-configured, dedi
cated setup containing several word/pulse generators for
stimulus, and logic analyzers, counters, and oscilloscopes
for analysis. Programming these different instruments re
quires long familiarization, and synchronizing the setup is
difficult and time-consuming.
The New Solution

The new HP 8180A/81A/82A Stimulus/Response System

(Fig. 1) provides an answer to the new requirements. It
offers both device stimulation and response analysis, and
consists of the 8180A Data Generator, the 8181A Extender,

Fig. 1 . The 81 80 A Data Generator

(top left), the 81 81 A Extender (not
shown), and the 8182A Data
Analyzer (bottom left) provide
high-performance, easy-to-use,
economical at-speed testing for
digital integrated circuits, boards,
and modules. The system can be
controlled by a desktop computer
(right) via the HP-IB (IEEE 488).

JULY 1983 HEWLETT-PACKARD JOURNALS

Copr. 1949-1998 Hewlett-Packard Co.

Parametric Characterization of
Digital Circuits
The representative timing parameters of synchronous ICs are
setup time, hold time, and propagation delays. During the setup
and hold times, data has to be stable before or after an active
clock be occurs. The propagation delay is the time be
tween stimulation and the output reaction of a device. This value
depends upon the complexity of the circuit and its technology,
and varies from less than 1 ns for an ECL gate to over 1 00 ns for an
MOS gate. Hence, the measuring instrument must have high
timing resolution. In addition, these timing parameters depend on
the dynamic input characteristic of a device. For example, the
reaction time of a comparator depends on the input overdrive.
At the output of the device under test, the variation of amplitude
with deliver is of interest. A driver may have the ability to deliver
adequate dc current under the specified fanout conditions, but be
unable to deliver adequate peak current to drive the parasitic
capacitance at high clock rates.
Applications of the New System
The 8180A/81 A/82A System is oriented predominantly towards
the solving of engineering measurement problems. 1C and board
designers need to measure parameters to specify the perfor
mance rate optimize the yield of ICs, or achieve a high turn-on rate
of boards. Digital 1C manufacturers do parametric characteriza
tion of ICs in the prototype phase, allowing the process to be
optimized. They also use parametric analysis to establish sales
specifications and for in-depth evaluation in the quality assurance
department.
The 8180A/81A/82A System's main applications for digital cir
cuit, module, and system manufacturers are found in R&D, pro
duction engineering, quality assurance, incoming inspection, and
materials engineering. For example, various propagation delays
must be matched for minimum skew. Logic functions must be
verified. A stimulus is required to simulate the interface pattern of
a board that is not yet built. Parametric circuit test at the board
level has the same requirements and objectives as at the 1C level,
differing in complexity and error sources. The cross talk on a bus
with around lines or data stability during the time window around
the system's synchronous clock are of interest. Other applications
include low-volume and at-speed testing in production, incoming
inspection, and depot maintenance, where traditional equipment
such perfor board testers offers insufficient speed and timing perfor
mance. High-reliability applications, such as aerospace equip
ment, also demand a 100% parametric test in addition to detec
tion ap soldering and component loading errors. Component ap
proval and production engineering departments are also applica
tion areas for the new system.

which expands the number of stimulus channels, and the

8 182 A Data Analyzer. Each of the units can be configured
for different numbers of channels. In addition, provision
has been made for the synchronous operation of two sys
tems in parallel. This expands the channel count with a
minimum of skew. Generator/analyzer synchronization and
interaction with the device under test are assured by vari
ous control inputs and outputs. Adaptability and clean
signals are provided by a range of useful accessories.

Although the instruments are designed as a system and

offer the same key specifications, the 8180A (with or with
out the 8181A) and the 8182A can also be operated indi
vidually. Each unit has its own control processor, display,
keyboard, and HP-IB (IEEE 488) interface. This modular
concept makes the instruments very adaptable to different
devices and measurement problems.
This compact system brings the performance of a big
tester to the bench. Its user-oriented softkey operating con
cept speeds the setup of timing and level parameters, either
manually or under HP-IB control. The same is true for test
pattern creation, which is further simplified by standard
data patterns and pattern editing features.
Modularity gives the user the means to adapt a system
exactly to requirements, thus keeping the cost low enough
to justify the expense. This is important in low-volume
testing, where the return from high throughput found in
production areas is not achieved.
The need for high speed is satisfied by the system's
50-MHz data rate. This is faster than most of the big 1C
testers and covers most applications, bearing in mind that
the overall speed of a device is much slower than a single
gate delay. The rate slows with increasing complexity, so
that even fast ECL devices can be tested by the system. In
addition to a fast rate, sophisticated timing capabilities are
provided for setting up various test patterns for an 1C. The
stimulus channels provide independent delay and data
width programming, with excellent resolution of 100 ps for
very tight adjustments. Operation of the 8182A Data
Analyzer is mostly synchronous, a technique similar to that
employed by more expensive 1C test systems . It is capable of
both synchronous (external clock) and asynchronous (in
ternal clock) operation, and it also offers 100-ps resolution,
never previously achieved in an asynchronous analyzer.
When the device under test and the analyzer are syn
chronized, automatic comparison of sampled data with ex
pected data is possible. The 8182A's threshold level range
is sufficiently wide to test all common logic families, and
logic levels are independently adjustable for mixed logic
circuits. The level resolution is 10 mV.
Data Generator Provides Stimulus

The 8180A Data Generator outputs a digital data pattern

according to the user's specifications, which may include
data rate, bits per word, number of data words, levels, tim
ing relationships between channels, and other parameters.
The data generator contains the following basic blocks:
Timing circuits to achieve specific timing relationships
between channels
Memory to hold the data pattern that is to be output in
parallel
Output amplifiers that determine the output voltage
swings
Control inputs and outputs
Central processor unit
HP-IB interface
Keyboard and display.
An add-on unit and pods provide for three-state opera
tion with programmable levels for testing bidirectional
buses.
The 8180A is designed to be the stimulus for parametric

4 HEWLETT-PACKARD JOURNAL JULY 1983

Copr. 1949-1998 Hewlett-Packard Co.

hardware analysis. Therefore, it provides precise level and

timing capabilities at high speed. This is in contrast to other
types of word generators, which often provide data patterns
only at low speed, with moderate or no level definition and
poor timing capabilities.
The 8180A can be configured as an 8-to-16-channel
generator in 4-channel increments. A maximum of eight
channels can be R Z (return to zero), that is, having inde
pendently programmable delay and width. All other chan
nels are NRZ (non return to zero) channels. The channel
count can be expanded by using up to two 8181A Extend
ers, each with a maximum of 24 NRZ channels with
programmable common delay. This configuration provides
64 channels. Synchronous parallel operation doubles this
number to 128.
Data Analyzer Measures Response

The 8 1 82 A Data Analyzer collects , analyzes , and displays

digital bit patterns. These tasks are performed according to
the user's specifications, which may include input
thresholds, data qualification, sampling point (the point in
time at which data is sampled, referenced to the external
clock), compare time window, and other parameters. The
8182A contains the following basic blocks:
Input comparators to determine the states of logic signals
(1, 0, or between high and low thresholds)
Trigger circuits to indicate the beginning or end of data
capture
High-speed memory to store sampled data
Low-speed memory to hold reference data
Timing circuits to determine precisely when data is to be
captured or compared
Control inputs and outputs

Central processor unit

HP-IB interface
Keyboard and display.
The 8182A Data Analyzer is designed for high-speed
functional and parametric hardware analysis of digital cir
cuits. The data analyzer can be configured from 8 to 32
channels in 8-channel increments, each channel having
identical analyzing capabilities. Synchronous parallel
operation doubles the number of channels to a maximum of
64.
User-Friendly Operation

The 8180A Data Generator and 8182A Data Analyzer are

designed to fulfill the requirements of digital circuit
hardware analysis. For this reason the usual data generator
and logic analyzer concepts have been replaced by a new
operating concept tailored to this application. The design
goals for this new concept were:
Fast, guided setup of the instrument, thereby dispensing
with the operating manual in a relatively short time
a Parameter changing during the instrument's active state
(live keyboard)
Comfortable data loading, edit, and display features
Simple HP-IB programmability for fast program genera
tion
Similar operation of both instruments.
To reduce the number of keys and the confusing use of
shift keys, a layered menu softkey concept was chosen. The
display is divided into three areas (see Fig . 2) . The top of the
screen next to the OPERATION keys gives status informa
tion. (These keys operate independently of all other keys on
the front panel). The remainder of the display is split into a
report area, showing one of seven selectable reports

Fig. 2. The 81 80A Data Generator

andthe8182A Data Analyzer have
similar operating concepts, dis
plays, and front panels. The dis
play is divided into status, report,
and softkey areas.

JULY 1983 HEWLETT-PACKARD JOURNALS

Copr. 1949-1998 Hewlett-Packard Co.

(parameter settings or data display), and the softkey area

containing the entry field, which shows the selected
parameter and the labels for the eight softkeys beside the
screen. Softkey and report areas can be selected indepen
dently or in common by the CALL keys. If the entry field is
not activated, the area can be used to extend the data display
in the report area. The parameters are selected by the
softkeys either directly (e.g., softkey label Cycle Mode to
access the cycle mode), or through a softkey tree (e.g.,
softkeys Trigger and Delay lead to trigger delay). After selec
tion, the parameters can be changed either by softkeys or by
data keys. Any softkeys that contribute nothing to the cur
rent operational status of the instrument are blanked to
avoid confusing the user.
This operating concept offers quick access to every
parameter, regardless of the contents of the report area.
Similar Operation

The 8180A Data Generator and 8182A Data Analyzer are

designed to operate similarly, so that a user who is familiar
with one instrument can use the other without a long
familiarization time. The keyboards differ in only three
keys. Softkeys with the same function for each instrument
are labeled identically and occupy the same position on
both keyboards (many softkey label sets on both instruments
are identical). The pages displayed are very similar, too,
with main settings shown on the control page. The timing
parameters, which appear on the analyzer's control page,
require a page of their own in the generator. The generator's
output page corresponds to the analyzer's input page in the
same way the 8180A output data page corresponds to the
8182A expected data page.
Both instruments have a miscellaneous page, which in
the 8182A also contains store and recall functions. A spe
cial feature of the generator is the macro data page, which
extends considerably the editing capabilities of the instru
ment. The data analyzer has three pages on which to display
received data in the form of the state list, the timing diag
ram, and an entirely new feature called the error map. The
8 1 SEA
Clock

error map gives a concise overview of the Compare Data

result. It displays one dot for each memory location. This
dot is replaced by a square if data received differs from
expected data. Furthermore, there are special features for
the individual masking of any bit or word (see article,
page 14).
The expected data page, the state list, and the data page in
the 8182A are user-defined. Channels can be deleted from
the display or grouped together in segments of up to 16. The
channels in each segment can be configured in a manner
that is totally independent of their hardware order, en
abling the user to assign a code to every segment. It is also
possible to display a channel in more than one segment
concurrently. The timing diagram offers a full set of fea
tures, such as zooming in both horizontal and vertical axes,
glitch display, delta measurement, and individual channel
arranging.
Special attention was paid to the editing features. The
8180A has memory, line, and channel editing features.
Memory editing consists of set and clear functions. Channel
editing provides clear, set, counter, PRBS (pseudorandom
binary sequence), and copy. Line editing functions include
insert, delete, copy, and move. It is also possible to define a
set of 16 words (macros) on a special page, and then copy
one or more of them to the data memory.
In addition to many of the above features, the 8182A has
other analyzer- oriente d editing features for expected data.
These include the setting and clearing of masks, and the
ability to copy received data to the expected data memory
for future comparison.
Live Keyboard

For fast parametric measurement it is often. necessary to

change instrument parameters during the measurement
phase. In the 8180A this is achieved by changing all analog
parameters without disturbing the generator output in any
way. If other changes (e.g., output data) are being made, the
instrument interrupts the cycle to make the change, and
then returns to the point at which the cycle was interrupted.

S t a t u s I B L E
Stored Wor d 040
T I M I N G
D I A G R A M S

Display Glitches

Window

U h a n n e I U u r s Address 0084

Fig. 3. When changing mea

surement parameters during a
measurement, the display can be
split into two areas, one of which
shows control page parameters in
the softkey area. Here the timing
page is displayed while the probe
threshold is being changed.

ENTER MEW VALUE

Label A

6 HEWLETT-PACKARD JOURNAL JULY 1983

Copr. 1949-1998 Hewlett-Packard Co.

This operation is performed without disturbing clocks or

data. Therefore, the device under test, which contains only
combinational and sequential logic, does not recognize the
interruption.
The 8182A Data Analyzer's input thresholds can also be
changed during the active cycle (data capture). Other
analyzer features facilitate fast measurements. The mi
croprocessor displays the received data not only when the
instrument stops, but also when the data rate falls below 10
MHz. This is a particularly useful attribute when dealing
with slow clock rates or clock bursts. Other instruments
usually have to abort a measurement to display the data
received. Another useful feature is auto-arming. This al
lows the user to preselect the point in time at which the
instrument will begin a new measurement, so that the result
of any change in the operating condition of the device
under test can be displayed immediately.
To ensure that the instrument is operable during a period
of processor activity, the keyboard is interrupt-controlled,
permitting the storage of up to eight keystrokes until the
processor has time to respond to them. When changing
measurement parameters during a measurement phase, it is
possible to split the display screen into two areas, one
showing the report area (perhaps the error map), and the
other showing control page parameters in the softkey area.
Fig. 3 shows an example; the timing page is displayed while
the probe threshold is being changed.

The HP-IB system is designed to give direct access to

parameters and a fast instrument response. The mnemonics
are easy to remember, avoiding the need to refer to the
manual continuously, making programming simple and
reducing familiarization time. To program the 8180A to a
frequency of 20 MHz, for example, the controller sends the
message FRQ 20 MHZ. Where the parameters on each in
strument are identical, so are the mnemonics.
For the transfer of high volumes of data from the control
ler to the instrument and vice- versa, a binary transfer mode
has been implemented. In this mode, every byte on the bus
is dedicated to eight bits of data. Data transfer rates of
12,000 bytes per second are achieved, giving a data reload
time of less than two seconds for the whole system. This
makes it possible to write fast-running programs.
Acknowledgments

We would like to thank Roberto Mottola for his outstand

ing contribution in defining the operating concept for
both instruments, Horst Link, who was responsible for
the mechanical design of all instruments and accessories,
Eckhard Hettlage, who designed the displays, Peter Stollenmaier, who contributed to the 8182A power supply,
Rainer Storn, who helped with the 8182A self-test software,
and last but not least, our thanks go to Gnter Riebesell and
Reinhard Falke for their help and guidance along the way.

A High-Speed Data Generator for

Digital Testing
by Ulrich Hbner, Werner Berkel, Heinz Nssle, and Josef Becker

THE NEW HP 8180A DATA GENERATOR represents

a significant step forward in the production of high
speed data generators. It offers high timing accuracy,
precise pulse-level definition, ease of operation with either
manual or remote control , and great flexibility on the bench
and in the rack.
The 8180A consists of a power supply module, a display
module, and eleven printed circuit boards, which plug into
a motherboard. Fig. 1 illustrates the various functional
blocks and main interconnections.
The 81 81 A Data Generator Extender is constructed along
the same lines as the 8180A with the exception of the
display and keyboard. All control functions are performed
by the 8180A, the 8181A being merely a system slave. Fig. 2
is a block diagram of the 8181A.
Internal System Organization

All high-speed control signals in the 8180A Data

Generator are generated on two boards, designated Address

Control 1 and Address Control 2 (ADCl and ADC2). The

logic consists of 10k and 100k ECL devices and is controlled
by the central processing unit by way of a high-speed
bidirectional bus.
On ADCl, three signals that synchronize the address
counters and signals for both clock channels are generated.
The board responds to the commands RUN, GATED, STOP,
and BREAK. There are three basic operating modes:
Single cycle: a RUN signal starts a single cycle from the
preset first address to the preset last address.
Auto cycle: after a RUN signal, data is generated repeti
tively from the first to the last addresses.
Initialization and auto cycle: a RUN signal starts data
generation at address 0000. Data generation continues to
the first address and then cycles between the first and last
addresses repetitively.
In these three operating modes, cycling is halted on receipt
of a BREAK or STOP signal.
Two further modes are controlled from this board:
JULY 1983 HEWLETT-PACKARD JOURNAL 7

Copr. 1949-1998 Hewlett-Packard Co.

Address Bus

Data Bus

Microprocessor Board
32K ROM
HP-IB Interface
Device Interface
Keyboard

Device Interface
Busto 81 81 A

Display Assembly
Control Logic
CRT Drive
CRT

Video Out

GEM Board
General Module
Interface
RAM, DACs
External
Clock
ADC1 -Address Control 1
Rate Generator
Control Logic
RUN

BREAK,
RUN, GATED

Sync Board
Sync Signals
Rates
Input Circuits
Amplifier
CLOCK 0
CLOCK CHANNEL 1,
LOAD

CLOCK CHANNEL 2
CLOCK CHANNEL 1

BREAK,
STOP

ADC2-Address Control 2
Strobe Channel
High-Speed Control
Circuits
Address Counter

Timing Board
Rate ICs
Sync Signals

Sync Master 2

Strobe Channel |
t BREAK, STOP

High-Speed Sync
to 81 81 A

Clocks
(CLK1, CLK2, CLK3)

Module Board
ECL RAM
Latches
Output Amplifiers

4 Data Channels
I

-To other Modules -

Fig. 1. Block diagram of the 81 80 A Data Generator.

by the microprocessor via a 10-bit DAC (digital-to-analog

converter). Decade counters for ranging are built into the
chip. The Rate 1C generates the system clock in both inter
nal and external clock modes.
Another important feature of ADCl is the address differ
ence counter. This 10-bit presettable counter operates con
tinually in the count-down mode, counting clock pulses.
Each clock pulse corresponds to an address. Whenever the
LOAD signal goes low (Fig. 3], indicating a counter state of
all zeros, the counter is preset to the difference between the
first and last addresses programmed by the user. LOAD then
goes high and the counter counts down until it again
reaches a state of all zeros. LOAD then goes low and the
cycle repeats.
A break at a defined address is performed by the strobe
counter. This device is loaded by the microprocessor with
the break address minus the first address.
The internal clock control generates two high-speed sig
nals, CLKO and CLOCK CHANNEL 1. These and the LOAD
signal control and synchronize all other high-speed circuits
in the 8180A and 8181A. By bypassing these signals, it is
possible to set up a master-master system with two 8180As
operating in parallel. In this case, only one instrument is
active. The LOAD, CLKO, and CLOCK CHANNEL 1 signals are
fed via a rear connector directly into the second 8180A.
Bypassing is achieved automatically when the intercon
necting cable is fitted.
Fig. 4 illustrates the principal signal paths on the ADC2
board. Signals CLKO, CLOCK CHANNEL 1, and LOAD are fed
from ADCl to ADC2 on twisted-pair lines by line drivers.
Line receivers pick up these differentially transmitted sig
nals, refresh them, and drive the delay lines. Another line
driver generates the differential signals for the 8181A Ex
tender. The CLKO and LOAD signals are fed to the 81 80 A' s
address counter delayed by 10 ns to compensate for the
delay introduced by the interconnecting cable to the 8181 A.
The address counters support all data channels in the
8180A and 8181A. Several delay lines are required to
achieve correct timing relationships with the module
boards that carry the data channels (four channels per
8180A Device
Interface Bus

m Gated cycle: a gate signal at the external RUN input causes

data generation for as long as the input remains true.
If the input goes false, the currently active cycle is
completed.
Initialization and gated cycle: operation is similar to the
initialization and auto cycle mode. The currently active
cycle is always completed when the gate input goes false.
In either mode, cycling ceases on receipt of a STOP signal.
Fig. 3 is a functional block diagram of ADCl. The basis of
all internal timing circuits is the Rate 1C. This specially
developed device is used in recent HP pulse, function, and
data generators.1 Features of the Rate 1C include a pro
grammable oscillator with a range of 1 Hz to 100 MHz,
programmable delay with a range of one second to 10 ns,
and resolution of 100 ps or three digits when programmed
8 HEWLETT-PACKARD JOURNAL JULY 1983

Copr. 1949-1998 Hewlett-Packard Co.

8180A High-Speed
Sync Signals

Clocks

To Other Module Boards

Fig. 2. Block diagram of the 81 81 A Data Generator Extender.

CLOCK CHANNEL 1

External
Clock
Input

Rate 1C
Clock Generator

To ADC2

Microprocessor
Signals

Load Signal

FAD
Set

board). One delay line generates the signal CLKi, which

guarantees sufficient setup time for the address latch.
Fig. 5 shows worst-case conditions for high-speed RAM
operation.
CLK2, delayed by 20 ns, latches the RAM's output data.
This data then passes to the output latch after a further 10-ns
delay, and finally to the XOR gate at the output amplifier
input. The extra delay is necessary for generating R Z data
when the instrument has been fitted with an additional
timing board. Data from the data latch is fed to the reset
flip-flop FFl where, dependent upon delay and width set
tings, it is clocked through to the Q output. The negativegoing edge resets the flip-flop in the RZ mode via the OR
gate, and the XOR gate at the input to the output amplifier
switches between normal and complementary operation.
In the 8181A the timing circuits on the ADC3 board are
very similar to those described above. The main difference
is in the programmable delay, which makes it possible to
compensate for delay variations between the 8181A NRZ
and 8180A data channels. The programmable delay is also
needed for setting up address and data buses that differ in
timing.
Timing Board

The timing board consists of four delay and width

generator channels. At the heart of each channel is the Rate
1C,1 which in this application is configured to function in its
timing mode. The use of conventional devices instead of
Rate ICs would have taken about five times the space to
achieve the same objective.
Fig. 6 shows the basic blocks of a timing channel. Inac
curacies in analog voltages generated by the DACs and
peripheral components necessitate a number of control cur
rent adjustments to achieve an optimal overlap of ranges

FAD

Set

Fig. 3. The Address Control 1 cir

cuit generates synchronizing sig
nals for the various 8780/4 operat
ing modes.

and the matching of clock inputs. The control for ranging

and current programming comes from microprocessor
commands sent along the bus. The function of the reset
flip-flop is described in the preceding section. 32 adjust
ments are necessary to set up a timing board for operation,
eight for each channel. Repeatability and stability results
are excellent because of noncritical design and well speci
fied components.
Module Board

The module board consists of four data channels im

plemented with high-speed ECL RAMs, latches for timing,
and an output amplifier for each channel. With the excep
tion of the output amplifiers, no setup or adjustments are
necessary on this board. This facilitates the addition of extra
boards to the 8180A or 8181A with a minimum of incon
venience. The layout of each functionally identical channel
is designed in such a way that compensation for skew,
introduced by the different cable lengths attached to the
rear connector, is achieved by striplines on the board.
A module board complete with cables and rear connector
can be fitted or replaced merely by removing two screws
from the instrument's rear panel and taking off the cover.
Each channel on the module board is individually screened
to avoid EMI problems from neighboring channels, and to
provide a heat sink for the transistors in its output amplifier.
Sync Board
The sync board consists of two clock channels, CLOCK
CHANNEL 1 and CLOCK CHANNEL 2. Both channels have
DACs for delay and width programming and use the same
output amplifiers as the module boards. For applications
that require refresh capability, CLOCK CHANNEL 1 is kept

JULY 1983 HEWLETT-PACKARD JOURNALS

Copr. 1949-1998 Hewlett-Packard Co.

Programmable
Delay Circuit

Normal/Complement

Output

Output Amplifier
Module Board 4 Channels/Board

Address Control 3 81 81 A
Twisted '
Pairs

C o m p e n s a t e s
Cable Delay

1 0

n s

Normal/Complement (N/C)

Output

Amplifier Output
Address Control 2

Module Board

Note: Strobe channel not shown.

Microprocessor

Timing Board

Fig. 4. 8180A/8181A timing and signal distribution.

U 1 ns
CLK1

Address Latch Output

xx xx :

( XX XX
RAM Address Valid -

(xxxxxxxx::

RAM Access Time

(XXXXX

-tAminRAM Data
Valid
Window for
Flip-Flop Clocking

Flip-Flop Hold Time Fi9' 5' The delayed Clock S'9nal

-Delay T2

10 HEWLETT-PACKARD JOURNAL JULY 1983

Copr. 1949-1998 Hewlett-Packard Co.

CLK1 guarantees sufficient setup

time for the address latch under
worst-case high-speed operating
conditions.

Microprocessor Bus

RZNRZ

Range
Control
Delay
Generator o
Rate I

Delay

Width
> Generator o
Rate II

' Flip-Flop Q
Module Board

Width

Fig. 6. The timing board consists

of four delay and width generator
channels (one shown).

running in STOP or BREAK; hence it is handled differently

from CLOCK CHANNEL 2.
Output Amplifier

The final circuit in every data and clock channel is the

output amplifier. It is here that timing information (highlow transitions) and dc voltages to program the high and
low levels are translated into the desired pulse stream at the
output. The functional diagram, Fig. 7, shows the
amplifier's four major blocks. The first two form a pulse
generator, and the last two a linear power amplifier.
Timing information in the form of switching commands
is applied in a symmetrical ECL logic format at inputs Q and
Q. The signal is amplified by symmetrical complementary
differential amplifiers, which control current sources in a
cascade arrangement. The outputs of the four switched
current generators are combined to form a pair of alternat
ing bidirectional current sources^ which drive a diode
clamping bridge at nodes Vc_and Vc.
The clamp voltages Vc and Vc are held equal to either the
upper or lower programmed voltage by unity-gain buffers
Ula and Ulb. In the Q=high state, the symbolic DPDT
switch takes the position shown in Fig. 7. Positive current
flows through D4 into the lower voltage follower Ulb, and
negative current through D 1 into the upper voltage follower

WVH HyA-f^ArVf
dc Control

.A.
Fig. 8. Upper branch of the linear broadband amplifier shown
in Fig. 7.

Programmed Level 1/2VLo

Current Switches
(complementary
push-pull cascode)

Clamping
Pulse
Generator

Programmed Level '/2VH

Linear Broadband
Amplifier
(dc to 200 MHz)

dc Offset and Gain

Control

dc ErrorCorrecting Current ID

Fig. translates information amplifier block diagram. The output amplifier translates timing information and dc
voltages into the desired pulse stream at the output pods. The first two blocks form a pulse
generator and the last two a linear power amplifier.

JULY 1983 HEWLETT-PACKARD JOURNAL 11

Copr. 1949-1998 Hewlett-Packard Co.

Ula. As long as Vra>VLO with at least 0.5V difference, D3

and D2 may be regarded as switched off.
Two more diodes, D5 and D6, compensate for the on
voltages and temperature dependence of D3 and D4. All
diodes are supplied with equal currents (Ic=15 mA). Minor
differences in forward voltage drop and series resistance
of the matched pairs D3, D5 and D4, D6 are corrected by
trimmers. Thus Vc is equal to exactly V^V^ when D4 is con
ducting (Q=high), and V2VLO when D3 is conducting. A
small error because of junction temperature differences re
mains in D 3 and D4 when they are switched on or off.
Schottky-barrier diodes are chosen for fast switching with
out storage delay.
The second clamping node, Vc, provides a complemen
tary pulse train to Vc. This is used for two dynamic pur
poses. To accelerate the high-low transitions of Vc, Vc is
high-pass coupled to the bases of switched current-source
transistors. The result of this additional drive is that the
voltage transitions of Vc are more like pulses with constant
rise time than the natural constant-slope behavior of a
current-steered voltage clamp.
The second purpose of the Vc branch is to provide con
stant current loading for voltage-followers Ula and Ulb. If
Dl and D2 were not present, these op-amps would be re
quired to deliver pulsed output currents. Frequency re
sponse limitations within the op-amps would lead to dutycycle-dependent level variations in some critical highfrequency regions. The symmetrical clamping bridge draws
dc current from Ula and Ulb, keeping pulse tops flat from
dc to the nanosecond region.
The clamp output signal Vc is amplified to the desired
amplitude and power by a class-A push-pull amplifier. In
Fig. 8, only the upper half of the amplifier, with compo
nents determining gain A0 and output impedance Z0, is
shown. The complementary symmetrical lower branch
completes the amplifier for dc operation, and acts as a
doubler for overall transconductance and output current.
This basic arrangement is noninverting.
Minor aberrations and tolerances are trimmed out with

Table I
Data Generator Output Amplifier Performance
Programmed load from SOU into 5011 from SOU into high
condition (voltages double if impedance ( -10 kil)
no load connected)
High level range

-1.5V to +5.5V

-1.0V to +17.0V

Low level range

-2.0V to +5.0V

-2.0V to +16. 0V

Resolution

3 digits (best case

10 mV)

3 digits (best case

20 mV)

Amplitude range

0.5V to 5.5V

1.0V to 17.0V

Level accuracy after

20 ns settling time 40 ns settling time

0.5% of level
3% of amplitude
30 mV (add
30 mV

for amplitudes
>0.7V)
Level accuracy after

Transition time
Typical transition
time for ECL levels
(20%-80%)
Preshoot, Over
shoot, Ringing

0.5% of level
3% of amplitude
30 mV (add
30 mV
for amplitudes

1 ms settling time

1 ms settling time

0.5% of level
30mV
(add 30 mV for
amplitudes <0.7V)

0.5% of level
60 mV
(add 60 mV for
amplitudes <1.5V)

<3.0 ns +
<3.0 ns +
| amplitude j xo. 2 ns amplitude! X0-5 ns
1.5 ns

<10% of amplitude <10% of amplitude

P4, for the amplifier unloaded, and again with P5, after
connection of a 50 fi load.
Output amplifiers commonly found in pulse and data
generators are usually designed as current or voltage
sources, both needing an internal 50 load to present the

* * 1 I I t 1 I '

Fig. 9. The tri-state pod makes

the data generator outputs com
patible with bidirectional bus and
I/O pin configurations. The control
and power unit on top of the 81 80A
Data Generator is part of the pod
system.

12 HEWLETT-PACKARD JOURNAL JULY 1983

Copr. 1949-1998 Hewlett-Packard Co.

correct Z0 (50fi) at their output. The amplifier concept used

here gives well-defined gain and output impedance with
out wasting 50% of the ac power on an internal 50ft ballast.
This is important in generators in which many data chan
nels are packed closely together, like the 8180A and 8181 A.
Internal feedback in even' amplifier stage (RE,RF) helps
ensure stability and allows the use of full circuit speed
(speed is ultimately limited by the propagation delay
around the feedback loop in the amplifier network). The
free-running amplifier produces an output signal with the
specified accuracy. A dc offset of up to several hundred
millivolts reflecting the thermal history of linearly operat
ing transistors may appear, however. A dc error can also be
introduced by varying the supply voltages to accommodate
different output amplitude requirements. To eliminate
these offsets, an additional dc control loop checks the out
put voltage and the output current simultaneously and en
sures precise operation under any loading, level, or supply
conditions.
To understand the operation of this dc control loop, look
at the inputs of error amplifier U 2 in Fig. 7. At the noninverting input, a voltage V+ =V2(V0+RCSI0) is compared
with V~=y2Vc+3/8V0 at the inverting input. The op-amp
forces a corrective current into the inverting input of the
main amplifier to keep V + -V~=0. The dc precision now
depends on the op-amp quality and the ratio of the resistors
in the voltage divider. When V + -V~=0, then V0+RCSI0 =
Vc+3/4V0. Since A0=4 and RCg=50il/A0, this expression

becomes V4V0+(50il/A0)I0=Vc, or V0=A0Vc-(50n)I0,

which describes a linear amplifier with voltage gain A0 and
an output impedance of 50O.
Table I shows the principal specifications of the output
amplifier.
Tri-State Pod

The tri-state pod is an accessory to the 8180A Data

Generator and 8181A Extender. Its function is to make the
generator outputs compatible with bidirectional bus and
I/O pin configurations.
The main design aims for the tri-state pod were to be able
to output any level in the generator's output range of -2V to
+ 5. 5V, and to function as an analog switch, enabling the
control of different logic families. At the same time, the
level programmability of the system had to be maintained.
Other objectives were fast switching, high output imped
ance with low capacitance in the off state, and low output
impedance with high drive capabilities in the on state,
giving high speed when driving high-capacitance loads.
The solution adopted uses the familiar diode bridge (Fig.
9). A disadvantage is that the power consumption of the
bridge reduces the ratio of output to input currents signifi
cantly. This problem was solved by a complementary trans
istor stage that reduces the bridge supply current.
In the left-hand side of the bridge, two pairs of series
diodes produce a drop of approximately 1.4V. This is al
ways greater than the drop across the Schottky diode and

Signal
Input

H-HTAVVH*

Fig. 10. Schematic diagram of

the tri-state driver.

JULY 1983 HEWLETT-PACKARD JOURNAL 13

Copr. 1949-1998 Hewlett-Packard Co.

the emitter junction of the output transistor in each line.

The series Schottky-diode-resistor configuration, in com
bination with an offset adjustment, permits the use of un
matched components without a reduction in level accu
racy. When the enable voltage meets or rises above the
-4.30V threshold, transistors Q5 and Q6 begin to conduct,
thereby turning Ql and Q2 off. In this way, the bridge

becomes reverse-biased, with the clamping diodes ensur

ing that the output circuit is cut off when the voltage at the
output is in the range -2V to +5. 5V.
Reference
1. C. for et al, "Designing Bipolar Integrated Circuits for a
Pulse/Function Generator Family," Hewlett-Packard Journal, Vol.
34, no. 6, June 1983.

High-Speed Data Analyzer Tests Threshold

and Timing Parameters
by Dieter Kible, Bernhard Roth, Martin Dietze, and Ulrich Schttmer
CONVENTIONAL DATA ANALYZERS fall generally
into two categories. First, there are the high-speed
asynchronous timing analyzers, which produce test
results similar to an oscilloscope on several channels.
These analyzers are used mainly for functional analysis of
random logic. Second, there are the synchronous state
analyzers , which are used mostly for the functional analysis
of microprocessor-based systems with real-time or other
complex sequential logic. Parametric measurements, on the
other hand, have to be made with oscilloscopes or timing
analyzers having fairly coarse time resolution (the fastest
timing analyzers available offer only 1.5 to 2 ns resolution,
with a significant additional error because of skew between
data channels).
The new 8182A Data Analyzer is optimized for testing
threshold and timing parameters at high speeds. It has two
innovative features programmable sampling-point delay
and real-time compare mode. New, specially-developed ac-

Compare

tive probes with a wide range of accessories facilitate the

capture of data with minimum influence on the device
under test. All analog parameters, such as threshold volt
ages and the sampling point delay, are implemented with
high resolution and accuracy. Instead of setup and hold
times, a sampling time accuracy with respect to the external
clock input is specified. This is more suitable for parametric
measurements of integrated circuits and printed circuit
boards.
The 8182A is controlled by a 6809 processor with 40K
bytes of ROM space and a 14K CMOS RAM that is batterypowered for nonvolatile data storage. All I/O interfacing is
memory-mapped to achieve fast access to all parameter
registers.
Three Operating Modes

The 8182 A has two logic analyzer modes of operation and

a comparator mode. In the trigger start analyzer mode, after
To Error Map
in Trigger Start
Compare Mode

To State and Timing

Diagrams in Trigger
Start/Stop
Analyzer Modes

External Arm
External Stop
Trigger Qualifier

14 HEWLETT-PACKARD JOURNAL JULY 1983

Copr. 1949-1998 Hewlett-Packard Co.

Fig. 1. Simplified block diagram

of the high-speed circuits of the
87 82 A Data Analyzer. New fea
tures are a programmable master
clock delay and a real-time win
dow comparison capability.

recognizing a trigger condition, the 8182A begins to accept

data into its high-speed memory. Stored data can be dis
played as either state or timing diagrams, during or after
completion of the measurement cycle. The microprocessor
compares the received data with stored expected data and
displays comparison failures on the state diagram or a con
centrated overview in the error map.
In the trigger stop analyzer mode, after arming , the 8 182 A
is always active, that is, it continually accepts data, filling
and overwriting its high-speed memory . After recognition
of a trigger event, it stops accepting data (after a selectable
delay) and displays the previously stored data. Hence the
events preceding and following the trigger event can be
recorded. The monitoring capabilities are the same as in the
trigger start analyzer mode.
The third mode, the trigger start compare mode, is an
entirely new feature. It permits a real-time comparison of
incoming data with a prestored data pattern contained in
the high-speed memory. This comparison takes place dur
ing a time window synchronous with the clock input. The
window is defined by a delay and a width. Every failure
during the compare window generates a high logic level at a
rear-panel output, and an error report can be displayed on
the error map. Either one-shot or continuous cyclical mea
surements are possible.
Hardware Architecture

With the exception of the power supply, the display, and

the microprocessor section, the 8182A hardware is to a high
degree implemented with ECL to achieve 50-MHz opera
tion and high timing accuracy. The eight-channel version
of the 8182A contains two boards with digital control cir
cuits, one board of clock generators, two boards containing
the microprocessor and its interfacing , and two data boards,
each with 4 channels. Six more data boards can be plugged
into the remaining slots, extending the 8182A up to 32
channels. Fig. 1 shows a simplified block diagram of the
hardware.

What Is Window Comparison?

When the 8182A Data Analyzer s in either of its logic analyzer
modes, performing like a conventional analyzer, it samples a data
stream on a point-by-point basis. Whatever is present between
two sampling events can be analyzed with the aid of a glitch
detector, but it is impossible to determine whether a glitch ap
peared at the beginning or end of a sampling period. Similarly,
determining the stability of a logic level during any part of the
sampling period s also a problem.

Incoming Data
Compare Result
Stored Data Pattern
CLK1 (Window Pulse)

Fig. 1. In comparator mode, the 8182A Data Analyzer com

pares incoming data with stored data during a time window of
programmable width.
How can the status of data on a microprocessor bus be deter
mined during the period beginning 50 ns before and ending 1 0 ns
after a clock edge?
When the 81 82A s operating in its comparator mode, a periodic
time window can be programmed. During this time, the incoming
data can be compared with an internally stored pattern. The time
window s synchronous with the clock connected to the 8182A
clock probe.
The comparison circuit consists of an XOR gate in each chan
nel, which compares the incoming data with a fixed data pattern.
A pulse of the desired window width (60 ns in the example given
here) is applied to the enable input of the gate (see Fig. 1).
Fig. 2 shows the gate output. It is assumed that the stored data
pattern consists of low levels, so an Incoming high generates a
positive pulse at the gate output while the window pulse s high. In
Fig. 2a and 2b a failure s recognized. Fig. 2c and 2d are border
line. The compare error may not be recognized, although one
<4 ns <4 ns
LK1 (Window
Pulse)
Incoming
Data-S

Compare
Result <
(b)

(c)

Fig. 2. Assuming that the stored data pattern consists of all

low levels, an incoming high level during the window causes a
positive compare result. Failures (a) and (b) are easily recog
nized, while cases (c) and (d) are borderline.
edge of the Incoming data lies within the time window. At the
8182A input, the incoming data edge must be within 4 ns of the
window edge for an error to be diagnosed correctly.
Fig. 2. High-impedance, high-frequency, dual-threshold
probing is provided by a special thick-film hybrid probe
circuit.

-Martin Dietze

JULY 1983 HEWLETT-PACKARD JOURNAL 15

Copr. 1949-1998 Hewlett-Packard Co.

Active Probes and Input Amplifiers

To achieve the design goals of high-impedance, lowcapacitance probing with high-frequency and dualthreshold capabilities, a special active probe circuit was
developed. It is realized as a thick-film hybrid circuit
(Fig. 2).
The dimensions of the hybrid are 29.7 mm x 10 mm x 2.4
mm. Special care had to be taken in designing the highimpedance input stage layout to achieve the small input
capacitance of less than 7 pF (the hybrid alone has only a
4.5-pF input capacitance). The hybrid is packaged in a
small aluminum case and connected to a special 1.5-m
cable consisting of two shielded cables, one of which is 501
coax, and two single wires for power and compensation f or
ground potential differences. Four of these active probes
and cables share a single connector, which can be con
nected to the rear panel of the instrument.
Fig. 3 is a diagram of the active probe and input amplifier
circuit. At the probe input, a spark gap (a laser-cut thickfilm resistor) limits electrostatic discharge to several hun
dred volts. The input information VIN is divided into an ac
path and a dc path. The ac signal is FET-buffered after
frequency-compensated voltage division by 5 and sent to
the analyzer by a line driver with a series termination. At
the analyzer end of the cable, the signal is inverted.
The dc path is a simple op-amp circuit with a gain of
-0.2. The input signal is level-shifted according to the
threshold voltage so the high-speed comparator always
compares its input signal with zero volts. Since the dc input
is brought to the instrument at virtual ground, the input
characteristics of the active probe remain constant with
frequency. The capacitance of the line has no influence, and
cable noise is significantly reduced.

After the incoming signal has been divided by 5 and

shifted according to the programmed threshold, it is
applied to a high-speed comparator, which has a few mil
livolts of hysteresis to suppress noise and high-frequency
oscillation.
The 8 182 A is capable of comparing the incoming data
with two thresholds simultaneously. For this, two
neighboring channels are combined by relays in the follow
ing manner. The threshold comparator of one channel is
disconnected from its input amplifier and connected to the
amplifier of the other channel, so that the second channel
sends its signal to two comparators. The first channel is
then switched off. One comparator is used to compare the
low-level threshold and the other to compare the high-level
threshold. The data from both comparators is stored in two
memory channels before the microprocessor combines
them into one report. Eight channels can always be selected
to perform dual-threshold measurements on four data
inputs.
Data Sampling in the Logic Analyzer Modes

In the trigger start analyzer mode and the trigger stop

analyzer mode, the data being digitized by the comparators
is sampled in the sample flip-flop by the master clock (Fig.
1). The sample flip-flop is the heart of the analyzer. It has
excellent sampling jitter characteristics (50 ps or less) and
low temperature drift. The sampling jitter of a D-type flipflop is the time interval between the specified setup and
hold times where the output is not defined.
Variable delay lines in the data channels are used to
compensate for the clock circuit's propagation delay and
the sampling flip-flop's setup and hold times (more about
this later).
v+

81 82A Data Analyzer

High-Speed
Comparator

Thick-Film Active Probe Circuit

Digital

V-

Fig. 3. Active probe circuit and analyzer input amplifier circuit.

16 HEWLETT-PACKARD JOURNAL JULY 1983

Copr. 1949-1998 Hewlett-Packard Co.

The sampled data is written into the high-speed RAM by

the gated write pulse generated by the trigger unit. The data
can also be compared by the microprocessor with a fixed
data pattern contained in the reference memory. A previ
ously received data pattern contained in the high-speed
RAM can be dumped into the reference memory. The refer
ence memory can also be loaded and modified via the
keyboard.
Interactive Read

A novel feature of the trigger start analyzer mode is the

interactive read circuit. It permits the microprocessor to
read the contents of the high-speed RAM cells during data
sampling without losing data. This means that at a slow
clock rate, the sampled data can be displayed immediately
on the screen (e.g., on the timing diagram) after every sam
ple. Therefore, it is unnecessary to wait until a measure
ment has been completed before seeing the result on the
screen, unlike almost all other logic analyzers.
A special circuit to perform the difficult task of timing the
interaction of the microprocessor and the asynchronous
sampling was developed. A simplified diagram is shown in
Fig. 4. A READ/STORE signal is generated from the master
clock by a retriggerable monostable multivibrator (oneshot) and applied to an address multiplexer. For 50 ns after
each positive transition of the master clock, the outputs of
the high-speed address counter are tied to the RAM address
inputs and a sampled word is stored. The multiplexer then
applies the microprocessor-controlled read address to the
RAM address inputs. If a read request (RDRQ) was initiated
by the microprocessor, the circuit now starts reading the
RAM contents at the read address. RDRQ is stored in FF1.
Data
to Microprocessor

External
Clock
(Active
Probe)

Internal
Clock
Oscillator

Clock
Qualifier
(BNC)
Clock and
Error Recognition

Fig. 5. Clock timing circuits produce the programmable delay

and the programmable width of the master clock.

The ANDing of the Q output of FF1 and READ/STORE gener

ates a read data (RDDA) signal, storing the contents of the
chosen RAM cell in FF3 after an appropriate delay (RAM
access time). If this read activity is not interrupted by the
next master clock, RDDA generates a read acknowledge
(RDAC) signal, which informs the microprocessor that valid
data has been stored in FF3, and FF1 is reset. Now the
microprocessor can read the content of FF3 and initiate
another read request.
If the read process has been interrupted by a new master
clock, which has a higher priority, the multiplexer switches
immediately to the address counter. Sampled data is then
written into the RAM. Meanwhile, the RDDA signal is ter
minated and RDAC is suppressed. After the one-shot's
period has elapsed, the signal RDDA is reinitiated. This
process is repeated until the RAM has been successfully
read.
Real-Time Comparison

Data
Read Address
from
Microprocesso

From
Address Counter

Microprocessor

Master
Clock

Fig. 4. The interactive read circuit lets the microprocessor

read the contents of the high-speed RAM during data sam
pling without losing data.

In the trigger start compare mode, the high-speed RAM is

used to read out expected data in real time for comparison
with incoming data. The same patterns are used in both
logic analyzer and comparator modes. The high-speed
RAM contents and the data from the active probes are sent
simultaneously to the window comparators (see Fig. 1).
There they are compared channel by channel during a time
interval determined by the delay and width of the master
clock. The compare results of all channels are then ORed.
The output of the OR gate is qualified with a status signal
and the word mask signal (discussed later). The result is
immediately available at the rear-panel PULSED ERROR
output. The result of every word comparison is stored in a
special RAM that has one bit for each address. The contents
of this RAM form the error map.
A second error output, LATCHED ERROR, is driven by a
flip-flop that is set by the first error during a comparison
sequence. This output remains high after the end of the
measurement for further processing. It is reset when a new
measurement is started. A character string in the display
labeled Compare Failed/Passed shows the status of the latched
error output.
Channels that have failed are marked on the displayed
error page. To obtain this information, the output of every

JULY 1983 HEWLETT-PACKARD JOURNAL 17

Copr. 1949-1998 Hewlett-Packard Co.

channel's window comparator is processed separately. The

result of any real-time comparison is two-dimensional,
with both faulty words (Y) and faulty channels (X) iden
tified. To analyze any failures that are discovered, the logic
analyzer modes can be used.
Sometimes there is no requirement to compare the results
of all channels or all words. The 8182A is equipped with a
means for disabling individual channels and inhibiting the
compare results for individual words by setting them to
"don't care." Individual words are inhibited by a separate
word mask memory, running in parallel with the high
speed RAM. Its output is gated with the real-time compare
result.

Clock Timing

Clock timing circuits on the clock board produce the

programmable delay and the programmable width of the
master clock. A simplified diagram is shown in Fig. 5. The
active clock CLKO is a combination of the external or inter
nal clock and the clock qualifier input.
The internal clock is generated by a 100-MHz oscillator,
divided down in 1-2-5 steps. CLKO triggers the delay circuit,
which sets the width flip-flop, which is reset by the width
circuit. The flip-flop output is the master clock, CLKi.
Delay and width generation is done mainly with custom
ICs, as it is in the 8180A Data Generator (see Fig. 6). How
ever, in the analyzer, it is of great importance to keep the

Generation of Analog Voltages

A multichannel Instrument with universal features, working with
several logic families, requires the ability to attach various pro
grammable thresholds to different probes. In addition, features
like delay and width, also programmed by analog voltages, are
required. Therefore, many programmable voltages had to be
made is in the 81 82A Data Analyzer. The actual number is
42. These voltages are generated cyclically by a single digitalto-analog converter (DAC) and sent serially to all the places where
they the required. On arrival, sample-and-hold circuits select the
desired level at the desired time (see Fig. 1).
Addresses and logic control signals are generated by the
counter. The RAM, which stores 1 1-bit words to be used in level
generation, delivers information to the DAC for conversion. The dc
levels are then passed to the sample-and-hold circuits, together
with a control signals and address information. Fig. 2 shows a
simplified sample-and-hold circuit. Amplifier 1 charges the
capacitor during the time that the enable signal is low, and
amplifier 2 tracks that voltage. The feedback in the circuit elimi
nates the offset voltage of amplifier 2.

Address

Enable
Amplifier 2

Amplifier 1

120 /

-Bernhard Roth

Fig. 2. Sample-and-hold circuits at various points in the

analyzer select the correct analog voltage at the correct time.

Amplitude

-120
- Time

Clock

Control Logic
and
Address Buffer

To Other Receivers .

Analog Voltage Generator

Sample/
Hold

Voltage
Analog Voltage Receiver

18 HEWLETT-PACKARD JOURNAL JULY 1983

Copr. 1949-1998 Hewlett-Packard Co.

Fig. 1. In the 8182A Data

Analyzer, 42 analog voltages are
generated by time-multiplexing a
single digital-to-analog converter
(DAC).

High-Speed
Delay
Generator

CLKO

Programmed Master ,
Clock Delays 0-6.6 ns

.0-5.9 ns
High-Speed
Delay
Generator

6-21.9 ns

, \

> 12.5ns

To Next Stage

Error Signal
(Clock Is
too fast)

Fig. 6. Custom ICs are used in the delay circuit (shown) and
the similar width circuit. Special high-speed delay generators
minimize the delay through the circuit.

fixed delay through the circuit to a minimum. The reason

for this is that every data channel also has to be delayed to
achieve the specified sampling accuracy. For this reason,
special high-speed delay generators implemented with
100k ECL are used. These have a minimum fixed delay and
a variation of up to 20 ns. A simplified circuit diagram is
shown in Fig. 7.
An incoming edge sets the flip-flop. Output Q opens the
switch, forcing current I to charge capacitor C, creating a
positive-going ramp. I is generated from the signal U by an
analog divider. The instant the threshold voltage is
reached, the flip-flop is reset and C is discharged. The pulse
width at the output of the comparator is determined by the
propagation delay through the components. The fixed
delay is roughly 6 ns and the variable delay is proportional
to U. Current ll is used to improve linearity.
Since the main application of the data analyzer is in
synchronous operation with programmed delay and/or
width, constant observation of the external clock in relation
to the actual delay and width is necessary to ensure that
every active input clock edge creates a correct master clock
pulse. When the time between two consecutive CLKO pulses
is too short to produce two correct master clock pulses, the
operator is informed. This task is accomplished by an error
recognition circuit (see Fig. 8).
Error recognition is performed along two paths. The first
is FF1/FF2, and the second is FF3. The delayed CLKO sets
FFl and samples FFl's output with FF2. The master clock

Clock
from Delay
Circuit Output

Fig. 8. An error recognition circuit signals the operator when

the time between two consecutive active clock pulses is too
short to produce two correct master clock pulses.

resets FFl, so that the whole process starts again. If two

CLKO pulses occur before a master clock edge, the second
CLKO pulse causes FF2's output to go high and hold that
status. The delay circuit output samples the master clock
with FF3. FFS's output goes high if the width circuit (see
Fig. 5) is triggered too fast. If one of the paths recognizes
that the clock is too fast, the corresponding message
appears on the display and all flip-flops are reset by the mi
croprocessor, which then continues to poll the error circuit.
The master clock is buffered and sent on five transmis
sion lines of equal length from the motherboard to all data
channels and control circuits.
Trigger Unit

Triggering of the 8182A depends upon a combination of

the states of the trigger arm input and the trigger qualifier,
the appearance of the trigger word at the data inputs, and
the state of the trigger event counter.
When the RUN key is pressed, the 8182A goes into the
armed state and begins to search for trigger events. The
trigger arm input is edge-triggered and serves as a preset
input. The trigger qualifier and trigger word must be true at
the same time to clock the trigger event counter, which runs
in two modes:

Data
Multiplexer
Glitch Detector off

Master
Clock
De!ay=Fixed Delay+^
=Fixed Delay+KCU

Fig. 7. Simplified circuit diagram of the high-speed delay

generator.

Shift
Register
Glitch Detector off 1
Glitch Detector on 1

Fig. 9. For glitch detection, the high-speed RAM is split into

data and glitch memories of equal size.

JULY 1983 HEWLETT-PACKARD JOURNAL 19

Copr. 1949-1998 Hewlett-Packard Co.

1. "Allow gaps yes." After the programmed number of true

trigger events has been counted, the trigger event
counter activates the trigger delay counter.
2. "Allow gaps no." Only true trigger events that occur one
after the other at every master clock time are counted . If a
false trigger event is recognized, the counter is reset to
zero.
When all trigger conditions are true, the trigger delay
counter, which consists of a count-down circuit that can be
preset to any value between 0 and 65535, is started. When
the delay counter has reached zero, the 8182A responds to
the trigger condition. In the trigger start analyzer mode, it
begins to write the incoming data into the high-speed RAM.
In the In stop analyzer mode, it stops data recording. In
the trigger start compare mode, it begins the comparison of
incoming data.

All trigger conditions can be set to "don't care," that is,

they appear to be always true. The 8182A then starts im
mediately with the first incoming active clock edge. The
instrument can also be started immediately by pressing the
SAMPLE key, which overrides all trigger conditions and
generates one clock pulse. Pressing the SAMPLE key re
peatedly or holding it down generates additional clock
pulses for storing or comparing data manually.
Glitch Detectors

Each channel of the 8182A has its own glitch detector. In

contrast to the latch mode found in many analyzers, the
glitch detector allows the 8 182 A to differentiate between a
data transition and a glitch, which is defined as more than
one data transition per sampling interval. To store the de
tected glitches in addition to the sampled data, the high-

Testing the Key Specifications of

the 8182A
Since the 81 82A is an analyzer, and key specifications such as
sampling accuracy (1 ns) and threshold accuracy (10 mV)
refer is inputs, direct measurement of these important values is
extremely difficult. Instead, signals must be applied and conclu
sions drawn about the performance of the instrument from its
response.
Fig. the illustrates a method by which it is possible to measure the
actual test point and threshold directly. The heart of this test
system is an electronic tool consisting of a simple integrating
op-amp circuit and a voltage-controlled delay generator. The
8182A can be programmed and configured to appear as a volt
age comparator followed by a D-type flip-flop. For this applica
tion, the 81 82A can be considered as a level comparator for the

data inputs, and as a phase comparator for data inputs with

reference to the clock input.
The closed loop consisting of the 8182A, integrator, and
voltage-controlled delay generator provides a data slope at the
active probe's input at the precise time of the actual sampling
point. The measurement of the sampling point can now be per
formed simply by a time interval counter.
The threshold measurement is performed using the closed loop
without the delay generator, thus passing the output of the inte
grator to the data inputs. The resulting voltage level is read with a
DVM.
-Bernhard Roth

Time Interval
Counter
81 82A Data Analyzer
VoltageControlled
Delay
Generator
Trigger
Out

Control
Voltage

Fig. make . measure the and a voltage-controlled delay generator make it possible to measure the
81 82 A Data Analyzer's sampling point and threshold directly.

20 HEWLETT-PACKARD JOURNAL JULY 1983

Copr. 1949-1998 Hewlett-Packard Co.

CLKO

Master Clock CLKI

(Minimum Delay)
Master Clock CLKI
(Maximum Delay)
Control Input ci
CICLKO CICLKI

(Minimum Delay)
CICLK1

(Maximum Delay)

Fig. 10. This correlator circuit attaches the trigger control

input sampled by CLKO to the appropriate master clock (CLK1)
pulse. Thus it prevents erroneous triggering resulting from the
variable sampling point delay.

speed RAM is divided into data memory and glitch mem

ory, each having a capacity of 512 bits per channel. There
fore, when the glitch detector is switched on, the memory
capacity of each channel is halved.
Fig. 9 shows the organization of the high-speed RAM.
Each channel stores its data in four 256 x 1-bit RAM arrays.
The CS signals are provided by a shift register. When glitch
detection is off, data is simultaneously switched by the
multiplexer to each of the four RAM fields. The shift regis
ter is preloaded with (1,0,0,0). After 256 master clocks, the

Fig. 11. Timing diagram for the correlator circuit of Fig. 10.

contents of the shift register are advanced one cell to the

right, resulting in (0,1,0,0). In this way, all four RAM arrays
are enabled in sequence and filled with data. When glitch
detection is on, the multiplexer separates the RAM inputs
so that data goes to RAM arrays 1 and 2, and glitches to
arrays 3 and 4. The shift register is then preloaded with
(1,0,1,0). Thus, arrays 1 and 3 are enabled at the same time,
with 2 and 4 following 256 clock pulses later.
The glitch detector used in the 8 182 A is a modified ver
sion of the well proven detector in the 1615A Logic
Analyzer,1 improved to perform at 50 MHz. In principle, the
glitch detector consists of a pair of edge- triggered flip-flops
(one for positive edges and one for negative edges) that

To Twin
8182A

Fig. clock Block diagram in logic analyzer modes. A multiphase clock system and data pipelining
are used.

JULY 1983 HEWLETT-PACKARD JOURNAL 21

Copr. 1949-1998 Hewlett-Packard Co.

serve as pulse stretchers, lengthening every glitch so that it

can be sampled by the master clock. To avoid dead times
during resetting of the glitch detectors, each edge-triggered
flip-flop is duplicated. While one pair of flip-flops is active
to receive a glitch, the other is reset after its glitch informa
tion has been written into the glitch memory. Obviously,
very careful design of the circuit and printed circuit board
was necessary to achieve accurate 50-MHz operation.
Control Input Correlators

A special problem arose because of the programmable

sampling point delay. With an increase in the sampling
point delay, the relationship between the master clock and
the initial CLKO is modified, and this can result in erroneous
triggering. To avoid this phenomenon, the sampling of the
control inputs (clock qualifier, trigger arm, trigger qualifier,
and external stop) is performed with the undelayed signal
CLKO. This gives an exact specification of the control input's
sampling point delay.
A correlator circuit (Fig. 10) , has the task of attaching the
control input sampled by CLKO to the appropriate master
clock (CLKl) pulse. This task could have been performed by
programmable delay lines, but the 8182A uses a simpler
way, a digital circuit that adapts automatically to different
delay settings.
The key idea is to store the sampling result of clock phase
CLKO in two input registers f or two clock periods, each with
one clock period offset.
The output register takes the appropriate information

from the input registers via a multiplexer, which is clocked

by the delayed clock phase CLKl. The whole process is
controlled by flip-flops FF7 and FF8, which have to be
initialized correctly to maintain the phase relationships.
The minimum propagation delay time between CLKO and
CLKl is 10 ns (with CLKl programmed to 0 ns added delay),
which is equal to the maximum allowable propagation
delay for FF2 and multiplexer MUX5 and the setup time of
FF6. If the CLKl delay is increased to more than one clock
period, no problem arises.
Fig. 11 is a timing diagram for the correlator circuit, Fig.
10.
Time Relationships in the Logic Analyzer Modes

In the logic analyzer modes (i.e. , trigger start analyzer and

trigger stop analyzer modes), a multiphase clock system
and data pipelining are used (see Fig. 12). The clock phases
are derived from CLKO via delay lines. The multiphase sys
tem makes it possible to sample, trigger, and store data on
the same clock pulse. Data is sampled on transitions of the
master clock CLKl, which can be delayed relative to the
clock input. The sampled data is captured by SYNC2. The
trigger word is detected simultaneously and stored by the
trigger flip-flop at clock phase CLK2. The trigger informa
tion is fed through a selector to the run flip-flop. Data and
memory addresses are sampled by CLK3. With the run flipflop active, write pulses are supplied and data is written
into the high-speed RAM. Memory addresses are supplied
by the address counter, which is incremented by the write

One Channel

Microprocessor

Fig. 13. Block diagram in comparator mode.

22 HEWLETT-PACKARD JOURNAL JULY 1983

Copr. 1949-1998 Hewlett-Packard Co.

pulses.
Since the maximum sampling rate is 50 MHz (20 ns
period) . the clock phases must be within 20 ns of each other
to achieve synchronous operation. On the other hand, the
propagation delay of the trigger word is roughly 15 ns
because of some gating operations and the motherboard
connections. In the twin or parallel operating mode, trigger
information sent to and received from the second 8182A via
a twisted-pair cable takes about 12 ns excluding gate propa
gation delays. With these values in mind, the timing of
clock phases CLKl to CLK2 and CLK2 to CLK3 becomes rather
critical. Therefore, delay times T2 andT3 are adjusted indi
vidually to 18 ns. To allow for the data and address setup
time specifications and the write pulse width delay of the
high-speed RAM, T4 is set at 7 ns and the write pulse width
is 7 ns.

Timing in the Comparator Mode

The timing system for the comparator mode (i.e., the

trigger start compare mode) is somewhat different from the
logic analyzer modes because data is not sampled and writ
ten into memory. Instead, incoming data is compared in
real time with data read from the high-speed memory. The
compare result cannot be pipelined because of its analog
character. Unfortunately, trigger word detection and pro
cessing takes time, as it does in the logic analyzer modes. To
avoid a trigger delay that is a function of frequency, the
clock inputs of the trigger and ran flip-flops are modified to
be synchronous with clock phase CLKl. This results in a
fixed minimum trigger delay of 1 clock period. In twin
operation, the minimum trigger delay is increased to two
clock periods.
Fig . 1 3 is a block diagram of the instrument in comparator
mode. After the run flip-flop is triggered, a clock pulse

Interfacing the Device Under Test

Grabber with ground lead (Fig. 1c)
BNC adapter (Fig. 1d)
SMB adapter (Fig. 1e)
Adapter for the HP 10024A 1C Test Clip (Fig. 1f)
Open-ended adapter for a custom connector or fixed wiring
(Fig. 1g)
Solder-in receptacles for printed-circuit boards (Fig. 1h).

Multichannel instruments with high-frequency capabilities such

as the 8180A Data Generator and the 8182A Data Analyzer re
quire special attention to the problem of DUT interfacing.
The design goals were easy adaptability, good signal perfor
mance, all interface capabilities, and suitability for all
generator and analyzer requirements.
A special connector is located at the front end of the analyzer
active probe (see Fig. la). At the generator end, the same con
nector type is molded directly onto a 50 coaxial cable (Fig. 1 b).
Six accessories for various interface possibilities were designed:

-Horst Link

tiff
I]

li
(e)

(9)

Fig. 1. 8182A Data Analyzer ac

tive probe and accessories.

JULY 1983 HEWLETT-PACKARD JOURNAL 23

Copr. 1949-1998 Hewlett-Packard Co.

Data 1

Fig. 14. In parallel or twin operation of two 8182A Data Analyzers, clock arbiter circuits syn
chronize the two units in the logic analyzer modes. In comparator mode, the arbiters are not used.

(DCLK) is generated at the end of the time window signal

(CLK1). This prepares the next data word for comparison,
supplies the next address to the RAM, and increments the
address counter and the word masking circuit, which pro
vides for the masking of errors at every user-selectable ad
dress. The compare result of every word is stored by the
address marking circuit. The channel marking circuit pin
points the channel or channels in which a compare failure
has occurred. Interrogation of the address marking and
channel marking circuits gives the microprocessor the in
formation required to generate the error map.
Parallel Operation

The 8182A is designed to run with another 8182A in

parallel. This configuration increases the number of chan
nels up to maximum of 64. For this purpose, both instru
ments are tied together with a multiple-conductor cable
which synchronizes the trigger conditions. The clock
probes of both instruments are connected to the same exter
nal clock source to maintain high sampling accuracy and
low skew.
The user must ensure that both instruments are pro
grammed compatibly. Parameters such as trigger condi
tions, sampling point delay, and thresholds are individu
ally programmable for both instruments. The trigger delay
counters of both instruments can be programmed indepen
dently. Hence the start of one 8182A can be delayed with
respect to the other for a defined number of clock periods.
The clock arbiter synchronizes clock phases CLK2 and
CLK3 of both 8182As in the logic analyzer modes. The

arbiter output pulse (CLK2) is delayed by 18 ns with respect

to whichever CLKi has the greatest delay (see Fig. 14).
Basically, the circuit consists of two high-speed 100k ECL
D-type flip-flops. The first is tied to internal clock phase
CLKl by a compensating delay line, and the second is tied to
CLKI of the second 8182A. An output pulse, whose width is
determined by propagation delay feedback, is generated
when the second flip-flop is clocked. Clock phases CLK2 of
both 8182As then have the same delay with respect to the
incoming clock within a few hundred picoseconds, the
sampling point delays being different.
Each 8182A's trigger event is sent to the other 8182A (IT
in Fig. 14) after being sampled by CLK2. The received trigger
signal RT and internally detected trigger signal IT are com
bined in an AND gate. If triggers are detected in both
8182As, the trigger delay counters are started in parallel.
For correct single-analyzer operation, the high-speed line
receiver for the twin trigger signal is de-biased so that its
input is always true when no twin interconnections are
present.
There are no interconnections between the microproces
sors during twin operation, so the selection of any key other
than RUN and STOP on one instrument has no effect on the
other. When the RUN key is pressed, the microprocessor in
that instrument sends an arming signal. The other micro
processor reads this signal and sends its own arming signal
in return. When both sent and received arming signals
become true, both 8182As go into an armed status simul
taneously.
The stop function is achieved simply by OR-gating the

24 HEWLETT-PACKARD JOURNAL JULY 1983

Copr. 1949-1998 Hewlett-Packard Co.

stop signals of both machines.

In the comparator mode, only clock phase CLKi is used,
so the clock arbiter is not required. Trigger word detection
and combinational functions are performed by the same
elements as in the logic analyzer modes.

Reference
1. J.A. Scharrer, R.G. Wickliff, Jr.. and W.D. Martin, "Interactive
Logic State and Timing Analyses for Tracking Down Problems in
Digital Systems." Hewlett-Packard Journal, Vol. 29. no. 6, February
1978.

Data Analyzer Software/Firmware Design

by Roberto Motto la and Eckhard Paul
LOCAL PROGRAMMING of the 8 182 A Data Analyzer
is done by means of front-panel softkeys. As selec
tions are made, the functions of the softkeys change
and are displayed on the CRT. The programming concept
provides eight pages of information, four pages for pro
gramming instrument parameters, three for displaying the
results of a measurement, and an eighth page for use as an
alphanumeric display to aid the operator a of computercontrolled system.
For remote programming, the 8182A is provided with an
HP-IB (IEEE 488) interface, giving it two remote operating
modes, ASCII and binary, with different addresses. The
binary mode is intended for fast transfers of binary data to
and from the expected data and received data sections of the
memory. The contents of the expected data section can be
transferred in 320 ms. The memory section (expected or
received data), the start address, and whether the 8182A is
to send or receive the data can be selected via an HP-IB
command in ASCII mode. In ASCII mode, each data byte
transferred via the HP-IB is treated as a seven-bit ASCII
character. If the 8182A has been addressed as a listener in
ASCII mode, the data parameters and mode settings may be
changed by sending the appropriate HP-IB command (a
sequence of ASCII characters). All commands start with a
header which consists of three characters. A real number,
an integer, or a formatted entry may follow.
Any HP-IB command can be transferred regardless of the
8182A's status, whether it be idle, armed, or active. Some
parameters (e.g., thresholds) can be programmed on the fly,
that is the measurement currently being performed is not
interrupted. A measurement is only interrupted or aborted
if the 8182A is unable to program a parameter on the fly, or
if it is impractical to continue the current measurement (e.g ,
the operating mode has been changed). The commands
received via the HP-IB are stored in a 256-byte buffer. When
the first character has been stored in the buffer and the main
program has time, the main program begins to interpret
characters stored in the buffer. Recognized commands are
executed. It is unnecessary for the HP-IB controller (e.g., a
computer) to wait until each command has been interpreted
before sending another, because the storing of characters in
the buffer is a priority operation.
To achieve synchronization between the controller and

the instrument, a holdoff handshake has been im

plemented. The 8182A holds off the HP-IB handshake when
the controller sends a synchronizing character, thus tem
porarily halting execution of the controller's program. The
holdoff is released when all bytes in the buffer have been
interpreted, all recognized commands executed, and the
hardware settled. For this reason no WAIT statements are
necessary when programming the controller.
The following example is based on an 8180A Data
Generator with HP-IB address 7, connected to the 8182A.
The programmable synchronizing character defaults to
(CR), carriage return.
Output 703; "run"
Output 707; "run"

Starts the 8182A Analyzer

Starts the 8180A Generator

Because a (CR)(LF) (carriage return, line feed) sequence

is normally transferred at the end of each Output statement,
the controller executes the second statement in the example
only after the 8182A has executed the RUN command. Thus
the generator cannot be started before the analyzer has been
set to capture data.
When the 8182A has been addressed as talker, it sends
status information, current parameter settings, received or
expected data, error information, or the contents of the
display, depending upon the talker mode. The nine talker
modes are programmable by HP-IB commands.
Also available is a status byte, which is read out via the
HP-IB after a serial poll has been sent by the controller. The
status byte contains the following status and error informa
tion:
HP-IB command either not recognized by interpreter, or
not allowed at this time
Attempt to program an erroneous value via the HP-IB
Context autoconnection made
SRQ key pressed
Stop routine executed
Fast clock detected
RQS (Request Service. See IEC 625.1)
Power-on self-test failed.
The 8182A can also generate a service request to interrupt
the controller to inform it that a measurement has been
completed.

JULY 1983 HEWLETT-PACKARD JOURNAL 25

Copr. 1949-1998 Hewlett-Packard Co.

Interrupt

Not Expected
Event

Fig. 1. operating 8182A Data Analyzer main program resembles a real-time multitask operating
system. CPU. skilled task dispatcher makes full use of the interrupt structure of the CPU.

Program Architecture

The structure of the program controlling the 8182A is

designed to obtain good performance from the 8-bit micro
processor under varying load conditions. The goal was to
guarantee no loss of commands and a prompt response,
even in extreme situations. For example, the 8182A might
be continuously displaying new data from a repetitive mea
surement with the operator varying some parameter using a
vernier key, while simultaneously logging the whole pro
cess on a computer using the HP-IB local talker capability.
The program is designed around a skilled task dis
patcher, which makes full use of the interrupt structure of
the 6809 CPU. The resulting structure closely resembles the
nucleus of a simple real-time multitask operating system.
As illustrated in Fig. 1, five tasks are performed by the
dispatcher. It uses a mixture of fixed priority and first come,
first served queue structures to select from competing tasks.
Priorities are dynamically variable, that is, the program
has the ability to decide when a task is becoming gradually
more important and requires extra attention, or when a task
can be delayed to conserve CPU resources. This feature
allows it to adapt to a wide range of controller speeds and
different operating conditions (perhaps a lot of display ac
tivity, or a CPU-intensive measurement).
The dispatcher also has the ability to recognize when the
operator or controller are sending commands that would
cause multiple repetitions of the same action. This situation
is flagged when recognized, and the action is executed only
once at the last possible moment.
The primary function of the interrupt structure is to inter
face the dispatcher to the outside world. This is made pos
sible via the keystroke and HP-IB acceptor routines. The
respective routines react to the asynchronous stimuli from
local or remote commands sent to the processor. A routine
checks the commands, then communicates with the dis
patcher, which is responsible for further processing. The
synchronization of different processes is assured by control

signals between the acceptors and the dispatcher, data

being exchanged between them in global areas.
The NMI line of the 6809 is connected to a 10-ms-period
time base which triggers the TIMERS module. This module
generates eight programmable delay timers and consumes
roughly 5% of the total CPU time. The dispatcher uses the
delay timers to perform time-related actions that have to be
independent of the CPU load, and to supply timing infor
mation to the dynamic priority algorithm.
Keystroke Operation

The keystroke handler and keystroke processor routines

together manipulate all the commands entered into the
8 182 A via the front-panel keypad (local commands). In the
event of no local activity, a simple circuit monitors the
keypad. If any key is pressed, an FIRQ interrupt, which calls
up the acceptor routine, is generated. The acceptor is re
sponsible for keyboard scanning and debouncing the
mechanical switches. If a key operation is recognized as
valid, the message is placed in a first in, first out (FIFO)
buffer, and the dispatcher is told to activate the keystroke
processor. The acceptor routine is then complete.
The eight-key FIFO buffer, combined with the interrupt
operation, ensures that key action is never lost, even in the
case of a heavy CPU load. When the processor takes more
time to complete its task the user may perceive a small delay
between giving a command and getting a response. The
keystroke data is taken from the FIFO buffer by the proces
sor. Different processes are selected and triggered from here
depending on the program status and the key function.
Typical actions are control of the softkey menus, program
ming the hardware, and activating the measurement con
troller.
By using the TIMERS module, the keystroke processor is
also able to reschedule itself periodically when keys calling
for an automatic repeat function are selected.

26 HEWLETT-PACKARD JOURNAL JULY 1983

Copr. 1949-1998 Hewlett-Packard Co.

HP-IB Operation

Display Updating

The HP-IB functions are realized with the acceptor and

processor routines shown in Fig. 1. The acceptor interrupt
routine handles all HP-IB activities, but does not interpret
ASCII characters received or execute HP-IB commands.
Changes of state or address (local, remote, listener, talker)
are recognized and stored in RAM. Characters received are
stored in a 256-byte buffer in ASCII mode, or in the desired
memory location in binary mode. Upon an interrupt, the
acceptor disables itself after a state change or if the receiver
buffer is full.
The processor executes the state changes, interprets
characters stored in the buffer, executes recognized HP-IB
commands, and enables the acceptor when the state
changes have been executed, or the receiver buffer is no
longer full.
The acceptor-processor structure is designed to handle
HP-IB activities very quickly. The acceptor stores charac
ters without interpreting or executing them. A slow control
ler can send the characters on an interrupt-controlled basis.
The processor interprets the characters stored in the buffer
interactively.
Measurement Controller

The measurement controller is responsible for all pro

cessing related to measurement cycles. The most important
activity is the programming of hardware for a RUN, and
reading back information generated in the hardware after a
STOP.
During measurements, the controller can read informa
tion from the hardware as soon as it is generated. This
information is also passed to the display updating routines,
giving the impression of a real-time instrument.
The controller recognizes when the instrument stops in
dependently of external commands. It also generates the
synchronization necessary for twin operation, propagating
the run and stop commands to the second 8182A. Using the
timers, the measurement controller schedules itself in the
case of automatically repeated measurement cycles (autoarming).
Periodic Tasks

Two categories of periodic tasks are recognized by the

dispatcher. The first is scheduled every 100 ms. Its primary
use is for display activities such as pulsing the clock sym
bol, flashing warnings, updating counters and so on. The
period of 100 ms is sufficient to assure good message visi
bility during a one-shot function.
The second category includes CPU-intensive tasks hav
ing a low priority. These are scheduled every 2.5 seconds to
keep the demand for CPU time down to 5% of the total. In
this category is a routine that checks the rear- panel settings,
allowing reaction to an HP-IB address change or a singletwin change without the necessity of a power-off power-on
sequence.
A routine for building a CRC (cyclic redundancy check)
signature on the 250 bytes of RAM containing the parameter
settings is also called every 2.5 seconds. This routine makes
it possible to use a CRC algorithm that gives excellent data
protection, but is so complex that it would take too much
time to compute the CRC after every data change.

The display updating routines are responsible for writing

information to the CRT. Static display changes caused by
operator commands and dynamic changes occurring when
the instrument is measuring and displaying the incoming
results in real tune require two different routine categories.
Self-Test

When the 8 182 A is powered up, it executes six self-test

routines. Any error encountered during these routines is
displayed as an error message. Tests performed are:
1. RAM/stack test
2. ROM test
3. Keyboard test
4. Clock test
5. Address counter test
6. Data board test.
If tests 1 or 2 fail, the self-test procedure stops because the
tested 3 is vital to the remaining tests. If any of tests 3
to 6 fail, the power-up sequence is stopped after completion
of the self-test. The program flow resumes after a stop if the
user presses the Continue softkey.
The RAM/stack test performs a read/write check of the
RAM stack area, then calculates the 16-bit CRC signature of
the remaining RAM, storing it repeatedly throughout the
stack area. Possible influences between the stack area and
the remaining RAM area are checked by building the signa
ture for the remaining RAM area again and comparing it
with the signatures stored in the stack area.
The ROM test calculates the signature for each ROM,
compares the value with the corresponding signature in the
self-test ROM, and displays the numbers of any ROMs in
which the comparison failed.
The keyboard test causes an error message to be displayed
if one or more keys are being pressed, or appear to be
pressed (indicating a short circuit), during this routine.
The clock test checks all paths a signal may take through
the timing circuits as a result of delay and width parameter
settings. Errors are recognized in the clock error circuit.
The address counter test simulates clock pulses and reads
the state of the address counter at several times to check that
pulses have been counted correctly and ensure that every
address bit is able to toggle.
The data board test checks the high-speed RAMs and
ascertains the number of data boards installed in the in
strument.
It is possible to execute further self-test routines by press
ing an unlabeled softkey and entering a password via the
keyboard. The tests have been implemented for service and
diagnostic purposes.
In addition to the self-tests described above, the HP-IB
commands MEW (memory write) and JSR (jump subroutine)
provide another powerful test facility. With the MEW com
mand, any data can be written into a sequence of RAM
addresses. If the data represents 6809 program code, the
program can be started via the HP-IB command JSR. In this
way, additional test programs can be loaded and executed
from a controller via the HP-IB for servicing purposes.
Initialization and Unexpected Events

The initialization block of the 81 82 A firmware checks the

JULY 1983 HEWLETT-PACKARD JOURNAL 27

Copr. 1949-1998 Hewlett-Packard Co.

data stored in the battery-backed RAM for integrity, when

selecting between a cold or warm start. It then transfers
control to the dispatcher. The unexpected event routine
permanently monitors the program flow and some critical
data areas to avoid the possibility of a system crash caused
by program bugs or a defective RAM.
Developing the Firmware

The 8182A control firmware is based on the microcom

puter board equipped with a 6809 CPU. The 64K address
space consists of 44K of ROM (only 40K is actually used),
14K of RAM for parameter and data storage, 4K of display
RAM, and 2K of memory-mapped I/O space.
The 6809 was found to be the most suitable 8-bit CPU for
this application. The powerful command set includes com
plete stack-relative addressing. The local data space of all
procedures is in the stack, giving total separation of code
and data as well as reentrant and recursive coding.

The 40K-byte program is written entirely in assembly

language to extract maximum power from the small CPU
and keep the amount of code to a minimum. Most of the
procedures are written as structured code and use standard
interfaces to communicate with each other. Some excep
tions are made in critical areas. Well known synchronizing
methods and data exchanging techniques (flags,
semaphores, mailbox, etc.) are used in the multitask area of
the program.
Most of the procedures dealing with programming or the
displaying of categories of similar information are written
as interpreters, working on information tables. The struc
ture of these tables is complex in some cases. For example,
the parameter updating procedure derives from its table the
command to call itself, and builds a new parameter set in a
recursive manner. Another typical example of a tabledriven interpreter is the procedure dealing with the softkey
menu tree.

Power Supplies for the Stimulus/Response

System
by Ulrich Otto and Horst Link
IN DESIGNING POWER SUPPLIES for the 8 180 A Data
Generator, 8181A Extender, and 8182A Data Analyzer,
the most important objectives were high load current
capability and serviceability within a restricted space.
The sizes of the instrument cabinets were limited by the
desire to produce a system suitable for benchtop operation.
To accommodate the generator, the analyzer, and perhaps
two extenders on a bench, the cabinet dimensions obvi
ously had to be limited.
The card rack in the generator holds eleven printed cir
cuit boards with a spacing of 20 mm between boards. The
analyzer has thirteen boards with 17-mm spacing. Because
both cabinets also accommodate a CRT display, the space
available for power supplies is severely restricted. The
outer dimensions for the power supplies were fixed at 254
x 152 x 139 mm.
The high load currents are required especially by the
-5.2V rails. In the 8180A Data Generator, the current taken
from the 5.2V supply by the Rate ICs alone is approxi
mately 7.1 amperes. The total current demanded of the
-5.2V supply is on the order of 18.5 amperes. Furthermore,
the numerous output amplifier circuits each demand 5.5W
from their +15V and -7.5V supplies. In the 8182A Data
Analyzer, the complex ECL logic circuits have a total re
quirement of some 28 amperes on the -5.2V rail, making
this supply the most heavily loaded in the system.
To aid production and serviceability, the decision was

made at an early stage to produce modular supplies consist

ing of plug-in units and boards, and to make the three
supplies as similar as possible, if not identical.
It soon became evident that the demands made on the
8182A power supply were unusually heavy, and although
the concept of switched-mode-with-flyback was adopted
for all three instruments' supplies, fundamental differences
exist in the 81 82 A supply.
Fig. 1 illustrates all of the main functional blocks of the
8180A supply. After rectification (or doubling and rectifica
tion, in the case of a 115V ac input), the value of the poten
tial at the switching transistor collector is between 200 and
360V, depending on the line voltage.
Starting Circuit

Fig. 2 shows the schematic diagram of the starting circuit

block shown in Fig. 1. A 1000-/uF capacitor is charged via
the resistive divider until the potential across it reaches
26V. At this point, Z24 becomes conductive and switches
on the drive circuit for the transistor switch. Energy is then
fed to the control circuit. When the potential across the
capacitor has decayed to 12 V, the transistor switch is turned
off and the circuit is ready to begin a new cycle. If no fault
condition is sensed by the control circuit, the power supply
begins operation before the end of the first start cycle. If, on
the other hand, a fault is detected, the start cycle is repeated
at three-second intervals. After the start cycle, all primary

28 HEWLETT-PACKARD JOURNAL JULY 1983

Copr. 1949-1998 Hewlett-Packard Co.

185 to 360 Vdc

Rectifier
and
Filter

PostRegulator

+5V, 4.5A

Rectifier
and
Filter

PostRegulator

-5.2V. 18.6A

Vcc (Supply for Primary Circuits)

+ 15V, 6.2A

Uo/Undervoltage Sensing
Overvortage Sensing
Current Sensing

Method: Flyback Regulator

Power Output: -260W
Efficiency: 57%
Load Variation: 0-100%
Frequency: 25 kHz

Negative
Supply Voltage

Rectifier
and
Filter

PostRegulator

-7.5V, 4.4A

Rectifier
and
Filter

PostRegulator

-15V, 0.3 A

Rectifier
and
Filter

PostRegulator

+23V, 2.3A

Fig. power supply. block diagram of the 8180A Data Generator power supply.

circuits in the supply draw their power from the flyback

transformer. The normally closed relay contacts remain
open, disconnecting the start circuit and conserving power
while the instrument is operational.
Control Circuit

A potential of approximately 25V, derived from the trans

former, is used to preregulate the power supply. This poten
tial is fed by a divider to the control 1C, which establishes
the duty cycle. Sensing resistors at various points in the
circuit protect components against the following condi
tions:
Collector potential of the switching transistor exceeding
1100V
Rectified line voltage exceeding 370V
Rectified line voltage falling below 185V
Switching transistor drawing in excess of 12A.
If the temperature inside the cabinet exceeds 75C, the
switching circuit cuts back its output and supplies a
minimum duty cycle. The fan in the power supply con
tinues to run, and all output supplies are shut down.

The positive bias current to the base of the switching

transistor is the algebraic sum of the currents from two
sources (see Fig. 3). The greater of the two currents (approx
imately 2.5A) is generated by transforming the 25V primary
supply into a pulsed waveform of 6.5V peak to peak. The
magnitude of the current is determined by the value of Rl.
This current goes to the base of the switching transistor via
Ql.

The lesser current, 50 mA, is drawn through R2 to the

base of Q2. The function of this current is to maintain Q2 in
its cutoff state until switch-off occurs. Switch-off is ac Rectified Line Voltage
Relay

12 kll.

Switching Transistor

The peak voltage at the collector of the switching transis

tor is in the order of 1000V. It was therefore necessary to use
a component with a very high V(BR)CEOand Icmax- When the
device is negatively biased in its cutoff range, it is possible
to attain the 1000V collector-emitter potential. However,
during the start cycle, negative voltages are unavailable
because of transformer output settling times. Therefore,
potentially hazardous situations cannot occur because
voltages greater than 700V are not generated during the
start cycle, and the transistor adequately handles every
thing below 700V without being negatively biased.

1000MF

* N T N *From

Tran
Transformer

Primary
Supply
Voltage

Fig. 2. Starting circuit schematic diagram.

JULY 1983 HEWLETT-PACKARD JOURNAL 29

Copr. 1949-1998 Hewlett-Packard Co.

~25V

R4 ^ 220U

<

KR- J 1 m

4H- 01

1 1:3.7

+IB
-15V

-^ Switching
KJ,r \ Transistor

VR1
Z6.8

-M-

-MR3

Fig. 3. Switching transistor base

control circuit.

-15V

complished in two stages. In the first, the main component

of the bias current at the base of the switching transistor is
withdrawn by disabling Q3. The second stage deals with
the delayed withdrawal of the lesser component. The delay
is achieved by the slow switch-off of Ql , governed by an RC
network at the base. R4 allows fine adjustment of the delay
time. The result of this delay is that Q2 is maintained in its
cutoff state long enough to sink any stored current present
at the base-collector junction of the switching transistor. A
delay of 2 /JLS is normally adequate to minimize switching
losses.
When the switching transistor is finally turned off, its
base potential falls to approximately -7.5V. The cycle is
then repeated continuously for as long as the instrument is
switched on.
Flyback Transformer

The complex requirements of the power supply, includ

ing the number of rails, the high currents, and the total
power consumption, made the transformer design a critical
concern and especially difficult.
As a starting point, a magnetic core that at the rated power
consumption could guarantee an acceptable magnetic flux
density of less than 220 mT was required. Another impor
tant design criterion was the stray inductance present in the
windings, which causes losses in the primary side of the
power supply. Furthermore, the triangle current waveforms
generated in a flyback system cause a high rms current. This
and the skin effect had to be taken into consideration when
selecting a material for the windings.
The transformer (see Fig. 4) consists of a U-shaped core
with both legs carrying a bobbin, giving a double-width
layer. This is of critical importance in dealing with stray

inductances. The primary winding is constructed by con

necting the windings of both bobbins in series. The second
ary windings are sandwiched between two primary wind
ing packages, each having an identical number of turns. All
secondary windings are connected in parallel to help deal
with high-density currents. These construction methods
result in a symmetrical flyback transformer having a stray
inductance of less than 3% of the primary inductance.
Further problems concerning high current densities and
skin effects were overcome by using multifilar windings of
high-frequency stranded wire, each strand having a crosssectional area of approximately 1.5 mm2. Special care in the
production of the transformer is necessary because of the
numerous bobbin-to-mounting-board connections. Finally,
the transformer is encased in a silicone rubber compound.
Postregulators

The 8180A Data Generator power supply provides eight

different rails. Six of these are dedicated to satisfying in
strument requirements and the remaining two are for sup-

30 HEWLETT-PACKARD JOURNAL JULY 1983

Copr. 1949-1998 Hewlett-Packard Co.

P = Primary
S = Secondary
CT = Center Tap

Bobbin I

Secondary
Output
Q

Out

Bobbin II

Fig. 4. Flyback transformer design reduces stray inductance

to less than 3% of the primary inductance.

plying the priman- circuits.

The 8180A power supply is also used in the 8181A Ex
tender. In each instrument the internal load varies accord
ing to the options supplied. Fine control of the output rails
is achieved by postregulators. each of which has its overcurrent sensing circuit.
A transformer sense coil in the primary winding preregulates all supply rails. Preregulation is determined by the
efficiency of the secondary voltage that takes the highest
loading. In the case of the 8180A, this is the 5. 2V rail. If an
overcurrent condition occurs at one of the outputs, the
common reference voltage for all postregulators is shut off.
In turn, all output rails decay to 0V. Because short-duration
overcurrents and capacitive loads are present from time to
time, shutdown of the output rails is delayed by 70 ms . This ,
however, results in extra stress on the series transistors in a
short-circuit situation. For this reason, if an overload does
occur, the postregulators remain down for approximately
two seconds, after which the reference voltage comes on
again. A check is then made to see if the overload condition
still exists, and if so, the reference voltage is again shut off.
8182A Power Supply

Because of the high current demands made on the -5.2V

rail, the concept of individually postregulated rails
couldn't be implemented in the 8182A because postregulator power losses became unacceptable. To reduce the
demands made on the - 5. 2V rail by the ECL circuits, a -2V

supply was implemented. This additional supply permits a

saving of over 5A in the -5.2Y supply.
A more direct method, in which the duty cycle of the
flyback transformer is controlled by an optocoupler, is used
in the 5.2V supply. The transformer is designed to allow
all rails other than -5.2V to be finely postregulated, as in
the 8180A and 8181A. Again, all supplies are protected
against short circuits, and if an overload occurs, the power
supply is switched off on the primary side by a second
optocoupler.
The switch-on phase of the power supply came under
close scrutiny. Limited space prevented the use of a line
transformer to supply the secondary regulating circuits, so
the start cycle is controlled as in the 8180A power supply,
using a high-resistance divider for preregulation. As the
secondary reference voltage rises, a low-resistance divider
fed by the optocoupler is paralleled with the highresistance divider and takes over control.
Circuitry for regulating the -5.2V rail in the 8182A
power supply is very different from that in the 8 180 A and
8181A. The configuration here consists of a circuit with
infinite gain and a very low cutoff frequency. The 100-to120-Hz ripple cannot be regulated in this circuit, so it is
inverted in a low-gain amplifier, then mixed with the
5.2V sense voltage in a unity-gain circuit. The output is
then passed to the emitting diode of the optocoupler. The
result is a 5.2V supply with a ripple content of less than 30
mV.

JULY 1983 HEWLETT-PACKARD JOURNAL 31

Copr. 1949-1998 Hewlett-Packard Co.

New Multi-Frequency LCZ Meters Offer

Higher-Speed Impedance Measurements
These instruments, combined with an optional interface and
a component handler, make production-line measurements
of the impedance parameters of discrete electronic
components rapidly and accurately at actual operating
frequencies.
by Tomio Wakasugi, Takeshi Kyo, and Toshio Tamamura
MANY INSTRUMENTS used for production-line im
pedance measurements have only one or two test
frequencies, usually 1 kHz and/or 1 MHz. But elec
tronic components, once installed in a piece of equipment,
must operate at other frequencies where their parameters
can be quite different. Therefore, component users want to
measure components at the actual operating frequencies.
Equally important, component manufacturers and end
users need low-cost, stand-alone systems capable of high
speed go/no-go measurements without a controller to im
prove throughput and reduce the cost of production-line
testing and incoming inspection of discrete LCR compo
nents.
HP's new LCZ Meters, Models 4277 A (Fig. 1) and 4276A
(Fig. 2), were developed to satisfy these requirements. They
can measure eight impedance parameters (L, C, ESR, G, D,

Q, | Z | , and 9) under a wide variety of test conditions,

including variable frequency and dc bias, and can be re
motely controlled via the HP-IB (IEEE 488). Features of
these two instruments include:
High speed (see Fig. 3) and high accuracy. Up to 4Vzdigit-resolution measurements with 0.1% basic accuracy
can be done in 70 ms. Fast C-only or L-only measure
ments are done in 30 ms, and packed binary output sends
the measurement data to a controller in 3.4 ms.
Component measurements at in-circuit operating fre
quencies. The 4276A measures at 801 frequencies from
100 Hz to 20 kHz and the 4277A measures at 701 frequen
cies from 10 kHz to 1 MHz.
Continuous memory. When turned off, or when the
power fails, both instruments automatically store all
front-panel settings except dc bias and comparator

Fig. 1. The HP 4277 A (shown

here) and 4276A (Fig. 2) LCZ
Meters perform accurate high
speed measurements of imped
ance parameters of components
and materials at frequencies
from 100 Hz to 20 kHz (4276 A) or
10 kHz to 1 MHz (4277 A). By add
ing an optional comparator/
component-handler interface
(shown in the foreground), they
can be operated in a stand-alone
configuration for low-cost, high
speed production-line testing or
incoming inspection.

32 HEWLETT-PACKARD JOURNAL JULY 1983

Copr. 1949-1998 Hewlett-Packard Co.

Fig. 2. HP 4276A LCZ Meter.

levels.
de bias. Measurements can be performed using an exter
nal dc bias from 0V to 40V. An optional internal dc bias
source (Option 001) provides the same range of voltages.
This internal option is HP-IB programmable with 10-mV
resolution from -10V to +10V.
Two ac signal levels: IV and 50 mV for the 4276A, and IV
and 20 mV for the 4277 A.
Residual compensation. The residual impedance of the
test fixtures and cables can be compensated (up to 20 pF
of open-circuit capacitance and 2 ohms of short-circuit
resistance).
Self-test. An automatic built-in functional test verifies
proper operation of analog and digital circuits.
Low cost and high reliability. Thick-film hybrid inte
grated circuits reduce cost and increase reliability.
Comparator (Option 002). This option allows a user to set
up ten pairs of high/low limits for L, C, and Z and, by
adding the HP 16064A keyboard (shown in Fig. 1), con
trol a component handler directly. This interface, which
uses optoisolators to provide high noise immunity, can
be programmed via the HP-IB.

play. The oscillator supplies the excitation to the bridge at

frequencies selected by pressing front-panel pushbuttons
or by HP- IB commands.
Bridge

Although the bridge sections in the 4276A and 4277A

operate in the same way, they use different approaches
suitable for covering the different measurement frequency
ranges. The bridge section used in the 4277A is the same
design as the bridge used in the earlier HP 4274A and
4275A LCR Meters.1 A modulation technique maintains
high gain and stability over the 4277A's full test frequency
200 -r

Design

The major sections of these LC Z meters are the bridge, the

vector ratio detector, the oscillator, and the digital circuitry
(see Fig. 4). The bridge provides a complex voltage Vx
across the device under test (DUT) and another complex
voltage Vr which is proportional to the complex current
flowing through the DUT as a result of the applied voltage
Vx. The vector ratio detector determines the precise com
plex voltage ratio between Vx and Vr and supplies this
information to the digital section for processing and dis

100

1k

I 1

10k 20k

Measurement Frequency (Hz)

Fig. 3. Typical high-speed measurement capability of the

4276 A for various component parameters as a function of test
frequency when the instrument is operated in the AUTO circuit
mode.

JULY 1983 HEWLETT-PACKARD JOURNAL 33

Copr. 1949-1998 Hewlett-Packard Co.

Comparator

Digital
Section

Fig. 4. Simplified block diagram of 4276A and 4277 A LCZ

Meters.

range from 10 kHz to 1 MHz, and a four-terminal pair

configuration eliminates errors caused by mutual induc
tance between cables.
A different design {Fig. 5} is used in the 4276A. Since the
test frequency range is lower than that of the 4277A, a
simple feed-forward differential amplifier is used for the
current- to- vol tag e converter. This amplifier is designed to
maintain high stability for various DUTs and measurement
cable lengths. The measurement error in the high-fre
quency range caused by the reduced gain of this amplifier
and the capacitance of the LCUR (low current) and LPOT (low
potential) cables is compensated by the error correction
program in the 4276A's digital section. The connections
between the DUT and the 4276A's measurement cables use
a five-terminal configuration consisting of HCUR, LCUR,
HPOT, LPOT, and Guard. This configuration eliminates er
rors caused by mutual inductance between cables because
the outer conductors of the HCUR and LCUR cables form a
current return path, as in the earlier HP 4274A and 4275A
LCR Meters.1
In both the 4276A and 4277A, the voltage at the virtual

When the 4276A is equipped with the optional comparator

(Option 002), a powerful, fully automatic system capable of 1 0-bin
sorting and go/no-go testing can be easily built by connecting a
component handler as shown in Fig. 1. Such a system can be
operated either with an external controller or, by using the HP
16064A keyboard, without a controller. All comparison informa
tion and settings are accessible via the HP-IB (IEEE 488).
To Improve sorting speed, two different output signals are
available to the component handler, the INDEX signal and the EOM
(end-of-measurement) signal. Total measurement time for a com
ponent consists of two periods: the time required to do the analog
measurement and the time required for the instrument's micro
processor to calculate the results. Many instruments output only
an EOM signal when the microprocessor completes its calcula
tions. However, since the device under test (DUT) must be con
nected to the meter only during the analog measurement period, it
can be disconnected during the calculation period and the han
dler can be moving to the next DUT while the microprocessor is
calculating the result. The INDEX signal output by the comparator
when the analog measurement is completed allows this, thus
increasing measurement speed by subtracting the calculation
time. For example, the time per C-D measurement at 1 MHz with
the 4277A can be reduced by 20 ms.
This interface has open-collector outputs, so the signal lines
can be driven from an internal TTL-level 5V supply (all outputs) or
from external voltage supplies up to 30V for decision outputs and
15V for control signals. The control lines TRIGGER, EOM, and
INDEX are optically isolated to provide ground isolation and pre
vent false triggering caused by noise.

V, = -Rr

High Current Lead

(HCUR)

High Potential Lead

(HPOT)

= 2,

Fig. 5. Bridge section of the 4276A.

Fig. 1. Option 002 provides simple go/no-go testing and

1 0-bin sorting by interfacing a 4276 A or 4277 A directly to a
component handler.

34 HEWLETT-PACKARD JOURNAL JULY .1983

Copr. 1949-1998 Hewlett-Packard Co.

Dual-Slope
A-to-D
Converter

nf

To
Digital
Section

Fig. 6. Block diagram of the vec

tor ratio detector used in the
4276A and 4277 A LCZ Meters.

Reference
Phase Detector

ground point (LpOT) is zero when the bridges are balanced.

Then

Therefore,
V

Therefore, all that is needed to calculate the complex

impedance of the DUT is the value of Rr and the vector ratio
between Vx and Vr.
The bridge sections in the 4276A and 4277A are fabri
cated using custom thick-film hybrid integrated circuits to
reduce cost and space.
Vector Ratio Detector

Fig. 6 shows a simplified block diagram of the high-speed

vector ratio detector section used in both LCZ meters. This

section detects the precise complex voltage ratio between
the two signals Vx and Vr from the bridge section. To obtain
good tracking characteristics and linearity for various sig
nal levels, the two signals are time-multiplexed by switch
Si. The process amplifier block contains attenuators and
variable-gain amplifiers to optimize the signal level to the
phase detector. The four-phase generator provides the pre
cision 90 phase rotations used to detect the orthogonal
components of the input signals to the phase detector.
Fig. 7 is a vector diagram showing Vx, Vr, and the x and y
coordinates as determined by the reference phase detector.
Here a, b, c, and d are the orthogonal components of Vr and
Vx detected by the phase detector and defined in the
following equation.

V,

c + jd

ac + bd

ad - be

a + jb

a2 + b2

a2 + b2

= x + jy

The vector ratio detector used in the earlier HP 42 74A and

4275A LCR Meters requires four integration cycles and two
dc reference current sources, and measures the four com
ponents a, b, c, and d separately by using dual-slope

Phase Detector 0

Fig. 7. Vector diagram showing the orthogonal components

of Vx and V, at the input of the phase detector. 81 is the
reference phase error.

Fig. 8. Comparison of the dc and ac discharge methods for

determining the orthogonal components a, b, c, and d of the
voltages Vx and Vr

JULY 1983 HEWLETT-PACKARD JOURNAL 35

Copr. 1949-1998 Hewlett-Packard Co.

The vector ratio x + jy is calculated as

Step:
S1

ac + bd y + a/3
a2 + b2 1 + a2

vr

and

ad - be ft - ay

y = 2 + b2 1 + a2

where a = b/a, /3 = d/a, and y = c/a. Then the unknown

impedance Zx is given by
B ay

1 + a2 1 + a2

Fig. 9. Measurement sequence for the vector ratio detector.

The steps are 1) reference phase detection, 2) autozero, 3)
first for conversion for a, 4) second A-to-D conversion for
ft, and 5) third A-to-D conversion for y. In the reference phase
detection step, the output signal Q of the flip-flop (see Fig. 6)
goes high coincidentally with the zero-crossing point of V,
thus detecting the reference phase 0

analog-to-digital conversion (dc discharge method).1 To

improve measurement speed, the dual-slope analog-todigital converter in the 4276A and 4277A measures the
scalar ratio between two components directly without any
dc reference source (ac discharge method). This increases
measurement speed by eliminating one integration cycle
and reduces cost by eliminating the dc reference sources. If
only the reactive component of impedance is measured, the
ac discharge method requires only one integration cycle,
further increasing measurement speed. Fig. 8 shows the
difference in steps between the two methods.

The measurement sequence of the vector ratio detector

consists of five steps which include reference phase detec
tion and autozeroing (see Fig . 9) . The reference phase detec
tor is used to minimize the reference phase error 0^ (Fig. 7).
Since the discharge period of the first integration cycle is
proportional to 0a, reducing 6l reduces measurement time.
To minimize Oit reference phase detection in the 4276A
uses high-speed clock signals (nf) which are multiples of
the test frequencies (see Fig. 6). The 4277A uses the
same automatic phase adjust technique used in the 4261A
LCR Meter.2
For high-speed inductance or capacitance measure
ments, only ft =d/a is detected. Equation (1) implies that
when the reference phase error 0a is negligible (i.e., a = b/a
is very small), the unknown impedance Zx is approxi
mately equal to -Rr(y + j/8). Thus the unknown reactance
is simply obtained by dividing or multiplying by the fac
tor 27rf.
The phase-detected signal of Fig. 6 has to be converted
into a dc voltage very quickly for high-speed measure
ments. In the 4277 A, a simple low-pass RC filter is used to
reject the ripple component of the phase-detected signal.
However, the 4276A's lower frequency range requires a
different approach. The low-pass RC filter, which would
have too long a settling time at the lower frequencies, is
replaced by a preintegrator (Fig . 10) whose integration time
is equal to a multiple of the test signal period. Therefore, the
output V0 is exactly proportional to the dc component of the
input signal. Thus, even at the lowest frequency of 100 Hz,

Dual-Slope
A-to-D
Converter

To Digital
Section

Fig. detector A preintegrator is used in the 4276A's vector ratio detector to convert the phasedetected signal to a dc voltage precisely proportional to the dc component of the input signal.

36 HEWLETT-PACKARD JOURNAL JULY 1983

Copr. 1949-1998 Hewlett-Packard Co.

10 kHz to 1 MHz
20 mVor Wrms

Power
Amplifier

Fig. 1 1 . Block diagram o- the sig

nal source used in the 4277 A LCZ
Meter.

- Microprocessor

only one test signal period of 10 ms is needed for the

conversion.
Oscillator

The signal source in the 427 7 A contains an 11.52-MHz

crystal oscillator, a conventional phase-locked loop, pro
grammable dividers, a voltage-tuned oscillator (VTO), and
an active low-pass filter output stage. This circuit, illus
trated in Fig. 11 by a simplified block diagram, outputs a
10-kHz-to-l-MHz sine wave with 2Vz-digit frequency reso
lution. The output filter consists of four active low-pass
filters that convert the square-wave output of the pro
grammable divider into the required sine wave. Each lowpass filter is realized in the form of a specially designed
hybrid circuit. Feedback is used to shift the cutoff frequency

of the filter to the fundamental frequency of the squarewave input. Total harmonic distortion of the output sine
wave is less than 60 dB over the 10 kHz to 1 MHz range.
Digital Section

The digital section is based on a Z80B microprocessor

clocked by a 6-MHz quartz oscillator. All of the high-speed
measurement calculations including error correction, data
control, and analog control are handled by this section in
addition to controlling the display and interfacing to the
front-panel and the HP-IB. A battery-powered memory
stores all front-panel settings except dc bias, the opencircuit and short-circuit zero offset values, and reference
values for deviation measurement for approximately two
weeks after the instrument is turned off or power fails.

High-Speed Programmable dc Bias Options

To evaluate varactor diodes or MOS devices, high-speed mea
surement of capacitance at a number of dc bias levels is essen
tial. These measurements are usually performed with an external
dc source controlled by a computer, but the H03 and H04 options
for the 4277A offer an inexpensive and simple solution to this
requirement.
The only differences between the two options are the test signal
levels. The H03 option supplies 20-mV and 1 V signal levels, and
the H04 option supplies 20-mV and 500-mV levels, which are
better suited for varactor testing. Both options use an HP-IB
programmable dc bias source and operate at 1 MHz. Other
common features include an automatic bias sweep, packed bi
nary data transfer, short settling times, and selectable delay.
Typical operation begins with the H03 or H04 4277A accepting
and memorizing up to 31 dc bias points from the HP-IB before
starling measurement. Upon receiving the measurement trigger
signal, the instrument begins its sweep, makes a measurement at
each preprogrammed dc bias level, and stores the measurement
data in its buffer memory. When the sweep is completed, the
stored data is dumped to the computer through the HP-IB in
packed binary format.
With these options, measurement speed is 1 .4 times faster than
with 25% standard 4277A because the settling time for dc bias is 25%
faster, and the settling times for function, ranging, and signal level
are 50% faster. For example, the capacitance for ten dc bias
levels can be measured in approximately 450 ms (see Fig. 1).
In addition, because the automatic bias sweep eliminates the

need for handshaking to set dc bias and to trigger measurements

during a sweep, one computer can control several H03 or H04
4277As through timesharing.

C-V

CHRRRCTERISTICS at 1MHz
DUT=VHRI-CRP DIODE

L
C

LJ
U
Z

.6

c
h

.4

u
rx
1

.2

C
U
3
D C

B I R S

15

C Vo 1 t 3

Fig. 1. The C-V characteristics of a varactor diode at 10 dc

bias points can be measured in approximately 450 ms using a
4277 A LCZ Meter equipped with an H03 option.

JULY 1983 HEWLETT-PACKARD JOURNAL 37

Copr. 1949-1998 Hewlett-Packard Co.

Acknowledgments

The 4276A and 4277A design team members who de

serve special recognition are Noriyuki Sugihara, project
manager, Toshio Ichino, Katsushi Aoki, and Masayoshi
Habu, analog section, Satoru Hashimoto, digital section,
Kiyoshi Suzuki, power supply, Kenji Shimada, self-test sys
tem, Hiroshi Shiratori and Tetsuya Shiraishi, mechanical
design, and Tsuneji Nakayasu, industrial design. Special
thanks also to Takeo Shimizu for his general management
and encouragement.
Besides those mentioned above, many other people in
cluding Seiichi Kikuta, Kohichi Yanagawa, and Hideki
Wakamatsu made significant contributions to the project.
Our thanks to them too.

References
1. K. Maeda and Y. Narimatsu, "Multi-Frequency LCR Meters Test
Components under Realistic Conditions," Hewlett-Packard Jour
nal, Vol. 30, no. 2, February 1979.
2. S. Hashimoto and T. Tamamura, "An Automatic Wide-Range
Digital LCR Meter," Hewlett-Packard Journal, Vol. 28, no. 1, Sep
tember 1976.

History from the Pages of the Hewlett-Packard Journal

More than a year ago, Roberto Albanesi, HewlettPackard's Country Manager in Italy, offered us a sug
gestion. Why not publish a book of articles from past issues
of the HP Journal? Select the articles on the basis that the
technologies or products they describe represented signif
icant advances for their time.
The same suggestion later turned up in responses to last
year's First HP Journal Reader Opinion Survey. Around
here, people seemed to like the idea, and the idea became
a project.
Twenty or so knowledgeable HP experts were polled for
suggestions of what articles to include. They had thirtythree years of HP Journals to pick from. The final selections
were made by William R. Hewlett, HP's cofounder, and
Bernard M. Oliver, former vice president for research and
development. Mr. Hewlett agreed to write an introduction
for the book.
We're now in the last stages of getting the book ready
for the printer, so when you read this, it may not be avail
able yet. But look for it, it will be along soon.

The name of the book is "Inventions of Opportunity:

Matching Technology with Market Needs." William R.
Hewlett's introduction describes the technique of "en
gineering of opportunity" that guided the company to many
successful products, especially in the early days.
Among the articles are those on the first high-speed fre
quency counter, the HP-35 Calculator that replaced the
slide rule in engineers' pockets, the beginnings of com
puter-controlled instrumentation systems, and many
others. The introduction comments on each article, provid
ing insights into the dynamics of innovation and showing
how a need, a technology, and creative people come to
gether to produce a successful invention.
The book is cloth-bound, has a dust cover, and contains
more than 350 pages. It will be obtainable through HP
sales offices around the world, or through HP's computer
supplies telephone order service, which is available in
many countries. The book has an HP part number: 92233B.
The cost will be modest.
-P.P. Do/an

38 HEWLETT-PACKARD JOURNAL JULY 1983

Copr. 1949-1998 Hewlett-Packard Co.

Authors

Josef Becker

July 1983

Klaus-Peter Behrens
I Klaus-Peter joined Bblingen Instruments Division
in 1978 after receiving his
engineering diploma from
the University of Stuttgart.
| He was responsible for the
design of the software and
digital control in the 81 80A
l 'f' I Data Generator and the
8181A Data Generator Extender. Klaus-Peter is
married and lives in Boblingen. In his spare time
he builds model railways.
Andreas Wilbs
Andreas Wilbs was born in
1953 in Heilbronn, West
Germany. He joined HP in
1980 as a sales support
I engineer for logic signal
sources, and since 1981
: has been a product market
ing engineer. He designed
I the specialist training pack
age for the 81 80A/81A/82A and is responsible for
the system's post-introduction activities. Andreas
is married and enjoys all kinds of outside activities
including portable video recording.

Jo Becker joined HP toI wards the end of 1979,

having previously worked
at the University of Stuttgart
as a biomedical engineer.
He was responsible for the
design of the 81 80A Data
1 Generator output
' amplifiers, and has now
moved to the fiber optics group at HP's Boblingen
Instruments Division, where he is involved in the
development of detector heads. Jo is married and
has three sons. His hobbies include amateur radio
and astronomy.
Heinz Nssle
Heinz Nssle was born in
Boblingen and came to HP
as an apprentice in 1967.
' After his apprenticeship,
j he studied at the University
j of Stuttgart, where he
gained his engineering
degree in 1 980. In 1 981 he
returned to HP and contrib
uted to the design of the 8180A Data Generator
and the development of the 15413A/15414A TriState Unit and Pod. Heinz is a member of the
German Red Cross and enjoys white-water
kayaking.
Ulrich Schottmer
Ulrich Schottmer joined HP
in 1981, and has contri
buted to the high-speed
digital control ofthe8182A
Data Analyzer. He received
his engineering degree at
the University of Bochum.
Ulrich lives in Boblingen
and enjoys radio-controlled
aircraft and skiing.

Werner Berkel
Werner Berkel was born in
Speyer and attended
i school in Karlsruhe, where
I he received his degree in
1 979. He contributed to the
design of the 8180A and
j supervised the design of
the 15413ATri-State Unit
I and the 15414A Tri-State
Pod. Werner enjoys music, soccer, and table ten
nis. He is married and lives in Boblingen.
Ulrich Hubner
Born and raised in
Stuttgart, Ulrich Hbner
studied at the technical
universities of Aachen and
Stuttgart. He joined HP's
m Boblingen Medical Division
* in 1967, then movedto the
instrument lab, where he
developed low-cost oscil
loscopes. He was responsible for the hardware
design of the 81 70A Logic Pattern Generator, then
served as project manager for the 8180A/8181A
Data Generator. Ulrich is married, has one child,
and enjoys swimming and skiing in his spare time.

Martin Dietze
Martin Dietze has been a
project engineer with HP
since 1980. He received
his engineering degree
' from the University of
Stuttgart in 1979. Martin
1 was involved in the de
velopment of the 8182A
Data Analyzer and de
signed the high-speed counters and glitch detec
tors. Born in Leipzig and now a resident of Bob
lingen, Martin enjoys traveling, backpacking, cy
cling, and playing the violincello.
Dieter Kible
Dieter Kible was born near
Lake Constance, and
graduated from the Univer
sity of Stuttgart in 1 975. He
joined HP shortly after
wards, contributing to
hardware and software
design of the 8165A Pro
grammable Signal Source.

He has been project leader for the 8182A Data

Analyzer since 1978. Dieter is a member of DKE.
the German counterpart of the IEC, working on
instrument bus systems. He is married, has two
sons, and spends much of his spare time working
on his house and playing tennis and cards
Bemhard Roth
Born and raised near Lake
Constance, Bernhard
earned his degree at Ulm
in 1978. He joined HP the
I same year, and began
work on the 8182A Data
Analyzer. He designed the
. active probe, input
it. amplifiers, clock timing
circuits, and log voltage generator for that
instrument. A member of the HP table tennis team,
he is single and enjoys rock music concerts and
motion pictures.
Roberto Mottola

I T- LLI I I Robert Motto|a s a natve

^Mk I of lta|y. and attended the

VVVU b University of Trieste,
J Pw -.-* graduating in 1976. He
moved to Germany and
joined HP in 1979, after
three years of research in
digital mage processing.
Roberto developed the
operating concept for the 8180A/81 A/82A Data
Generator and Analyzer system, and part of the
81 82A control program. He is married and has two
children. He spends his spare time with his family,
swimming, listening to classical music, and build
ing a model railway.
Eckhard Paul
I Eckhard Paul was born
near Flensburg in Northern
Germany, and graduated
from the Fachhochschule
of Kiel in 1981. He joined
HP the same year and has
I been responsible for the
HP-IB and self-test softI ware in the8182A Data
Analyzer. Eckhard is unmarried, and enjoys play
ing cards, skiing, camping, and cycling.
Horst Link

Horst Link joined HP in

1970, and spent his early
years on the mechanical
design of tape punches
and readers for calculators
and data acquisition sys
tems. He later became
| responsible for the
mechanical design of vari
ous pulse and data generator mainframes and
modules for use in 1-GHz pulse generators and
the 8170A Logic Pattern Generator. Horst is mar
ried, has a ten-year old daughter and lives in Bob
lingen. He spends much of his free time working
around his house and hiking. He is also an ac
complished photographer.

JULY 1983 HEWLETT-PACKARD JOURNAL 39

Copr. 1949-1998 Hewlett-Packard Co.

Ulrich Otto

Toshio Tamamura

Tomio Wakasugi

Ulrioh Otto joined HP in

t 1979. He was responsible
! forthe design of the power
supplies for the 8180A,
1 81 81 A, and 8182A. He
I received his BSEE from the
I School of Engineering in
I Esslingen in 1976, and
I went on to gain his MSEE
at the University of Stuttgart. A native of Stuttgart,
Ulrich now lives in Ehningen, near Boblingen. He
enjoys tennis, swimming, running, and reading,
and spends much of his spare time hiking in the
mountains in summer and downhill skiing in the
winter.

Since joining YokogawaHewlett-Packard in 1971,

Toshio Tamamura has
been a principal contrib* utor to the design of several
i HP LCR meters. He de
veloped the analog section
circuitry for the 4277A LCZ
Meter and designed some
of the hybrid integrated circuits. He has a BSEE
degree received in 1971 from the University of
Electro-Communications. He is married and the
father of two children, and likes playing tennis.

After graduating with a

a BSEE degree from Waseda
I University in 1976, Tomio
Wakasugi worked for three
years in telecommunica
tions before joining
Yokogawa-Hewlett-Packard in 1979. He worked on
the 4192A LF impedance
Analyzer and designed the self-test systems for
the 41 93A Vector Impedance Meter. He designed
the analog section for the 4276A LCZ Meter. Be
sides being captain of VHP's volleyball team, his
interests include tennis, jazz, and amateur radio.

Takeshi Kyo
Takeshi Kyo joined
Yokogawa-Hewlett-Packard in 1970 after receiving
his BEE degree from Iwate
University. He contributed
! to the design ofthe4271A
1-MHz LCR Meter and the
[ 4274A and 4275A LCR
Meters. He developed the
software for the 4276A and 4277A LCZ Meters. He
is married, has one daughter, and enjoys skiing
and playing GO, a Japanese game similar to
chess.

Address Correction Requested

Hewlett-Packard Company, 3000 Hanover
Street, Palo Alto, California 94304
H E W L E T T - P A C K A R D

Bulk Rate
U.S. Postage
Paid
Hewlett-Packard
Company

J O U R N A L

JULY 1983 Volume 34 . Number 7

Technical Information from the Laboratories of
Hewlett-Packard Company
Hewlett-Packard Company. 3000 Hanover Street
Palo Alto. California 94304 U.S.A.
Hewlett-Packard Central Mailing Department
Van Heuven Goedhartlaan 121
1181 KK Amstelveen. The Netherlands
Yokogawa-Hewlett-Packard Ltd.. Suginami-Ku Tokyo 168 Japan
Hewlett-Packard (Canada) Ltd
' Goreway Drive. Mississauga. Ontario L4V 1M8 Canada

0200020707 fcC PL AC eCAOO

MR C A BLACKELRN
JOHN HOPKINS UNIV

APPLIED PHYSICS LAB

JOHNS HOPKINS RO

L A U R E L

M D

2 0 7 0 7

/"* LJ A mailing send C Send p A r\ r\ CD fi O O To change your address or delete your name from our mailing list please send us your old address label. Send
changes to Hewlett-Packard Journal, 3000 Hanover Street, Palo Alto. California 94304 U.S.A. Allow 60 days.
5953-8513

Copr. 1949-1998 Hewlett-Packard Co.