Impedance Measurement Comparison - Part 1

by Langston Holland
Background
It took almost a month to pull off, but predictably several of my assumptions went down in flames and
the guys with the physics backgrounds proved right. Imagine that. I am very thankful for the
significant time and effort put into the SAC list impedance thread by the usual suspects. What a gift
you guys are. :)
My goal was to find the most accurate way to measure audio related electrical impedance that was as
simple as possible to execute. I also wanted to understand it. I do have some weird applications in
mind, but that's for another day.
The Players
DATS, CLIO, Audio Precision APx515 with the LinearX VI-Box and a couple of other boxes I made.
Summary
The DATS tester very accurately covers 99% of loudspeaker impedance measurement needs at $100,
about 40dB less than the Audio Precision based methods I tried. You should buy this thing.
CLIO is almost as convenient, extremely accurate, wildly more powerful since it's a full audio
measurement system and reasonably costs about 30dB more than DATS. As you've probably noticed
in my prior posts, CLIO is my primary measurement platform.
I bought the Audio Precision APx515 about a year ago because I wanted to make higher resolution
electrical measurements in both analog and digital domains. The hardware performance of this thing
baffles me - on one hand it can show low level detail I didn't know was possible outside of a simulator.
On the other hand the philosophy behind the software interface is very restrictive for someone used to
EASERA or CLIO for example. Accuracy with the AP is state of the art if you setup the test correctly.
For impedance measurement I used two-channel setups that divided voltage by current represented
by a sense voltage. Z = V/I.
Note
Anyone interested in using one of the two-channel methods that I used with the AP can do so easily
with either the old version of Smaart v5.4 or the current version of SysTune Pro v1.2. Both achieved
very high accuracy in my tests using the high impedance TRS inputs of the Smaart IO hardware
interface. Smaart has by far the easiest calibration setup, but SysTune Pro obviously offers much
more power and has the slight advantage in that you can buy it. :)
Myth Buster 1
On the SAC list I speculated that voltage and current source methods would yield different results due
to: "Effectively zero electrical damping provided to the reactive loudspeaker load, which exaggerates
resonance peaks and general sensitivity relative to an actual connection to an amplifier." WRONG. I
found no difference between methods that had direct connections between amplifier and loudspeaker
vs. methods with up to 660 ohm generator source impedances. Jay Mitchell wrote "Measured
correctly, the impedance of a loudspeaker will indeed be the same when measured with different
methods." I didn't believe this until I saw it. The key appears to be keeping drive voltages low enough
that everything that moves remains linear.
Dale Shirk’s contribution to the thread clarified the reason that we see the same result from either
type of measurement. We are dividing voltage by current to calculate impedance. The lower source
resistance constant voltage type methods will show a decrease in voltage variation and an increase in
current variation, while the opposite occurs with the higher source resistance constant current type
methods. The V/I quotient remains the same in either case. This will be demonstrated in Part 2 where
differing V and I plots are shown generating virtually identical Z curves.
Take away (Pat™): Measure impedance any way you want to, just do it correctly with proper cabling
and connections. It's only when you want to measure at higher levels that you'll need to minimize
source (amplifier) to loudspeaker resistances.

1 of 2
Myth Buster 2
In reference to the constant current method, I stated “Very low voltage delivered to the loudspeaker,
thus high sensitivity to environmental noise induced errors.” DECEPTIVE. The fact is you have to keep
drive levels low regardless of method in order to stay within the linear region, you can’t just choose a
high drive level with S/N as your only consideration. Drive voltages at the loudspeaker terminals begin
to skew the impedance measurement results when you exceed very low average voltages (approx.
50mV) with constant current type methods. With constant voltage type methods, you can get away
with higher drive voltages prior to skewing the results (approx. 300mV). Methods that land in the
middle of this range, such as in DATS, show good S/N in normal indoor environments.
I adjusted each measurement method to about 40mV across a 10 ohm resistance to keep the results
from being biased due to drive voltage differences. The one exception to this was the VI-Box with the
0.01 ohm current sense resistor switched in which required nearly 300mV to get reasonable S/N.
For sake of discussion, I now define "constant voltage type" and "constant current type" methods as
those having resistances between source and loudspeaker of 1 ohm or less and 500 ohms or more,
respectively. Those landing in the middle of this range were called "voltage divider" methods by Bill
Whitlock and that is exactly what is going on.
Finally, as mentioned by both Bill and Jay, loudspeaker temperature makes a difference. Heating from
use can cause significant changes, but I also saw minor variations due to the temperature changes
throughout the day in my home (sorry - I mean laboratory).
lh
Soundscapes <><
15 November 2013

2 of 2
 Impedance Measurement Comparison - Part 2

by Langston Holland
In Part 1 I gave an overview of the impedance measurement methods we would be comparing. More
importantly, we learned that we should expect equivalent results from each when driving the DUT at
low enough levels to stay within its linear region. Exceeding the linear region is easiest to determine
by increasing levels until the peaks at resonance begin to compress. In Part 2 we begin by describing
the measurement setup and conclude with a thorough look at each of the methods including actual
results into standard resistances as well as a loudspeaker.
Setup
Reference resistances were provided by a GenRad Type 1432M decade resistor. This eBay sourced
golden oldie is a work of art. It has extremely low reactance and measures as a perfect resistor per
the indicated value within the tolerances of my Fluke 8060A multimeter with the test lead resistances
subtracted out with the "REL" function.

GenRad Type 1432M - External View

1 of 23
 GenRad Type 1432M - Internal View
The reference loudspeaker used was a fairly new Danley SM100 known to be in good working order. In
regard to the loudspeaker, tests were made both without and with an obstruction.

2 of 23
Without Obstruction

3 of 23
 With Tool Box as Obstruction
This obstruction may seem severe, but little effect was visible with smaller items due to the fact that
there is very little horn-loading in this loudspeaker. You'll also notice on the plots that the traces
showing the obstruction are somewhat different between measurement methods. Given that the non-
obstructed measurements are virtually identical, we must conclude that I placed the obstructing box
slightly differently between measurements. Next time I'll go to the trouble of leaving the obstruction in
place and running through the different methods before moving anything. Bad scientist! :)
An overview of the mess:

4 of 23
 Light Bulbs?!
DATS
Great engineering is the art of using the least resources to get the job done with the necessary
accuracy, reliability, etc. DATS is a home run in this dept. I've used DATS for several years - my
software was upgraded from when it was called WT3 while the hardware continues to work perfectly.
Though the manufacturer doesn't include this information, measurements show that the small epoxy
encapsulated USB based hardware interface uses a 100 ohm generator output impedance to drive and
measure the DUT. The unit boots up with a default indicated +5dBu output into no load regardless of
the setting you may have selected in a previous session. The 1.40V no load voltage drops down to
125.54mV when loaded with a 10 ohm resistor. The math to calculate the output impedance of a
circuit measured in this way is:

!"#$#%& − !"#$%
!"#$ = !"#$%
!"#$%
The circuit schematic probably looks like this:

5 of 23
 DATS Schematic
This 100 ohm resistance allows a higher drive level than typical current source type methods. I
reduced it to an indicated -5dBu (616mV peak) during testing to make it equivalent to CLIO and the
other methods, but normally I'd leave it alone for improved S/N.
Resistive loads of 1, 5, 10, 50, 100 ohms measured 1.02, 4.99, 9.96, 49.81, 99.63 ohms at 1kHz with
flat traces. Measurements up to 3,000 ohms were good to within 1% at 1kHz with a slight HF rolloff.
Above that wasn't reliable.
SM100:

6 of 23
 DATS SM100
SM100 with obstruction:

7 of 23
 DATS SM100 with Obstruction
CLIO
CLIO also is a great work of engineering and more than justifies its cost for those who's work includes
a broad cross section of loudspeaker design and testing, acoustic work and electrical domain
measurement. CLIO offers four impedance testing methods, each of which can use four very different
stimulus types and associated calculations. Though they all yield the same answers when used
correctly, the non-FFT sine sweep and gated sine sweep methods have slightly lower frequency
resolution than the two FFT based methods of sine sweep and MLS. I chose FFT sine sweep for
loudspeaker measurements to keep apples to apples. CLIO's next version due out very soon will add a
5th method with a simultaneous division of voltage by current sense like SysTune Pro offers.
CLIO's "internal mode" impedance method was used in this test, which simply involves connecting a
XL cable to one of its outputs and attaching the DUT between pins 2 and 3.
The circuit schematic looks like this:

CLIO Schematic
This 660 ohm resistance provides a typical current source method low level drive to the DUT. You do
need a quiet environment during the test for adequate S/N.
Resistive loads of 1, 5, 10, 50, 100 ohms measured 1.01, 5.01, 10.01, 50.01, 99.98 ohms at 1kHz
with flat traces using the non-FFT based sine sweep method. Measurements up to 50,000 ohms were
good to within 1% at 1kHz with a slight HF rolloff. Above that wasn't reliable.
SM100:

8 of 23
 CLIO SM100
SM100 with obstruction:

9 of 23
 CLIO SM100 with Obstruction
APx515 with ZBox
This may sound crazy, but even at the 40dB price the APx515 is a very good deal in my opinion and a
stunning piece of engineering. Apart from all the fancy things it can do and its extreme accuracy, what
I really like is that you don't have to be careful to get dead-on repeatability.
The thing I call a ZBox is the result of an effort to get a highly accurate impedance measurement
method that didn't need an external amplifier and could be used for both loudspeakers and electronic
testing up to at least 50,000 ohms. It uses the V/I differential method nicely documented by Paul
Henderson and Chris Strahm. One of the lovely things about the V/I method is that the division
cancels any like values in the numerator and denominator, thus transfer function source imperfections
are removed.
The circuit schematic looks like this:

10 of 23
 ZBox Schematic
The 100 ohm current sense resistor attached to circuit ground proved to be the best overall
compromise for the "one size fits all" design goal.

ZBox - External View

11 of 23
 ZBox - Internal View
The 50 ohm generator output impedance added to the 100 ohm current sense resistor provides a
strong output level if desired given the APx515's capabilities.
Resistive loads of 1, 5, 10, 50, 100 ohms measured 1.03, 5.02, 10.01, 49.96, 99.85 ohms at 1kHz
with flat traces using a 5 second swept sine FFT with no averaging. Measurements up to 50,000 ohms
were extremely flat. Up to 100,000 ohms was within 1% at 1kHz with HF rolloff beginning above
2kHz.
SM100:

12 of 23
 ZBox SM100
SM100 with obstruction:

ZBox SM100 with Obstruction

The nominally 8 ohm loudspeaker results in more voltage swing than current given the 150 ohm
source:

13 of 23
 ZBox SM100 - Voltage and Current
APx515 with VI-Box at 1 Ohm
The circuit schematic can be seen in the VI-Box manual linked to in the ZBox section.
The 1 ohm current sense resistor is attached to circuit ground given that the majority of voltage
across this resistor will be the same at either end, thus common, thus stressing the CMR ability of the
analyzer hardware. It is far better to place the loudspeaker load across the floating differential circuit
given that it will rarely get below 2 ohms and will mostly be much higher, resulting in a much greater
voltage drop across the differential input relative to the common voltage.
This will be 100 times as true in the next section that covers the VI-Box using the 0.01 sense resistor.

14 of 23
 LinearX VI-Box with Joe Nickell Modification

LinearX VI-Box - Internal View

We now must use an amplifier to drive the 1 ohm resistor into the load and this amplifier must have
it's negative terminal in common with the circuit. A bridged type amplifier will overwhelm the CMR

15 of 23
requirements of many analyzers, and switching to the 0.01 ohm sense resistor with a bridged amp will
overwhelm the CMR of any analyzer.
The amplifier used in each of the following constant voltage type measurements was a Stewart Audio
PA-50B in common ground mode. Although the amp is not rated for use below 4 ohms, the low levels
used in these tests allowed operation into 1 ohm for the duration of the sweep without problems.
Again, the V/I division is our friend as it removes transfer function imperfections from the measured
results.
Resistive loads of 1, 5, 10, 50, 100 ohms measured 1.04, 5.03, 10.01, 49.89, 99.78 ohms at 1kHz
with flat traces using a 5 second swept sine FFT with no averaging. Measurements above 100 ohms
are somewhat impractical due to the diminishing output from the current sense resistor. With greater
current capability in my resistors, I could have measured above 100 ohms - but this isn't the right
hammer for high Z measurements - it's designed for higher power, in-use type measurements which it
does beautifully.
SM100:

LinearX VI-Box SM100 - 1 Ohm Sense

SM100 with obstruction:

16 of 23
 LinearX VI-Box SM100 - 1 Ohm Sense with Obstruction
The nominally 8 ohm loudspeaker results in much less voltage swing than current given the 1 ohm
source:

LinearX VI-Box SM100 - 1 Ohm Sense Voltage and Current

APx515 with VI-Box at 0.01 Ohm
We again use an amplifier to drive the 0.01 ohm sense resistor into the load and it's more important
than ever that this amplifier have it's negative terminal in common with the circuit. Given the 100 fold
decrease in the current sense voltage, I was forced to increase the DUT drive voltage to about 300mV

17 of 23
to get adequate electrical S/N. Fortunately, the nearly direct connection between amp and
loudspeaker kept it within its linear region, thus the measurements are again virtually identical to all
the other methods.
Resistive loads of 1, 5, 10, 50, 100 ohms measured 0.96, 4.98, 9.96, 49.93, 99.30 ohms at 1kHz with
flat traces using a 5 second swept sine FFT with no averaging. Measurements above 100 ohms are
completely impractical due to the diminishing output from the current sense resistor at the chosen
drive level.
SM100:

LinearX VI-Box SM100 - 0.01 Ohm Sense

SM100 with obstruction:

18 of 23
 LinearX VI-Box SM100 - 0.01 Ohm Sense with Obstruction
The nominally 8 ohm loudspeaker results in nearly zero voltage swing. Almost all the action is in the
current realm with the amp's 0.04 ohm source impedance rating combined with the 0.01 current
sense resistor:

LinearX VI-Box SM100 - 0.01 Ohm Sense Voltage and Current

APx515 with IBox
I call this an IBox to highlight the fact that it uses a Pearson 411 current monitor to directly measure
the current going through an 8" piece of 12 awg wire. The goal of this project was to eliminate the
need to use common ground amplifiers and to increase the maximum voltage and current the device
could handle beyond the capabilities of any bridged audio amplifier for the foreseeable future. Gene
Brandt sent me one of his two model 110 CM's which are identical to the 411 excepting size and
somewhat higher RMS current handling capability. After using Gene's for about an hour I wanted one
really bad and got the 411 because it's smaller, yet still rated at 50 amps continuous and 5,000 amps
peak. I think it'll handle audio amplifiers into the next century.
I called Pearson and talked to a guy that I could tell knew his stuff. I asked him if the wire needed to
be run through the middle of the eye, if a metal box would affect it, etc., and he said no - all the
action is in the eye and it doesn't matter where you route the wire through it. That sounded too good
to be true, so I took Gene's 110 with 2" eye and placed the current wire in the middle, flush against
the inside wall, even diagonally from one side of the eye wall to the other and got the same
measurements within 10nV! (0.00001V) Then I laid it on top of steel and aluminum boxes - again the
measurement was unchanged and virtually identical to what Ohm's Law said it should be with the 4
ohm power resistor I was driving. After installing the 411 inside the final IBox, measurements again
were perfect.
The circuit schematic looks like this:

19 of 23
 IBox Schematic
The physical unit:

IBox - External View

20 of 23
 IBox - Internal View
We have to use an amp of course, but this time it doesn't matter which type as long as it can drive the
loudspeaker successfully to the level of interest. An interesting observation with the CM is that both
the voltage and current outputs of the IBox show identical THD on the AP analyzer no matter what the
amp's output is. This simply means the CM is tracking flawlessly. It also means that the V/I division
will again remove any transfer function source imperfections.
Resistive loads of 1, 5, 10, 50, 100 ohms measured 1.04, 5.01, 9.98, 49.71, 99.34 ohms at 1kHz with
flat traces using a 5 second swept sine FFT with no averaging. Measurements above 100 ohms at low
drive levels are again impractical due to the diminishing output from the CM.
SM100:

21 of 23
 IBox SM100
SM100 with obstruction:

IBox SM100 with Obstruction

The nominally 8 ohm loudspeaker results in the least possible voltage swing from this setup given the
0.001 ohm resistance added by the 8" piece of 12 awg cable:

22 of 23
 IBox SM100 - Voltage and Current
lh
Soundscapes <><
15 November 2013

23 of 23