# Complex Load for Power Amplifier torture testing

#### pma

##### Major Contributor
Complex load for power amplifier torture testing

It has been discussed on the forum that we should test power amplifier not only to purely resistive load, but also some kind of the complex load should be used as well to verify amplifier stability and ability to drive complex load that may occur with different loudspeaker types and designs.

This is my take on the possible complex load circuit. It takes into account that some ESL speakers impedance falls deep below 1 ohm at very high frequencies and has capacitive character at middle frequencies.

The circuit is as follows, it is a 4 ohm resistor in parallel with a 33 uF capacitor which is in series wit a small 0.44 ohm resistor.

The real thing looks like this

and this is its impedance measured by LIMP software

We can see that the impedance falls to 0.5 ohm at 20kHz and has 1 ohm and -45° at 5kHz.

I have made some tests with the small linear amplifier described in this thread
https://www.audiosciencereview.com/...vintage-kenwood-ka-5010-amplifier-clone.9902/

This linear class AB amplifier has very good parameters so it was a good candidate to me.

First, I made measurements of responses to square waves with 1kHz and 10kHz repetition frequency, into this complex load. The measurements showed that the amplifier is stable with this complex load. There is a small decrease of impulse top seen in the 10kHz square response, for the reason that the initial impedance is close to 0.44 ohm value. But, no oscillations and overshoots.

1kHz square response

10kHz square response

Next, there is a measurement of THD vs. frequency at output amplitude that corresponds to 5W/4ohm/50Hz power. At 50Hz, the impedance is still resistive.

First plot shows distortion together with output voltage level

Second plot shows distortion vs. frequency in %. We can see that distortion rises above 500Hz, where the impedance starts to fall down quickly.

To compare, I am showing distortion plots of the same amplifier loaded with a purely resistive 4ohm load. We can see that the distortion is quite flat with frequency and stays below 0.007% up to 20kHz.

Distortion in % vs. frequency into resistive 4ohm load

Not sure if it is clear at 1st view, but during this test done at output voltage usual for the 5W/4ohm test (4.47Vrms) when the 4ohm resistive load would draw only 1.12Arms from the amp, this complex load draws 9Arms at 20kHz!! So what was the power? 40W/20kHz!

Simulated output current

So this test is much more severe and stressful than the resistive test.
Same applies for stability test with squares.

To show that <1ohm at 20kHz is not unusual, please see impedance measurements from Stereophile:

If you imagine 4ohm woofer connected with the ESL, we have similar impedance as that of this test.

Last edited:
An ESL will have a step-up transformer before the stators.

At least, mine do.

An ESL will have a step-up transformer before the stators.

At least, mine do.

And that's how it is measured, you can see phase turning up and crossing 0° below 20kHz (Martin Logan). The panel without transformer is a pure capacitor. You see real-thing impedance plot, with transformer contribution.

https://www.stereophile.com/content/martinlogan-montis-loudspeaker-measurements

John A. wrotte
As with other MartinLogan electrostatic loudspeakers, the Montis's impedance drops to a very low value at the top of the audioband. The Montis is specified as having an impedance of 0.52 ohm at 20kHz. My measurement (fig.1), taken with an Audio Precision System One, gave a figure of 0.7 ohm at 20kHz, but this included 15' of speaker cable. Repeating the measurement using the DRA Labs MLSSA system, which compensates for the speaker-cable impedance, gave a figure of 0.55 ohm at 20kHz. Either way, that this speaker is a difficult load for the partnering amplifier to drive is compounded by the high electrical phase angle, and ameliorated only by the fact that music rarely has high levels of energy in the top octaves.

I am interested since I have used stats for decades. First Quad ESL 57s and more recently Quad 2805s. My original amplifier was a Quad 303 and more recently I have used a refurbished Quad 606-2. Both were of course designed with stats in mind and are supposed to be ultra stable. So I would love to see one of the recent Quad current dumpers tested.
My hunch is that I could do with more power than the Q606-2's 2x140 watt hence my interest. Do I go for Quad monoblocks at quite serious expense or do I have alternatives that may be both cheaper and better.

My hunch is that I could do with more power than the Q606-2's 2x140 watt hence my interest. Do I go for Quad monoblocks at quite serious expense or do I have alternatives that may be both cheaper and better.

I have had here the NC400 monoblocks only twice in last 6 years. Therefore I am sorry that I cannot answer your question and am not able to repeat my test with the class D. However, I assume that @amirm or @March Audio should be able to replicate my test, because the setup is so simple, as can be seen in the post #1 of this thread. Just higher power rating of the 0R22 resistors should be used, these are overloaded even at test output voltage (4.47Vrms), however they survive the stepped frequency sine test, because the big stress is only in the top octave.

I would also be interested in a test of a Quad current dumper like the 909 or QSP/QMP.

Last edited:
A load like this doesn't tell us anything new. Here is the NC400 with a resistive load:

We see that as frequencies go up, distortion does too.

Now, if you make the load capacitive, then the impedance gets lower with frequency resulting in more measured distortion. No new insight is given.

A load like this doesn't tell us anything new. Here is the NC400 with a resistive load:

We see that as frequencies go up, distortion does too.

Now, if you make the load capacitive, then the impedance gets lower with frequency resulting in more measured distortion. No new insight is given.

You have kindly omitted the fact that capacitive reactive impedance adds complex imaginary impedance part and this may radically change stability of the amplifier. This is something you do not test and you seem to refuse to test it. You would see the result after addition of capacitance and applying square wave. I thought you were Electrical Engineer, were not you? Or you came from IT industry? That would explain it.

Not sure if it is clear at 1st view, but during this test done at output voltage usual for the 5W/4ohm test (4.47Vrms) when the 4ohm resistive load would draw only 1.12Arms from the amp, this complex load draws 9Arms at 20kHz!! So what was the power? 40W/20kHz!

View attachment 41954
Simulated output current

So this test is much more severe and stressful than the resistive test.
Same applies for stability test with squares.

To show that <1ohm at 20kHz is not unusual, please see impedance measurements from Stereophile:

View attachment 41955

View attachment 41956

If you imagine 4ohm woofer connected with the ESL, we have similar impedance as that of this test.

One comment I would take issue with here is "> 1 ohm is not unusual".

It is very unusual and very bad design. That sort of load will cause issues for many amps. It's not an accepted/normal characteristic/issue of esl speakers. Quad manage to have quite sensible impedance levels, so it's all about design.

You have to ask yourself when does it stop being just a "challenging" load an amplifier should handle and just plain pathological. A design that should be criticised and avoided. Why does anyone think its acceptable to present an amplifier with what is effectively a short circuit????

Second thing to mention is that power requirements at high frequencies are actually very low. Above 15kHz the signal level of real music might be 50dB or more lower than say 1kHz. So your 4.45 v rms 5 watt example becomes 14mV rms. With a pathalogical 0.5 ohm load current is 28mA and power is 392mW. So even with the extra dissipation in the amp due to phase angle it's not a big challenge.

We know distortion will rise with load, nothing new to be learnt there.

So the only real issue here is stability.

I'm happy to do some testing, however might be next week due to getting parts in and workload.

BTW, isnt the soundlab inductive at 20kHz?

Last edited:
I am interested since I have used stats for decades. First Quad ESL 57s and more recently Quad 2805s. My original amplifier was a Quad 303 and more recently I have used a refurbished Quad 606-2. Both were of course designed with stats in mind and are supposed to be ultra stable. So I would love to see one of the recent Quad current dumpers tested.
My hunch is that I could do with more power than the Q606-2's 2x140 watt hence my interest. Do I go for Quad monoblocks at quite serious expense or do I have alternatives that may be both cheaper and better.
All the quad impedance measurements I have seen show them to be not too difficult. You need to find the plots for the specific model but going on the ones below any decent amp can drive them.

You have kindly omitted the fact that capacitive reactive impedance adds complex imaginary impedance part and this may radically change stability of the amplifier. This is something you do not test and you seem to refuse to test it. You would see the result after addition of capacitance and applying square wave. I thought you were Electrical Engineer, were not you? Or you came from IT industry? That would explain it.
Watch your language. I am not here as your test slave to run and do things just because you demand.

I am an Electrical Engineer and more importantly started to learn about analog electronics in 1960s and have worked on hundreds of amplifiers.

Fact is that an amplifier becoming unstable is not remotely a common consumer problem with amplifiers. Even cheap ones. I have already explained to you why some of your assumptions are completely wrong about ultrasonic content, step function testing, etc. You don't seem to want to listen and keep pushing for new tests that are simply of no value to membership.

If you think people love to see such things, then setup the tests and run them. See how many people will loan you amplifiers to test for stability.

I've owned several ESL's and used quite a few amps with them. The toughest loads I've had on hand were the Soundlabs I have, and some older Acoustats. Both have a variable resistance added to the upper frequency band. Turned in one direction you get less than an ohm near 20 khz. Turned the other way you get 1.5 or 2 ohms.

Most ESL's are a tough load because they are inefficient and need more power. The only one which seemed to be bothered by the upper frequency low impedance I've had were Tripath based amps. Though I may have been lucky it is true some amps can become unstable with such loads even with the low power requirements above 10 khz. They must not be very common.

I've had amps of low power that just couldn't push the ESL's without distorting. But the same would happen with a purely resistive early Magnepan. Most ESL's have very large impedances at the lower frequencies where most power is needed. You can't get much power thru that without a lot of voltage. A side effect is high power amps have high voltage capabilities. Unlike conventional speakers you'll need to push a hot signal to your powerful amp to get enough voltage to put much power thru the ESL. Current usually isn't a problem.

The most easily disturbed amp I have owned was a Pass Aleph 0. It had 30 watts at 8 ohms, and was impossible for even an instant for it to put out more than 8 amps current with even a dead short circuit. I forget its maximum voltage, but it wasn't much. It was probably unconditionally stable. It would play my ESL's fine just not very loud. Same on some Maggies. Same on some closed box speakers. Ported speakers with a fair phase angle and moderately low impedance in the range where real power is needed? It would almost collapse and sounded awful, awful, awful. Even with small ported speakers.

So something I'm not sure of, can we get about the same useful information to test amps at 2 ohms as we could get by using a reactive load that has combinations of phase angles out of sync with impedance? What typically seems to be the case is such conditions use several time more current for a given power than a resistor needs. Maybe amps should be tested at 1 ohm. I think most all will show less power capability there, but for comparison maybe that shows us something nearly as useful as using highly reactive loads. It won't quite stress the output stage in dissipation quite the same way as reactive loads.

If an amp isn't being asked for more current and voltage than it can supply, the reason complex reactive speaker loads sound different with various amps is most likely the interaction of output impedance and its result on frequency response. With most class D amps having such a lower output impedance that probably is no longer a problem either.

Last edited:
One comment I would take issue with here is "> 1 ohm is not unusual".

It is very unusual and very bad design. That sort of load will cause issues for many amps. It's not an accepted/normal characteristic/issue of esl speakers. Quad manage to have quite sensible impedance levels, so it's all about design.

You got there before I did. I note a few other things: the capacitance in Pavel's load is much higher than even his pathological Martin Logan speaker, and significantly higher than Quads or Acoustats (which, although uncommon, were still made in higher quantities than the Martin Logan or Sound Lab). So yeah, if you're in the 0.01% who owns such things, the amp's tolerance of a capacitive load is of interest. It's just not for the other 99.99% of us who have better-engineered speakers.

That said, if no-one else has done it by this weekend, I'll do some measurements with the AP of a good quality Class D amp I have on hand using a load like this.

You got there before I did. I note a few other things: the capacitance in Pavel's load is much higher than even his pathological Martin Logan speaker, and significantly higher than Quads or Acoustats (which, although uncommon, were still made in higher quantities than the Martin Logan or Sound Lab). So yeah, if you're in the 0.01% who owns such things, the amp's tolerance of a capacitive load is of interest. It's just not for the other 99.99% of us who have better-engineered speakers.

That said, if no-one else has done it by this weekend, I'll do some measurements with the AP of a good quality Class D amp I have on hand using a load like this.
Yes please, I am strapped for time to get some tests done, but will follow up yours when I get a chance.

Oh also the 2 speakers noted above, Sound labs and Martin Logan, one is actually inductive and the other capacitive at 20kHz.

Complex load for power amplifier torture testing

It has been discussed on the forum that we should test power amplifier not only to purely resistive load, but also some kind of the complex load should be used as well to verify amplifier stability and ability to drive complex load that may occur with different loudspeaker types and designs.

This is my take on the possible complex load circuit. It takes into account that some ESL speakers impedance falls deep below 1 ohm at very high frequencies and has capacitive character at middle frequencies.

The circuit is as follows, it is a 4 ohm resistor in parallel with a 33 uF capacitor which is in series wit a small 0.44 ohm resistor.
View attachment 41939

The real thing looks like this
View attachment 41940

and this is its impedance measured by LIMP software
View attachment 41941
We can see that the impedance falls to 0.5 ohm at 20kHz and has 1 ohm and -45° at 5kHz.

I have made some tests with the small linear amplifier described in this thread
https://www.audiosciencereview.com/...vintage-kenwood-ka-5010-amplifier-clone.9902/

This linear class AB amplifier has very good parameters so it was a good candidate to me.

First, I made measurements of responses to square waves with 1kHz and 10kHz repetition frequency, into this complex load. The measurements showed that the amplifier is stable with this complex load. There is a small decrease of impulse top seen in the 10kHz square response, for the reason that the initial impedance is close to 0.44 ohm value. But, no oscillations and overshoots.

View attachment 41942
1kHz square response

View attachment 41943
10kHz square response

Next, there is a measurement of THD vs. frequency at output amplitude that corresponds to 5W/4ohm/50Hz power. At 50Hz, the impedance is still resistive.

First plot shows distortion together with output voltage level
View attachment 41944

Second plot shows distortion vs. frequency in %. We can see that distortion rises above 500Hz, where the impedance starts to fall down quickly.
View attachment 41945

To compare, I am showing distortion plots of the same amplifier loaded with a purely resistive 4ohm load. We can see that the distortion is quite flat with frequency and stays below 0.007% up to 20kHz.
View attachment 41946

View attachment 41947
Distortion in % vs. frequency into resistive 4ohm load

Typically for Audio I test with NHS ... I believe they are bifilar or Ayrton-Perry winding and limit the inductive component quite well. Might be flat wound too..
https://en.wikipedia.org/wiki/Bifilar_coil
http://www.resistorguide.com/wirewound-resistor/

https://www.mouser.com/ProductDetail/TE-Connectivity-CGS/NHSC2001R5J?qs=sGAEpiMZZMvNd0dY0Kymzkh87IulMj8ARSXBl6LIsCU=
https://www.digikey.com/product-det...assive-product/NHSC2001R0J/A139121-ND/2055337

Standard HS Series 4.7 ohm, 200W

NHS Low Inductance 8 ohm, 200W

#### Attachments

• 4ohm-hs-200w.jpg
437.4 KB · Views: 163
Last edited:
I once had an idea to test how an amp reacts when it is fed back EMF.
Planned to use an 8 Ohm resistive load and feed the amplifier 100Hz and then use another amplifier coupled with 33 Ohm to the same load and put a much higher frequency (and lower level) on it and measure the output voltage of the DUT and see how much of the 'back EMF" would be visible at the output. Never got around to actually do this though.
In total the DUT would 'see' a 6.4 Ohm load with back EMF and the higher freq. amp would 'see' an 33 Ohm load with a low 'freq. back EMF.
The 'back EMF amp could be the other channel of the DUT or a separate amp.

You got there before I did. I note a few other things: the capacitance in Pavel's load is much higher than even his pathological Martin Logan speaker, and significantly higher than Quads or Acoustats (which, although uncommon, were still made in higher quantities than the Martin Logan or Sound Lab). So yeah, if you're in the 0.01% who owns such things, the amp's tolerance of a capacitive load is of interest. It's just not for the other 99.99% of us who have better-engineered speakers.

That said, if no-one else has done it by this weekend, I'll do some measurements with the AP of a good quality Class D amp I have on hand using a load like this.

@SIY Im sure Im telling you something you already know (this is really for others reading), but if you manage to do some testing, there is one issue you need to be careful of not conflating with another.

If using square waves and comparing amplifiers (any class), you have to be aware of the bandwidth limit of said amplifiers. In a word Gibbs effect. A 50 kHz BW amp will show this ripple more pronounced than a 100kHz BW amp.

So it can be misleading. Not any indication of a problem, its just what a bandwidth limited square wave does/is.

This is a pertinent issue if looking at class D amps. If you use a wide bandwidth measurement (scope) that includes the switching frequency you are not going to see the wood for the trees. If you bandwidth limit and filter it to remove the switching frequency then depending where the filter cut off point is it could be an issue re Gibbs effect and leading edge steepness or peaking. Also most class D is limited bandwidth anyway, maybe 50 to 60 kHz.

its just a case of making sure of comparing apples with apples.

For those who may not be aware of this, some more reading

https://en.wikipedia.org/wiki/Gibbs_phenomenon

Last edited:
Quad amplifiers always were deliberately bandwith limited, and the effect on square waves was easily visible.

Quad amplifiers always were deliberately bandwith limited, and the effect on square waves was easily visible.
Yes, so the idea that square waves should come out square and flat is quite misleading (indications of instability/oscillation not withsatanding)

5 harmonics

25 harmonics

125 harmonics

Replies
133
Views
12K
Replies
15
Views
1K
Replies
21
Views
1K
Replies
4
Views
551
Replies
0
Views
309