Comparison of power amplifier distortions and FR into complex speaker dummy load

[pma]

Comparison of AIYIMA A07 with DIY class AB amplifier into complex speaker dummy load (and more amps added soon)

This is a measurement comparison of AIYIMA A07 amplifier (powered by SMPS Mean Well RS-100-24 27V)

https://www.audiosciencereview.com/...-100-24-smps-power-supply-measurements.19507/

with my new 250W/4ohm class AB amplifier

https://www.audiosciencereview.com/...on-blameless-topology-and-measurements.21542/

with respect to operation into complex dummy load that simulates a 2-way real speaker.

I have built a dummy load to simulate speaker impedance years ago. It simulates a simple 2-way box and uses highly nonlinear ferrite 18mH coil to simulate woofer impedance nonlinearity.

The load looks like this

This is the circuit schematics of my dummy load

and this is a measurement of the dummy load impedance

The plots were shown in the DIY amp thread but I re-post them for easier orientation.

First, let's see frequency response of both amplifiers into the dummy load. We can see quite considerable modulation of the AIYIMA frequency response due to its output LC filter interaction with the complex load impedance.

Frequency response of two amps into dummy load


The complex load includes highly nonlinear ferrite core coil of 18mH, which simulates quite well the nonlinearity of the real woofer near resonance. So let's measure THD vs. output voltage at 70Hz first. At 70Hz, we have 8ohm magnitude and -45° phase.

We can see quite similar distortion of both amps at this frequency, with AIYIMA having slightly lower distortion between 2V and 8V and of course a big difference in output power. Please note the fast rise of ferrite coil nonlinearity effect above 10Vrms.

Now let's make measurements at 70, 130, 400 and 1300 Hz. Why these frequencies? Because there is a non-comfort combination of impedance magnitude and phase. 70 Hz was already mentioned, 130 Hz has 6 ohm minimum of magnitude with 0° phase, 400 Hz has 12 ohm magnitude with +45° phase and 1300 Hz has 15 ohm impedance with -40° phase.

Class AB distortion vs. amplitude into dummy load at 70, 130, 400 and 1300 Hz

AIYIMA distortion vs. amplitude into dummy load at 70, 130, 400 and 1300 Hz

Distortion at 70Hz is similar, at 130Hz and 400Hz AIYIMA has lower distortion but at 1300Hz AIYIMA has considerably higher distortion.

Now let's compare distortion vs. frequency, it was measured at 9Vrms which is still comfortable and almost optimum for AIYIMA.

One can see fast rise of distortion below 80Hz, which reflects high nonlinearity of the ferrite core 18mH coil, this is reflected in nonlinear load current and this again in voltage distortion at amp terminals due to its finite output impedance.

AIYIMA has a bit lower distortion near woofer resonance (due to lower output impedance at lowest frequencies), slightly lower distortion between 80Hz and 1kHz, however much higher distortion above 1kHz.

Conclusion
So it goes, there are hints that the level matched AB listening test will bring differences. Mostly because of the differences in frequency response, but even the distortion differences might be interesting.

 

[pma]

I am starting to believe that this is the method that may explain why amplifiers sound different and why they should be tested into complex loads, not only to resistive load.

I have added one more amplifier into the comparison plot. The results quite precisely reflect what is heard during the level matched AB listening test.

 

[ebslo]

Great stuff, I've often wondered if this kind of test would expose audible differences between amps that resistive loads do not show. I also appreciate how Soundstage! measures amps with an actual speaker in addition to resistive loads like here, second plot under heading "RMS level vs. frequency vs. load impedance (1W, left channel only)", although only measuring at 1W doesn't show what happens when the driver goes highly non-linear as you show here. A counter argument, however, could be that if the driver's electrical response has gone non-linear it's acoustic response has probalby also degraded and that likely dominates acoustic error.

 

[pma]

A counter argument, however, could be that if the driver's electrical response has gone non-linear it's acoustic response has probalby also degraded and that likely dominates acoustic error.

That's absolutely correct, the acoustical distortion will always be greater than the electrical distortion. This is because the speaker impedance near resonance is highly nonlinear and even if the speaker was driven from an ideal voltage source with zero output impedance, the speaker current would be distorted due to Ohm's law (linear voltage divided by nonlinear impedance makes nonlinear current). However, it is still interesting to observe how speaker nonlinearity is reflected at the amplifier output.

BTW very good set of measurements at Soundstage.

 

[Head_Unit]

Very interesting! I had thought of the "real" impedance affecting most the clipping but of course distortion can change. And probably the frequency response can modulate somewhat as for instance the woofer impedance changes instantaneously. Somewhere I had seen a multi-tone plot both of real speakers but also of tests into a simulated load. Of course the big question is how audible is this? Here there are some variations up to 12 dB which is a big change-but are the absolute levels audible on music program? I'm not sure what I think about that.

 

[audio2design]

Do you have comparative data with a 4/8 ohm purely resistive load? We would expect some of the same issues to be evident, even with a fixed resistive load. You can tell that the Aiyima has insufficient gain-bandwidth.

There appears to be something wrong with your testing. What is going on here? Why the stair-steps prior? Those look to be faults in the measurement/analysis system.

 

[levimax]

Thank you! Very interesting information. A question I have regarding the impedance / phase graph is if you have an active speakers (one amp per driver with active crossover) does that eliminate the "load" and "phase" issues for the amp? If not what would the graph look like for only one driver driven directly by an amp?

 

[thulle]

There appears to be something wrong with your testing. What is going on here? Why the stair-steps prior?

Limited measurement bandwidth. If you read the graph from the right the first step is around 7350 Hz telling us that the measurement probably is sampling at 44100Hz, making 22050Hz the highest frequency the measurement sees. For the test tone at 7350 Hz we measure the 2nd harmonic at 14700 and the third at 22050. As the test tone goes higher in frequency the third harmonic falls outside of the measurement bandwidth, and since it's suddenly only 2nd harmonic and not 2nd+3rd being counted in the THD, it steps down. Continuing reading to the left, next step occurs at 22050/4=5512.5Hz, where the 4th harmonic is suddenly added. Next at 22050/5=4410 where THD is 2nd to 5th harmonic, and so on.
Checking the slope of the line between the steps we see it increasing from left to right, so we can suspect that harmonic distortion is actually increasing continously with frequency, we just add fewer harmonics to the THD.

 

[pma]

Do you have comparative data with a 4/8 ohm purely resistive load? We would expect some of the same issues to be evident, even with a fixed resistive load. You can tell that the Aiyima has insufficient gain-bandwidth.

There appears to be something wrong with your testing. What is going on here? Why the stair-steps prior? Those look to be faults in the measurement/analysis system.

View attachment 155285

Yes with resistive load you can see a part of behaviour, but resistive load has no imaginary impedance part and lacks the speaker nonlinearity so the extrapolation of results is difficult.
Re your second question, it was already explained and I would prefer for you to study on the subject before accusing on the faulty measurement.

 

[Dal1as]

Interesting. I always wondered if the difference in sound between amplifiers was due to interaction with the loudspeaker. Good work.

 

[Chr1]

Just discovered this thread and wish I had found it earlier.

Phew!

I have always thought that I could hear audible differences between some amplifiers and found it weird that it is so vigorously denied by the vast majority here on ASR.

As these differences can clearly be shown given the right measurements, surely this rather unscientific mantra should now be dropped here?

 

[JayGilb]

This is the definition of science. Tests being developed to try and quantify what is being observed.

 

[Chr1]

Indeed. But clearly more comprehensive testing of amplifiers makes sense. I was referring to the common statement that all amplifiers sound the same being rather unscientific in nature.

 

[SIY]

As these differences can clearly be shown given the right measurements, surely this rather unscientific mantra should now be dropped here?

The effect of amplifier (and wire) source impedance on frequency response is discussed regularly here. There's nothing to drop.

I was referring to the common statement that all amplifiers sound the same being rather unscientific in nature.

No one ever said that, so no problem.

 

[Dal1as]

Indeed. But clearly more comprehensive testing of amplifiers makes sense. I was referring to the common statement that all amplifiers sound the same being rather unscientific in nature.

Well this effect is difficult to test I believe. At least from what I've gathered from comments here.

I do know that I've experienced differences with amplifiers in a system that I know and live with and play with in a room that I know.

The thing is I didn't notice much until I started dsp tuning to an SPL I like and am used to when jamming to music. I especially notice it when listening to pronounced bass lines. Geddy Lee, Heaven and Hell, and Steve Harris especially.

Room is 15 ×22. Been using it for music and home theater for 20 years now. Back when RPTVs were the thing with a Boston Acoustic T1030 and 5.1 THX system with a Lexicon processor. Progressed through a few different 5 and 7 channel systems. REVEL 5.1 system, Then the Paradigm Signature S8 7.1 system, I now have a Klipsch Reference RF-83 7.1 system which I liked better. I know the 83s very well.

A few times using different amps setting up the same speaker setup with the same processor and dsp I noticed frequency differences when played at a level I like. I noticed this before and even after redoing dsp.

1st time was with the Paradigm system. I bought a Denon 5 channel amp at this time and was going to use a Carver TFM35 for the front towers but it was getting recapped. Used a 2 channel Crown amp for the rear 2 channels and the Denon 5 channel for the other 5. Setup dsp and wasn't liking the Paradigm speakers. I had a few weeks before I could return them so started playing around with different amps as I had been running 3 zones and had a few laying around and the Carver and Sunfire 5 channel were coming back from being refurbished. Only difference I noticed was the Crown and Denon sounded different in the bass lines than the other amps I ended up using. It was real pronounced when not changing the dsp settings. Less so after redoing them but still noticable.

I ended up returning the Paradigm system, trying some Revels, and settled on the Reference 83 Klipsch system which I've stuck with for a long time. Over the years with the same speakers and room I noticed the same differences in a couple of different amps. The Carver sounded different then pretty much all others I tried.

The differences were there. Noticed mostly when just switching out the amp, level setting, but also a couple times after redoing dsp and listening at my desired spl.

At lower levels I didn't really ever pay attention but I dont feel that any of them sounded differently except the Crown. It was just noisy.

Something is going on there which causes the speakers to react different. All amps were probably only being pushed 50 to 70% if that much so not clipping.

Just playing music which sounded different.

 

[dominikz]

Well this effect is difficult to test I believe. At least from what I've gathered from comments here.

Hm, have to say it is a bit strange to me that amplifier FR load dependence is considered novel or difficult to test.
I posted about this recently.

In my understanding, knowing the amplifier output impendance vs frequency function enables us to completely predict the amplifier frequency response variation for any known load impendance. It is a simple voltage division circuit in principle, just frequency-dependent.
Measuring the amplifier output impedance characteristic is also not difficult - just rarely done (for some strange reason).

Of course that also means that these FR deviations, if significant at all, can be flattened by means of EQ.
In my understanding this is where the often misquoted "all amplifiers sound the same" argument comes. Most of non-broken amps will indeed sound the same with a little corrective EQ to flatten FR, assuming identical (and non-clipping) levels.

Non-linear distortion differences (if any) are another matter, of course - but audibility of those is questionable (barring severe ones).

Anyway - just my 2c

 

[Chr1]

I agree but surely if EQ frequency correction is assumed it should be stated. Differences in FR are clearly fairly important differences.

 

[Dal1as]

Hm, have to say it is a bit strange to me that amplifier FR load dependence is considered novel or difficult to test.
I posted about this recently.

In my understanding, knowing the amplifier output impendance vs frequency function enables us to completely predict the amplifier frequency response variation for any known load impendance. It is a simple voltage division circuit in principle, just frequency-dependent.
Measuring the amplifier output impedance characteristic is also not difficult - just rarely done (for some strange reason).

Of course that also means that these FR deviations, if significant at all, can be flattened by means of EQ.
In my understanding this is where the often misquoted "all amplifiers sound the same" argument comes. Most of non-broken amps will indeed sound the same with a little corrective EQ to flatten FR, assuming identical (and non-clipping) levels.

Non-linear distortion differences (if any) are another matter, of course - but audibility of those is questionable (barring severe ones).

Anyway - just my 2c

I know sound but the testing aspect of amplifiers and such is not my realm although I'm learning. Can't remember what the Lexicon dc2 had. I had Audyssey pro on the Integra but it only did so much.

Thing is the amps did sound distinctively different and if they all sound the same that shouldn't be true.

 

[Chr1]

Am I right in thinking that some amplifiers will have varying FR depending on the load? If so, does this not mean that it cannot be simply sorted out via EQ? ie if you have a changing FR, you need a corresponding changing EQ correction?

 

[SIY]

Am I right in thinking that some amplifiers will have varying FR depending on the load? If so, does this not mean that it cannot be simply sorted out via EQ? ie if you have a changing FR, you need a corresponding changing EQ correction?

In order:

Sorta. The overall frequency response varies with load if the source impedance is not relatively low (say, 10 or more times lower than the load impedance).

No, assuming you're not changing amps or speakers. If you are, you have to change the EQ.

It might be helpful to go read Fred Davis's papers on amp-wire-speaker interface from about 40 years ago. They are very accessible for non-engineering types.