On Class D Amplifiers Measurements

[amirm]

This is the purifi thd+n against frequency. Its superb.

You need to include the full graph and notes:

As you see, the measurement bandwidth was 20 kHz. This means no harmonic distortion is captured at 20 kHz which is the reason it sinks so low. Even at 11 kHz no second harmonic is included since that would be at 22 kHz and hence filtered.

My measurements on the other hand go up to 45 kHz bandwidth so second harmonic of 20 kHz is captured. And for frequencies below 15 kHz, I capture third harmonic and so on.

Distortion then likely will rise at higher frequencies with Purifi.

 

[Blumlein 88]

It is rather music than sine waves listening that shows the perceived difference. It is then much more difficult to prove by ABX. I do not think that the reason is full power response, rather the content of high harmonics and intermodulation products.

Either way, at this point we need some doable listening tests. So much the better if the issue is more clearly shown by music.

Music has many frequencies, but the level of each one is usually so low that even with many, many IMD products the levels of those products get lost in the noise floor. If we can get some listening tests which uncover it with music then wonderful. How do we get those files?

 

[March Audio]

It is rather music than sine waves listening that shows the perceived difference. It is then much more difficult to prove by ABX. I do not think that the reason is full power response, rather the content of high harmonics and intermodulation products.

Which we have shown to be vanishingly small (with a decent class d amp)

 

[pma]

My final contribution to this debate. I have made a thorough THD vs. amplitude measurement of my AIYIMA A07 TPA3255 amplifier, with LM4562 opamps (reduces noise and distortion slightly compared to original NE5532), powered from Mean Well RS-100-24 SMPS set to 27Vdc voltage. One channel driven, 4ohm load. A set of frequencies 100Hz-1kHz-3kHz-6kHz-12kHz. Measurement bandwidth 48kHz. I am trying to be as accurate as possible.

IMO the behaviour measured can be extrapolated to all amplifiers with TPA325X circuit and without PFFB. Probably to most class D amplifiers with the LC filter outside feedback loop. I am not saying that to class D generally. At least we know that Hypex NC400 (and similar Hypex products) and Purifi modules behave much much better, without measured issues.

But there is a plenty of class D amplifiers with similar design as this AIYIMA A07 that will behave very similarly. And this non-linear behaviour, together with load-dependent frequency response, give a possibility of audible sound differences on music signals.

 

[Wombat]

Technical Measurement,

not listening tests, is what this thread is about. Let's focus on the raw data to allow electrical performance comparison.

Listening sessions are not relevant re the OP.

I can't see that anyone is making rash-generalisations based on the OP 'sample-of-one'.

 

[March Audio]

My final contribution to this debate. I have made a thorough THD vs. amplitude measurement of my AIYIMA A07 TPA3255 amplifier, with LM4562 opamps (reduces noise and distortion slightly compared to original NE5532), powered from Mean Well RS-100-24 SMPS set to 27Vdc voltage. One channel driven, 4ohm load. A set of frequencies 100Hz-1kHz-3kHz-6kHz-12kHz. Measurement bandwidth 48kHz. I am trying to be as accurate as possible.

IMO the behaviour measured can be extrapolated to all amplifiers with TPA325X circuit and without PFFB. Probably to most class D amplifiers with the LC filter outside feedback loop. I am not saying that to class D generally. At least we know that Hypex NC400 (and similar Hypex products) and Purifi modules behave much much better, without measured issues.

But there is a plenty of class D amplifiers with similar design as this AIYIMA A07 that will behave very similarly. And this non-linear behaviour, together with load-dependent frequency response, give a possibility of audible sound differences on music signals.

View attachment 107520

So in terms of audible distortion we can ignore the 12kHz line as its 2nd harmonic is at 24kHz and 3rd at 36kHz.

So for a cheap amp is it anything horrendous? It can bang out 60 watts ish at less than 0.1% THD. For $72 thats probably quite good!

 

[March Audio]

Either way, at this point we need some doable listening tests. So much the better if the issue is more clearly shown by music.

Music has many frequencies, but the level of each one is usually so low that even with many, many IMD products the levels of those products get lost in the noise floor. If we can get some listening tests which uncover it with music then wonderful. How do we get those files?

We need to record the output of the amplifiers.

 

[Blumlein 88]

My final contribution to this debate. I have made a thorough THD vs. amplitude measurement of my AIYIMA A07 TPA3255 amplifier, with LM4562 opamps (reduces noise and distortion slightly compared to original NE5532), powered from Mean Well RS-100-24 SMPS set to 27Vdc voltage. One channel driven, 4ohm load. A set of frequencies 100Hz-1kHz-3kHz-6kHz-12kHz. Measurement bandwidth 48kHz. I am trying to be as accurate as possible.

IMO the behaviour measured can be extrapolated to all amplifiers with TPA325X circuit and without PFFB. Probably to most class D amplifiers with the LC filter outside feedback loop. I am not saying that to class D generally. At least we know that Hypex NC400 (and similar Hypex products) and Purifi modules behave much much better, without measured issues.

But there is a plenty of class D amplifiers with similar design as this AIYIMA A07 that will behave very similarly. And this non-linear behaviour, together with load-dependent frequency response, give a possibility of audible sound differences on music signals.

View attachment 107520

Can you convince me that 1% is audible with music? I know you can't convince me that .1% is. Load dependent is usually FR differences which can be audible without any distortion otherwise. Looks like 60 watts or so will be audibly blameless.

 

[Blumlein 88]

We need to record the output of the amplifiers.

That would be one good approach which I had in mind.

 

[pma]

https://www.audiosciencereview.com/...iyima-a07-amplifier-abx-listening-test.19352/

I was doing that (amplifier output record), as you know, and also posted the ABX tests. Thinking deep about the test, it would be optimal to record somehow IxV interaction with real speaker, however difficult.

 

[restorer-john]

Thinking deep about the test, it would be optimal to record somehow IxV interaction with real speaker, however difficult.

Pad it down over the speaker terminals and feed it into your A/D interface & computer. That is what I do for amplifier ABX comparisons after level matching electrically at an arbitary frequency (1kHz).

 

[Matias]

When we are making blanket statements about what a technology can do, shouldn't we at least consider the best of them? In the case of class D, Purifi and Hypex NCore? I think it is unfair to make points based on a cheap, mid performing amplifier.

Or would it also be fair to categorize all class AB based on this example below?

 

[pma]

Pad it down over the speaker terminals and feed it into your A/D interface & computer. That is what I do for amplifier ABX comparisons after level matching electrically at an arbitary frequency (1kHz).

Yes I know, and it was already done, however with the complex dummy load and not with the speaker, because it is really annoying to listen to the roaring speaker during recording.

The test was posted, however I was the only one to post the ABX result

https://www.audiosciencereview.com/...iyima-a07-amplifier-abx-listening-test.19352/

 

[pma]

When we are making blanket statements about what a technology can do, shouldn't we at least consider the best of them? In the case of class D, Purifi and Hypex NCore? I think it is unfair to make points based on a cheap, mid performing amplifier.

Sir, there are no blanket statements made. Please be so kind and so honest to read what is exactly written.

https://www.audiosciencereview.com/...ss-d-amplifiers-measurements.9940/post-644545

Exactly this

IMO the behaviour measured can be extrapolated to all amplifiers with TPA325X circuit and without PFFB. Probably to most class D amplifiers with the LC filter outside feedback loop. I am not saying that to class D generally. At least we know that Hypex NC400 (and similar Hypex products) and Purifi modules behave much much better, without measured issues.

But there is a plenty of class D amplifiers with similar design as this AIYIMA A07 that will behave very similarly. And this non-linear behaviour, together with load-dependent frequency response, give a possibility of audible sound differences on music signals.

P.S.: I am not sure why Amir's ASR measurements of those amplifiers have so much and so high mains components. Is it the measurement grounding system issue, are are these implementations of power supplies used really so bad?

 

[Matias]

Sir, there are no blanket statements made. Please be so kind and so honest to read what is exactly written.

https://www.audiosciencereview.com/...ss-d-amplifiers-measurements.9940/post-644545

Exactly this


P.S.: I am not sure why Amir's ASR measurements of those amplifiers have so much and so high mains components. Is it the measurement grounding system issue, are are these implementations of power supplies used really so bad?

I meant John, sorry, should have quoted him.
https://www.audiosciencereview.com/...ss-d-amplifiers-measurements.9940/post-644413

"All class D unfortunately gets tarred with the same brush due to implementations such as the amplifier reviewed. Sure, that's not fair to the good examples, but it's reality, nonetheless."

 

[restorer-john]

The test was posted, however I was the only one to post the ABX result

I was on self imposed sabbatical. Full time carer to my lovely partner who badly broke her leg bones. Audio, measurements and HiFi in general takes a back seat.

Didn't log in or see it.

 

[restorer-john]

I meant John, sorry, should have quoted him.

John (me) was specific also. Not sure what you are driving at.

 

[Matias]

John (me) was specific also. Not sure what you are driving at.

It's early here, I definitely should take my coffee before posting, as it is causing lots of miscommunication. Sorry again! Here is the exact quote.

"All class D unfortunately gets tarred with the same brush due to implementations such as the amplifier reviewed. Sure, that's not fair to the good examples, but it's reality, nonetheless."

 

[restorer-john]

It's early here, I definitely should take my coffee before posting, as it is causing lots of miscommunication. Sorry again! Here is the exact quote.

"All class D unfortunately gets tarred with the same brush due to implementations such as the amplifier reviewed. Sure, that's not fair to the good examples, but it's reality, nonetheless."

Basically, I am saying that for every review that technically castigates typical Class D implementations, the quality performing products get dragged along for the ride.

As I said, it's not fair, but until the excellent performers outweigh the vast majority, the overall reputation will not improve. Instead of attacking the messengers of truth, proponents should be encouraging other players in the field to lift their collective game, improve their offerings and be honest for everyone's benefit.

That is what happened post 1974 thanks to our friends at the FTC. The current atmosphere is ripe for change and improvement.

 

[pma]

I was on self imposed sabbatical. Full time carer to my lovely partner who badly broke her leg bones. Audio, measurements and HiFi in general takes a back seat.

Didn't log in or see it.

No problem. There are so many posts and threads that it is not easy. Hope your partner will recover soon!