Class D A07 x class D UcD180 and class AB amp - there are measurable differences on acoustical side

[mcdn]

I’m still not getting how this is a class AB vs class D thing then. It might be an issue with the A07’s power supply, or even with the PC or interface, but the amp itself works perfectly well it seems.

 

[SIY]

I’m still not getting how this is a class AB vs class D thing then.

It's not. The major logical error is generalizing from a specific example. There may also be test setup issues, but I'm in no position to examine that. I can say that with NCore driving the tweeters in my speakers, and using lab-grade measurement equipment, I'm not seeing anything like this.

 

[pma]

It's not. The major logical error is generalizing from a specific example. There may also be test setup issues, but I'm in no position to examine that. I can say that with NCore driving the tweeters in my speakers, and using lab-grade measurement equipment, I'm not seeing anything like this.

As explained, the issue was detected as SMPS leakage current and addition of interference voltage on a screen of SE link cable (plus possible intermodulations of this interference signal in the amp circuits). This was identified beyond any doubts. This can be generalized to SMPS use and SE link interconnects.
Lab-grade measurement equipment does not mean that the method or user is free of mistakes. The ownership of AP of yours or @amirm does not automatically guarantee you do everything right. It is rather the qualification, knowledge and experience of the user that defines the result. And the system must be measured under real conditions, not separated to single components measured at clinical "lab-grade" conditions. You are saying that you are "in no position to examine that", but you do so, though indirectly however very clearly. I am not examining your measurements as I have not been present when you did them.

P.S.: SMPS leakage current corresponds with SMPS ripple which has repetition frequency of about 65kHz with high dv/dt. Maybe the reviewers could concentrate to this as well. I know that @KSTR is well aware of this.

 

[SIY]

As explained, the issue was detected as SMPS leakage current and addition of interference voltage on a screen of SE link cable.

Specific to ONE amplifier example. Generalizing it to "Class D" is ridiculous and unsupported by data.

The argument against using lab-grade equipment for measuring a consumer-grade amp and speaker is... novel, I'll grant you that. I do like how you slipped in the entirely incorrect "separated to single components measured at clinical "lab-grade" conditions."

 

[Absolute]

It's not. The major logical error is generalizing from a specific example. There may also be test setup issues, but I'm in no position to examine that. I can say that with NCore driving the tweeters in my speakers, and using lab-grade measurement equipment, I'm not seeing anything like this.

Can I see, please?

 

[SIY]

Can I see, please?

What would you like to see?

 

[KSTR]

I'm completely with @pma here.
His point about "isolated bench" AP etc measurements is that they depict best-case scenarios for the item under test. This fully OK with me as we want to know the limits of what a device is capable of in the first place. But to which amounts these results are achieved in a "typical" compound user setup is a different matter and this doesn't get enough spotlight here. Some devices will hold their specs, other will degrade severely.
I have suggested sort of a cross-current generation and immunity testing here before, looking at a) how much cross current (eg mains leakage currents) a device emits and b) how much impact a forced cross current through the device has. There are quite some surprises here.

 

[Absolute]

What would you like to see?

The measurements of your tweeter with Ncore with a single ended connection like pma describes.

I mean, if you have measured it and not found the same as he has, why not share it with us?

 

[antcollinet]

Obviously for a fair comparison (if you are interested in comparing amp classes) - the comparison should be at a similar design quality/build quality and/or price level.

Obviously the Class D you are measuring is at the bargain basement end even of class D amps, and certainly of amps generally.

Recognising that the AB amp you have compared is home built from an adapted design, what do you think the retail price of an amp with similar build quality/performance would be? Including all the separated PCB's, wiring looms, large toroid and very nice looking case?

(really nice job on that build by the way)

 

[SIY]

The measurements of your tweeter with Ncore with a single ended connection like pma describes.

I mean, if you have measured it and not found the same as he has, why not share it with us?

I'll post some spectra after I get home from work today. Bottom line: the spectra are dominated by the usual 2H and 3H, with none of the artifacts Pavel saw.

 

[Blumlein 88]

What microphone is being used here,?

 

[SIY]

The measurements of your tweeter with Ncore with a single ended connection like pma describes.

I mean, if you have measured it and not found the same as he has, why not share it with us?

Here's the spectrum with the NCore (D) and for comparison an Adcom GFA-555 (AB). Main difference is that the Adcom is slightly noisier (not audibly from 1 meter, but I have the mike close like in Pavel's photo). Tweeter is a SEAS aluminum dome in my main speakers.

 

[Absolute]

Excellent @SIY. Thank you

 

[jan.didden]

I'm completely with @pma here.
His point about "isolated bench" AP etc measurements is that they depict best-case scenarios for the item under test. This fully OK with me as we want to know the limits of what a device is capable of in the first place. But to which amounts these results are achieved in a "typical" compound user setup is a different matter and this doesn't get enough spotlight here. Some devices will hold their specs, other will degrade severely.
I have suggested sort of a cross-current generation and immunity testing here before, looking at a) how much cross current (eg mains leakage currents) a device emits and b) how much impact a forced cross current through the device has. There are quite some surprises here.

But if you are interested in the difference in performance between the class AB and the class D, why hook it up to a crappy SMPS and a pathetic cable that skew all your measurements? Sounds like exactly the wrong way to do this.

 

[pma]

But if you are interested in the difference in performance between the class AB and the class D, why hook it up to a crappy SMPS and a pathetic cable that skew all your measurements? Sounds like exactly the wrong way to do this.

How many customers who have bought the amplifier that was used, AIYIMA A07, which has only single-ended RCA input, will feed it from a "non-crappy" SMPS (your own words used) with low leakage or put the PSU behind an isolation transformer with low stray capacitance as I did? This forum is mostly about common users and they will use something as Amir in his review

AIYIMA A07 TPA3255 Review (Amplifier)

This is a review and detailed measurements of the AIYIMA A07 desktop amplifier based on TPA3255 class D amplifier chip. It was kindly purchased new by a member and drop shipped to me. The cost is US $72 on Amazon including prime shipping. And as of this shipping, it also included a 32 volt, 5...

www.audiosciencereview.com

Look at his measurements https://www.audiosciencereview.com/...asurements-32v-5-amp-power-supply-png.102788/
and you can see the SMPS current leakage effect, exactly what was shown here in this thread then as an effect at acoustical side.

Look at my measurement with a linear regulated PSU, as a comparison

https://www.audiosciencereview.com/forum/index.php?attachments/a07_4562_5w_4r__30vregpsu-png.104317/

Similar result as with the linear regulated PSU is obtained with the SMPS + isolation transformer. Again, how many common users will do it? They have to rely on what they buy on the consumer electronics market.

Yes there is no question that the issue can be fixed. But such information is normally NOT provided to the common user.

 

[jan.didden]

How many customers who have bought the amplifier that was used, AIYIMA A07, which has only single-ended RCA input, will feed it from a "non-crappy" SMPS (your own words used) with low leakage or put the PSU behind an isolation transformer with low stray capacitance as I did? This forum is mostly about common users and they will use something as Amir in his review

AIYIMA A07 TPA3255 Review (Amplifier)

This is a review and detailed measurements of the AIYIMA A07 desktop amplifier based on TPA3255 class D amplifier chip. It was kindly purchased new by a member and drop shipped to me. The cost is US $72 on Amazon including prime shipping. And as of this shipping, it also included a 32 volt, 5...

www.audiosciencereview.com

Look at his measurements https://www.audiosciencereview.com/...asurements-32v-5-amp-power-supply-png.102788/
and you can see the SMPS current leakage effect, exactly what was shown here in this thread then as an effect at acoustical side.

Look at my measurement with a linear regulated PSU, as a comparison

https://www.audiosciencereview.com/forum/index.php?attachments/a07_4562_5w_4r__30vregpsu-png.104317/

Similar result as with the linear regulated PSU is obtained with the SMPS + isolation transformer. Again, how many common users will do it? They have to rely on what they buy on the consumer electronics market.

Yes there is no question that the issue can be fixed. But such information is normally NOT provided to the common user.

But the question was about the difference between the amplifiers. If you now add additional boundary conditions, the question should have been put differently.
It might, for instance, have been posed as follows: given that users of x and y normally use crappy SMPSs and pathological cables, is there a measurable difference between x or y in such a situation? That might lead to a useful answer.

Not that it would be good research to try to answer a multiple variable question - it normally only leads to confusion because of the intertwined variables.

 

[pma]

But the question was about the difference between the amplifiers. If you now add additional boundary conditions, the question should have been put differently.
It might, for instance, have been posed as follows: given that users of x and y normally use crappy SMPSs and pathological cables, is there a measurable difference between x or y in such a situation? That would lead to a useful answer.

I agree that the original question might be misleading and I have just corrected the title.

On the other hand, I stand firmly at my position that states that any audio system must be considered and measured as a complete chain. And that it is the designer who is responsible for the behaviour of components that are interconnected, when something like two class I components connected by a single ended link is inevitable. It is the designer who should have adopted the rules to prevent issues.

We know from the user reactions at this forum that even the best components (individually measured under clinical conditions) are not free from buzz issues when interconnected. Including Topping, NCore, Purifi especially when class I PC is a sound source.

Last, "the pathological cable" you speak about is 75cm of RG-59.

 

[jan.didden]

Yes I agree, for the user, a system should be considered and judged as a system. I'm not sure a component designer should be responsible for the complete system. If I design an amp, I can't reasonably guarantee the performance no matter what supply is used. What I should do is make sure that my design has a reasonable PSRR, and is reasonably insensitive to EMI for instance. But often, if you buy say a streamer, you have a choice of different (outboard) power supplies, with different performance of the unit (or so we are lead to believe). It's not a perfect world Pavel as you know!

Then again, my Purify set plays completely hum free from my PC, right now ...

 

[pma]

I have modified the thread title and made more measurements as I have now not only the AIYIMA A07, but also the amplifier based on Hypex UcD180HG module.
Measurements were made with Quadral Ascent 90 speaker, on tweeter axis, microphone placed 50mm from tweeter. At 2.8V amplifier voltage. Both acoustical frequency response and amplifier output electrical frequency response were measured, together with distortions at electrical and acoustical side as well. Especially an investigation of AIYIMA A07 LC filter interaction with speaker was of my interest.

Acoustical frequency response

We can see that frequency response of Quadral Ascent driven from A07 is about 1dB and more above response with UcD180 starting from 10kHz.

To get more precise view, let's make A07 over UcD180 comparison plots

A07 over UcD180 at acoustical side

Gradual rise of acoustical output of A07 can be seen, resulting in more bright sound.

A07 over UcD180 at electrical side

Same rise we can see at electrical side. It proves that the difference in frequency response measured at electrical side is directly transformed into speaker acoustical output.

Distortion

The image below shows distortion measured both at acoustical and electrical side with the two amplifiers under test. We can see that though AIYIMA has "better" SINAD result (one number at one frequency and resistor load, 5W/1kHz/4ohm, as used in ASR reviews), its electrical distortion with Quadral speaker measured at 2.8V is higher than that of UcD180. Acoustical distortion is very similar, with A07 a little bit worse, however this difference would be negligible.
Please note that the jump down in A07 distortion at 7kHz is only because of the fact that the measuring bandwidth was 22kHz and this is the point where 3rd harmonic measurement is just finished.

So the main difference remains in frequency response affected by class D amplifier circuit topology. UcD has flat response independent of load, A07 response is load dependent. Might be noted that UcD distortion is also less affected by load and frequency. Might be also noted that none of these features we may read from SINAD 1kHz 5W 4ohm comparative chart.

 

[SIY]

Is this just an output impedance issue? Most speakers have a rising impedance at increasing frequency, so that’s the kind of frequency response aberration one expects irrespective of the amp class or topology if source impedance is not low. I’ve seen similar stuff with AB amps with reduced open loop gain in the top octaves.