• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

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

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,592
Likes
10,728
Location
Prague
Following the assumptions on possible influence of ultrasonic noise and carrier frequency residuals to tweeter output I have prepared an experiment where the tweeter is driven by 2.82V sine tone of 6kHz from 2 different amplifiers. One of the is AIYIMA A07, see my review

the other one is a bridged amplifier based on my re-designed Sinclair Z-30 circuit

Both amplifiers work deeply below clipping, as the output voltage was set to 2.82V.
In addition, 19+20kHz CCIF distortion test was made as well.

This is the measurement setup
classDxAB_test.JPG

Focusrite 2i2 Gen3 soundcard was used, set to 96kHz/24bit sampling.


The results

1. Distortion of sine 6kHz


Class AB amplifier
Magnat_tweeter_6k_Z30B.png



Class D amplifier
Magnat_tweeter_6k_A07.png



2. Distortion of 19+20kHz twin tone

Class AB amplifier
Magnat_tweeter_19+20k_Z30B.png



Class D amplifier
Magnat_tweeter_19+20k_A07.png


Conclusion

It is above any doubts that the spectrum at acoustical side measured from the class AB amplifier under test is more clean. The spectrum from the class D amplifier under test is messy and contains strange non-harmonic components.
 

mcdn

Addicted to Fun and Learning
Forum Donor
Joined
Mar 7, 2020
Messages
559
Likes
780
Hardly a surprise though, is it? Of course a switching amp has artifacts from intermodulation of a signal with the switching frequency, but they're all tiny. The additional artifacts are at -80dBc compared with the primary IM distortion which is at -29dBc for both amps.
 

oivavoi

Major Contributor
Forum Donor
Joined
Jan 12, 2017
Messages
1,721
Likes
1,935
Location
Oslo, Norway
Thanks, really interesting! For absolute fidelity it does seem like class AB and/or SOTA class D (like Hypex/Purifi) is better for tweeter duty than cheaper class D. At the same time, I'm also left wondering about the audible importance of these distortions... they do seem to be very low in level?
 

mcdn

Addicted to Fun and Learning
Forum Donor
Joined
Mar 7, 2020
Messages
559
Likes
780
The TPA3255 can do better given a better implementation (mainly the output inductors and circuit layout). TI’s eval board is very cheap and did a bit better than the results above in a test by Quantasylum: https://quantasylum.com/blogs/news/ti-tpa3255-evm. It uses a higher default rail voltage than the A07, and might use better inductors etc. TI have a whole spreadsheet on various inductors‘ performance with the TPA3255.

Anyway, assuming an 85dBSPL/1m/2.83V speaker efficiency, these distortion products would be a whole 5dB in volume at 1m at the tested output level. I don’t think there’s any research showing audibility of any artefacts on any signal at that level.
 

pjug

Major Contributor
Forum Donor
Joined
Feb 2, 2019
Messages
1,775
Likes
1,561
This is interesting. I think you previously measured the A07 and these artifacts are not present in amp's electrical signal. So then this is evidence that Benchmark may have a valid a point when they say "Tweeters can demodulate this ultrasonic noise and fold it into the audible band."

It would be interesting for someone to do this with a top nCore or Purifi amplifier.

edit: Do you see anything with silence input to the A07, where it would still be putting out the ultrasonics?
 
Last edited:

Matias

Master Contributor
Forum Donor
Joined
Jan 1, 2019
Messages
5,031
Likes
10,806
Location
São Paulo, Brazil
I wish you could measure those with an NC252MP, would be so interesting.
 

pjug

Major Contributor
Forum Donor
Joined
Feb 2, 2019
Messages
1,775
Likes
1,561
There is more dirt in the class d. But its far under the dirt of the tweeter. So?
Fair point for an inexpensive amp. But if we are going for a high end Class D amp that is expected to be very clean, and what comes out of the tweeter is not as clean as Class AB, then that's a big deal. Of course I don't know whether you'd see this in a better Class D amp; it would be interesting to find out.
 

tomtoo

Major Contributor
Joined
Nov 20, 2019
Messages
3,607
Likes
4,514
Location
Germany
Fair point for an inexpensive amp. But if we are going for a high end Class D amp that is expected to be very clean, and what comes out of the tweeter is not as clean as Class AB, then that's a big deal. Of course I don't know whether you'd see this in a better Class D amp; it would be interesting to find out.

Sure from a theoretical standpoint we shoul find out. But its more complicated. Can we hear it?
 

MaxBuck

Major Contributor
Forum Donor
Joined
May 22, 2021
Messages
1,515
Likes
2,118
Location
SoCal, Baby!
Following the assumptions on possible influence of ultrasonic noise and carrier frequency residuals to tweeter output I have prepared an experiment where the tweeter is driven by 2.82V sine tone of 6kHz from 2 different amplifiers. One of the is AIYIMA A07, see my review

the other one is a bridged amplifier based on my re-designed Sinclair Z-30 circuit

Both amplifiers work deeply below clipping, as the output voltage was set to 2.82V.
In addition, 19+20kHz CCIF distortion test was made as well.

This is the measurement setup
View attachment 160284
Focusrite 2i2 Gen3 soundcard was used, set to 96kHz/24bit sampling.


The results

1. Distortion of sine 6kHz


Class AB amplifier
View attachment 160285


Class D amplifier
View attachment 160286


2. Distortion of 19+20kHz twin tone

Class AB amplifier
View attachment 160287


Class D amplifier
View attachment 160288

Conclusion

It is above any doubts that the spectrum at acoustical side measured from the class AB amplifier under test is more clean. The spectrum from the class D amplifier under test is messy and contains strange non-harmonic components.
You've compared a single example of an AB amp to a single example of a D amp. Get back to me with more data (more amp models) before I am persuaded you've identified anything meaningful.
 

audio2design

Major Contributor
Joined
Nov 29, 2020
Messages
1,769
Likes
1,830
Anyway, assuming an 85dBSPL/1m/2.83V speaker efficiency, these distortion products would be a whole 5dB in volume at 1m at the tested output level. I don’t think there’s any research showing audibility of any artefacts on any signal at that level.

And this post pretty much renders most else said irrelevant, but let's throw in equal loudness contours for good measure :)
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,592
Likes
10,728
Location
Prague
Another question I have: is it possible that the artifacts are not acoustic signals, but rather aliases of electrical pickup of the ultrasonics.

No. They really exist and they are inside acoustical spectrum, beyond any doubts. No electrical pickup, the pickup of interferences is thoroughly tested as this is the main issue I am concentrated on, and almost nobody else is, because the components are tested here in ASR and magazine reviews in a "clinical" setup.

The reason of the mess in acoustical signal with AIYIMA A07 was found and is interesting, though not unexpected. It is the SMPS leakage current and how it adds on the A07 single ended RCA input cable. What is interesting to me is that it intermodulates with the useful signal differently for every type of input signal (silence, sine, twin-tone etc.) and its frequency, most probably related to SMPS changing load and thus switching operation and leakage.

First, a comparison with no input signal, only soundcard connected to amp input and microphone measures acoustical output. I must emphasize that this is an acoustical measurement, not electrical. And I must also add that the gain of both amplifiers is made same, so the amplification of the soundcard output noise goes to same values.

Magnat_tweeter_background_Z30B_A07_dBFS.png

Green plot is with AIYIMA A07, violet plot with Z30B. A07 is worse, however not that much as with the 19+20kHz test.
Just for a comparison, review the #1 post and results for 19+20kHz IMD.

Then, before the SMPS power supply that feeds the AIYIMA, the isolation transformer 1:1 (230V/230V) was connected, with stray capacitance only of 120pF. And this has fixed the issue. The messy spectral lines have disappeared and this is the result:
Magnat_tweeter_19+20k_A07_no PE.png


It is good that the issue was fixed however it is not a good news. The input signal for AIYIMA A07 was sent from the notebook and external soundcard, A07 was connected only to a speaker and nothing else. Still, we had a messy spectrum due to SMPS leakage currents and worse, extremely dependent on the input signal shape. The only possible fix was to use an isolation transformer before SMPS, which is definitely not what the common user is doing.
This story shows again that measurements of single isolated audio components are insufficient and that surprisingly low level signals are successfully radiated by the speaker on its acoustical side.
It is also interesting to see that the speaker easily radiates the signal noise shaping that originates from the DAC, up to the frequency limit when the frequency response of the tweeter starts to fall rapidly.
 
Last edited:

MakeMineVinyl

Major Contributor
Joined
Jun 5, 2020
Messages
3,558
Likes
5,871
Location
Santa Fe, NM
Not surprising results; the ultrasonic out of band noise from a class D amplifier is going straight to the tweeter, and it has to have some effect, even if the result is below the level of audibility. Personally, I'd never choose a class D amplifier for my ultra-sensitive compression drivers, but with less insane speakers, there would undoubtedly be no issue.
 

DVDdoug

Major Contributor
Joined
May 27, 2021
Messages
2,920
Likes
3,835
a measurable difference
Usually we can measure things that we can't hear.... I'm at work right now looking at an oscilloscope that goes-up to 100MHz. An amplifier that goes to 100mHz isn't necessarily "better" than one that goes to 20kHz.

Comparing one random class A/B amplifier to another random class-D amplifier doesn't prove anything although I would generally expect a class-D amplifier to be noisier, especially in the ultrasonic range. But if the noise is below audibility it's it has no effect on sound quality.
 

pjug

Major Contributor
Forum Donor
Joined
Feb 2, 2019
Messages
1,775
Likes
1,561
Following the assumptions on possible influence of ultrasonic noise and carrier frequency residuals to tweeter output I have prepared an experiment where the tweeter is driven by 2.82V sine tone of 6kHz from 2 different amplifiers. One of the is AIYIMA A07, see my review

the other one is a bridged amplifier based on my re-designed Sinclair Z-30 circuit

Both amplifiers work deeply below clipping, as the output voltage was set to 2.82V.
In addition, 19+20kHz CCIF distortion test was made as well.

This is the measurement setup
View attachment 160284
Focusrite 2i2 Gen3 soundcard was used, set to 96kHz/24bit sampling.


The results

1. Distortion of sine 6kHz


Class AB amplifier
View attachment 160285


Class D amplifier
View attachment 160286


2. Distortion of 19+20kHz twin tone

Class AB amplifier
View attachment 160287


Class D amplifier
View attachment 160288

Conclusion

It is above any doubts that the spectrum at acoustical side measured from the class AB amplifier under test is more clean. The spectrum from the class D amplifier under test is messy and contains strange non-harmonic components.
So this is a power supply issue and nothing to do with Class D ultrasonics? Would a Class AB amplifier do any better with a poor SMPS ?
 

mcdn

Addicted to Fun and Learning
Forum Donor
Joined
Mar 7, 2020
Messages
559
Likes
780
Then, before the SMPS power supply that feeds the AIYIMA, the isolation transformer 1:1 (230V/230V) was connected, with stray capacitance only of 120pF. And this has fixed the issue. The messy spectral lines have disappeared and this is the result
Well that's interesting. What mechanism do you think is at work? The transformer should really only remove a path to ground for the SMPS (including via the neutral wire which also ultimately runs to ground) and any DC offset on the mains supply. Everything else from the 50Hz signal up should pass through the transformer unchanged.

So it's great that adding the transformer removes the spuriae, but I'm not knowledgable enough to figure out how it does it!
 

audio2design

Major Contributor
Joined
Nov 29, 2020
Messages
1,769
Likes
1,830
This story shows again that measurements of single isolated audio components are insufficient and that surprisingly low level signals are successfully radiated by the speaker on its acoustical side.

Preaching to the choir, even if a choir of one (me). I have said many times w.r.t. measurements on ASR. The Audio Precision devices are designed to isolate their performance as much as possible from the system, i.e. they are relatively immune to ground loops and the like. i.e. They won't add anything that is not there initially. When you test a DAC or other device into the AP, the AP is very good at not adding in any other noise sources. That is not true of other audio devices in the chain.
 

AnalogSteph

Major Contributor
Joined
Nov 6, 2018
Messages
3,338
Likes
3,278
Location
.de
So it's great that adding the transformer removes the spuriae, but I'm not knowledgable enough to figure out how it does it!
The ground loop currents are also a function of amplifier power supply current. In ham radio setups with substantial RF power amplifiers, the added distortion can be so strong that you'd think there was RF getting into the microphone input when in fact it's just a bog standard ground loop involving e.g. a PC audio connection.
 
Top Bottom