Blind test - can you tell the class D amp from original loop in ABX

[pma]

This is the test of the looback file vs. the file recorded through a class D amplifier. The files for download are here:

Mutter.zip

Shared with Dropbox

www.dropbox.com

If you are interested, please download the zip file, unzip it and listen. Dropbox account is not needed, just click somewhere outside the login box and it will disappear. I would like for you to support your result by a protocol from the ABX test, like provided in Foobar plug-in, e.g.

P.S.: the music in the test is not an "Easy listening". It is from the Beethoven Violin Concerto, movement 3.

 

[pma]

To encourage those who might hesitate a bit, this is what the DeltaWave SW by @pkane is saying:

 

[sq225917]

Is that measured at the output of the class d amp when driving a speaker vs the input, or is no speaker connected?

 

[pma]

It is loaded with 6.8 ohm/200W resistor and output power is up to 118W in loudest passages.

P.S.: I would like to ask those who claim to hear a difference and have the ABX protocol to post it here. Otherwise I cannot take the vote into account and consider it invalid.

P.S2: I tried the test and achieved 7/16 score.

 

[sq225917]

Not to labour the point, but surely that's not representative of the reactive load presented by a speaker to the amp?

 

[Blumlein 88]

Not to labour the point, but surely that's not representative of the reactive load presented by a speaker to the amp?

Hey this is a good first step. Don't let perfect be the enemy of good.

 

[pma]

Not to labour the point, but surely that's not representative of the reactive load presented by a speaker to the amp?

Yes it is not, but this amplifier has frequency response independent of load impedance. There is just some small dependence of distortion on load impedance.
I would not be able to survive if speaker next to me would play at the level used in the test. And my complex dummy load would also not survive, so the resistor was the only chance and quite representative in case of this amp. This would not be the same if it was the TPA32XX or similar amp, however TPA was tested into dummy load earlier and posted here. But at much lower power.

 

[sq225917]

Gotcha, thanks for additional info

 

[Pdxwayne]

Anything you are trying to prove with this abx?

 

[pma]

Anything you are trying to prove with this abx?

Maybe .

 

[pma]

The amp under test is this one (as many of you probably already have guessed):

https://www.audiosciencereview.com/...r-amplifier-module-analysis-and-review.27852/

From the review, one can see how high or low is the “SINAD” and also the amp frequency response into 4 and 8 ohm load and into a loudspeaker.

In the ABX test here, a classical music sample was used, with high dynamics, big differences between silent and loud passages and playing both a single instrument (violin) and a big philharmonic orchestra. The silent passage between 16 s and 28 s has average level only -44dBFS. The loud passage between 32 s and 60s has average level of -21dBFS, and it is quite a difference. The highest peak is at -3dBFS. So a kind of music like this checks very well amplifier linearity and dynamic range, the complex character of the philharmonic orchestra reveals possible “melting” of sound by intermodulations. At the peak, amplifier output is driven to 41Vpeak output voltage, this to 6.8 Ohm/200W resistor. Peak power is 247W, equivalent sine power would be 123.6W. The recorded voltage from amplifier output was compared to loopback record of the same music sample. The comparison was made in DeltaWave and shown in post #1. Pkmetrics says -80.2 dBr(rms). The DeltaWave difference file is almost inaudible. I made an analysis in Adobe Audition, where the difference (“delta”) file is amplified 100x (40dB) and shown with the original data. Even if 100x amplified, the level of the delta file is very low compared to original, only in case of several peaks, where the output voltage exceeds some 33Vpeak (23.5Vrms eq., 81W/6.8 Ohm), the amp nonlinearity and proximity to clipping plays its game and the deviation from original and rise of delta signal is visible. Visible, but not audible, the plot below shows the highest “distorted” peak, again 100x amplified, but its width is only 317 microseconds. So as long as we are below some 80W, there is no chance of audible differences. This is my point, keep the frequency response invariant of load and then the SINAD number is almost pointless, of course if kept below a reasonable level better than 60 dB. Rather take care about noise measured over the audio band, to be able to play well the silent passages.

 

[Pdxwayne]

With Topping E30, Topping L30, and AKG371, I could only do 17/30. I voted I can't hear a difference. I can't try my stereo system as it would drive my family members crazy....

 

[Pdxwayne]

Yesterday, I focused on other sections of the song and I did not do better than guessing.

Earlier today, I focused on 45 second to 1 minute of song and I got 8/10, 4/10, 5/10 as indicate in my post above. Not that good. Before I gave up, I decided to try one more time (focus on 45 seconds to about 50 seconds of song) and I got 8/10.

For song section from 45 seconds to about 1 minute, I got total of 25/40. I think probability of 7.69% is better than guessing, right?

 

[Triliza]

Did two tests, got 9/16 and 10/16. Sometimes I thought I could hear some differences (like one of the tracks being slightly louder), but from the results it seems I was just guessing/imagining more or less. So I'll vote no.

 

[Pdxwayne]

Did two tests, got 9/16 and 10/16. Sometimes I thought I could hear some differences (like one of the tracks being slightly louder), but from the results it seems I was just guessing/imagining more or less. So I'll vote no.

Track two is slightly louder. You sensed that even when using replay gain to volume match in foobar2000?

 

[Triliza]

Track two is slightly louder. You sensed that even when using replay gain to volume match in foobar2000?

This was the first time I used the ABX plugin on foobar, I didn't use the replay gain option, should I have used it? Should I always use replay gain for some other tests I have in mind (mp3 vs flac and such)?

 

[Pdxwayne]

This was the first time I used the ABX plugin on foobar, I didn't use the replay gain option, should I have used it? Should I always use replay gain for some other tests I have in mind (mp3 vs flac and such)?

I am pretty new user too. Just started doing abx with foobar200 this year. In this instance where you compare a captured song with original, I think it is best to volume match. For others, I guess it depends?

 

[Blumlein 88]

This was the first time I used the ABX plugin on foobar, I didn't use the replay gain option, should I have used it? Should I always use replay gain for some other tests I have in mind (mp3 vs flac and such)?

No you do not want to use Replay gain for ABX tracks. You do want matched levels, but Replay gain more or less matches general perceived loudness which is a subtle difference between level matching. The matching needs doing to the tracks before you ABX them in Foobar.

 

[Pdxwayne]

No you do not want to use Replay gain for ABX tracks. You do want matched levels, but Replay gain more or less matches general perceived loudness which is a subtle difference between level matching. The matching needs doing to the tracks before you ABX them in Foobar.

Hmm, in this particular abx, file 2 is slightly louder. Did you volume match before abx? If you did, which app did you use to adjust volume?

 

[Pdxwayne]

Yesterday, I focused on other sections of the song and I did not do better than guessing.

Earlier today, I focused on 45 second to 1 minute of song and I got 8/10, 4/10, 5/10 as indicate in my post above. Not that good. Before I gave up, I decided to try one more time (focus on 45 seconds to about 50 seconds of song) and I got 8/10.

For song section from 45 seconds to about 1 minute, I got total of 25/40. I think probability of 7.69% is better than guessing, right?

View attachment 165439View attachment 165440View attachment 165441
View attachment 165442

Since pma likes large number of trials, I decided to do one more. Only got 6/10.

Cumulative so far is 31/50. Cumulative probability now lower a little to ~5.94% of guessing.