• 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!

Listening test of 2 power amplifiers - files recorded for download - disclosed

Can you hear a difference and which file do you prefer

  • I can hear a difference but have no ABX result

    Votes: 4 14.3%
  • I cannot hear a difference but have no ABX result

    Votes: 14 50.0%
  • I can hear a difference and have an ABX result

    Votes: 2 7.1%
  • I cannot hear a difference and have an ABX result

    Votes: 4 14.3%
  • I prefer witch1 file

    Votes: 1 3.6%
  • I prefer witch2 file

    Votes: 8 28.6%

  • Total voters
    28
  • Poll closed .
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
There are tons of wrong assumptions and false conclusions in this thread (of course not only in this thread). Thanks to @pkane , @Blumlein 88 and few others for keeping the right track. The poll expires tomorrow and I will bring the disclosure and also one of the test files x the original file. It will be a bit shorted and renamed so the direct listening comparison with the witch1 and witch2 files will be impossible, to prevent speculations.
 

pjug

Major Contributor
Forum Donor
Joined
Feb 2, 2019
Messages
1,775
Likes
1,561
The track I used is freely available on HDTracks in their 2020 Sampler, it's called I Put a Spell on You by Chantal Chamberland.
I tried comparing the original track and the AHB2 recording, and I can't get a very good match. Maybe I am doing something wrong. Has anyone else tried this?
Edit: Even your DAC3-APx515 Loopback recording does not match as well as I would expect. I did the same with a $30 Behringer UCA202 to compare (note that the UCA202 max sample rate is 48KHz so ignore ultrasonics in this). I thought the reason your matching was not better was the clock issue, but maybe it is something else. Or maybe I'm doing something wrong but the DW settings were the same for these two runs (attached).
 

Attachments

  • Spell_UCA202.jpg
    Spell_UCA202.jpg
    267.2 KB · Views: 63
  • Spell_DAC3_APx515_loopback.jpg
    Spell_DAC3_APx515_loopback.jpg
    305 KB · Views: 71
Last edited:

Pdxwayne

Major Contributor
Joined
Sep 15, 2020
Messages
3,219
Likes
1,172
Now I see the software in post 5 calculated the actual value with 0,1213dB.

btw: one thing that cought my atention in the null audio is the esses; very strong. it should be possible to abx focussing on them
Interesting, initially I was focusing on the brightness/sharpness of the vocal (didn't do that well, only 10/16 correct), then I sensed a slight shift of vocal to the right when playing file 2. That was when I was able to pass last 15 of 16 ABX tests.
 
Last edited:

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
Interesting, initially I was focusing on the brightness/sharpness of the vocal (didn't do that well, only 10/16 correct), then I sensed a slight shift of vocal to the right when playing file 2. That was when I was able to pass last 15 of 16 ABX tests.
So yes a channel imbalance would be audible. If that channel is trimmed up in level, then you are back to the test files being inaudibly different.
 

Pdxwayne

Major Contributor
Joined
Sep 15, 2020
Messages
3,219
Likes
1,172
So yes a channel imbalance would be audible. If that channel is trimmed up in level, then you are back to the test files being inaudibly different.
Yeah, the files not of great quality anyway...Both sounded a bit dull. I wonder how the original would sound in my setup.

Anyway, I will be curious about differences between class A and class AB. At least in my living room, my floor stander sounded more 3d with my Parasounnd A21 than class AB Yamaha. No, not going to do blind tests.
: P
 

dasdoing

Major Contributor
Joined
May 20, 2020
Messages
4,209
Likes
2,674
Location
Salvador-Bahia-Brasil
Interesting, initially I was focusing on the brightness/sharpness of the vocal (didn't do that well, only 10/16 correct), then I sensed a slight shift of vocal to the right when playing file 2. That was when I was able to pass last 15 of 16 ABX tests.

yea, witch1 compresses the highs,
I can make the null significantly quieter aplying a high shelf of -0.40dB at 6540Hz on witch2.
I could probably make it even quieter with another high shelf.

[edit] multiband compression in that range gives even deeper null. amp1 is compressing [/edit]

the test in post 5 isn't realy usefull in this case. the null is very low level, but it will add on top. we are not talking about noise here
 
Last edited:
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
Test disclosure
First I would like to thank everyone who participated in the test, sent the vote into the poll and posted comments here in the thread.

The same test was also posted in our local audio forum so I have prepared a summarized result from both.
Witch test results.png

So we have 7 people who voted that they could hear a difference (either with or without an ABX result) and 23 people who voted they could not. This is quite a significant number of participants and quite a significant result, IMO.

Now the technical disclosure. The “witch1” sample was recorded through my A250W4R amplifier that was also described in its thread here in the ASR:
https://www.audiosciencereview.com/...on-blameless-topology-and-measurements.21542/
and on my web
http://pmacura.cz/DIY_250W_4ohm_amplifier.html

This amplifier is capable to give 2x250W into 4ohm load at reasonably low distortion and it is a dual-mono design with two separate power supplies and minimized crosstalk between the 2 channels.

The “witch2” sample was recorded through the modified, now 52 years old topology of the Sinclair Z-30 kit. I posted a thread about this kit and my circuit modifications at

https://www.audiosciencereview.com/...-amplifier-and-40-years-of-my-diy-mods.20198/

It is the non-bridged version and is able to give relatively clean power up to 2x40W/4ohm, however with higher distortion than the A250W, worse PSR (power supply rejection ration) and more interactions with speaker complex impedance between 50Hz and 500Hz.

Because of the big difference of the amplifiers power the files were recorded at peaks up to some 13Vp to keep the smaller amplifier well inside a linear zone. It was not the aim to test audibility of clipping.

As we can see from the poll results, “witch2” was preferred for those who posted a subjective comparison vote, which is in fact a preference for the amplifier with higher distortion. I wanted to comment on this based on many results of my previous tests but then I decided not to do so.

Last, I have prepared a new set where one of the files is the original that was used in the test and the other one is one of the test files (witch1 or witch2) used. The files were a bit shorted and renamed with the aim not to tell easily which file is which. So if you are interested, try this new challenge, but no poll is available now.

https://drive.google.com/file/d/13UpCCZT9glrFH8YNfUTb9oXxADY2hoWd/view?usp=sharing


Test setup
2amp_test_scheme.png


witch1+2_THD_1k.png

THD at 1kHz / 4ohm for both amplifiers

witchFR_matching.png

Deviation of frequency responses of the amplifiers used in the test measured into speaker complex dummy load
 
Last edited:

pogo

Major Contributor
Joined
Sep 4, 2020
Messages
1,241
Likes
382
As we can see from the poll results, “witch2” was preferred for those who posted a subjective comparison vote, which is in fact a preference for the amplifier with higher distortion.
Was it really just the higher distortion that made the difference, nothing else!?
Would be interesting to know if I and of course some others really like more distortion and this also leads to a 'more correct' sound.
But that doesn't make any sense, does it?
But maybe it was due to the recording quality of the example...
 
Last edited:

Pdxwayne

Major Contributor
Joined
Sep 15, 2020
Messages
3,219
Likes
1,172
Test disclosure
First I would like to thank everyone who participated in the test, sent the vote into the poll and posted comments here in the thread.

The same test was also posted in our local audio forum so I have prepared a summarized result from both.
View attachment 147754
So we have 7 people who voted that they could hear a difference (either with or without an ABX result) and 23 people who voted they could not. This is quite a significant number of participants and quite a significant result, IMO.

Now the technical disclosure. The “witch1” sample was recorded through my A250W4R amplifier that was also described in its thread here in the ASR:
https://www.audiosciencereview.com/...on-blameless-topology-and-measurements.21542/
and on my web
http://pmacura.cz/DIY_250W_4ohm_amplifier.html

This amplifier is capable to give 2x250W into 4ohm load at reasonably low distortion and it is a dual-mono design with two separate power supplies and minimized crosstalk between the 2 channels.

The “witch2” sample was recorded through the modified, now 52 years old topology of the Sinclair Z-30 kit. I posted a thread about this kit and my circuit modifications at

https://www.audiosciencereview.com/...-amplifier-and-40-years-of-my-diy-mods.20198/

It is the non-bridged version and is able to give relatively clean power up to 2x40W/4ohm, however with higher distortion than the A250W, worse PSR (power supply rejection ration) and more interactions with speaker complex impedance between 50Hz and 500Hz.

Because of the big difference of the amplifiers power the files were recorded at peaks up to some 13Vp to keep the smaller amplifier well inside a linear zone. It was not the aim to test audibility of clipping.

As we can see from the poll results, “witch2” was preferred for those who posted a subjective comparison vote, which is in fact a preference for the amplifier with higher distortion. I wanted to comment on this based on many results of my previous tests but then I decided not to do so.

Last, I have prepared a new set where one of the files is the original that was used in the test and the other one is one of the test files (witch1 or witch2) used. The files were a bit shorted and renamed with the aim not to tell easily which file is which. So if you are interested, try this new challenge, but no poll is available now.

https://drive.google.com/file/d/13UpCCZT9glrFH8YNfUTb9oXxADY2hoWd/view?usp=sharing


Test setup
View attachment 147756

View attachment 147755
THD at 1kHz / 4ohm for both amplifiers

View attachment 147757
Deviation of frequency responses of the amplifiers used in the test measured into speaker complex dummy load
Just by using deltawave, I can guess that #3 is the amp and #4 is the orig.

Based on Deltawave info, can I assume that the amp is ~0.1db lower than orig for left channel, and ~0.3db louder than orig for right channel?

In that case, the channel imbalance of the amp is ~0.4db between left and right channel?
 

Langston Holland

Member
Forum Donor
Joined
Aug 5, 2021
Messages
58
Likes
119
Location
Pensacola FL
It wasn't the distortion, it was the increased dynamic range* in the right channel that made witch2 sound better.

This is also the give-away clue that witch3 is the original file in the second test set (in addition to the fact that some of the metadata got through with the original file).

/end golden ears/ : )

* Loudness actually. See my post below for correction.
 
Last edited:

Langston Holland

Member
Forum Donor
Joined
Aug 5, 2021
Messages
58
Likes
119
Location
Pensacola FL
I do not think that the right channel in witch2 has increased dynamic range compared to witch1.

Love your efforts on this - you're teaching a lot of folks important stuff here. : )

You're absolutely right, the interchannel dynamic range (difference between the loudest and quietest portions of the song) is nearly identical. I was wrong to use that term, but the right channel of the better sounding track in both cases has higher peaks and averages a bit louder. This is also the case with witch3 compared to witch4. I clearly heard this on my Klipschorns when comparing witch1 to witch2, but with the witch3 and witch4 files I simply looked at the data you presented above, but for tracks 3 and 4 (below).

I still say that witch3 is the original and if I'm right I want an invitation to Prague! : )

God bless you and your precious family - Langston

PS: I wasn't able to hear such small differences until I started using horn loaded loudspeakers.

3.png


4.png
 
Last edited:

pjug

Major Contributor
Forum Donor
Joined
Feb 2, 2019
Messages
1,775
Likes
1,561
Witch1 and Witch4 are almost a perfect match although not bit perfect. Essentially the same, though. That would leave Witch3 as the original.
 

Pdxwayne

Major Contributor
Joined
Sep 15, 2020
Messages
3,219
Likes
1,172
Witch1 and Witch4 are almost a perfect match although not bit perfect. Essentially the same, though. That would leave Witch3 as the original.
Did you compare #2 and #3?
 
Top Bottom