• Welcome to ASR. 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!

DAC listening test files - updated with new filters files

Depends.

To make it clear about the test files that I linked in post #1, they are level matched and the clock drift is corrected in DW. The clock drift correction was necessary to get reasonable result when the files are analyzed by our members. On the other hand, there was no correction of amplitude/phase response and no correction of nonlinear distortion. Just to put one number, THD/1kHz/-1dBFS is -130.8 dB and THD+N/1kHz is -118.5 dB. Still, these numbers are pointless. And a note - the files are NOT two consecutive recordings from one and same DAC. There is a BIG HW difference.
Very interesting. I think I’ll be busy next week :)

Did you try to compute 1kHz sine tone before and after level matching, clock drifting correction? I mean does the processing introduce other type of distortion or random/correlated noise to the signal?
 
BTW: 2/16 is far beyond statistical noise: are you not concerned you simply did A for B ?
It is just a chance and statistics. I would not try to find any deeper background.
 
PKMetric as a single number is an RMS (average) over the whole recording, so there could be much larger components to it than -100dB - a graph of PKMetric is a better way to look at it. In fact, -100dB PKM speaks pretty poorly of the A/B channel matching in the ABX box.
Two consecutive recordings with the same circuit and a consistent clock drift between DAC and ADC should produce PKMetric in the -120 to -130dB range assuming the noise level isn't high and doesn't fall in the more "sensitive" frequencies where it would affect the result more.
Interesting, thanks for sharing. And that means I’ll have plenty new stuff to play with when I get back home. Nice :p
 
Did you try to compute 1kHz sine tone before and after level matching, clock drifting correction? I mean does the processing introduce other type of distortion or random/correlated noise to the signal?
Yes I did both. The DW result is almost same in both cases. Level matching does nothing, it is calculated with high precision. Without level matching, one evaluates possible level differences only, which might be something like 0.1-0.2dB without matching. And this would be a fertile soil for further speculations.
 
So, we do not seem to have a valid positive ABX result? Shall I disclose or let you suffer a bit for some time?
 
So, we do not seem to have a valid positive ABX result? Shall I disclose or let you suffer a bit for some time?
I say disclose. I was going to try a few more trials tonight, but I am almost sure I am just guessing and luck will even out.
 
I would suffer and try an ABX tomorrow, equally to suffer for a disclosure before it tomorrow. :cool: ;)
 
Here's mine:
Code:
foo_abx 2.1 report
foobar2000 v1.6.9
2026-08-07 21:26:50

File A: file_1.wav
SHA1: 21b718084cbefae84cc1723ec49976b3c9ffa9ee
File B: file_2.wav
SHA1: 53949d842b273ee3b08edcc880344359b6e6ad50

Used DSPs:
MathAudio Headphone EQ

Output:
WASAPI (push) : Headphones (Meizu HiFi DAC Headphone Amplifier PRO), 32-bit
Crossfading: YES

21:26:50 : Test started.
21:27:41 : 00/01
21:28:12 : 01/02
21:29:07 : 02/03
21:29:42 : 02/04
21:30:11 : 02/05
21:31:29 : 02/06
21:31:44 : 03/07
21:31:53 : 04/08
21:32:07 : 04/09
21:32:37 : 05/10
21:33:07 : 05/11
21:33:21 : 05/12
21:33:39 : 05/13
21:33:54 : 05/14
21:34:07 : 05/15
21:34:21 : 06/16
21:34:33 : 06/17
21:34:47 : 07/18
21:34:56 : 08/19
21:35:20 : 08/20
21:35:20 : Test finished.

 ----------
Total: 8/20
p-value: 0.8684 (86.84%)

 -- signature --
a6c1de95d3405798d1fff432ac2f331bc9ea56fd
 
  • Like
Reactions: pma
I might give it a try just to get some more statistics but not tonight. So wait a day or two. I would need some time to train on the tracks.
 
BTW: 2/16 is far beyond statistical noise: are you not concerned you simply did A for B ?
I do ABX just as I listen to music, no more effort, that would defeat the purpose as I see it, even if I'm trained.

Did a second one, 3/16 this time.
Every time I think I'm hearing some more grain from the one of the two, turns out I choose the opposite.
No error, that's for sure, or swap them, just like any other ABX I have done and posted here.
Grateful Dead, 1970. People are never satisfied with music choices :);).

1970 is a good year for recordings, what happened to this one? :confused:
Ok, I know Grateful Dead and I'm sure I like some work I don't remember by name.
But boi, they are boring, what happened to drama? (spoiled by classical, next time do Rachmaninoff :cool: )
 
I do ABX just as I listen to music, no more effort, that would defeat the purpose as I see it, even if I'm trained.

Did a second one, 3/16 this time.
Every time I think I'm hearing some more grain from the one of the two, turns out I choose the opposite.
No error, that's for sure, or swap them, just like any other ABX I have done and posted here.


1970 is a good year for recordings, what happened to this one? :confused:
Ok, I know Grateful Dead and I'm sure I like some work I don't remember by name.
But boi, they are boring, what happened to drama? (spoiled by classical, next time do Rachmaninoff :cool: )
So you have 5/32. You need to reverse your thinking.
 
So you have 5/32. You need to reverse your thnking.
It's the second time someone tells me that this week, for different reasons :p
I'm absolutely sure it's pure luck (or un-luck) , I wouldn't give it a second thought.

There was two things I was searching for, one was wide vs narrow, accurate or similar effects (so some phase change) and the usual grain, etc.
I did felt one of the two was somehow lower in level though (the second one, file_2) , but not every time, and that was strange.
 
I've tried a few would-be "tells" but haven't gotten a strong positive result yet, my ears are tired now.
 
I did felt one of the two was somehow lower in level though (the second one, file_2) , but not every time, and that was strange.
I had that impression too, even though I knew they were corrected. And I was listening only to the same few seconds each time. And also not most of the time but a few times. Surely a brain thing.
 
It's the second time someone tells me that this week, for different reasons :p
I'm absolutely sure it's pure luck (or un-luck) , I wouldn't give it a second thought.

There was two things I was searching for, one was wide vs narrow, accurate or similar effects (so some phase change) and the usual grain, etc.
I did felt one of the two was somehow lower in level though (the second one, file_2) , but not every time, and that was strange.
If it had been a positive result, i.e. 27/32, we would have considered it statistically significant, but apparently either you have a real ability to identify and reverse the differences or we simply need to realize how, on a few attempts, luck can be a factor, both positively and negatively.
 
Here's mine:
Code:
foo_abx 2.1 report
foobar2000 v1.6.9
2026-08-07 21:26:50

File A: file_1.wav
SHA1: 21b718084cbefae84cc1723ec49976b3c9ffa9ee
File B: file_2.wav
SHA1: 53949d842b273ee3b08edcc880344359b6e6ad50

Used DSPs:
MathAudio Headphone EQ

Output:
WASAPI (push) : Headphones (Meizu HiFi DAC Headphone Amplifier PRO), 32-bit
Crossfading: YES

21:26:50 : Test started.
21:27:41 : 00/01
21:28:12 : 01/02
21:29:07 : 02/03
21:29:42 : 02/04
21:30:11 : 02/05
21:31:29 : 02/06
21:31:44 : 03/07
21:31:53 : 04/08
21:32:07 : 04/09
21:32:37 : 05/10
21:33:07 : 05/11
21:33:21 : 05/12
21:33:39 : 05/13
21:33:54 : 05/14
21:34:07 : 05/15
21:34:21 : 06/16
21:34:33 : 06/17
21:34:47 : 07/18
21:34:56 : 08/19
21:35:20 : 08/20
21:35:20 : Test finished.

 ----------
Total: 8/20
p-value: 0.8684 (86.84%)

 -- signature --
a6c1de95d3405798d1fff432ac2f331bc9ea56fd
Thanks!
 
If it had been a positive result, i.e. 27/32, we would have considered it statistically significant, but apparently either you have a real ability to identify and reverse the differences or we simply need to realize how, on a few attempts, luck can be a factor, both positively and negatively.
I am sure you are right, but still it would be interesting for him keep to going!
 
I will not disclose completely, but I'll give a strong hint.

file_1 against original data file:
file_1 vs original.png
(please note it is bit perfect)

file_2 (that is level matched and drift corrected) vs. uncorrected recorded file:
file_2_driftcorr.png

Level matching, drift correction (and trim):

Initial peak values Reference: -0.1dB Comparison: -0.007dB
Initial RMS values Reference: -13.578dB Comparison: -13.586dB

Final peak values Reference: -0.1dB Comparison: 0.01dB
Final RMS values Reference: -13.59dB Comparison: -13.581dB
Gain= -0.0169dB (0.9981x) DC=0 Phase offset=-0.00005ms (-0.005 samples)
 
Last edited:
Back
Top Bottom