DAC listening test files - updated with new filters files

[pjug]

Any connection to audible differences between DACs is separate. I was conjecturing that the reported differences in cymbals might relate to this -- in the end I have no idea

I think you are talking about the cymbals in Test #2, where @pma was hearing a difference at ~10 seconds in the file? This won't be because of lack of bandwidth since the files are 96kHz and DeltaWave shows the high frequencies rolling off the same in both files. What we know is that the difference in these files is the flatness of the amplitude frequency response, and most of that difference is passband ripple. So audibility of passband ripple is what we should be investigating if we want to know the how and why of how he was able to hear a difference.

Having said that, I don't think passband ripple can explain all of what @Josh83 might have heard. It is pretty unlikely that all of his comparisons would have had this kind of passband ripple difference.

 

[jabbr]

I think you are talking about the cymbals in Test #2, where @pma was hearing a difference at ~10 seconds in the file? This won't be because of lack of bandwidth since the files are 96kHz and DeltaWave shows the high frequencies rolling off the same in both files. What we know is that the difference in these files is the flatness of the amplitude frequency response, and most of that difference is passband ripple. So audibility of passband ripple is what we should be investigating if we want to know the how and why of how he was able to hear a difference.

Having said that, I don't think passband ripple can explain all of what @Josh83 might have heard. It is pretty unlikely that all of his comparisons would have had this kind of passband ripple difference.

We are crosscutting different threads on different sites. I think these are all good possibilities to look at. The cymbals was what Josh reported in reference to his original article. That then led to a discussion (at AS) of whether there could be more objective human tests than ABX and both fMRI and BAER was suggested with literature on BAER dating back to the 1970s. I conjectured that cymbals might invoke the rise time pathway (localization) in the brainstem as distinct from the tone audibility pathway analogous to bat clicks for localization. Agree that 96kHz is a reasonable bandwidth to look at. All that said if anyone with a handy BAER showed a rise time sensitivity in the 2 us - 10 us range then the required bandwidth would rise to 384kHz +/- but that is a theoretical argument at this point. My point of these back of napkin calculations is that the ADI-FS 'new' has the measurement bandwidth to handle necessary bandwidth depending...

 

[pma]

Cymbals were asked as a test signal, so here we go:

https://pmacura.cz/cymbals.zip

There is an effect (quite huge) of 15kHz Bessel LPF in one of the files, so it (should) be audible, only ABX report counts.

There is a lot of silence (transient fade out) between the hits, so be careful with the volume knob.

 

[pjug]

I thought I hear something but I did not score well. I will keep trying because I do think I hear a little difference. There is some construction outside so I will try at night.

foo_abx 2.2.3 report
foobar2000 v2.25.10
2026-09-10 12:28:21

File A: cymbal1.wav
SHA1: e6d2dd8a9ffbf01d03692ecb2c3f482c0b0fc56d
File B: cymbal2.wav
SHA1: 8546a3ec2d321d35b9f9d75a3dd4d18ce6b43ef9

Output:
Default : Headphones (Realtek(R) Audio)
Crossfading: NO

12:28:21 : Test started.
12:29:08 : Test restarted.
12:29:08 : 00/01
12:29:25 : Test restarted.
12:29:25 : 01/02
12:29:43 : Test restarted.
12:29:43 : 01/03
12:30:06 : Test restarted.
12:30:06 : 02/04
12:30:22 : Test restarted.
12:30:22 : 02/05
12:30:43 : Test restarted.
12:30:43 : 02/06
12:31:11 : Test restarted.
12:31:11 : 02/07
12:31:22 : Test restarted.
12:31:22 : 03/08
12:31:31 : Test restarted.
12:31:31 : 04/09
12:31:51 : Test restarted.
12:31:51 : 05/10
12:32:04 : Test restarted.
12:32:04 : 06/11
12:32:21 : Test restarted.
12:32:21 : 07/12
12:33:00 : Test restarted.
12:33:00 : 08/13
12:33:11 : Test restarted.
12:33:11 : 09/14
12:33:19 : Test restarted.
12:33:19 : 09/15
12:33:40 : Test restarted.
12:33:40 : 09/16
12:33:40 : Test finished.

----------
Total: 9/16
p-value: 0.4018 (40.18%)

-- signature --
855c065ae698dd47864285329869ae56a84204f5

 

[pma]

I thought I hear something but I did not score well. I will keep trying because I do think I hear a little difference. There is some construction outside so I will try at night.

Thank you for taking the test.

BTW, below is the "rise time" of the LPF filter used. Distance between the samples is 10 us approx. It should be noted that the rise time debate does not make much sense, it is rather the frequency response to be noted. 15kHz cut-off frequency, 6th order Bessel (math of course).

 

[pjug]

That filter makes me doubt I was picking up on some difference. Doesn't do much below 13kHz where I can hear.

 

[j_j]

How loud (in flat measurement SPL terms) is this at the entrance to the eardrum? (assuming not in-ear headphones)
It would be interesting to run this test at 96k with wideband input, and run a variety of filters on it. If somebody has a wideband (at least 30khz bandwidth) sampled at 96k, CLEAN (no clipping!) 16/20/24 bit recording ( I doubt bit depth is an issue here), I could run some filters. I'll need some place to drop a variety of .wav files.

 

[pjug]

How loud (in flat measurement SPL terms) is this at the entrance to the eardrum? (assuming not in-ear headphones)
It would be interesting to run this test at 96k with wideband input, and run a variety of filters on it. If somebody has a wideband (at least 30khz bandwidth) sampled at 96k, CLEAN (no clipping!) 16/20/24 bit recording ( I doubt bit depth is an issue here), I could run some filters. I'll need some place to drop a variety of .wav files.

Someone made a castanet recording at 96K if that would work. I have the recording but I can't remember who made it (@Blumlein 88 ? @solderdude ?) so I feel bad about sharing (I could probably find the link where I got it). And I don't know if that is the type of sound you are looking for.

 

[j_j]

Someone made a castanet recording at 96K if that would work. I have the recording but I can't remember who made it (@Blumlein 88 ? @solderdude ?) so I feel bad about sharing (I could probably find the link where I got it). And I don't know if that is the type of sound you are looking for.

I think something a bit more sustained would be a lot better than a castanet recording. I have one of those, it's really brutal to listen to, and does not sustain long enough for good bandwidth sensitivity.

 

[pjug]

I think something a bit more sustained would be a lot better than a castanet recording. I have one of those, it's really brutal to listen to, and does not sustain long enough for good bandwidth sensitivity.

Would snipping a section of this work? It is 192K.

Chesky Records - Dr. Chesky's Ultimate Headphone Demonstration Disc

Playlist - Chesky Records - Dr. Chesky's Ultimate Headphone Demonstration Disc - 59 items

tidal.com

 

[j_j]

Would snipping a section of this work? It is 192K.

Chesky Records - Dr. Chesky's Ultimate Headphone Demonstration Disc

Playlist - Chesky Records - Dr. Chesky's Ultimate Headphone Demonstration Disc - 59 items

tidal.com

View attachment 557901

Maybe ask David to ok the use?

 

[pjug]

Maybe ask David to ok the use?

Not sure I know how to contact but I will try. 5 or 10 sec would be enough?

 

[danadam]

It would be interesting to run this test at 96k with wideband input, and run a variety of filters on it. If somebody has a wideband (at least 30khz bandwidth) sampled at 96k, CLEAN (no clipping!) 16/20/24 bit recording ( I doubt bit depth is an issue here), I could run some filters. I'll need some place to drop a variety of .wav files.

There is 24/176.4k recording linked in this article:

Paiste 24" Giant Beat Multi-Functional (Ride) Cymbal - HomeTheaterHifi.com

Paiste's Giant Beat series consists of a 24", 18", and 20" "Multi-functional", and 14" and 15" Hi-Hats. They have a very distinctive sound that one might call "Retro Rock", as the ping is crisp. The 18", 20", and 24" are labeled as "Multi-functional" because they can be used as rides or crashes...

hometheaterhifi.com

 

[j_j]

10-20 seconds cut with some sensible onset.

 

[pma]

It would be interesting to run this test at 96k with wideband input, and run a variety of filters on it. If somebody has a wideband (at least 30khz bandwidth) sampled at 96k, CLEAN (no clipping!) 16/20/24 bit recording ( I doubt bit depth is an issue here), I could run some filters. I'll need some place to drop a variety of .wav files.

I can do it. 30s sample, 96/24, filters as needed - Butterworth, Bessel, Chebyshev, all what is in AA. I have space for upload.

 

[pma]

I can do it. 30s sample, 96/24, filters as needed - Butterworth, Bessel, Chebyshev, all what is in AA. I have space for upload.

I would start with this, the files can be downloaded from the link below:

dchestest.zip

drive.google.com

20kHz brickwall added

dchesF4.zip

drive.google.com

The original 96/24 is from the source discussed, 30s sample. Spectrum shown in the post above. There are 4 files in the zip, the original and 3 filters. I am not using brickwalls. An example of spectra of the original file and with filters are shown below. So, good luck in your ABX testing! Based on the results, we may add another filters.

A hint: most of the ultrasonic content is in the right channel, time 25.48 s. The trumpet. Much more than for the cymbals. Below a spectrum after 8kHz HPF.

 

[Thomas_A]

F3 is very easy:
foo_abx 2.2.3 report
foobar2000 v2.25.10
0126-09-11 09:33:10

File A: dchesorig1.wav
SHA1: 9578d6c3b0a59c80ea046a0ecf3fa6688a3e2aff
File B: dchesF3.wav
SHA1: 1ab5f8073b5a97ec8726ed5b7cf6daa5f74e8195

Output:
Core Audio : Default, 16-bit
Crossfading: NO

09:33:10 : Test started.
09:33:41 : Test restarted.
09:33:41 : 01/01
09:33:51 : Test restarted.
09:33:51 : 02/02
09:33:58 : Test restarted.
09:33:58 : 03/03
09:34:03 : Test restarted.
09:34:03 : 04/04
09:34:08 : Test restarted.
09:34:08 : 05/05
09:34:14 : Test restarted.
09:34:14 : 06/06
09:34:23 : Test restarted.
09:34:23 : 07/07
09:34:29 : Test restarted.
09:34:29 : 08/08
09:34:34 : Test restarted.
09:34:34 : 09/09
09:34:39 : Test restarted.
09:34:39 : 10/10
09:34:43 : Test restarted.
09:34:43 : 11/11
09:34:47 : Test restarted.
09:34:47 : 12/12
09:34:51 : Test restarted.
09:34:51 : 13/13
09:34:55 : Test restarted.
09:34:55 : 14/14
09:35:00 : Test restarted.
09:35:00 : 15/15
09:35:04 : Test restarted.
09:35:04 : 16/16
09:35:08 : Test restarted.
09:35:08 : 17/17
09:35:12 : Test restarted.
09:35:12 : 18/18
09:35:16 : Test restarted.
09:35:16 : 19/19
09:35:21 : Test restarted.
09:35:21 : 20/20
09:35:21 : Test finished.

----------
Total: 20/20
p-value: 0 (0%)

-- signature --
194af3f30e6edbd8a4199a37ff9800c54d363d7f


F1 easy as well

foo_abx 2.2.3 report
foobar2000 v2.25.10
0126-09-11 09:45:51

File A: dchesorig1.wav
SHA1: 9578d6c3b0a59c80ea046a0ecf3fa6688a3e2aff
File B: dchesF1.wav
SHA1: bcb137c93675b778a0a6f61425b7217d7080b6a3

Output:
Core Audio : Default, 16-bit
Crossfading: NO

09:45:51 : Test started.
09:46:16 : Test restarted.
09:46:16 : 01/01
09:46:32 : Test restarted.
09:46:32 : 02/02
09:46:44 : Test restarted.
09:46:44 : 03/03
09:47:03 : Test restarted.
09:47:03 : 04/04
09:47:10 : Test restarted.
09:47:10 : 05/05
09:47:14 : Test restarted.
09:47:14 : 06/06
09:47:33 : Test restarted.
09:47:33 : 07/07
09:47:41 : Test restarted.
09:47:41 : 08/08
09:48:07 : Test restarted.
09:48:07 : 09/09
09:48:21 : Test restarted.
09:48:21 : 10/10
09:48:29 : Test restarted.
09:48:29 : 11/11
09:48:34 : Test restarted.
09:48:34 : 12/12
09:49:02 : Test restarted.
09:49:02 : 13/13
09:49:09 : Test restarted.
09:49:09 : 14/14
09:49:15 : Test restarted.
09:49:15 : 15/15
09:49:19 : Test restarted.
09:49:19 : 16/16
09:49:39 : Test restarted.
09:49:39 : 17/17
09:49:47 : Test restarted.
09:49:47 : 18/18
09:49:57 : Test restarted.
09:49:57 : 19/19
09:50:05 : Test restarted.
09:50:05 : 20/20
09:50:05 : Test finished.

----------
Total: 20/20
p-value: 0 (0%)

-- signature --
1c7b9651c030c032f3db8b1e68fb6bc785b54098

 

[pma]

Output:
Core Audio : Default, 16-bit
Crossfading: NO

The results seem to be very good. The core audio on Mac is not well set. Please make sure your setting is to 96/24, to prevent re-sampling. Do not use "default" setting. Below how to do it in Foobar. 96kHz to be set in Audio-Midi setup.

 

[Thomas_A]

The results seem to be very good. The core audio on Mac is not well set. Please make sure your setting is to 96/24, to prevent re-sampling. Do not use "default" setting. Below how to do it in Foobar. 96kHz to be set in Audio-Midi setup.

View attachment 557931

Saw that now, I've switched mac and use my M5 more and forgot to change settings. But the two steep filters are easy anyway, no need at least to me to do the third filter; I'm no bat... My hearing goes abruptly down past 14-15k.

 

[Thomas_A]

That filter makes me doubt I was picking up on some difference. Doesn't do much below 13kHz where I can hear.

It is also SPL-dependent. My hearing drops like a stone past 15k, but enough SPL difference 15-25 kHz might be audible. Problem is that there are no music sounds that you would like to hear at that SPL...without risking damaging the ears...