DAC listening test files - update

[pedrojoaquimborges]

Anyone want to suggest a track to use for making recordings? Josh said he used 96K recordings of Steely Dan “Negative Girl” and “Cousin Dupree”, so maybe one of those? He also listed a couple Van Morrison tracks he said were 96K but I only see them in 44.1K in Tidal. I haven't looked but I would buy those for the testing if I can find them in 96K. And the other ones he used he said were upsampled to 96K. I'd rather use something released at 96K.

Music taste aside, I'd personally go for something like Daft Punk's Random Access Memories. Older hifi-ish tracks, like Steely Dan are great for system testing, but with modern production and mixing standards we've reached a type of sound that is very useful for pushing systems towards their limits in frequency and dynamics/transient response.

 

[pedrojoaquimborges]

Online test Original data vs. Marantz SA7001 output

Marantz SA7001 CD player output vs. original CD data · FunkyABX

abx.funkybits.fr

This is an online test that I prepared in March 2026. Wav files. Anyone can try it online, the test interface avoids cheating. Test files are also attached at the bottom of this post.


View attachment 551187


View attachment 551182View attachment 551183View attachment 551184View attachment 551185View attachment 551186

To be frank, most classical/concert program material is not perfect for system testing if you're looking for a more all-round picture. These audio files, for example, don't have much information nor complexity above 8kHz. Dynamics-wise, they also barely test components' ability to reproduce extreme sudden changes in amplitude.

 

[bennetng]

I don't have stamina to follow the the whole drama (Josh, AS etc...) but if the topic is about ABXing transparently measured DACs, how about using the track below?

Let's develop an ASR inter-sample test procedure for DACs!

Source: https://www.sfu.ca/sonic-studio-webdav/cmns/Handbook Tutorial/Audiology.html Thanks for reminding this book! Assuming that music has similar effect like noise, after 15 minutes of listening at 85dB SPL average, we will temporarily lose ability to hear noise at 8dB SPL. Does nature abhor...

www.audiosciencereview.com

@KSTR prepared some test files in the replies, and subsequently two members posted some successful ABX results, but @pma or others can also download the original audio file I uploaded and design a different test.

 

[pjug]

Music taste aside, I'd personally go for something like Daft Punk's Random Access Memories. Older hifi-ish tracks, like Steely Dan are great for system testing, but with modern production and mixing standards we've reached a type of sound that is very useful for pushing systems towards their limits in frequency and dynamics/transient response.

Yeah, I was just trying not to deviate too much from what Josh was doing. So 96KHz to minimize reconstruction fiter effects. And maybe using one of his tracks. But I agree better to use some other track if it will make it more likely for someone to pass an ABX.

 

[pjug]

One thing I want to try...
I want to see if I can get around the timing drift problem by recording just one channel of each of two DACs at the same time. I don't think this is quite as good as a synchronous DAC and ADC, but it seems like it might be a good improvement. Thoughts?

 

[NTTY]

Online test Original data vs. Marantz SA7001 output

Marantz SA7001 CD player output vs. original CD data · FunkyABX

abx.funkybits.fr

5/10, perfect guessing

Test results - Marantz SA7001 CD player output vs. original CD data · FunkyABX

abx.funkybits.fr

 

[pma]

To be frank, most classical/concert program material is not perfect for system testing if you're looking for a more all-round picture. These audio files, for example, don't have much information nor complexity above 8kHz. Dynamics-wise, they also barely test components' ability to reproduce extreme sudden changes in amplitude.

Signals like pure tones or narrow band noises are best if we try to find some tiny audibly differences. Anyway everything is masked by driver (speaker, phones) distortions and inaccuracies. Music of any kind tells much less than generated signals. These debates are long, never ending, boring and bring nothing new. Including this thread - nothing new is the outcome.

 

[pma]

To be frank, most classical/concert program material is not perfect for system testing if you're looking for a more all-round picture.

To me, classical music says most for testing purposes. Artificial studio music or electronically generated music tells least. I had a chance to spend some time in Supraphon recording studios of classical music during my university studies and it gave me most, as an eye opener.

Attached is a DR from my own recording of a wind quintet recorded 2 years ago:

 

[pma]

I can see 10/10 in the online test results, is it from anybody here?

 

[danadam]

I can see 10/10 in the online test results, is it from anybody here?

It was me, but with cheating. You can disregard that.

Test results - Marantz SA7001 CD player output vs. original CD data · FunkyABX

abx.funkybits.fr

 

[pma]

Thanks for confessing that, really appreciated, as we do not need false results . So it is possible to cheat the test? Consecutive output monitoring? Or SW code trick? It is interesting.

 

[danadam]

So it is possible to cheat the test? Consecutive output monitoring? Or SW code trick? It is interesting.

It's quite easy to capture what's being sent through pulseaudio and then it's just a matter of comparing two captures.

 

[pma]

It's quite easy to capture what's being sent through pulseaudio and then it's just a matter of comparing two captures.

I think that something like that was mentioned some time ago at diyaudio. One of the reasons of former Stuart’s (SIY) skepticism to any kind of remote, thus uncontrolled test.

 

[Holy Spirit]

Attached is a DR from my own recording of a wind quintet recorded 2 years ago:

I guess 16dB is dynamic range, what is that 22dB RMS value?

 

[pma]

It is more complicated.

https://arxiv.org/pdf/1312.0538

 

[Holy Spirit]

It is more complicated.

Thanks! I have read the first four pages. I think it is enough for me.

 

[pma]

Online test Original data vs. Marantz SA7001 output

Marantz SA7001 CD player output vs. original CD data · FunkyABX

abx.funkybits.fr

This is an online test that I prepared in March 2026. Wav files. Anyone can try it online, the test interface avoids cheating. Test files are also attached at the bottom of this post.

I have closed that online test as there was a confessed cheating here. I am glad of the confession (the courage to confess), however I cannot let the test continue regarding collecting further results. The files are still working and the test files can be listened to, just without a possibility to get results.

Before closing, the test was played 70 times and the results were stored 12 times. From those 12 results, I have discarded the cheated 10/10 result and also the worst 3/10 result. 10 results remaining give this:

The 8/10 result that would start considerations on a successful audible discernment has not been reached.

 

[Thomas_A]

I have closed that online test as there was a confessed cheating here. I am glad of the confession (the courage to confess), however I cannot let the test continue regarding collecting further results. The files are still working and the test files can be listened to, de just without a possibility to get results.

Before closing, the test was played 70 times and the results were stored 12 times. From those 12 results, I have discarded the cheated 10/10 result and also the worst 3/10 result. 10 results remaining give this:

View attachment 551349

The 8/10 result that would start considerations on a successful audible discernment has not been reached.

View attachment 551351

Why did you exlude the 3/10 result(s)? At least one could infer whether the total results are normally distributed. From the graph it seems maybe or maybe not, but I have not made any calculations. Since testing was performed many more times than stored it would most likely skew results upwards. Or… it depends on how the test was played but not stored…

 

[bennetng]

I have closed that online test as there was a confessed cheating here. I am glad of the confession (the courage to confess)...

I suppose "courage" is not the appropriate description, it is not the first time @danadam expressed an opinion that these tests can be cheated:

Is 22kHz brickwall cut off audible? - Listening test

10 trials is not enough but 2x10 trials is. It is an equivalent to 20 trials within some time interval. You may sum it and get 8/10 + 8/10 = 16/20. This is perfectly valid in probability theory and one may find what 16/20 means. BTW, I have a further 8/10 result. Why not 16 or 20 trials in one...

www.audiosciencereview.com

I remembered the post above because I defended you right below in the reply.
I would treat danadam's action as a proof to inform others about the possibility to cheat, because you specifically mentioned about it:

the test interface avoids cheating

I had a similar experience in exposing Archimago's blind test project because he also believed that he had taken enough measures to prevent cheating:

RESULTS: Internet Blind Test of 24-Bit vs. Dithered 16-Bit Part Deux - Daft Punk Edition

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

I am not convinced about the bragging from "The other person submitted from Slovakia" and left a comment below to inform Archimago that a software I wrote two years earlier was able to de-blind his test files:

RESULTS: Internet Blind Test of 24-Bit vs. Dithered 16-Bit Part Deux - Daft Punk Edition

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

 

[chris256]

Hi,
you may digitally compare both files provided by pma and hear and see the diff (in audacity e.g.):
sox -m -v 1 file_1.wav -v -1 file_2.wav file_diff.wav

sox -m -v 1 DAC1.wav -v -1 DAC1.wav DAC1_1-diff.wav

If the compared files would be equal, the diff would be deep silence (you can test it easily by comparing same files).

The DAC-files differ extremely compared to the file_x - files above.

What does the sox command do here? It mixes two files together - but the second file is inverted. If the second file is identical, it completely cancels out each sample of the first file and deep silence is the result. If there are any differences, the samples can't cancel each other out any more and you get an audible diff.

You now may argue if this diff is audible or not if you compare each of the test files serially with your ears.