DAC audible differences are finally proven with ABX?

[pma]

Josh on the other hand, got paid for article/testing by his boss at AS

Reading the latest posts there, now it all makes sense. (to translate - the result must have been what it was)

 

[solderdude]

AFAIK the situation for the recordings Josh has (to check the AB box performance and determining volume matching was) was:
DAC -> box -> Benchmaark HPA-4 -> (SE in) Cosmos.
So in no case the Cosmos was ever connected to the output of the ABX box (which would be an incorrect load asking for trouble).
Also AFAIK the only recordings that were made were done after the turmoil started and he wanted to null A and B difference.
Sure he could have uploaded the original test file and the recordings of A and B which would have given us valuable info about the performance of the ABX box.

This would only be done for one and the same volume setting (maybe even at max volume) so would only give us a part of the actual changes to the original.

Not needed anymore as Amir gets to measure it and hope explores all angles.... (this should keep Amir busy for a few days !)

 

[pma]

So in no case the Cosmos was ever connected to the output of the ABX box (which would be an incorrect load asking for trouble).

That’s correct, though that has never been confirmed in my tests, like those in the parallel thread. Only one of four tests posted there uses a preamp (with multiple inputs) between the DAC under test and the Cosmos.

 

[RandomEar]

Also AFAIK the only recordings that were made were done after the turmoil started and he wanted to null A and B difference.

There's definitely also the multitone recordings he talked about which, as far as I remember, were done before he even started ABX testing. He may have deleted them later, but they did exist.

 

[solderdude]

That’s correct, though that has never been confirmed in my tests, like those in the parallel thread. Only one of four tests posted there uses a preamp (with multiple inputs) between the DAC under test and the Cosmos.

I don't see that as much of an issue (DAC-> Cosmos) which in worst case could possibly increase distortion a little as DACs usually have a low output resistance and will have no issue driving 2kohm loads.
In Josh's case, where basically a 20k passive volume control is used with just a little attenuation so still less than 2k added series resistance, a good (which the Benchmark HPA4 is) preamp between the passive attenuator (which the box basically is) and cosmos was the right call.

I also think Amir's testing (at different attenuation levels) should be done using RCA cables (for the capacitance) into a 10k load or so (not 600 ohm or 100k)

On appearance Josh's test and setup seems to be done properly, alas there is no way to verify by lack of (sufficient and quality) evidence.
The surprising results will need to get replicated using recordings and foobar with minimal 20 attempts.
Preferably also by Josh.

 

[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, but the test has been closed for the reason of cheating.

 

[SSS]

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.

View attachment 551188

Was the SA7001 analog or digital output used? If analog, what was the wav recorder?

 

[pma]

Was the SA7001 analog or digital output used? If analog, what was the wav recorder?

Analog, of course. More in this post. Together with files. So you can try online test or download the files. Online test avoids any kind of cheating and the files are the same 44.1/16 wav format. One of the files is the CD rip and the other one is recorded via my special (active balanced) preamp with DACT stepped attenuator inside + Cosmos ADC. It is all unimportant, the results you or me or anyone gets are important. Remember one of the files is an original rip. If you want to check the setup, so you see it below. The power amp was of course not used.

 

[pjug]

In addition to Josh's amazing success rate reported in his ABX, another thing I find very surprising is that he would be using 6 different tracks to help hear differences. Maybe I am drawing the wrong conclusion, but to me it makes it sound like DACs cause a number of types of audible differences. That creates more doubt in my mind about the whole thing.

 

[solderdude]

Assuming this is the case then one could argue the DACs can be told apart in certain conditions with specific recordings and with a lot of training.

That does not mean that 'well measuring, performing and not 'defective' DACS do not all sound different 99.99% of the time but only that in specific conditions/recordings it is still possible to tell them apart.

For that reason a recording of both DACs (direct or via the ABX box) is pretty essential as the ABX box (very likely) applies the exact same distortion via A and B input (see Josh's null at -100dB).

For me the most likely scenario is that Josh could reliably tell the compared DACs apart and that it is not caused by the ABX box nor level matching ... but this can't be ruled out until measured.

The best thing is that when the exact same recordings are used (they are known), the same DACs are 'recorded' directly (thus NO AB box), level matched and prepared for Foobar ABX comparator and @Josh83 could hint which recording and time stamp he was able to tell them apart then all interested parties can do this at home.
Of course with their DAC (that may or may not reproduce the files equally well) on their transducers with their ears.

Would be fun if more people could pass the test... no ABX box needed.
Just proper level matched recordings (say at 1kHz -3dB) and a Foobar comparator.
When no one can pass the test and Josh also cannot pass the test 19/20 or 20/20 then most likely there was a circumstance that remains unknown.

 

[Roland68]

Assuming this is the case then one could argue the DACs can be told apart in certain conditions with specific recordings and with a lot of training.

That does not mean that 'well measuring, performing and not 'defective' DACS do not all sound different 99.99% of the time but only that in specific conditions/recordings it is still possible to tell them apart.

For that reason a recording of both DACs (direct or via the ABX box) is pretty essential as the ABX box (very likely) applies the exact same distortion via A and B input (see Josh's null at -100dB).

For me the most likely scenario is that Josh could reliably tell the compared DACs apart and that it is not caused by the ABX box nor level matching ... but this can't be ruled out until measured.

The best thing is that when the exact same recordings are used (they are known), the same DACs are 'recorded' directly (thus NO AB box), level matched and prepared for Foobar ABX comparator and @Josh83 could hint which recording and time stamp he was able to tell them apart then all interested parties can do this at home.
Of course with their DAC (that may or may not reproduce the files equally well) on their transducers with their ears.

Would be fun if more people could pass the test... no ABX box needed.
Just proper level matched recordings (say at 1kHz -3dB) and a Foobar comparator.
When no one can pass the test and Josh also cannot pass the test 19/20 or 20/20 then most likely there was a circumstance that remains unknown.

Personally, I found it very odd that Josh didn't support other users conducting the test themselves—for example, using the recordings.

 

[Purité Audio]

That would be the very last thing he would do.
Keith

 

[welwynnick]

Assuming this is the case then one could argue the DACs can be told apart in certain conditions with specific recordings and with a lot of training.
................
For me the most likely scenario is that Josh could reliably tell the compared DACs apart and that it is not caused by the ABX box nor level matching ... but this can't be ruled out until measured.

That's a turn-around for this thread. Are you preparing for the eventuality that Amir finds nothing wrong with the AVA box?
I'm very much looking forwards to seeing Amir's test results, and I wouldn't be surprised if they look good. THEN this thread would get interesting.

 

[solderdude]

Personally, I found it very odd that Josh didn't support other users conducting the test themselves—for example, using the recordings.

From what I got from his article is that the attempt was mostly to satisfy his own curiosity and to prove or disprove (to himself) his idea that he could hear differences between DACs using the ABX box in a more scientific way without the help of others.

That's a turn-around for this thread. Are you preparing for the eventuality that Amir finds nothing wrong with the AVA box?

Amir will find increased distortion that may or may not reach audible levels.
The real question, however, is whether or not there is a noticeable difference between A and B input (including L-R matching) that may be attenuation and or frequency dependent for different attenuation levels.
As Josh reported not seeing numbers below 90 (-1dB attenuation difference) between DACs it would be interesting to see if there are audible 'tells' with different attenuation.

Personally I do not expect to see much of a difference between A and B input which basically is the only thing that matters, not so much the inevitable added distortion from the volume control IC. But... the box could have issues that might explain what Josh heard.
If so a recording with music could be posted so people can replicate.
In such a case AB files could be prepared and uploaded (without AB box) which folks (including) Josh could test.

So not a turnaround for my part. This has always been my position.
There are many aspects. Source files, DACs, AB box differences between A and B (unlikely) and above all the listener.

Unless we get to the bottom of this it still remains a good thing to question the found results as they don't jive with the consensus here.
Needs more proof than just Josh's results as reported.
Hopefully more clarity is coming...

 

[welwynnick]

Amir will find increased distortion that may or may not reach audible levels.
The real question, however, is whether or not there is a noticeable difference between A and B input (including L-R matching) that may be attenuation and or frequency dependent for different attenuation levels.
As Josh reported not seeing numbers below 90 (-1dB attenuation difference) between DACs it would be interesting to see if there are audible 'tells' with different attenuation.
Personally I do not expect to see much of a difference between A and B input which basically is the only thing that matters, not so much the inevitable added distortion from the volume control IC. But... the box could have issues that might explain what Josh heard.
If so a recording with music could be posted so people can replicate.
In such a case AB files could be prepared and uploaded (without AB box) which folks (including) Josh could test.
So not a turnaround for my part. This has always been my position.
There are many aspects. Source files, DACs, AB box differences between A and B (unlikely) and above all the listener.

Thank you for the considered response (one of many). I agree. We need to see how distortion varies with different attenuations, and between channels. If the AVA looks good, I hope people will accept that Josh wasn't just listening to a comparator drowning out the differences between sources. Then I hope that Josh might re-enter the discussion and we can take it forwards and actually learn something.

Although the AVA is going to Amir, that doesn't stop him taking good E1DA recordings from a selection of DACs. The measurements and recording he's taken so far and kept to himself were taken under non-optimal conditions, and would probably not been of much use anyway. That doesn't mean they can't be done better. It might take cables, connectors, training, constructive support (from ASR!) or even an E1DA Scaler. Whatever it takes, I think it would be worth it.

That way, Josh might be able to overcome his scepticism about using an ADC + DAC to evaluate DACs, and then yes, everyone would be able to hear the same things. Of course that might not happen for various reasons, and it would be back to using a comparator to compare DACs. However it would be a great data point to establish DAC differences C AN be captured using an ADC.

 

[rjp]

You know what's really funny? my inbox, and probably most of ours, is filled with hundreds of emails with this title. Maybe that was the goal

DAC audible differences are finally proven with ABX?​

 

[antcollinet]

You know what's really funny? my inbox, and probably most of ours, is filled with hundreds of emails with this title. Maybe that was the goal

DAC audible differences are finally proven with ABX?​

Mine isn't. I turned off email post notifications aprox 1 day after joining ASR.

 

[PristineSound]

I’m happy to confirm we worked out the details, and we have a deal with Josh

Special thanks to Flo!

Now the pressure is on Amir to do a speedy measurement in the face of a mob but I have no doubt Amir will rise to the challenge and hit it out the ballpark.

Curiously waiting for the results.

 

[krabapple]

People who express any doubts about some of the agreed upon assumptions at this site are relentlessy attacked or ridiculed. For instance, there is actual science showing issues with using SINAD as a one stop shop for judging audio equipment. SINAD isn't flawless. Why is it important if it's in the "inaudible" range among various pieces of equipment? Your certainly know this, but the hidden assumption here is that it is the be and end all of rating equipment. I assume you will deny this, but that's what comes thru here and what many people take away. It's reflected in many posts here.

And many posts on ASR , from Amir too, not at all hidden, have plainly warned against using SINAD as a "be-all and end-all".

IME when people posit 'hidden assumptions', it's often a tell. On them.

 

[krabapple]

Instead of testing thousands of random people, why don’t the people that claim they can hear the differences back up their claims by doing the tests. Seems like a much faster and more efficient way to come at the desired results.

Bingo!

The data we have..which is good, regardless of what anklebiter joe, who seems a stranger to 'doing the research' unless its handed to him, thinks ...gives us every reason to believe that very high frequency hearing is rare. And that very small differences are hard to hear, with music. And because of the way DACs work, they can be assumed to be 'transparent' if they hit those measured marks.

So, anyone who claims they inhabit the far tail of the distribution, should step forward. It's just probabilities, man. You aren't impossible, but a reasonable person should consider you improbable. So step forward and take your shot. The more prima facie improbable your claim, the stronger your data will have to be for it to hold up. It's just science, man.

And....heeeeeere's Josh!
But alas Josh won't allow us to test the strength of his remarkable, nay, paradigm-shifting, data.
Ah well.