ABX DAC comparison 100% Accuracy... HELP!

[Sokel]

These recorded files have issues that don't make sense. Some other issue is going on here. Or something has been done to the files.

Threw all of them in REW's tool to see.

And allthough peaks are almost identical...

..average spectrum varies wildly:

 

[Old_School_Brad]

I’d suggest investigating impedance mismatch, given that the difference spans the entire spectrum.
I think otherwise the test setup is good.

 

[Sokel]

I’d suggest investigating impedance mismatch, given that the difference spans the entire spectrum.
I think otherwise the test setup is good.

The weirdest is D400EX and D10 balanced.
The later is consistently 5-7dB (!) louder from 200Hz to 10kHz,that's outrageous.
Its peaks though are the same.

 

[Old_School_Brad]

Topping D90 MkIII: 150 ohm
SMSL D400 EX: ?
Topping D10: 88 ohm
Denafrips Pontus 15th: 1250 ohm
Sugden DAP 800: ?
Roland R-88 (input, line 1-8): 5.6 kohm

 

[Blumlein 88]

Topping D90 MkIII: 150 ohm
SMSL D400 EX: ?
Topping D10: 88 ohm
Denafrips Pontus 15th: 1250 ohm
Sugden DAP 800: ?

Sorry these are from the wrong files. So disregard the charts here.

That would not cause what Sokel has shown which is also what I'm seeing. FR that varies wildly over short spans of frequency. Loudness varying over time. Phase all over the place. What you show would only cause a minor level difference and might cause a bit of FR droop at the upper end of one vs the other.

Here is the relative FR of two devices that null moderately well.

Here is the same thing showing the FR differences between the Sugden and Topping D90. Notice the widened scale and the wide fluctuations.

 

[Old_School_Brad]

That would not cause what Sokel has shown which is also what I'm seeing. FR that varies wildly over short spans of frequency. Loudness varying over time. Phase all over the place. What you show would only cause a minor level difference and might cause a bit of FR droop at the upper end of one vs the other.

Here is the relative FR of two devices that null moderately well.
View attachment 405848

Here is the same thing showing the FR differences between the Sugden and Topping D90. Notice the widened scale and the wide fluctuations.
View attachment 405849

I stand corrected.

 

[Old_School_Brad]

The files were produced directly in the Roland R-88, right?

 

[Blumlein 88]

The files were produced directly in the Roland R-88, right?

All I can tell you is something is wrong with them. He made them 16 bit so I wonder if whatever resampling he did caused issues or if he used the R88 in 16 bit for this and it does something odd.

I also might believe he recorded playback with a microphone. You would see something like this, but that isn't what he described doing. I've actually used a microphone close to a speaker and gotten better results than that. I notice some files are labeled main listening room and some office desktop.

 

[Old_School_Brad]

All I can tell you is something is wrong with them. He made them 16 bit so I wonder if whatever resampling he did caused issues or if he used the R88 in 16 bit for this and it does something odd.

I also might believe he recorded playback with a microphone. You would see something like this, but that isn't what he described doing. I've actually used a microphone close to a speaker and gotten better results than that. I notice some files are labeled main listening room and some office desktop.

Yes, I was curious about that too, because in the video’s introduction, he mentions recording the outputs from the DAC using the R-88, but towards the end, he demonstrates in-room recordings. In the end, I decided the choice didn't make much sense if it were files of in-room recordings, so I opted to believe it was as he initially described.

 

[Sokel]

All I can tell you is something is wrong with them. He made them 16 bit so I wonder if whatever resampling he did caused issues or if he used the R88 in 16 bit for this and it does something odd.

I also might believe he recorded playback with a microphone. You would see something like this, but that isn't what he described doing. I've actually used a microphone close to a speaker and gotten better results than that. I notice some files are labeled main listening room and some office desktop.

At the video he says that the DACs are straight to Roland.
He is a recording engineer though (says for classical,I have to search him! ) .one would expect that he knows what he's doing.
On the other hand Roland is not exactly top with it's stated 0.02% THD+N (I think)

 

[Sokel]

Yes, I was curious about that too, because in the video’s introduction, he mentions recording the outputs from the DAC using the R-88, but towards the end, he demonstrates in-room recordings. In the end, I decided the choice didn't make much sense if it were files of in-room recordings, so I opted to believe it was as he initially described.

In room recordings on the other hand would not have so precise peaks.
The whole story is strange.

 

[Blumlein 88]

If you look at 2:18 he shows the download page from his web site. And it isn't the one we have been downloading. The files from the Roland are not available and I do believe the others have been recorded over a microphone and used in the video.

Sorry, I was using the wrong files, the files are linked in the first post on the thread.

 

[SIY]

He is a recording engineer though (says for classical,I have to search him! ) .one would expect that he knows what he's doing.

Yes, which leads to an unpleasant conclusion.

 

[Old_School_Brad]

Screenshot from video:

Screenshot from link under the video in the OP: "V2"

 

[Sokel]

Screenshot from video:

View attachment 405863

Screenshot from link under the video in the OP: "V2"

View attachment 405864

The ones I have used in all comparisons at this thread are the V2 ones which are allegedly corrected.

 

[Old_School_Brad]

The ones I have used in all comparisons at this thread are the V2 ones which are allegedly corrected.

What does it mean that they are corrected? Something he says in the video?

 

[Sokel]

What does it mean that they are corrected? Something he says in the video?

It means that the whole thing is a mess
One cannot even remotely conclude something useful out of it.

 

[Blumlein 88]

The ones I have used in all comparisons at this thread are the V2 ones which are allegedly corrected.

Well I watched the video and looked on their website, not noticing the link in the OPs post. Sorry for the mistake.

 

[davmol]

Now,that guy that did the recordings is an advocate for no differences between DACs and it seems it has done a solid job recording them.

I've watched a lot of this man's videos and think he has a really great channel but I would not say he is an advocate for no differences between DACs, he often talks of one DAC sounded better than another. I like what he is doing here, he is responding to issues with previous tests and creating new ones, so if there is an issue with this one, I'm sure he'll address it.

 

[Blumlein 88]

So the two Topping DACs look believable. I used the "I've got you under my skin" file. Better than -60 db null, PK metric of -78 db. Unlikely to sound different. There is a large amount of drift at nearly 40 ppm though not going to be audible by itself. The files were not well aligned and one was over 20 seconds longer so it needed trimming at the end. Ditto for the Toppings vs the Denafrips. The Toppings vs the SMSL is pretty good too with a PK Metric of better than -73 db, but a bit odd on phase. Phase gradually differed by a few degrees over the whole audio band.

The Toppings vs the Sugden are all messed up like files I tried earlier. The FR is all up an down with ridiculous phase vs the Topping. Something is still wrong with this one.

You do have to trim the rear of the files. They fade out at different rates and different times. Once you do this they all look reasonably similar other than the Sugden. I doubt you'll hear differences. Some minor differences below 10 hz, and in the upper half octave no doubt from the different filtering used.

With the different length of the different files that alone may make them discernible in Foobar ABX.