ABX DAC comparison 100% Accuracy... HELP!

[HiFiJosh]

Hi all,

Inspired by a recent Pearl Acoustics video (below), I thought I'd test the claims that two 'transparent' DACs should be indistinguishable. I downloaded the five .wav files (below) from the five different DACs along with foobar and its ABX comparator.

Loading the D90 III and D10s, the differences were very apparent and I could ID which DAC was which immediately. To avoid HiFi phrases, I'll just say that the D10s was significantly worse. I was shocked.

I'm making this post to see if someone with more knowledge could help me understand what's happening here. I tried various other comparisons and there were significant differences between all the files and I could ABX between them accurately.

Video:

Files:

DACs: The Ultimate Comparison - Pearl Acoustics Shop

A selection of samples comparing a range of DACs

shop.pearlacoustics.com

 

[kemmler3D]

Well, they are recorded at slightly different levels, for one. Funny enough the D90 came out at a lower RMS than the D10. The D10 is about 0.2dB louder RMS, which is a marginal difference, I am not sure if it's a smoking gun but it does hint at something.


On the D10 it almost looks like the transients are clipping, you have what looks like bursts of white noise on some of them that aren't there on the D90.

So my guess is they made some kind of mistake recording these clips, possibly multiple mistakes.

No, I was just looking at the wrong peaks for comparison. After lining them up properly they don't look different to my eye.

 

[Jimbob54]

There may be some clues in this thread https://www.audiosciencereview.com/...ng-two-dacs-proper-process.57890/post-2116638

 

[Doodski]

I used headphones for auditioning this video and could not hear a difference other than variations in the band's playing.

 

[Sokel]

The way I see it is phase and filtering:



(that's for "Under My Skin" D10 (blue) vs D90 (purple),both balanced)

 

[SIY]

Hi all,

Inspired by a recent Pearl Acoustics video (below), I thought I'd test the claims that two 'transparent' DACs should be indistinguishable. I downloaded the five .wav files (below) from the five different DACs along with foobar and its ABX comparator.

Loading the D90 III and D10s, the differences were very apparent and I could ID which DAC was which immediately. To avoid HiFi phrases, I'll just say that the D10s was significantly worse. I was shocked.

I'm making this post to see if someone with more knowledge could help me understand what's happening here. I tried various other comparisons and there were significant differences between all the files and I could ABX between them accurately.

Video:

Files:

DACs: The Ultimate Comparison - Pearl Acoustics Shop

A selection of samples comparing a range of DACs

shop.pearlacoustics.com

Did you run them through DeltaWave?

edit: looks like Sokel was right ahead of me.

 

[HiFiJosh]

Well, they are recorded at slightly different levels, for one. Funny enough the D90 came out at a lower RMS than the D10. The D10 is about 0.2dB louder RMS, which is a marginal difference, I am not sure if it's a smoking gun but it does hint at something.


On the D10 it almost looks like the transients are clipping, you have what looks like bursts of white noise on some of them that aren't there on the D90.

So my guess is they made some kind of mistake recording these clips, possibly multiple mistakes.

No, I was just looking at the wrong peaks for comparison. After lining them up properly they don't look different to my eye.

Thanks for taking the time to analyze both recordings! I could bring the level down of the D10 file and try again, but the differences didn't seem to be volume related.

 

[Sokel]

Any suggesting for an interesting combination now that I have DeltaWave open?

That's Pontus (blue) vs D400Ex (purple),same song:



Edit: original spectrum added.

 

[kemmler3D]

Thanks for taking the time to analyze both recordings! I could bring the level down of the D10 file and try again, but the differences didn't seem to be volume related.

How would you describe the difference?

Level matching is very important in these situations because small differences in level sound like small differences in quality due to loudness effects. If the difference is more than 1-2dB one will sound louder, but if the difference is between 0.1-1dB a level mismatch can sound like a subtle change in quality.

So finding a >0.1dB difference in level shows the test wasn't conducted 100% perfectly.

 

[Sokel]

How would you describe the difference?

Level matching is very important in these situations because small differences in level sound like small differences in quality due to loudness effects. If the difference is more than 1-2dB one will sound louder, but if the difference is between 0.1-1dB a level mismatch can sound like a subtle change in quality.

So finding a >0.1dB difference in level shows the test wasn't conducted 100% perfectly.

The way I see the original spectrum above (the first one) is next to impossible to level match as they present different amplitudes at different freqs and not proportional.
One is louder up high and low and the other one is louder at 200-3000Hz.

 

[HiFiJosh]

Sorry I really don't have the technical knowledge to analyze the files to check RMS, phase, filtering etc.

The difference I am hearing is in the sustain, overtones, and reverb. I'm attaching some of my ABX results here as proof. Interestingly, the difference between the Sugden and the D10s was the most apparent, whereas the SMSL v D90 were a lot more subtle

 

[kemmler3D]

Sorry I really don't have the technical knowledge to analyze the files to check RMS, phase, filtering etc.

The difference I am hearing is in the sustain, overtones, and reverb. I'm attaching some of my ABX results here as proof. Interestingly, the difference between the Sugden and the D10s was the most apparent, whereas the SMSL v D90 were a lot more subtle

Well, that's a pretty solid result...

This plus the deltawave results makes me think the recorder they were using had something funky going on with input impedance or something. DACs are not supposed to sound that different nor are they supposed to show these sorts of differences in difference tests.

I don't mean that "according to objectivist orthodoxy", I mean even if you had deliberately voiced your DAC with audible coloration it probably wouldn't look like this. The weird group delay spike at 400hz, for example, doesn't really make sense to me.

 

[SIY]

This plus the deltawave results makes me think the recorder they were using had something funky going on with input impedance or something.

Or... a subtle little tweak to the recordings.

 

[kemmler3D]

Or... a subtle little tweak to the recordings.

It almost looks like comb filtering from crosstalk or something in some of those deltawave results... I am having a hard time coming up with a normal tweak you could intentionally apply to the recordings to make it look like that.

 

[HiFiJosh]

So finding a >0.1dB difference in level shows the test wasn't conducted 100% perfectly.

I agree this is not a 100% perfect test. The motivation for this post was to hopefully see if other members were able to tell me if there was some sort of inconsistencies between the files disproving that differences I am hearing was purely the DACs.

However, unless all these files are unequally level matched, I don't believe we can attribute my preference simply to that.

I tested all the DACs in the OP and my (personal, subjective, sighted ABX)* preference is Sugden > D90 > SMSL >>> D10s. The difference between the Sugden, D90, and SMSL were a matter of detail retrieval.

I know this is ASR, BUT SUBJECTIVELY..... The Denafrips v D90 was more so that the decay of the Denafrips was 'smoother', but more detailed on the D90. I just thought it was an interesting observation given it falls in line with a lot of subjective impressions of R2R DACs in general. I only scored 7/8 (last one wrong, tired/end of the day?) p = 0.0352 on this comparison so take those impressions with a massive grain of salt, just thought I'd share

 

[HiFiJosh]

Or... a subtle little tweak to the recordings.

It almost looks like comb filtering from crosstalk or something in some of those deltawave results... I am having a hard time coming up with a normal tweak you could intentionally apply to the recordings to make it look like that.

Thank you for investigating the files! I am still open to the fact that there is something more going on here than simply the DACs sound different.

One of the reasons I used these files for testing was that the man in the video is a recording engineer and founder of a loudspeaker company. He has praised all the products in previous videos and it would be very damaging to his company's image to intentionally manipulate these files for ulterior motives.

 

[SIY]

it would be very damaging to his company's image to intentionally manipulate these files for ulterior motives.

That has stopped no-one in the fashion audio market segment.

 

[HiFiJosh]

I don't mean that "according to objectivist orthodoxy", I mean even if you had deliberately voiced your DAC with audible coloration it probably wouldn't look like this. The weird group delay spike at 400hz, for example, doesn't really make sense to me.

Besides the Denafrips, I really wouldn't call any of the differences I'm hearing as 'colorations'. The differences I am hearing are simply in terms of detail retrieval.

 

[Sokel]

Ok,I did ten comparisons cause things are weird here.
They all show differences,either in FR or phase,filtering,etc. which is strange (ok,filtering is not strange,it's there a lot of times due to different settings,etc)

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.

Can we attribute part of what we see at Roland's R88 5.6kOhm input impedance?Given that some new DACs are not exactly low output impedance champions someone must run the math I think.

 

[Blumlein 88]

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.