Just got a Yggdrasil!

[manisandher]

26 minutes ago, jabbr said:

That aside are we really comparing 2 state of the art DACs using a non state of the art ADC?

 

Dennis believes the ADC is pretty good:

 

18 hours ago, esldude said:

This capture of the Yggy matches in frequency response to better than .1 db until more than 12 khz (mostly it matched closer than .01 db once volume was matched).  It then is down only .25 db at 20 khz.  It would seem to be of high fidelity to the original file in terms of response.  Doesn't sound different either.

 

But I agree, it'd be much better if I'd had to hand an ADC with substantially higher resolution than any of the DACs.

 

Mani.

[manisandher]

33 minutes ago, jabbr said:

This is like trying to compare a Zeiss vs Leica lens by shooting a test pattern through an iPhone...

 

I would expect the reproductions of the Zeiss and Leica test patterns through the iPhone to be indistinguishable due to the relatively low resolution of the iPhone lens. I would be amazed if the result were anything but this.

 

The Yggy and the control DAC captures through the ADC are nowhere near indistinguishable. This implies to me that the ADC had enough resolution. It has the following specs (as used) in balanced mode:

 

THD+N: 0.003% or less (Ref:-16dB/BALANCED,JEITA)

S/N ratio: 113dB or higher (Ref:-20dB/BALANCED,JEITA)

 

Every ADC I've ever tried sounds different. Most have decent specs, and this one is no different. Of course, whether these specs are really correlated with performance is debatable. I suspect they don't tell the whole story.

 

Mani.

[Jud]

9 minutes ago, manisandher said:

The Yggy and the control DAC captures through the ADC are nowhere near indistinguishable. This implies to me that the ADC had enough resolution.

 

Or that, to use jabbr's language, one or more non-linear interactions between the ADC and Yggy diverged significantly from the non-linear interactions between the ADC and the other DAC.

 

Does this overcomplicate things?  Possibly.  On the other hand, you have two files whose common measurements diverge very little that sound very different to you.  So it's difficult to say that what's causing the difference that isn't being measured is definitely *not* some non-linear interaction (or something to do with the differences in USB inputs, more prosaically).

[mansr]

32 minutes ago, Jud said:

Or that, to use jabbr's language, one or more non-linear interactions between the ADC and Yggy diverged significantly from the non-linear interactions between the ADC and the other DAC.

I seriously doubt there's any interaction between the DAC and the ADC at a level detectable by the ADC.

[Jud]

7 minutes ago, mansr said:

I seriously doubt there's any interaction between the DAC and the ADC at a level detectable by the ADC.

 

I guess then we'll vote "Yes" on "Does this overcomplicate things?" 

[jabbr]

1 minute ago, mansr said:

I seriously doubt there's any interaction between the DAC and the ADC at a level detectable by the ADC.

Well that's the other point --  the testing equipment is not adequate to reasonably compare the devices under test

[mansr]

3 minutes ago, jabbr said:

Well that's the other point --  the testing equipment is not adequate to reasonably compare the devices under test

That's not at all what I meant.

[manisandher]

1 minute ago, jabbr said:

Well that's the other point --  the testing equipment is not adequate to reasonably compare the devices under test

 

What would constitute 'adequate' testing equipment in this case? I mean, how would I know that my replay/recording chain was adequate or not?

 

Mani.

[Ralf11]

Summary:

 

Mani hears:

grating opening cymbals

overall weird holographic sound

metallic bass 

hollow piano

HF junk in background

 

George hears at least some of that to a lesser extent as well.

 

All on a single test file.

 

Do I have that more or less correct?

 

Can you guys try a different test file, esp. with cymbals (and maybe snares)?  And maybe soe female vocals?

 

And does anyone else hear these differences?

[manisandher]

39 minutes ago, Ralf11 said:

Summary:

 

Mani hears:

grating opening cymbals

overall weird holographic sound

metallic bass 

hollow piano

HF junk in background

 

George hears at least some of that to a lesser extent as well.

 

All on a single test file.

 

Do I have that more or less correct?

 

Can you guys try a different test file, esp. with cymbals (and maybe snares)?  And maybe soe female vocals?

 

And does anyone else hear these differences?

 

Unfortunately, I only used one 'test' file - what I've referred to as the 'original file'. I totally agree, it would have been useful to have used a variety of different types of music. But the track I used does have a lot of cymbals all the way through. I find the quality of the piano and bass quite telling too.

 

When you summarised what I heard, I think it's important to stress that these things are not in isolation. For example, I found the opening cymbals in the Yggy file to be grating compared to the opening cymbals in the original file. I found the opening cymbals in the control DAC file to be slightly subdued (and not grating at all) compared to those in the original file. Etc, etc. In my mind, by doing this, I'm eliminating the effects of the playback/ADC chain. But there are a lot of smart people here telling me that that's not a valid approach.

 

Mani.

[manisandher]

52 minutes ago, Ralf11 said:

And does anyone else hear these differences?

 

Do you?

 

Mani.

[semente]

16 minutes ago, manisandher said:

 

Unfortunately, I only used one 'test' file - what I've referred to as the 'original file'. I totally agree, it would have been useful to have used a variety of different types of music. But the track I used does have a lot of cymbals all the way through. I find the quality of the piano and bass quite telling too.

 

When you summarised what I heard, I think it's important to stress that these things are not in isolation. For example, I found the opening cymbals in the Yggy file to be grating compared to the opening cymbals in the original file. I found the opening cymbals in the control DAC file to be slightly subdued (and not grating at all) compared to those in the original file. Etc, etc. In my mind, by doing this, I'm eliminating the effects of the playback/ADC chain. But there are a lot of smart people here telling me that that's not a valid approach.

 

Mani.

 

I find the digitised output of the Yggy too bad. I should have listened to the Control file to see if there was a consistency in badness but to be honest I don't believe that the Yggy can be that bad.

[esldude]

15 minutes ago, semente said:

 

I find the digitised output of the Yggy too bad. I should have listened to the Control file to see if there was a consistency in badness but to be honest I don't believe that the Yggy can be that bad.

Now wait a minute.  Are you deciding this without hearing the original file?  That would make little sense to made any judgement about the Yggy, or the ADC without reference to the source signal.

[manisandher]

11 minutes ago, semente said:

I find the digitised output of the Yggy too bad. I should have listened to the Control file to see if there was a consistency in badness but to be honest I don't believe that the Yggy can be that bad.

 

Some others here are not finding large differences between the Yggy file and the original.

 

Does the original sound OK to you, at least? In any event, certainly nothing can be concluded without taking a listen to the control DAC file too. If what you're finding as "too bad" exists exactly the same in the control DAC file too, then you can call it a day. But if you can hear any obvious differences between the Yggy file and the control DAC file (e.g. in my case, grating opening cymbals in one and not the other), these can only be due to differences in the outputs of the DACs. I can't see how any other mechanism could be responsible.

 

Mani.

[manisandher]

2 minutes ago, esldude said:

Now wait a minute.  Are you deciding this without hearing the original file?  That would make little sense to made any judgement about the Yggy, or the ADC without reference to the source signal.

 

I agree.

 

Dennis, any thoughts on the control DAC file? If you've done any analysis on it, it should be very similar in FR to the Yggy file. And yet it sounds very different to my ears.

 

Mani.

[jabbr]

4 hours ago, manisandher said:

 

What would constitute 'adequate' testing equipment in this case? I mean, how would I know that my replay/recording chain was adequate or not?

 

All test systems need to be validated before use. Not a simple answer. Sure I could say that you need 27 bit samples at 4x768 ksps(!!) in order to entirely capture the "DAC" (assuming 768 khz is the highest sample rate) but obviously that's impractical. So what do people do? They often use different measurements, each looking at some aspect. We could ask @PeterSt what he would accept as a measurement that .e.g "proved" that the output of the Yggy and the NOS1a were identical? Or what measurement does he use to determine the linearity of his DAC i.e. how many significant bits?

You can input known signals and then do high resolution frequency analysis, you can look at linewidth.

Now for recoding analysis I'd use at a minimum, the maximum input frequency e.g. 768 kHz x 24bits. 

But in general the measurement equipment needs to me more "accurate" than the device under test. Not saying this is easy ...

[SuperRoo]

Mani,

We all hear you. Is there a better DAC in that price range we should consider instead?

Richard

[semente]

1 hour ago, esldude said:

Now wait a minute.  Are you deciding this without hearing the original file?  That would make little sense to made any judgement about the Yggy, or the ADC without reference to the source signal.

 

The original file is the CD rip, I mean the second DAC

[esldude]

1 hour ago, manisandher said:

 

I agree.

 

Dennis, any thoughts on the control DAC file? If you've done any analysis on it, it should be very similar in FR to the Yggy file. And yet it sounds very different to my ears.

 

Mani.

If I am clear on which files we are talking about.  The Yggy file recorded by the ADC and the original wav file right?

 

Yes the frequency response is very, very close.  I mentioned earlier they match to less than .1 db until about 10 khz and the Yggy dips eventually to -.25 db by 20 khz.  Below 8 khz they match to within .01 db in basic frequency response.  I had to get a close volume match to determine this.  I did the volume matching and listening first.  Fortunately the match from that was good enough for the analysis. So it sounded the same or very, very close before I knew the frequency response.  My hearing is pretty solid to 12 khz in one ear, and 15 khz in the other.  Age is not kind to our hearing.

 

If I zoom in and look at sample points there is no apparent ringing at frequencies that would show up this way in the recorded file.  The ADC and DAC were good enough not to contribute any noise at levels that are going to be heard in normal playback.  The piano sounds distorted in both files over headphones and speakers.  Apparently in the recording.

 

I also slowed the files down by 50%.  Sounds weird, but low frequencies are lower and high frequencies too.  So my bad ear should be good to 24 khz.  I still don't hear any certain difference. Definitely no large differences.

 

So I had one final idea.  I was impressed that the drift between ADC and DAC were just about 1 ppm.  You were very lucky to get that close a match unless you had locked clocks. So I upsampled both files to 96/24.  This is after matching volume.  I took the left channel from each and made a stereo track.  There is close to a perfect match at one point.  So you can listen to several seconds in synch.  Over speakers everything images dead center except some cymbal splash was one sided.  I reversed channels and it stays to that side so a reflection issue in my setup. I then listened over headphones and everything images dead center and you hear no oddities from quality differences in the files.  So I don't know what you folks are hearing.

 

 

[esldude]

Here is a 5 second snippet of the Yggy in one channel and the original file in the other.

 

Yggy composite snippet.wav.zip

 

 

It will unzip into a 96/24 wav.  Should be just the ticket for Jud to listen to.  No echoic memory involved. Put it on loop play and listen over and over.

[jhwalker]

21 minutes ago, esldude said:

If I am clear on which files we are talking about.  The Yggy file recorded by the ADC and the original wav file right?

 

Yes the frequency response is very, very close.  I mentioned earlier they match to less than .1 db until about 10 khz and the Yggy dips eventually to -.25 db by 20 khz.  Below 8 khz they match to within .01 db in basic frequency response.  I had to get a close volume match to determine this.  I did the volume matching and listening first.  Fortunately the match from that was good enough for the analysis. So it sounded the same or very, very close before I knew the frequency response.  My hearing is pretty solid to 12 khz in one ear, and 15 khz in the other.  Age is not kind to our hearing.

 

If I zoom in and look at sample points there is no apparent ringing at frequencies that would show up this way in the recorded file.  The ADC and DAC were good enough not to contribute any noise at levels that are going to be heard in normal playback.  The piano sounds distorted in both files over headphones and speakers.  Apparently in the recording.

 

I also slowed the files down by 50%.  Sounds weird, but low frequencies are lower and high frequencies too.  So my bad ear should be good to 24 khz.  I still don't hear any certain difference. Definitely no large differences.

 

So I had one final idea.  I was impressed that the drift between ADC and DAC were just about 1 ppm.  You were very lucky to get that close a match unless you had locked clocks. So I upsampled both files to 96/24.  This is after matching volume.  I took the left channel from each and made a stereo track.  There is close to a perfect match at one point.  So you can listen to several seconds in synch.  Over speakers everything images dead center except some cymbal splash was one sided.  I reversed channels and it stays to that side so a reflection issue in my setup. I then listened over headphones and everything images dead center and you hear no oddities from quality differences in the files.  So I don't know what you folks are hearing.

 

 

 

I'm not sure what others are hearing, either - once the two files are matched volume-wise, they sound virtually identical to me.  Certainly neither is "unlistenable" or whatever - pretty sure I couldn't ABX them.

[jabbr]

7 minutes ago, jhwalker said:

 

I'm not sure what others are hearing, either - once the two files are matched volume-wise, they sound virtually identical to me.  Certainly neither is "unlistenable" or whatever - pretty sure I couldn't ABX them.

Ok here's my point about whether this is an acceptable test:

 

Let's say we get 10,000 people to ABX and "prove" beyond a reasonable doubt that there is no observable difference between the files. Does that mean there is no sonic difference between the DACs? Or would this mean that all DACs sound the same?

[esldude]

8 minutes ago, jabbr said:

Ok here's my point about whether this is an acceptable test:

 

Let's say we get 10,000 people to ABX and "prove" beyond a reasonable doubt that there is no observable difference between the files. Does that mean there is no sonic difference between the DACs? Or would this mean that all DACs sound the same?

You are forgetting it doesn't prove it fails to reject the null hypothesis.

 

It definitely would not prove that all DACs sound the same. 

 

It would indicate the recording via TEAC/Yggy vs the original file is close enough any difference was not heard by those taking part in the test. With their varied hearing and systems.  The two could sound different on better gear with better listeners.  Yet until you show that you haven't rejected the null hypothesis.

 

To make it more meaningful we need to vet the listeners and have them all listen over gear vetted to be of rather higher than average fidelity. 

 

It would seem unlikely there are large differences in the two (though not proven) if you had 10,000 results to work with.

 

Now if you also did measurements on the gear, and found distortions of all known kinds to be below known thresholds of audibility you would have two pieces of evidence pushing toward the idea the TEAC/Yggy was audibly extremely good.  We do have something like that though perhaps not to the level of accuracy you have been writing about recently. 

 

Any problems with those statements in your opinion?

[jabbr]

31 minutes ago, esldude said:

You are forgetting it doesn't prove it fails to reject the null hypothesis.

 

It definitely would not prove that all DACs sound the same. 

 

It would indicate the recording via TEAC/Yggy vs the original file is close enough any difference was not heard by those taking part in the test. With their varied hearing and systems.  The two could sound different on better gear with better listeners.  Yet until you show that you haven't rejected the null hypothesis.

 

To make it more meaningful we need to vet the listeners and have them all listen over gear vetted to be of rather higher than average fidelity. 

 

It would seem unlikely there are large differences in the two (though not proven) if you had 10,000 results to work with.

 

Now if you also did measurements on the gear, and found distortions of all known kinds to be below known thresholds of audibility you would have two pieces of evidence pushing toward the idea the TEAC/Yggy was audibly extremely good.  We do have something like that though perhaps not to the level of accuracy you have been writing about recently. 

 

Any problems with those statements in your opinion?

In order to validate the test I would have the group (eg 10,000) ABX NOS1 vs Yggy. If there is a demonstrable difference, then I would test the same group against the files, and look for intraobserver variability.

 

That would be an empirical way of validating the recordings as a test of NOS1 vs Yggy but really then you'd need to repeat for different recordings and then between different DACs.

 

This would be a huge and costly undertaking. Easier to do real measurements using real test equipment. What is still needed is to validate those measurements against sound but that's actually much easier!

[Ralf11]

Mani,  Could you list DACs that like better for redbook than the Schiit?