Review and Measurements of Schiit Yggdrasil V2 DAC

[Dro]

I don't know about how well speakers do in terms of distortion, but it will be hard to find a headphone that consistently does less than -60 dB for the second harmonic distortion. In that case, having -80 or -110 dB distortion come from your DAC and amp makes very little difference.

 

[Grave]

By the way, I never listened to any Schiit or Topping DAC in any system (I have Schiit Eitr USB/SPDIF converter). But what intrigues me is that regardless of really good measurements, some listeners are not satisfied with Topping's sound. Is this a bias, or those listeners actually do hear something being wrong?

They are just idiots.

The last three audibly transparent DAC/amps I have owned have sounded like nothing but extremely clear. If extremely clear sounds bad to you then idk what to say to that.

They are certainly not hearing anything being wrong, since Topping DAC/amps are easily audibly transparent, as are worse performing DAC/amps.

 

[audiobill]

They are just idiots.

The last three audibly transparent DAC/amps I have owned have sounded like nothing but extremely clear. If extremely clear sounds bad to you then idk what to say to that.

They are certainly not hearing anything being wrong, since Topping DAC/amps are easily audibly transparent, as are worse performing DAC/amps.

Schiit uses fecum to make their DACs sound better.

 

[Blumlein 88]

I suspect there's something uncovered by objective measurements as being done today, which would strictly relate objective to blind test listening. And I don't just mean audible or not. I suspect testers might prefer some equipment that measure poorer at least in some setups.

Yggdrasil vs Topping DX7s might be a perfect experiment. Objective side makes it pretty clear, and expected possible results from the objective perspective is either Topping wins (distortions audible on the 'Yggy' side) or without conslusion (distortions on both proved to be inaudible in practice). But what if Yggdrasil won the blind test?

Only one way to find out: do the blind test comparison.

By the way, I never listened to any Schiit or Topping DAC in any system (I have Schiit Eitr USB/SPDIF converter). But what intrigues me is that regardless of really good measurements, some listeners are not satisfied with Topping's sound. Is this a bias, or those listeners actually do hear something being wrong?

You might be onto something. Then again maybe not.

What has sometimes happened in such comparisons is an example of just how powerful our eyes are over our hearing. A good story, a great built, beautiful device, and us knowing which is which and one unit seems to sound so much better. Beyond simple transparency and adds something more. Or so the story goes. Then if the identities are removed said wonder device is sometimes found to simply be inferior with a sound of its own added to real fidelity.

Now mind you I've see it go the other way too. The difference was heard, and preferred over simple fidelity.

My actual expectation for the Yggy vs some other good DAC is blind it will not be detectable. Without the story, the knowledge of the price and reputation, it will fair no better than another cheaper better measuring DAC. While it has its measured faults, I don't know if they are all that audible. I would expect not.

We've read of the Circle of Confusion in loudspeakers. We have something of a circle of confusion of the opposite extreme with regard to DACs. Mostly if FR is flat they all sound the same even though we can measure differences. The confusion is where an inferior measured performance becomes audible.

 

[andreasmaaan]

Ok, didn't plan to get into another debate about distortion audibility thresholds before, but it seems I've blundered into one. So here's the study I had in mind when I said the threshold might be lower than 0.01%. I've definitely posted this one before here and I think it's been discussed a bit on another thread. So nothing new...

The thing is that the researchers in this cases were using DSP to emulate THD+N caused by capacitors and and opamps. So they ended up with some nonlinear noise/distortion patterns that were quite strange. Listeners were also trained and were being tested on the mere ability to detect differences, not on preference (although incidentally the study also found that in many cases listeners expressed a preference for slightly distorted signals).

Anyway, some of the findings were (p. 171):

In the listening tests performed for this dissertation, participants were able to detect levels of capacitor distortion as low as .002% THD+N and levels of op amp distortion as low as .005% THD+N... These distortion percentages are expressed as integrated averages over the 48KHz bandwidth of the audio samples used in listening tests.

The overall distortion perception thresholds observed in this dissertation were low relative to the thresholds for distortion perception on program material and complex signals reported in many past studies. Still, these past studies have varied significantly in their criteria for testing perceptual thresholds for distortion. The listening tests performed for this dissertation work used trained listeners in a controlled experimental environment. In these experiments, participants were asked to distinguish even the slightest difference between two audio samples. Many of the above-referenced studies were designed to determine when the distortion had reduced the quality of a signal.

 

[andreasmaaan]

Looking back at the Schiit distortion figures, it seems to be at about 0.005% at max output. The above study would predict there is at least some chance this distortion would be audible.

More interestingly, if the Schiit's distortion is audible, this study would suggest there is some chance it would be preferred over the DX7s (which should not produce any audible distortion).

Not necessary my personal view, just comparing @amirm's measurements against the study's findings...

 

[zalive]

You might be onto something. Then again maybe not.

What has sometimes happened in such comparisons is an example of just how powerful our eyes are over our hearing. A good story, a great built, beautiful device, and us knowing which is which and one unit seems to sound so much better. Beyond simple transparency and adds something more. Or so the story goes. Then if the identities are removed said wonder device is sometimes found to simply be inferior with a sound of its own added to real fidelity.

Now mind you I've see it go the other way too. The difference was heard, and preferred over simple fidelity.

My actual expectation for the Yggy vs some other good DAC is blind it will not be detectable. Without the story, the knowledge of the price and reputation, it will fair no better than another cheaper better measuring DAC. While it has its measured faults, I don't know if they are all that audible. I would expect not.

We've read of the Circle of Confusion in loudspeakers. We have something of a circle of confusion of the opposite extreme with regard to DACs. Mostly if FR is flat they all sound the same even though we can measure differences. The confusion is where an inferior measured performance

I'm not much in knowledge of scientific research results. I read hear and there about some general conclusions, so I might know this or that but I never directly studied those. From what I understand, it's the profile of distortion that can make it generally sound worse, not change much, or sound even nicer. Aren't tubes the perfect example? They generally have a higher distortion than a correct solid state design, yet it's the profile of distortion typical to tubes which sounds more pleasing than a profile of distortion typical to a solid state. If we describe high fidelity idealistically as the truth and distortion as a lie, then both tubes and solid state lie, where tubes typically might lie more...however they use pleasing lies, while solid state uses harsh lies.

Now the question is this: do listeners like more the truth, or pleasing lies? It would be much easier to answer this if we had a system which completely tells the truth. However no such. Lies may be to a lesser degree as systems designs and components used get better, but every system lies to a certain degree. So we don't know for sure whether truth is pleasing enough, or some beautifying helps to make it sound better. But in realistic world...it's a type of lie that might decide the outcome, and it's quite possible that bigger but pleasing lie would be assessed by listening panel (blindly, of course) as sounding better than smaller but harsher lie (to 'ears').

Now as with Yggrdrasil...I don't think what shows in measurements as being bad (namely, issues recognized by Amir like power supply noise leakage, or a faulty USB clock at 1.008 Mhz) are what was intended by its design, or something which was done on purpose to create a more pleasing to the ear kind of distortion. Looks like me more like design flaws which originate from cutting production cost (which is very bad for a digital device costing as much. I don't see justification). Now, as for not using dither in 24 bits format and going for rounding bits...this is possible that Schiit designer thinks it creates a more pleasing to the ear result. From my point of view, it's not impossible. After all, a decent designer should ensure listening tests during development phase in which listeners can evaluate different implementation ideas, so quite possible that dither vs no dither (in different combos like truncation or rounding) was tested through listening, and no dither was picked as one which was voted to sound better, even if it doesn't measure better.

Generally I think Amir made a great job in exposing weaknesses of what looks to be an overpriced product. It's great to tell the truth. But from the objectivists vs subjectivist side, I'm actually more interesed in Topping DACs and how are them when it comes to pleasing sound. If I listen to objectivists they will tell something like 'it measures better...of course it sounds better!', however that assesment didn't come from their ears if it's what they hear as well, it might be the truth...however, what they hear might be biased as well because of measurements result influence. You're told what to hear, so you hear it exactly like this...right? Bias is a two sided blade.

This is why I tend to believe my own ears the most. Of course I'm not immune to biases. However as long as they are mine (biases), I can battle them myself. And there are situations where there is simply no bias. For example, same musical chain, everyhing the same including laptop PC instalation and kernel and setup, parameters available set to the same values...the only difference, two different SW player. Environment was Audiophile Linux, non rt kernel, direct ALSA on both. I didn't hear anything previously on any of them from anyone, and for what I know, both were chosen to represent audiophile players by AP Linux distro designer. So both should be good. However, they didn't sound the same, and I was able to recognize and describe the difference. Mostly, one of them (DeadBeeF) handled highs better, with more subtlety and reality, even a bit better resolution. Sound was milder and more pleasing but with better realism. So, if there's no bias and one was obviously more pleasing to me, then the difference is real, I guess? Also, there are other situation where bias tells me one thing, yet I hear something different. The same example environment: the AP Linux author convinced that RT kernel session was the better sounding one because of minimum latency. This is what I expected to hear. This is not what I heard. It was actually clearly audibly harsher, and I wondered what the AP Linux author heard when he thought it was better. Anyway as for examples: as long as I keep hearing things differently to what my bias should tell me, or am able to hear difference in sound (and also locate and describe the difference) in a situation where there is no bias like price, brand or comments of others (read or heard), I don't have a reason to doubt that I do hear something real. Now, whether that liking comes from better fidelity or a more pleasing lie, that is less concern to me - I cannot command to myself what's more and what's less pleasing to hear to me, it's just what it is. What's beneath is less important to me, as I'm not an audio gear designer, I'm just a consumer.

 

[Blumlein 88]

Ok, didn't plan to get into another debate about distortion audibility thresholds before, but it seems I've blundered into one. So here's the study I had in mind when I said the threshold might be lower than 0.01%. I've definitely posted this one before here and I think it's been discussed a bit on another thread. So nothing new...

The thing is that the researchers in this cases were using DSP to emulate THD+N caused by capacitors and and opamps. So they ended up with some nonlinear noise/distortion patterns that were quite strange. Listeners were also trained and were being tested on the mere ability to detect differences, not on preference (although incidentally the study also found that in many cases listeners expressed a preference for slightly distorted signals).

Anyway, some of the findings were (p. 171):

Your link doesn't seem to work. I do remember this being discussed elsewhere. There were plenty of issues with how that was conducted if I'm thinking of the correct one. Is this it?

http://digitool.library.mcgill.ca/webclient/StreamGate?folder_id=0&dvs=1536138657281~476

 

[andreasmaaan]

If I listen to objectivists they will tell something like 'it measures better...of course it sounds better!', however that assesment didn't come from their ears

I consider myself 100% an objectivist in these matters, but I extend that objectivity to being open to the possibility that objective data (obtained via properly controlled listening tests) could lead to conclusions that audible distortion is preferred by some or many listeners.

Indeed, I don't think this topic has been adequately explored by the research, but I do think there are enough data in the studies to suggest that it is at least a possibility.

Note that this is a different thing from "believing my own ears the most"

And there are situations where there is simply no bias. For example, same musical chain, everyhing the same including laptop PC instalation and kernel and setup, parameters available set to the same values...the only difference, two different SW player.

I disagree. Wherever the test is sighted, there is the possibility of bias. There's plenty of research on this (mostly not in the field of audio), so much so that I take it to be an accepted fact. Note that this is not necessarily bias in the sense that you go into the test having a prejudged preference, but it is bias nonetheless. This format of test just cannot produce reliable results.

 

[zalive]

Looking back at the Schiit distortion figures, it seems to be at about 0.005% at max output. The above study would predict there is at least some chance this distortion would be audible.

More interestingly, if the Schiit's distortion is audible, this study would suggest there is some chance it would be preferred over the DX7s (which should not produce any audible distortion).

Not necessary my personal view, just comparing @amirm's measurements against the study's findings...

What interests me is following.

Usual DA equipment (say, not the highest class available) acquire more or less a sound which would be described by many as 'digital', and which is caracteristic for less than realistic reproduction of highs, where there's significant audible distortion present. Obvious example: applause. On which digital equipment applause sounds like applause should sound? It's far away from what it should sound like. It's nowhere near as realistic as in analogue phonogram reproduction (even when far away from high end), if you ask me.

What in objective measurements correlates to this particular distortion, as heard in applause recordings? Which measurement graphs shows it, and how can it be recognized? And which is the threshold for this particular distortion for audible?

@amirm I'm hoping for your reply on this

 

[andreasmaaan]

Your link doesn't seem to work. I do remember this being discussed elsewhere. There were plenty of issues with how that was conducted if I'm thinking of the correct one. Is this it?

http://digitool.library.mcgill.ca/webclient/StreamGate?folder_id=0&dvs=1536138657281~476

Yeh, I remember that some links work for some people and other links work for others. If you google the title of the study you'll find a link that works for you:

Robert-Eric Gaskell, "Subjective and Objective Evaluation of Distortion in Analogue Electronics: Capacitors and Operational Amplifiers"

I would be really interested to hear your thoughts on the flaws in this study @Blumlein 88. I recall having a conversation with you about another similar study back then, but I'm not sure if we ever specifically discussed possible flaws in this one

 

[zalive]

I disagree. Wherever the test is sighted, there is the possibility of bias. There's plenty of research on this (mostly not in the field of audio), so much so that I take it to be an accepted fact. Note that this is not necessarily bias in the sense that you go into the test having a prejudged preference, but it is bias nonetheless. This format of test just cannot produce reliable results.

I'm not saying it's the proof (it isn't), or a format for objective tests (it isn't), but you must accept the possibility that I'm right neverhteless, and that what I heard in a particular situation was real and not biased.

 

[andreasmaaan]

I'm not saying it's the proof (it isn't), but you must accept the possibility that I'm right neverhteless, and that what I heard was real and not biased.

Yes, I accept that possibility. Unfortunately though, the possibility that your experiences are influenced by your knowledge of what you are listening to (and even the knowledge that you are listening to two different things) means that your belief is unreliable, which you also must accept.

 

[andreasmaaan]

On which digital equipment applause sounds like applause should sound?

The goal of a transparent system is not to make recorded applause sound "like applause should sound".

It should make the recorded applause sound like the recording of the applause. That's a subtle but very important distinction.

 

[Blumlein 88]

Yeh, I remember that some links work for some people and other links work for others. If you google the title of the study you'll find a link that works for you:

Robert-Eric Gaskell, "Subjective and Objective Evaluation of Distortion in Analogue Electronics: Capacitors and Operational Amplifiers"

I would be really interested to hear your thoughts on the flaws in this study @Blumlein 88. I recall having a conversation with you about another similar study back then, but I'm not sure if we ever specifically discussed possible flaws in this one

Well it will take a while with 170 pages. I'm not sure it is the one I was thinking of either. I believe the one I had in mind was only about op amps and not capacitors.

 

[zalive]

Yes, I accept that possibility. Unfortunately though, the possibility that your experiences are influenced by your knowledge of what you are listening to (and even the knowledge that you are listening to two different things) means that your belief is unreliable, which you also must accept.

There was no previous knowledge about two SW players I was comparing in my case, nor I heard or read any opinions on them.
So it' unreliable in your view of an objectivists as it's not objective, however completely reliable to me.

When you drive and see a car heading in your direction straight towards you, do you find what your vision is telling you reliable, or you suspect you might be biased by your fears or a possible hallucination? And was there objective proof that car was heading your way? Or is your assesment good enough...for you? On the objective side there's no proof it headed at you, yet you consider it good enough for your personal use, right?

 

[Krunok]

What interests me is following.

Usual DA equipment (say, not the highest class available) acquire more or less a sound which would be described by many as 'digital', and which is caracteristic for less than realistic reproduction of highs, where there's significant audible distortion present. Obvious example: applause. On which digital equipment applause sounds like applause should sound? It's far away from what it should sound like. It's nowhere near as realistic as in analogue phonogram reproduction (even when far away from high end), if you ask me.

I don't agree, to me it sounds pretty much the same. If your claim is to be valid it should be tested with the same recording in the controlled environment, otherwise it is just subjectve impression.

 

[andreasmaaan]

There was no previous knowledge about two SW players I was comparing in my case, nor I heard or read any opinions on them.
So it' unreliable in your view of an objectivists as it's not objective, however completely reliable to me.

When you drive and see a car heading in your direction straight towards you, do you find what your vision is telling you reliable, or you suspect you might be biased by your fears or a possible hallucination? And was there objective proof that car was heading your way? Or is your assesment good enough...for you? On the objective side there's no proof it headed at you, yet you consider it good enough for your personal use, right?

In the example you give, my subjective opinion is not good enough (it is not fully reliable; I may be mistaken).

Rather, it is merely the best information available on which to form a judgement. Relying on the information is my only option.

In audio, on the other hand, there is better information available than my own unavoidably biased perceptions obtained through sighted tests.

Please, the argument I'm making about bias is a basic tenet of psychological research that is accepted by all psychologists and scientists who work with subjective data. It is 100% uncontroversial.

 

[zalive]

The goal of a transparent system is not to make recorded applause sound "like applause should sound".

It should make the recorded applause sound like the recording of the applause. That's a subtle but very important distinction.

I understand that and I agree. However no one knows for sure how a certain record should sound, as we deal with reproductions. On the other hand, comparison with live event is a natural reference, regardless of what record should sound, because when you don't know how record should sound, live event sound is the best reference point you can set. The ideal of fidelity is...fidelity. It should apply both on sound technicians and on hi-fi consumers.

 

[andreasmaaan]

I understand that and I agree. However no one knows for sure how a certain record should sound, as we deal with reproductions. On the other hand, comparison with live event is a natural reference, regardless of what record should sound, because when you don't know how record should sound, live event sound is the best reference point you can set. The ideal of fidelity is...fidelity. It should apply both on sound technicians and on hi-fi consumers.

Sure, nobody knows how the recording should sound by stabbing in the dark.

That's why so much research has been done to determine the thresholds at which distortion becomes audible.

We can listen to the recording on a system that we have measured and determined to not exceed any of these thresholds. Only by listening to the recording on such a system can we satisfy ourselves that we are (likely to be) listening to the recording as it should sound.

EDIT: I say "likely to be" because room always needs to be left for a degree of experimental error in the studies that determine these thresholds to begin with.