Measurements and Review of Schiit Yggdrasil DAC

[Jimster480]

The gallery comments are rather funny. The other day we were talking about classic dacs from the past and how sought after some were. How the Berkley has stood the test of time and still sounds good. Mike Moffat is the pretty much the first designer of the stand alone dac, and some of his designs are sought after and still recognized as excellent sounding. You really can't have it both ways. If you can talk about sound quality of classic equipment it would only be fair listen to current equipment and factor that in to get a total assessment.

Just becuase he designed things a long time ago doesn't mean that his designs were good.
People are seeking after schiit products that sound like garbage and measure like garbage too.... does that make them good? No it makes the people seeking after them clueless or enticed by fancy marketing and "group think".

 

[rebbiputzmaker]

Just becuase he designed things a long time ago doesn't mean that his designs were good.
People are seeking after schiit products that sound like garbage and measure like garbage too.... does that make them good? No it makes the people seeking after them clueless or enticed by fancy marketing and "group think".

You speak for tiny number of haters which if fine for you but the majority of users are pretty satisfied with their products. You can scream as much as you want but people will in the end trust their judgement.

 

[DonH56]

I was discussing the issue of bit depth bought up here with the yggy apparently acheiving 16 bits using the linearity measurement. I was told that linearity had almost nothing to do with determining bit depth. It was S/N ratio that determines it. With the Yggy having 119 db which produces the 21 bits claimed.
I would like to know
1. What you think of that.
2. Where are the S/N tests?

Here's where I'm getting it from.

There are some errors in this article but it provides a good overview: https://en.wikipedia.org/wiki/Audio_bit_depth

Linearity and bit depth are coupled but you can have great linearity with few bits and many bits but with low linearity.

SNR is the ratio of the signal to the total of all the noise (integrated; "summed", though it is not a linear sum). SNR = 6N+1.8 for an ideal N-bit converter assuming quantization noise only. That is, the noise because the signal is sampled to N bits instead of an infinite number of bits, which causes the values to be resolved to 1 part in 2^N (about). The noise floor, or spurious-free dynamic range, is around 9N and is the "noise floor" expected for an ideal converter. Noise comes from circuit elements like resistors, transistors, and such that also contribute noise so is more than just quantization error. A real DAC will always have just a little lower SNR than ideal though good ones come close.

 

[Jimster480]

You speak for tiny number of haters which if fine for you but the majority of users are pretty satisfied with their products. You can scream as much as you want but people will in the end trust their judgement.

In the end sadly most people have no point of reference. Audiophiles are the small % in the first place.
Most people are completely content with their $5 JVC Gummy earbuds or $10 Panasonic RP-TCM125 or apple earpods....

Those people think those headphones are completely fine and "awesome", so what does that say? Does it make those headphones awesome for $5? You would look like a complete fool for buying even a Modi at $99 when a $5 headphone played out of your phone is fine right?
Subjectivity fails every single time, because its all perspective. The majority of Schiit customers are people who have no basis of good Audio, they keep buying the better and better tier to hear an "upgrade" but in the end Yggdrasil is as good as a $90 DAC from a slew of other companies....
Had you started somewhere OTHER than Schiit you would probably be able to tell that all their products perform horribly.

As with the studies into how our brains process sounds.... people cannot trust their ears because after some days you get "used to" the crappy sound. And then everything is "fine"... also depends on how musically inclined you are. Personally I bought a FiiO K1 first and then a Fulla2 and the Fulla2 sounded noticeably worse....
Infact in my list trip I brought my Fulla2 and actually found out that it sounds worse than my laptop headphone output.... making it a pointless item to bring to listen to my Aeon or PM-2 or AH-MM400.....

So once again subjectivity just fails entirely as its seeded in lack of experience and also ego and denial. The same applies to cars, if you were into cars you would understand what I mean..... since most car marketing makes claims to every car feeling great and performing incredibly well with some nice shots of a car flying through a city or on a track.... but that doesn't mean that those cars are objectively good.... and usually they are not and are beaten by cars that are a fraction of the cost even though their owners don't want to believe it or want to try to justify their purchases.

 

[garbulky]

There are some errors in this article but it provides a good overview: https://en.wikipedia.org/wiki/Audio_bit_depth

Linearity and bit depth are coupled but you can have great linearity with few bits and many bits but with low linearity.

SNR is the ratio of the signal to the total of all the noise (integrated; "summed", though it is not a linear sum). SNR = 6N+1.8 for an ideal N-bit converter assuming quantization noise only. That is, the noise because the signal is sampled to N bits instead of an infinite number of bits, which causes the values to be resolved to 1 part in 2^N (about). The noise floor, or spurious-free dynamic range, is around 9N and is the "noise floor" expected for an ideal converter. Noise comes from circuit elements like resistors, transistors, and such that also contribute noise so is more than just quantization error. A real DAC will always have just a little lower SNR than ideal though good ones come close.

Thank you. I appreciate your feedback. I just learned of the SNR figure from head-fi as well. I have to play around with it. Like solving for Number of bits, I imagine would be SNR/6 + 1.8?

How is the value of 16 bits for the Yggy derived from a linearity measurement rather than a SNR measurement because it seems like things can vary a bit few bits, great linaerity, many bits, low linearity? Wouldn't an SNR measurement be a more complete measurement?

 

[DonH56]

SNR encompasses more than just linearity since it is usually actually SINAD (signal to noise and distortion) that is reported. SNR/SINAD are measured using dynamic signals (even if just a single tone). So in that sense it includes more things. SNR alone would not include distortion terms but makes for a much harder measurement since you have to separate noise from distortion terms. SINAD is generally easier and more useful; measure the peak signal level, then subtract it and whatever is left is noise and distortion. That is fairly easy to do mathematically using an FFT, curve-fitting program, or whatever. (Aside: Because most analyzers measure SINAD, and not just THD, you see a rise in "distortion" at low levels in amplifiers as noise begins to dominate the measurement.)

Linearity is the deviation from ideal so a 4-bit converter could be considered to have 16-bit linearity if each one of the 16 steps is "perfect" to 16 bits. Linearity plots are often at "DC" using a very slow sweep.

The linearity plots show how the DAC's output deviates from ideal as the signal level is reduced. A perfect DAC with infinite resolution would be a flat (straight) line to -infinity. As the output deviates from ideal, that is becomes nonlinear, it shows up in the plots. Amir is plotting linearity vs. dBFS, dB relative to a Full-Scale output, and equating that to the ideal SNR so in some sense shows the same thing.

All these different measurements show different facets of performance. Some may "click" with some readers more than others. SNR/SINAD includes a lot of things but make it harder to separate say DC linearity from AC (dynamic) performance degradation. Different measurements provides clues into the performance and limitations of the DAC and help isolate the different sources of error.

Since SNR = 6N+1.8 dB (approximately) then N = (SNR-1.8)/6 bits.

 

[amirm]

@amirm I was discussing the issue of bit depth bought up here with the yggy apparently acheiving 16 bits using the linearity measurement. I was told that linearity had almost nothing to do with determining bit depth. It was S/N ratio that determines it. With the Yggy having 119 db which produces the 21 bits claimed.
I would like to know
1. What you think of that.
2. Where are the S/N tests?

Here's where I'm getting it from (headfi).: The poster has a doctorate in electronic engineering
"I've been trying to stay out of the argument/discussion about someone's measurements on some other forum I know nothing about, and what they might or might not mean. But I will say one thing concerning bit depth and the discussion above. The practical effect of bit depth is it determines the dynamic range of the signal and the S/N of the device. In theory, a 24-bit digital audio signal has a maximum dynamic range of a little over 144db, and 16-bit is approx. 96db. The S/N of a digital device demonstrates it's practical bit depth capability. According to the Schiit website Yggdrasil is rated (I have not measured it) at >119db S/N, which calculates to approximately 21 bits. Linearity is not a determinant of bit depth."

I wrote a complete response to this in Mike's thread and it was this reply that was deleted and I was given a reply ban. So I say it is safe to say what I had to say was not to some people's liking.

I did not save the reply so here is another version.

The motivation for both running the linearity test and determining bit depths from it actually came from tests that Stereophile used to run. Here is an example: https://www.stereophile.com/digitalprocessors/187/index.html

If I gave you this graph, and I asked you where there is too much error, you would probably draw a line a bit after the curve gets wiggly, say, around -110 dB. That is exactly what I am doing with my 0.1 dB variation from linearity.

While linearity is shown as a figure related to input value (X axis), it is very common to instead talk about it in the order of bits. See: https://en.wikipedia.org/wiki/Differential_nonlinearity

And much more deeper treatment from the manufacturer of the DAC chip used in Schiit Yggy: http://www.analog.com/media/en/trai...ndbooks/Data-Conversion-Handbook/Chapter5.pdf

LSB is least significant bit. The error can be 1, 2, 3 or more of LSB, reducing the effective resolution of the DAC by the same number of bits. So we are talking about bits in linearity measurements.

Here is a partial spec from Analog Devices on this front:

As you see they show 20 bits as the maximum resolution of the DAC, but then show under different conditions how much that is degraded in units of LSB (bits). In the example I have shown the error can be +- 1 dB or 2 dB total making the effective, error free resolution 18 bits. Worst case actually brings the resolution way down to 15 (-2.5 to + 2.5 bits). Don't run with these numbers as these are just the example from above. I am showing them just to demonstrate the common interpretation of linearity in the form of bits.

It is also true that we can also convert signal to noise ratio into bits. Here ENOB (effective number of bits) is a simple way to express something as is how it is used above. Here we are trying to determine how noisy the DAC is. In linearity we are trying to determine how distorted it is.

As to Schiit spec, this is all they say:

What don't know what frequency was used for this measurement. Or whether they used weighting which filters a lot of the noise making the number a lot higher. It is also possible this is measured with no signal, and then a full amplitude signal. A proper signal to ratio would be looking at the noise in presence of our full amplitude signal as the noise level can increase with signal. Without this kind of detail or a spectrum graph as I post, there is just no way of knowing what this number is.

SNR is such an important marketing number that manufacturers resort to every trick in the book to come up with the best number they can.

Summary
ENOB or effective number of bits can be applied to both distortion and noise measurements. My use of it with respect to linearity is absolutely consistent with how DACs are rated and measured. It is however not a common practice or term in audio reviews and hence lack of familiarity with it even from that Phd engineer.

Ultimately you can ignore my ENOB computation and just look at the graphs. The visual completely makes the same point when the DAC is compared to another showing much less error.

 

[DonH56]

Nice job Amir. I decided to not bring DNL and INL in but that is fundamentally what most mean by (DC, or static) linearity. ENOB is used by the IEEE so I would think a designer would have at least a copy of the Standard sitting around... Outside the audio world ENOB is in widespread use as it encompasses essentially all deviations from ideal (noise or linearity). ENOB is usually a dynamic (AC) measurement.

My data converter programs all generate DNL, INL, SNR, SINAD, THD, and ENOB since those are the most common parameters reported. I rarely present them all because it's too much work to explain them and I am lazy (busy, take your pick).

 

[rebbiputzmaker]

In the end sadly most people have no point of reference. Audiophiles are the small % in the first place.
Most people are completely content with their $5 JVC Gummy earbuds or $10 Panasonic RP-TCM125 or apple earpods....

Those people think those headphones are completely fine and "awesome", so what does that say? Does it make those headphones awesome for $5? You would look like a complete fool for buying even a Modi at $99 when a $5 headphone played out of your phone is fine right?
Subjectivity fails every single time, because its all perspective. The majority of Schiit customers are people who have no basis of good Audio, they keep buying the better and better tier to hear an "upgrade" but in the end Yggdrasil is as good as a $90 DAC from a slew of other companies....
Had you started somewhere OTHER than Schiit you would probably be able to tell that all their products perform horribly.

As with the studies into how our brains process sounds.... people cannot trust their ears because after some days you get "used to" the crappy sound. And then everything is "fine"... also depends on how musically inclined you are. Personally I bought a FiiO K1 first and then a Fulla2 and the Fulla2 sounded noticeably worse....
Infact in my list trip I brought my Fulla2 and actually found out that it sounds worse than my laptop headphone output.... making it a pointless item to bring to listen to my Aeon or PM-2 or AH-MM400.....

So once again subjectivity just fails entirely as its seeded in lack of experience and also ego and denial. The same applies to cars, if you were into cars you would understand what I mean..... since most car marketing makes claims to every car feeling great and performing incredibly well with some nice shots of a car flying through a city or on a track.... but that doesn't mean that those cars are objectively good.... and usually they are not and are beaten by cars that are a fraction of the cost even though their owners don't want to believe it or want to try to justify their purchases.

You are welcome to believe what you want. Thankfully there are people who know better and will continue to use proper judgment. Have you ever met or spoken to real designer, or audio engineers? I have been lucky over the years to meet and speak to a good number of famous ones. Invariably they used all the tools at hand INCLUDING listening in their designs.

 

[Purité Audio]

Daniel Weiss famously uses measurements , one of his DACs was the finest measuring piece of equipment John Atkinson had reviewed in his 25 years.
Keith

 

[Jimster480]

You are welcome to believe what you want. Thankfully there are people who know better and will continue to use proper judgment. Have you ever met or spoken to real designer, or audio engineers? I have been lucky over the years to meet and speak to a good number of famous ones. Invariably they used all the tools at hand INCLUDING listening in their designs.

Those who know better are on forums like this, watching the measurements and buying what does perform as the mfg claims.

Those who wish to justify their waste of money stay on other forums where information that proves certain brands to be lying about their performance or deliberately deceiving their customers for additional profit will be removed.

 

[garbulky]

@amirm and @DonH56 . Thank you for the detailed explanation. To be honest I'm still confused as I'm hearing two different sides from two different people. But still appreciate the time you took to try and explain this to me. Hopefully others can benefit from your explanation. Perhaps in the future you can also include an SNR measurement and then compare it from the linearity measurement to see if they agree and if they don't, perhaps a discussion.

Fwiw, at head-fi where I posted a summary of your explanation best as I could understand it, somebody mentioned that John Atkinson had this to say about judging bit resolution on the yggy,

"John Atkinson gave this response to why he didn't include the linearity test in his Yggdrasil measurements.

I didn't offer that judgement because the Yggdrasil is not like conventional D/A processors that use 24-bit D/C chips. With the latter, the resolution will be limited by the thermal noise of the overall DAC+analog circuitry, which produces a random noisefloor. With the Schiit, however, while the analog noise is very low, as I mentioned in the review, the noisefloor will rise with 24-bit audio due to what appears to be truncation of the LSBs. As the noisefloor will therefore be related to the encoded signal, my usual estimate of the DAC resolution will be misleading."

Meanwhile to make things even more confusing Jason says that the Yggy does not truncate.

 

[amirm]

Meanwhile to make things even more confusing Jason says that the Yggy does not truncate.

I tend to agree with him there. Not dithering the source would be a criminal offence in DSP circles.

I have read JA's paragraph above and it doesn't compute for me. Measurements are measurements. If there are design errors, it doesn't matter. He should still interpret the numbers as I do. Maybe he did and it was too low and he didn't want to say it.

 

[Blumlein 88]

Hopefully this isn't too out of context. It goes to Don saying each type of measure illuminates some different aspect of performance.

Lets take a theoretical multi-bit DAC. One that is linear for the first 10 bits and then the level from the lower bits goes hard in the downward direction. By this I mean lower bits 10-20 will register much lower than they should. Maybe bit 20 should result in -120 db, but instead outputs -160 db. Forgive me for my simplifications here.

If we do a SNR, max signal vs silence, the result is very nifty indeed. Being way, way down in the analog noise floor. If we do a dynamic range test, -60 db tone, notched out to show what is remaining. Again you'll get a wonderful dynamic range number for the spec sheet. Yet said DAC is highly non-linear in the lower bit ranges. So countering a poor linearity test with good SNR or dynamic range is to muddy the water and direct attention away from a problem.

I also fail to see how looking at INL and DNL for this Schiit DAC paints a prettier picture.

 

[March Audio]

@amirm and @DonH56 . Thank you for the detailed explanation. To be honest I'm still confused as I'm hearing two different sides from two different people. But still appreciate the time you took to try and explain this to me. Hopefully others can benefit from your explanation. Perhaps in the future you can also include an SNR measurement and then compare it from the linearity measurement to see if they agree and if they don't, perhaps a discussion.

Fwiw, at head-fi where I posted a summary of your explanation best as I could understand it, somebody mentioned that John Atkinson had this to say about judging bit resolution on the yggy,

"John Atkinson gave this response to why he didn't include the linearity test in his Yggdrasil measurements.

I didn't offer that judgement because the Yggdrasil is not like conventional D/A processors that use 24-bit D/C chips. With the latter, the resolution will be limited by the thermal noise of the overall DAC+analog circuitry, which produces a random noisefloor. With the Schiit, however, while the analog noise is very low, as I mentioned in the review, the noisefloor will rise with 24-bit audio due to what appears to be truncation of the LSBs. As the noisefloor will therefore be related to the encoded signal, my usual estimate of the DAC resolution will be misleading."

Meanwhile to make things even more confusing Jason says that the Yggy does not truncate.

I think in the simplest terms, even if the terminology can be misinterpreted, the basic message is still the same. The iggy is not capable of properly and accurately resolving signal levels below approximately -90dBFS which equates to approx the lowest signals in a 16bit system. This is very poor linearity by todays standards. Alternatively, assuming a 24bit recording has any useful audio below 16bits (cough cough thats a big if) this dac will not correctly reproduce it.

 

[Jimster480]

I think in the simplest terms, even if the terminology can be misinterpreted, the basic message is still the same. The iggy is not capable of properly and accurately resolving signal levels below approximately -90dBFS which equates to approx the lowest signals in a 16bit system. This is very poor linearity by todays standards. Alternatively, assuming a 24bit recording has any useful audio below 16bits (cough cough thats a big if) this dac will not correctly reproduce it.

Yes exactly, that is the whole point.
Regardless of how you want to skew the numbers, the DAC isn't capable of resolving anything past 16 bit and in some cases it is on the edge of 16 bit..... but there are other DAC's at 1/20th of the cost which can achieve similar performance.

 

[Deleted member 65]

@amirm and @DonH56 . Thank you for the detailed explanation. To be honest I'm still confused as I'm hearing two different sides from two different people. But still appreciate the time you took to try and explain this to me. Hopefully others can benefit from your explanation. Perhaps in the future you can also include an SNR measurement and then compare it from the linearity measurement to see if they agree and if they don't, perhaps a discussion.

Fwiw, at head-fi where I posted a summary of your explanation best as I could understand it, somebody mentioned that John Atkinson had this to say about judging bit resolution on the yggy,

"John Atkinson gave this response to why he didn't include the linearity test in his Yggdrasil measurements.

I didn't offer that judgement because the Yggdrasil is not like conventional D/A processors that use 24-bit D/C chips. With the latter, the resolution will be limited by the thermal noise of the overall DAC+analog circuitry, which produces a random noisefloor. With the Schiit, however, while the analog noise is very low, as I mentioned in the review, the noisefloor will rise with 24-bit audio due to what appears to be truncation of the LSBs. As the noisefloor will therefore be related to the encoded signal, my usual estimate of the DAC resolution will be misleading."

Meanwhile to make things even more confusing Jason says that the Yggy does not truncate.

If you browse back a number of pages in this thread I've already asked this question (on JA's lack of revealing measurements). A couple of insightful answers there.

 

[Dimitrov]

Hi Amir,

I remember you did a review of the Behringer UMC204HD last year which you classified as the king of low cost DACs. Not sure if it's still king or not, but how would the Behringer UMC204 compare to the Schitt Yggdrasil in terms of its technical performance?

Put another way, does the Yggdrasil have any advantages over the Behringer?

 

[rebbiputzmaker]

I wrote a complete response to this in Mike's thread and it was this reply that was deleted and I was given a reply ban. So I say it is safe to say what I had to say was not to some people's liking.

I did not save the reply so here is another version.

The motivation for both running the linearity test and determining bit depths from it actually came from tests that Stereophile used to run. Here is an example: https://www.stereophile.com/digitalprocessors/187/index.html

View attachment 10764

If I gave you this graph, and I asked you where there is too much error, you would probably draw a line a bit after the curve gets wiggly, say, around -110 dB. That is exactly what I am doing with my 0.1 dB variation from linearity.

While linearity is shown as a figure related to input value (X axis), it is very common to instead talk about it in the order of bits. See: https://en.wikipedia.org/wiki/Differential_nonlinearity

And much more deeper treatment from the manufacturer of the DAC chip used in Schiit Yggy: http://www.analog.com/media/en/trai...ndbooks/Data-Conversion-Handbook/Chapter5.pdf

View attachment 10766

LSB is least significant bit. The error can be 1, 2, 3 or more of LSB, reducing the effective resolution of the DAC by the same number of bits. So we are talking about bits in linearity measurements.

Here is a partial spec from Analog Devices on this front:

View attachment 10767

As you see they show 20 bits as the maximum resolution of the DAC, but then show under different conditions how much that is degraded in units of LSB (bits). In the example I have shown the error can be +- 1 dB or 2 dB total making the effective, error free resolution 18 bits. Worst case actually brings the resolution way down to 15 (-2.5 to + 2.5 bits). Don't run with these numbers as these are just the example from above. I am showing them just to demonstrate the common interpretation of linearity in the form of bits.

It is also true that we can also convert signal to noise ratio into bits. Here ENOB (effective number of bits) is a simple way to express something as is how it is used above. Here we are trying to determine how noisy the DAC is. In linearity we are trying to determine how distorted it is.

As to Schiit spec, this is all they say:

View attachment 10765

What don't know what frequency was used for this measurement. Or whether they used weighting which filters a lot of the noise making the number a lot higher. It is also possible this is measured with no signal, and then a full amplitude signal. A proper signal to ratio would be looking at the noise in presence of our full amplitude signal as the noise level can increase with signal. Without this kind of detail or a spectrum graph as I post, there is just no way of knowing what this number is.

SNR is such an important marketing number that manufacturers resort to every trick in the book to come up with the best number they can.

Summary
ENOB or effective number of bits can be applied to both distortion and noise measurements. My use of it with respect to linearity is absolutely consistent with how DACs are rated and measured. It is however not a common practice or term in audio reviews and hence lack of familiarity with it even from that Phd engineer.

Ultimately you can ignore my ENOB computation and just look at the graphs. The visual completely makes the same point when the DAC is compared to another showing much less error.

I believe the 20bits in the AD spec sheet is specified a minimum, not maximum. There are a couple of notes mentioned, is there any more useful info in them, and as you said there is no frequencies, filters etc listed.

 

[Sal1950]

Put another way, does the Yggdrasil have any advantages over the Behringer?

It will empty your bank account at a much higher rate.