Schiit Yggdrasil Stereophile Review + Measurements

Discussion in 'Digital: DACs, USB converters, decrapifiers' started by purr1n, Jan 20, 2017.

  1. OJneg

    OJneg The Most Insufferable

    Pyrate BWC
    Joined:
    Sep 25, 2015
    Likes Received:
    3,923
    Trophy Points:
    113
    Location:
    Grand Rapids, MI
    Wouldn't show up in a static test, as has been said. Does seem to be audible.

    precision vs accuracy.....now one wonders how this precision might vary away from accurate. Might explain why DS chips all tend to have their particular sounds, eh?
     
  2. schiit

    schiit SchiitHead

    Pyrate
    Joined:
    Sep 28, 2015
    Likes Received:
    9,937
    Trophy Points:
    93
    Location:
    Texas and California
    Home Page:
    Whether you agree with Mike or not, here's what really matters: we're doing multibit with a unique closed-form filter because we believe in it.

    Not because we need another delta-sigma DAC at a different price point.
    Not because we need another delta-sigma DAC with different features.
    Not because we need another delta-sigma DAC based on the same manufacturer's reference design.
    Not because we need another delta-sigma DAC with heroic power supplies and a milled chassis for a 5-figure price tag.

    We're doing it because we believe in the technology, we think it's the right way to do things, and we think it sounds better. And you know what, that's perfectly fine. Just as it's perfectly fine for Chord to believe in their million-taps FPGA tech and PS Audio is to believe in their 10X DSD tech, and MSB to believe in their discrete ladders, and DCS to believe in their ring DAC topologies. The important thing is that we are all doing something different, and something we all believe in. And we are all 100% insane in a world that has standardized on compressed streaming to reference-design $0.35 "24-bit" phone DACs as its primary music reproduction system.

    Regardless, this is what drives progress—different approaches duking it out in the market. It's up to you to decide which approach or approaches you prefer, for your particular system, and your particular taste.

    If you like our flavor of crazy, cool. if you don't, that's perfectly fine, too.
     
  3. DigMe

    DigMe Friend

    Pyrate
    Joined:
    Aug 25, 2016
    Likes Received:
    8,802
    Trophy Points:
    113
    Location:
    Tejas
    Feed me.
     
  4. anetode

    anetode Friend

    Staff Member Pyrate
    Joined:
    Sep 25, 2015
    Likes Received:
    479
    Trophy Points:
    93
    Location:
    How did I get here
    Home Page:
    In practice those discarded LSBs would be buried in the noise floor anyway. Even without dither there will be enough residual noise at the recording/ADC stage to whiten the quantizing noise (mask its relation to the signal).

    As for a 1-bit DAC x 8, then yes, you'll only get that 1-bit resolution, albeit a cleaner rendering of it. A cleaner rendering of the bits that count counts more than a dirty rendering of the bits that don't :p

    I doubt it, though Stereophile certainly catalogs a few DS DAC implementations with inferior monotonicity. You know my view though: transducers+room/ears+brain are ultimately the limiting factors. JA's complaints about how the noisefloor manifests in a different pattern of signals at well below -100db amount to bitching about the rearrangement of the deck chairs on the Titanic.
     
  5. OJneg

    OJneg The Most Insufferable

    Pyrate BWC
    Joined:
    Sep 25, 2015
    Likes Received:
    3,923
    Trophy Points:
    113
    Location:
    Grand Rapids, MI
    False meme. Plz put it to rest.
     
  6. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    I've not yet designed a bunch of audio DAC boards with a dozen commercial-off-the-shelf DS DAC chips. But I have been involved in the design of a few DS ASIC chips, though not for audio.

    As far as perception and the physical-universe/electrical-energy, I think the discussion about the DAC "missing codes" seems to belong to the electrical energy or digital domain.

    If we are going to talk about the physical-universe, vibrations in the atmosphere, and our reality perception, I think both currently available DS and multi-bit DACs are pretty good, just perhaps different. And that is my real perception to which I have a right as well.

    I think missing codes means the DAC will skip levels and will not be monotonic, which would produce jumps in the reproduced waveform, static or dynamic. Unless I'm missing something, I think this should result in noise and/or non-linear distortion depending on the statistics of the missing codes problem.

    I have no problem with any of the perceived performance discussions. Specially from the many folks here that have ample experience with all manner of DACs and equipment. But if we are going to talk about "missing codes" and how optimal the super-burrito-combo is, then I think we are moving away from the perceived performance discussions, and into that "electrical energy" domain which I think characterization can help in shedding some light.
     
    Last edited: Jan 22, 2017
  7. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    I think Stereophile indeed had a discussion about monotonicity and DS here:

    http://www.stereophile.com/content/pdm-pwm-delta-sigma-1-bit-dacs#1QG6LrBB7QrsiEoG.97

    According to the article, such errors would manifest themselves in distortion.

    I found their "zero-crossing distortion" or "MSB error" discussion very interesting and relevant, in light of JA complains about the Yggdrasil. Those blips which we saw in the tone characterization plots do not suggest the Yggdrasil is perfectly monotonic or accurate at the zero crossing. It could be crossover distortion, but the discussion in the article suggest another mechanism related to how the DAC handles low level digital signals.

    It is possible that JA attributes the Yggdrasil blips to the "MSB error" issue discussed in the article above, and therefore the DAC.

    There is also discussion about temperature effects on monotonicity.

    Like you said though, well below -100 dB issues IMO amount to TItanic deck chair rearrangement issues.
     
  8. OJneg

    OJneg The Most Insufferable

    Pyrate BWC
    Joined:
    Sep 25, 2015
    Likes Received:
    3,923
    Trophy Points:
    113
    Location:
    Grand Rapids, MI
    Well if you imagine a static 1k test tone (which is the norm for measuring THD+N performance) the DS converter will be able to converge on that signal value via its feedback mechanism. For any given static signal level actually (within its nominal output range). So there would be no jumps in a static waveform loopback test, as far as I can see.

    Now if the "missing codes" theory of Schiit's is true, we would in fact see errors in linearity during musical playback or any sort of dynamic signal activity. But no one is measuring this here so you shouldn't wonder that the measurable evidence is not seen. I'm not sure how we measure such dynamic activity reliably.

    I'm also not sure how pertinent the burrito filter is to the discussion. The encoding/decoding inherent to our AD DA conversions seems more basic. I imagine you could reprogram the Yggdrasil filter chips with an alternate filter but still see the same linearity performance.
     
  9. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    Most of the data sheets indeed do 1k tests. But they also do some basic two tone IMD tests. Other folks have done multi-tone tests. I've yet to see the "missing codes" mechanism on modern DS DACs and their significance. As you said, to me it seem like a theory. Limit cycles and other issues I think have been greatly mitigated in modern architectures.

    I do see the zero crossing "issue" in the Yggdrasil though. And that is not monotonic behavior.

    Sorry about the confusion about the burrito filter. It seems like it is directed to you, but it's more of a general statement more directed towards Mike given his previous statement:

    To me his "missing codes" and previous "burrito combo" filter discussions are opinions about things that never existed in the physical universe but only as electrical energy (or even before that). And even then, so far just in theory.
     
  10. OJneg

    OJneg The Most Insufferable

    Pyrate BWC
    Joined:
    Sep 25, 2015
    Likes Received:
    3,923
    Trophy Points:
    113
    Location:
    Grand Rapids, MI
    Would be most interesting if there was something like this rigged up in the Schiit-cave. Would seem to put the skeptics to rest.

    Could run any data into it, preferably Daft Punk - Get Lucky :D

    fRcXBue.jpg
     
  11. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    EDIT: I see what you did there. That would be awesome.

    BTW, different phase and time alignment would need to be applied. The delay between the two DACs might not be the same. From what I remember, the Schiit DACs tend to have a bit of delay given their FIR size. But that might work!

    Another thing. More than one type of song should be used. Some with lots of dynamics and some with not. I don't think the products will behave the same way with different material.

    EDIT2: Also, whatever the test shows, people are going to like what they like. If anything, it might show if the "missing codes" issue is really an issue. But I dunno how much folks really care in the end. Like @Marvey says, this is sort of a subjective hobby.
     
    Last edited: Jan 22, 2017
  12. OJneg

    OJneg The Most Insufferable

    Pyrate BWC
    Joined:
    Sep 25, 2015
    Likes Received:
    3,923
    Trophy Points:
    113
    Location:
    Grand Rapids, MI
    A two-tone IMD test is still a static test.

    The notion of having missing codes with the DS DAC (again if true) should not seem insignificant for us listeners. We are after all, processing that transient information and we should prefer to have more precision and accuracy if fidelity is the goal.

    On second thought, I'm not quite sure why a DS DAC would be more precise in decoding the digital signal, although I see the lack of resistor precision inherent to the resistor ladder DACs (of all types) as an obvious impediment for the precision of the analog signal that that class of DACs is attempting to render. @anetode
     
  13. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,594
    Trophy Points:
    113
    Location:
    near Munich, Germany
    I've seen tests where they compare the original file with the one that went through a DAC/ADC chain. I'd be worried about ADC precision in that case, though. Also if you record the same song twice the result won't be the same at all. Maybe there's less random variance with multibit DACs/ADCs. There's always a difference, I don't think you can just compare the original file against one that went through the DAC. I think the difference will be akin to a 128kbit/s mp3, or maybe slightly better.

    EDIT: Essentially in such a test one would look at the noise added by the DAC/ADC chain and how it correlates to the signal. I think with good converters you can get numbers that look much better than 128kbit/s mp3 (say -80db), but the sound will still suck.
     
    Last edited: Jan 22, 2017
  14. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    I think most DS DACs these days use multi-bit and are optimized to minimize resistor precision and other issues. I also think delta sigmas use a clock, register the steady signal, and avoiding glitches at the output. But I think it depends on the specifics of the architecture.
     
  15. msommers

    msommers High on Epipens

    Pyrate
    Joined:
    Nov 6, 2016
    Likes Received:
    2,750
    Trophy Points:
    113
    Location:
    Calgary, Alberta
    Home Page:
    I have nothing really to add but I am appreciative. This whole discussion has brought forth much more information than I've ever known about DACs.
     
  16. Psalmanazar

    Psalmanazar Most improved member; A+

    Pyrate Slaytanic Cliff Clavin
    Joined:
    Sep 27, 2015
    Likes Received:
    5,345
    Trophy Points:
    113
    Analog-> ADC -> DAC compared to an actually neutral (to what is on the tape or LP, not what your brain thinks sounds more neutral by modifying the sound of the recording) analog playback chain would be a good test along with a loop back from the DACs into the ADC to compare the results. This final bit would show if we as humans only perceive the errors as linear distortion or if they are in fact other errors manifesting themselves as perceived linear distortions.

    The more neutral DS DACs tend to have more correct "volume of bass" compared to even the best R2R ones, even the Yggdrasil. What DS misses out on is dynamics and treble timbre for most implementations and chip sets. Not that the better DS implementations can't approach the better R2R in more realistic treble timbre or that all DS DACs chipsets have a neutral tonality down below either: Sabres and Wolfsons tend to be warm, the dynamic and "off" timbre AD1955 has that off-timbre frequently hidden by added warmth, and then you have implementations even at the pro level that straight up just boost the bass to get bassheads to buy them even though their bassy recordings would turn out better if they made them instead with more neutral gear.
     
  17. baldr

    baldr Schiit-sterer

    Pyrate
    Joined:
    Oct 9, 2015
    Likes Received:
    2,758
    Trophy Points:
    93
    I may have muddied the discussion.... with my comments about music which existed in the physical universe. I was trying to state my self-perceived mission as a person who engineers/builds hardware designed to reproduce musical information. The key is how one defines reproduce. Music which once existed in the physical universe, using the atmosphere as a medium, is very different than technical or electronic music which was generated on a music computing device.

    It is for the former type where reproduction ideally conveys not only the physical energy characteristics of the singer, the instrumental player (be that instrument acoustical or electric), and the ensemble as a whole, but the emotion as well. Those who stand by passively who have never become ecstatic or misty eyed at a concert will not understand.

    The latter type, music which was generated purely by technological means (preset/arrange the melody, the harmonies, the tempo), to me lacks any emotion whatsoever. I view this as arbitrary in a musical sense. There are no human imperfections whatsoever. It is fine, I suppose, if you want to go dance until you drop or mark time or screw by it in the background. It's like a blow-up doll in the context of a real woman. The emotion (good and bad) is gone.

    That is why I believe in multi-bit. It is inherently better at conveying emotion, exceeded only by analog. With delta sigma, there is so little feeling that all remnants are the technical imperfections of the device nagging at you.

    I believe in science, but if one prostates himself at the altar of scientism he will deny fundmental imperatives.

    So this is what I believe and why I do what I do. It's why I think delta sigma sounds like ass in the context of multibit. Because it is a great starting point for many to follow our path, I will continue to make them at lower price points.

    If you disagree, that is fine. This is neither politics nor religion. There is a place for all.
     
    Last edited: Jan 22, 2017
  18. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    Well, I can say I recently went to a live Harry Potter live concert with the kids at the Segerstrom Center in Costa Mesa. I felt blood rushing to my head during the performance more than once. And I saw my son's reaction to the performance as well and I went misty eyed a few times there. I vicariously live through my kids. I also love and get moved by random performances. Sometimes unexpectedly so.

    I get what you are saying. However, how I personally connect to the music is not always straight forward. It is possible that certain equipment will allow me or other folks to connect to music better than others. But I don't think multi-bit or delta-sigma is a rule of thumb answer. Some may feel different of course.

    The day I became most emotional about music was actually in my car (through my POS stereo) listening to an old song I performed with a bunch of other kids. It was proly 8 or 9 years ago. I was 5 or 6 when I sang this song with my elementary friends. And the song is available on-line here:



    As you can see, it's not an audiophile song. Not even close. The recording is, well, sub-optimal. And the equipment used when I heard it was not even mediocre-grade. It was a miserable piece of fail. I don't know if it was multi-bit, but my best guess is that it was the cheapest thing Ford could find.
     
    Last edited: Jan 22, 2017
  19. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    Anyhow. I just want you (Mike) to know, that I'm appreciative of your efforts. And love your work and products. Whether I believe or agree with everything or somethings you say or not, is besides the point.

    Hope to see more great Schiit products. They are good products. Well made and well supported.
     
  20. iDesign

    iDesign Almost "Made"

    Joined:
    Oct 26, 2015
    Likes Received:
    230
    Trophy Points:
    63
    This really was a great post. I think what should not be missed here is that Jason Stoddard and Mike Moffat are a small, passionate team working together to build to products that deliver outstanding value and performance for enthusiasts like us, not Stereophile. Measurements don't always determine how a DAC will sound and soul is something that can't be measured. What's more is that they both take time away from their business and families to contribute to communities like SBAF, freely divulging product information and listening to this community's feedback. Their goal to do something different, to pursue the technologies they believe in and compete against other approaches from companies like Chord, PS Audio, and MSB are all the reasons I support Schiit.
     

Share This Page