Schiit Yggdrasil MIL discussion (and measurements eventually)

Discussion in 'Digital: DACs, USB converters, decrapifiers' started by purr1n, Feb 3, 2022.

  1. Cspirou

    Cspirou They call me Sparky

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    8,200
    Trophy Points:
    113
    Location:
    Northwest France
    Holy crap! that's a lot of amirbits!
     
  2. nishan99

    nishan99 Friend

    Pyrate
    Joined:
    Dec 1, 2019
    Likes Received:
    1,617
    Trophy Points:
    93
    Location:
    Layla
  3. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    upload_2022-2-5_13-7-25.png

    upload_2022-2-5_13-13-33.png

    x4 oversampling, not x8. This could be related to the sample and hold to eliminate glitch. At too high sampling rates, may not be possible to do a good sample and hold.
     
  4. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    upload_2022-2-5_13-39-50.png

    With a full scale signal, we get rising THD+N toward as go up in frequency past 8kHz. Could this be the cause of the "spitty" highs? In theory it shouldn't. The distortion is still too low and I'm betting much of this distortion is higher order past the range of human hearing. This concern is somewhat alleviated when we look at slightly less than full scale signals -5db and -10db. Here we see distortion in the highs become more even with the best of the band.

    Let's take a look at what happens on the analyzer for 8kHz, 14kHz, and 19kHz for a full scale signals

    8kHz
    upload_2022-2-5_13-45-6.png

    14kHz
    upload_2022-2-5_13-51-19.png

    19kHz
    upload_2022-2-5_13-51-49.png

    In theory, can't hear any of this. Too far up in frequency and in most instances too low in level.
     
  5. Jh4db536

    Jh4db536 Friend

    Pyrate
    Joined:
    May 25, 2016
    Likes Received:
    926
    Trophy Points:
    93
    Location:
    Los Angeles
    Home Page:
    Did they just drop in 4x single channel TI DAC11001A QFP48, disable the DSP, and voltage output straight into the output stage?

    Guessing that would result in a DAC that measures close to the chips datasheet specs and good enough as any respectable chip from 2019.
     
  6. roshambo123

    roshambo123 Friend

    Pyrate Contributor
    Joined:
    May 26, 2018
    Likes Received:
    2,763
    Trophy Points:
    93
    Location:
    Los Angeles
    In theory, but I recall you saying something with respect to EC amps that what cannot be heard can influence what can be heard in some cases.
     
  7. Cspirou

    Cspirou They call me Sparky

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    8,200
    Trophy Points:
    113
    Location:
    Northwest France
    Jason from Schiit mentioned this specifically for high frequency noise of an SMPS

     
  8. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    I'd assume do. Of course they've got the megacomboburrito filter, and then custom code to send data to the chip (it's not an audio). Datasheets says voltage out, so going to be not much after it.

    In theory yes. How the AP measures stuff is via FFTs (a fast compute algorithm for DCTs) which are used to create the nice spectrum analyzer visualizations. In order to calculate these presentations, we must take a certain amount of data points, 32k to 1M. By data points, we mean samples. In addition to this, to eliminate noise, we take another 3 to 10 or so averages of this.

    The limitation is spectrum analysis is that they don't tell us about transient performance. In fact, distortion is transient performance would likely be buried in all that averaging and and those samples. From the attack and decay burst envelope measurements that I've been doing (headphones for now), we know that initial attacks and decays can look substantial more distorted than after signal achieves a steady-state.

    So yes, in theory... Sometimes I wonder if little oddities, even if in theory we cannot hear them, may give us a slight hint of grosser errors in transient performance.

    The scientific approach is to only take things at face value. That is understand that what we are seeing is reflective only of steady state sine waves.

    --


    IN A NUTSHELL:

    The Schiit Yggdrasil More is Less performs spectacularly in AmirSINAD, or more precisely in THD+N when using a 1kHz steady state sine wave at full scale into a load of 200-ohms (balanced). If you believe this measurement to be the primary determinant of state-of-the-artness, performance, sound quality, then by all means, the Schiit Yggdrasil MiL must be be considered and would be highly recommended.

    From a subjective point of view, the Yggdrasil MiL is in my opinion, the worst of the Yggdrasils in that it lacks the emotion, lacks the resolution, and imparts an odd spittyness to the highs (which attempts to component match with laid-back amps and headphones could not resolved). It does have the most neutral tonal balance however. Many have asked me if the MiL should be considered if one wanted a neutral sounding R2R DAC (instead of LIM which is slightly dark or OG/A2 which is slightly V-shaped). Perhaps, but I feel the MiL is a poor value for its subjective technical performance. I'd say consider the Soekris DAC2541 (be aware of complaints that it lacks extension at the extremes or kick in the lows - I did not have any issue with it however) or even delta-sigma DACs such as the ECP Audio Walnut X.3. Alternatively, if one could opt for an Yggdrasil GS/A1 (not to be confused with the new OG/A1).
     
    Last edited: Feb 6, 2022
  9. roshambo123

    roshambo123 Friend

    Pyrate Contributor
    Joined:
    May 26, 2018
    Likes Received:
    2,763
    Trophy Points:
    93
    Location:
    Los Angeles
    Such thing as a CSD using a 1khz tone (or other tones) instead of a full sweep?

    Wasn't sure if you could see that far down in the mud with that kind of measurement
     
    Last edited: Feb 6, 2022
  10. Philimon

    Philimon Friend

    Pyrate Contributor Banned
    Joined:
    May 6, 2016
    Likes Received:
    2,092
    Trophy Points:
    113
    Location:
    Portland, OR
    Isnt that what a burst response is? Measurement of ‘growth to full amplitude then decay after signal stops’ using a single frequency.

    I had a question about burst response. 10 cycles means 10 bursts? Why not just 1 cycle for a longer period? Off-topic.
     
  11. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    10,961
    Trophy Points:
    113
    Location:
    Philippines, The
    Further off topic (sigh, I need to get out of that habit) but I do think there's value in the multiple burst approach since, at least with the orange trace version which I find easier to read and generally more intuitive than the far more aesthetic multi-colour lines, you can see some deviations in the amplitude and sometimes shape of individual blobs. Of course the analysis becomes more qualitative at that point since the measurement is relatively nascent and observable sonic aspects are still being tentatively correlated therewith, but more data to parse is generally more fun if the topic is interesting.

    e.g. RS2 behaviour at 4800Hz:
    https://www.superbestaudiofriends.o...e-hd800-sr-207-hd650.3688/page-11#post-296382
     
  12. roshambo123

    roshambo123 Friend

    Pyrate Contributor
    Joined:
    May 26, 2018
    Likes Received:
    2,763
    Trophy Points:
    93
    Location:
    Los Angeles
    I meant CSD mapping not just the the pulsed tone but the attack and decay of all the generated harmonics in the 20 khz bandwidth
     
  13. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Yes. 10 cycles for means 10 bursts at x set of frequencies

    1 cycle for longer = 1 cycle at lower frequencies.

    The trick with the bursts I've been doing is the transducers have gross errors. DACs and amps have errors several magnitudes below that of transducers for steady-state signals, and then magnitudes below that for transient signals. I'll have to work on better visualizations.
     
  14. Philimon

    Philimon Friend

    Pyrate Contributor Banned
    Joined:
    May 6, 2016
    Likes Received:
    2,092
    Trophy Points:
    113
    Location:
    Portland, OR
    I see. Overexcited frequencies (inducing extra harmonics etc) and ringing I imagine could correlate to certain characteristics like imaging or delineation and texture, rich or syrupy, and colorations like Grado electric guitair (its already hypothesized fast attack and slow decay / ringing at certain frequencies as partly responsible), etc. I cant see why you couldnt measure decay with just a single tone but graphed like a CSD instead of how Marv does the burst responses on a single plane. That would be a lot of data gathering. Marv would have to write a program to automate. Then figure out how to display results condensed into fewer graphs. Then find patterns.

    Im still trying to figure why there is not a CSG (cumulative spectral growth). Ive been googling but its either too technical, non-existent, and/or searching wrong terms.

    rtings.com uses a measurement for correlating imaging. Within their methodology is some links to research.:
    https://www.rtings.com/headphones/tests/sound-quality/imaging
     
  15. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Yup. It's basically a research assignment. I've laid a lot of groundwork, ideas, etc. Next step is for some grad student who has some time and funding to do the step next. (For example, I'm glad oratory on reddit nows includes an Optimum HiFi Target which is far better target IMO than Sean Olive's consumer target).

    @baldr and have discussed these ideas over the phone. We think that the errors are in the transients, and that we would look to a transient HD.equivalent of 50-70db (which isn't that high but also not that low either). The problem as I've mentioned in the past, we can't exactly do an FFT on a transient that consists of a handful of points.
     
  16. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    LOL @Grattle! Misinformation. I love it! :D
     
  17. Scott Kramer

    Scott Kramer Friend

    Pyrate
    Joined:
    May 3, 2016
    Likes Received:
    1,446
    Trophy Points:
    93
    Random quick (dumb?) thoughts: wonder if the megacomboburrito can be applied to 176.4/192 input (nowhere to go via upsampling and how the modiMB/BF1 behaves). Is LiM also 4x? If that's the case @Garns 64M tap filter or HQPlayer might be more effective which is interesting. Have you considered putting the case back on and letting it warm up like A1/A2? I've heard those cold, and they actually sound like ass!

    (looks like the TI has some sort of temperature drift fix, doubtful warmup will have an effect but what the heck.)
    Screen Shot 2022-02-07 at 5.30.35 AM.jpg
     
    Last edited: Feb 7, 2022
  18. Baten

    Baten Friend

    Pyrate
    Joined:
    Mar 18, 2018
    Likes Received:
    1,130
    Trophy Points:
    93
    Location:
    EU
    If my memory is worth anything, iirc LIM was 8X whereas MIL was 4X.
     
  19. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,778
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    LiM is 8x. Mil is 4x. This means using HQ player should completely override the built-in filter when feeding the MiL 176/192 signals. As for "warm-up", in my experience, it's more time left on, rather than actual temperature. Some folks in Schiit feel that the DACs sound better without the chassis. (I'm not sure I can tell the difference).
     
  20. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Pyrate MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    12,285
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    The chassis shields against one's ability to impact the quantum performance of the DAC through telekineses.
     

Share This Page