Schiit Yggdrasil Stereophile Review + Measurements

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

  1. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,549
    Dislikes Received:
    16
    Trophy Points:
    113
    Location:
    Irvine CA
    There are no billion (exaggeration) taps. There are thousands of taps. And the audio delay should definitively not be minimal. This per Jason:

    http://superbestaudiofriends.org/index.php?threads/schiit-multibit-bifrost.105/page-6#post-2450

    There are a lot of claims there about how awesome the filter is. But here is my take:
    1. Parks-McClellan optimizes time and frequency domain.
    2. Schiit's filter is an approximation to the ideal filter, closed-form solution or not. This because the ideal filter is not realizable.
    3. Optimization is not an "educated guess". That is a gross miss-representation.
    4. Thousands of taps will have audible delay.
    5. Parks-McClellan likely has lower error bounds than the Schiit's closed-form filter when approximating the ideal filter for a given filter order (which is kind of academic given the number of taps).
    There is a lot of discussion about this in that thread.

    I feel it is pointless to put down the Yggdrasil on the account of it's digital filter performance (other than delay - if that affects at all your application). It's indeed fairly good. It is also pointless to call Yggdrasil a flawless piece of perfection. I would encourage people to do their own research and think for themselves whenever possible.
     
    Last edited: Aug 1, 2017
  2. TonyNewman

    TonyNewman Validated by Tyll removing Utopia from WOF

    Joined:
    Jul 16, 2016
    Likes Received:
    257
    Dislikes Received:
    8
    Trophy Points:
    63
     
  3. AllanMarcus

    AllanMarcus Friend

    Friend
    Joined:
    Oct 23, 2015
    Likes Received:
    2,113
    Dislikes Received:
    21
    Trophy Points:
    113
    Location:
    Los Alamos, NM
    Home Page:
    I'm a little confused. Darren was praising the Yggdrasil. You make it sound like you are arguing with his praise. Are you saying the delay is significant? Didn't you suggest the delay could be a small as .1 ms?
     
  4. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,549
    Dislikes Received:
    16
    Trophy Points:
    113
    Location:
    Irvine CA
    You are not confused. The Yggdrasil has many things going for it. But minimal delay is not it. I'm definitively not praising that. And for some applications minimal delay may be a must. For some others probably not so much.

    As far as 0.1 ms, do you mean this? > http://superbestaudiofriends.org/index.php?threads/schiit-multibit-bifrost.105/page-6#post-2490

    That is for a minimum phase filter where the main tap is 10 taps in @ 96kHz sampling rate. Turns out the Yggdrasil uses a linear phase filter where the main tap is 9000 taps in (assuming 18000 taps). For a 96 kHz sample rate one gets 94 ms of delay.

    I think for lip synch things have to be around +/- 22 ms before corrections need to be applied.

    Something is gotta give. You want a linear phase, brickwall filter, with -120 or more dB of rejection, uber flat passband, and some random optimization that keeps the original samples the same and not much degradation on the interpolated ones? You need a large filter and you are going to pay with delay.
     
    Last edited: Aug 1, 2017
    Dino, AllanMarcus and landroni like this.
  5. frenchbat

    frenchbat BritishBat's arch enemy - Friend

    Friend
    Joined:
    Sep 26, 2015
    Likes Received:
    1,360
    Dislikes Received:
    3
    Trophy Points:
    93
    @ultrabike isn't the Yggdrasil and Gungnir Multibit's tap length 18000 as per Jason's post you have linked ?

    EDIT : Thanks for clarifying.
     
    ultrabike likes this.
  6. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,549
    Dislikes Received:
    16
    Trophy Points:
    113
    Location:
    Irvine CA
    Yup. For a 18000 symmetric linear phase filter, the main tap is 9000 taps in. Transients die out when the delay line is half full for a linear phase symmetric FIR. Delay is also half the size of such a type of filter.
     
    Dino and frenchbat like this.
  7. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,549
    Dislikes Received:
    16
    Trophy Points:
    113
    Location:
    Irvine CA
    Note I'm assuming a 96 kHz filter sampling rate. If the filter is run slower for certain selected rates, the delay will be worse (44.1 kHz would be about 2x worse). If the filter is run faster for certain selected rates, the delay will be more benign.

    It will be a large delay relative to your everyday DAC though. But just as there is no such thing as a free lunch, you also get some benefits for what you pay in this case: Linear phase, better rejection, your home-cooked optimization deal, flat pass band, and so on.
     
    Dino likes this.
  8. cskippy

    cskippy Creamy warmpoo

    Friend MZR
    Joined:
    Oct 14, 2015
    Likes Received:
    3,784
    Dislikes Received:
    1
    Trophy Points:
    113
    Location:
    Tempe, Arizona
    Sorry for the ignorance but @ultrabike I am officially confused. I use Gungnir Multibit, for all things audio and as long as it's not piped through JRiver, the delay to me is negligible, and I am very sensitive to A/V delay. Am I losing my touch or have I just gotten used to it? I do notice a slight delay using Eitr vs my Steinberg MR816x both coax to Gungnir Multibit. I also have a Creative Soundblaster Z that seems to give the least delay with all processing off.
     
    Darren G likes this.
  9. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,549
    Dislikes Received:
    16
    Trophy Points:
    113
    Location:
    Irvine CA
    What is the audio sampling rate being feed to the Gungnir Multibit?
     
  10. cskippy

    cskippy Creamy warmpoo

    Friend MZR
    Joined:
    Oct 14, 2015
    Likes Received:
    3,784
    Dislikes Received:
    1
    Trophy Points:
    113
    Location:
    Tempe, Arizona
    44.1/24
     
  11. dmckean44

    dmckean44 In a Sherwood S6040CP relationship

    Friend
    Joined:
    Nov 17, 2016
    Likes Received:
    1,230
    Dislikes Received:
    7
    Trophy Points:
    93
    Location:
    San Diego
    Most modern displays have 40-150ms of delay depending on the display mode and what video processing options are selected. With a CRT TV or monitor, you'd notice the delay though.
     
    Dino and ultrabike like this.
  12. cskippy

    cskippy Creamy warmpoo

    Friend MZR
    Joined:
    Oct 14, 2015
    Likes Received:
    3,784
    Dislikes Received:
    1
    Trophy Points:
    113
    Location:
    Tempe, Arizona
    I use mine with a 144Hz TN monitor and previously a CRT monitor which this particular TN monitor rivals with blur reduction enabled.

    I can't play on any TV due to the input lag but thanks for the reminder about most TVs.
     
  13. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,549
    Dislikes Received:
    16
    Trophy Points:
    113
    Location:
    Irvine CA
    I dunno. Maybe it's something like @dmckean44 mentions.

    I do know that the filter response has been measured:

    https://www.stereophile.com/content/schiit-audio-yggdrasil-da-processor-measurements

    However, JA ran out of space there. He usually displays a window of 5 ms and that was insufficient for the Yggdrasil (unlike pretty much every other random DAC). Meaning, indeed the delay will be unusually large (as in way larger than 0.1 ms, and maybe in the several tens of ms depending on the size of the post cursor output).

    Don't be fooled by the transients dying out. The delay will still be there, even if the coefficients are nano-insignificant as they move away from the main tap (as it should be).

    One could also measure the delta time when recording.
     
    Last edited: Aug 1, 2017
    landroni likes this.
  14. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,549
    Dislikes Received:
    16
    Trophy Points:
    113
    Location:
    Irvine CA
    So I don't have a Yggdrasil, but if you do:
    1. Connect a Yggdrasil and your other random DAC to the same signal source.
    2. Connect the right channel output of your Yggdrasil to your ADC capture card left channel.
    3. Connect the right channel of your random DAC to your ADC capture card right channel.
    4. Run silence followed by a pulse on the right channel.
    5. Compare delays on the recorded waveform.
    Maybe that would work.

    Play around with test signal sampling rate too.
     
    Mshenay likes this.
  15. cskippy

    cskippy Creamy warmpoo

    Friend MZR
    Joined:
    Oct 14, 2015
    Likes Received:
    3,784
    Dislikes Received:
    1
    Trophy Points:
    113
    Location:
    Tempe, Arizona
    I have a Gungnir Multibit but the results should be similar. That's definitely something I'll try as I've been curious about it but never bothered quantifying it.
     
  16. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,549
    Dislikes Received:
    16
    Trophy Points:
    113
    Location:
    Irvine CA
    Reading on ASIO, it seems depending on buffer, one may be looking at a latency of a few ms (2 to 6 ms depending on sampling rate).

    Other interfaces may inject even more latency. Still, I would be surprised if the latency on the Yggdrasil comes up to 6 or so ms. This given the sharp cut off and stop band rejection while maintaining linear phase. I would expect quite a bit more than that for that performance.
     
  17. cskippy

    cskippy Creamy warmpoo

    Friend MZR
    Joined:
    Oct 14, 2015
    Likes Received:
    3,784
    Dislikes Received:
    1
    Trophy Points:
    113
    Location:
    Tempe, Arizona
    Yup, I always run as low a buffer as possible, 32 samples at 44.1kHz when gaming. When working in the DAW I use 128 to avoid stuttering with a ton of effects and VSTis.
     
  18. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,549
    Dislikes Received:
    16
    Trophy Points:
    113
    Location:
    Irvine CA
    Darren G and cskippy like this.

Share This Page