Schiit (MultiBit) Bifrost

Discussion in 'Digital: DACs, USB converters, decrapifiers' started by FlySweep, Oct 2, 2015.

  1. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    Alright, after doing full bandwidth signal into Parks-McClellan the input to output error (when the signals are lined up) rises to about -87 dB.

    Here are the error power (difference between input and output samples relative to the input samples) results for different filters:

    500-tap Parks-McClellan = -87 dB (cut off @ 0.968*0.5)
    500-tap Least-Squares = -134 dB (cut off @ 0.968*0.5)
    9000-tap Classic window method (Hamming) = -105 dB (cut-off design to 0.5)
    9000-tap Classic window method (Kaisser, Beta=4) = -104 dB (cut-off design to 0.5)
    9000-tap Classic window method (Cheby, 120 dB atten stop band) = -172 dB (cut-off design to 0.5)
    9000-tap Sampling method (Matlab's fir2 default) = -69 dB (cut off @ 0.968*0.5)

    Note Octave's fir2 and Matlab's fir2 are not the same. Octave's fir2 seems to do pretty good in the difference between input and output, but I'm not confident that filter design function is working properly. Octave is free and it's not the first time I get weird results with it.

    Anyhow, best results with the above filters was the Classic Chebyshev windowing.

    If what Mike Moffat is saying is that the Burrito DAC filter preserves samples from input to output, then the closest thing I can find in Matlab that does that so far is this filter:

    18000-tap Classic window method (Cheby, 180 dB atten stop band) = -233 dB (cut-off design to 0.5)
    h = fir1(18000, 0.5, chebwin(18001,180));

    ... or Octave's questionable fir2.

    One thing I did notice is that the impulse responses of the classic Chebyshev windowed response and the questinable Octave fir2 looked closest to an ideal sinc. So I tried using a straight up truncated 18000-tap sinc to filter:

    h = sinc(-4500:0.5:4500)/2;
    Truncated 18000-tap Sinc filter = -286 dB (cutt off @ 0.5)

    So that may or may not be a hint of what is needed to get things out untouched, and how that relates to the ideal sinc filter... Usually something is gonna give. For example, the classic Cheby approach offers better stop band rejection than the turncated sinc deal.

    As far as time, frequency, phase, or whatever optimization, in the end different approaches will yield different filters. If Schiit is able to do a decent job at stop band rejection, nice and flat pass band, well behaved and short transition band, have linear phase, and preserve the samples from input to output, that all spells out AWESOME to me.
     
    Last edited: Oct 7, 2015
  2. lm4der

    lm4der A very good sport - Friend

    Pyrate
    Joined:
    Sep 26, 2015
    Likes Received:
    1,466
    Trophy Points:
    113
    Location:
    Seattle, WA
    Well, I have no idea how those guys do it, but the Bifrost Multibit is the real deal. At the risk of repeating myself and others, it gives so much more space to the HD650s; this was really my only nitpick about theses cans - the sound stage felt a bit close to my ears - but it is much better with Bifrost Multibit. Also, just generally more alive and refined/resolving. This is a huge upgrade over Modi 2.
     
  3. Tuco1965

    Tuco1965 Suffring from early onset Alzheimer's - Friend

    Pyrate
    Joined:
    Oct 4, 2015
    Likes Received:
    151
    Trophy Points:
    43
    Location:
    Just East Of West
    I'm going nuts watching the tracking for my Bifrost's return to Schiit. As much as I like my Modi, it's just not cutting it on my speaker system. It's okay with my HD600s, but not so much with my HE400is.
     
  4. lm4der

    lm4der A very good sport - Friend

    Pyrate
    Joined:
    Sep 26, 2015
    Likes Received:
    1,466
    Trophy Points:
    113
    Location:
    Seattle, WA
    Heh, yeah, nothing worse than waiting for new gear.
     
  5. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,945
    Trophy Points:
    113
    Location:
    On planet
    That is the Magic Magneto Music Multiplier(TM). This is where the dimension and depth is added to the music. All kidding aside, Schiit has created what I consider some of the most important developments in DACs and made these developments readily accessible by reasonable pricing.
     
  6. Armaegis

    Armaegis Friend

    Pyrate BWC
    Joined:
    Sep 27, 2015
    Likes Received:
    7,583
    Trophy Points:
    113
    Location:
    Winnipeg
    Guys guys... clearly that straw is for drinking the koolaid.
     
  7. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,945
    Trophy Points:
    113
    Location:
    On planet
    Well I voted pro-Schiit with my wallet. I already own the Yggdrasil and GuMB, thought I might as well have a BiMB in my lab too. Still want EricR to bring his upgraded unit for comparison measurements.
     
  8. schiit

    schiit SchiitHead

    Pyrate
    Joined:
    Sep 28, 2015
    Likes Received:
    9,974
    Trophy Points:
    93
    Location:
    Texas and California
    Home Page:
    LOL, that straw is a center tap--the transformer has quite a few power supplies to feed, and we needed a flying wire to do it. Not sexy, but there you go.
     
  9. Pyruvate

    Pyruvate Friend

    Pyrate Contributor
    Joined:
    Sep 29, 2015
    Likes Received:
    1,425
    Trophy Points:
    93
    Location:
    Here
    Jason, I never noticed this wire on the old bifrosts. So for those of us sending in our bifrosts for a multibit upgrade, will you guys be adding the wire too?
     
  10. Andre Y

    Andre Y Friend

    Pyrate
    Joined:
    Oct 4, 2015
    Likes Received:
    220
    Trophy Points:
    43
    Location:
    Southern California
    Looking forward to seeing your measurements!
     
  11. Luckbad

    Luckbad Traded in a unicorn for a Corolla

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    3,409
    Trophy Points:
    113
    Location:
    Holly Springs, NC
    Is it worth me selling an IEM I really like and a DAC that I'm using a stopgap to get the Bifrost Multibit? I'm seriously considering it. Such praise, and I could return it if I don't like it as much as I assume I will.
     
  12. elnero

    elnero New

    Joined:
    Oct 1, 2015
    Likes Received:
    6
    Trophy Points:
    3
    Location:
    Prince Edward Island, Canada
    First impressions of the Bifrost Multibit / Asgard 2 combo, they completely embarrass the Concero HP. Switching between the two listening to The Cowboy Junkies "Live at the Ark" is like listening to two completely different recordings. I'm a little dumbfounded to be honest.
     
  13. Luckbad

    Luckbad Traded in a unicorn for a Corolla

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    3,409
    Trophy Points:
    113
    Location:
    Holly Springs, NC
    Cool. That is a combo I'm actually quite interested in for work because it's pretty compact.
     
  14. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    BTW guys, before I stop technically beating the Mega Burrito Combo filter to dead I would like to say the following:

    All these filters (Parks-McClellan, Least-Squares, Window this, Window that...) are approximations to the ideal filter, including Mega Burrito Combo. All of them are trade-offs (pass-band, stop-band, transition-band, phase, numerical issues, ...).

    Furthermore, the constraint that an interpolation filter for up-sampling matches it's outputs to it's inputs may actually slightly degrade performance in it's stopband (see page 160 and 164 of "A Digital Signal Processing Approach to Interpolation" by Schafer and Rabiner). The idea it's obviously not new either.

    Note also that performance is frequency dependent: Samples further away from the cut-off of the filter (and depending on possible passband ripple) will likely be better off than the ones closest to the cutoff of the filter. This was indirectly shown (to some extent) in the Parks-McClellan example if one compares performance between 12.25 kHz BW input signals vs. 22.5 kHz BW input signals in previous posts in this thread.

    So what am I getting at?

    1) All filters are approximations AFAIK (including the Burrito Combo - interpolated values may not be optimum in this, that or whatever sense).
    2) All filters have trade-offs. You lose some, you win some.
    3) Parks and McClellan designed optimal filters (in some sense), and should not be IMO dismissed as POS.
    4) Burrito Combo is IMO not a bad filter, as I said before a few times.
    5) Some things are more expensive than others not necessarily because they are better across the board, but because they are unique.
     
    Last edited: Oct 7, 2015
  15. Sylafari

    Sylafari Acquaintance

    Joined:
    Oct 6, 2015
    Likes Received:
    43
    Trophy Points:
    18
    I'll be getting mines tomorrow. Just wondering I can connect with USB or Optical, which would be preferred? What's funny is I'm asking even though my major is Electrical Engineering. :)
     
  16. The Alchemist

    The Alchemist MOT: Schiit - Here to help!

    Pyrate
    Joined:
    Sep 24, 2015
    Likes Received:
    1,189
    Trophy Points:
    113
    Location:
    The SchiitR
    Home Page:
    So how much better is the multibit upgrade over the Uber? Obviously there is going to be more clarity. I guess my question is, what are the sample rates? Will they be the same as the Uber as in 32/192kHz or has it been improved to say 32/384KHz? If it is in this thread, I am sorry I missed it, I read all the post and didn't see anything like that. The specifications on the multi-bit upgrade really aren't explained on Schiit's website. It states:

    "This upgrade changes any existing Bifrost or Bifrost Uber into Bifrost Multibit, a radically different DAC using the medical/military grade AD5547 D/A converter and Schiit’s proprietary closed-form, time- and frequency-domain optimized digital filter. "

    Which is great info, and like I stated, no doubt a huge upgrade to the uber, just wondering if there were any specifications for it or if they will be posted soon. I know it is a new upgrade, but I guess I would like to know I little more about the specs on it.
     
  17. The Alchemist

    The Alchemist MOT: Schiit - Here to help!

    Pyrate
    Joined:
    Sep 24, 2015
    Likes Received:
    1,189
    Trophy Points:
    113
    Location:
    The SchiitR
    Home Page:
    P.S. Has anyone come up with a name for the new Bifrost Multi-bit like they did with the Bifrost Uber (Uberfrost)? Will it be something like Multifrost or something to that degree?
     
  18. Armaegis

    Armaegis Friend

    Pyrate BWC
    Joined:
    Sep 27, 2015
    Likes Received:
    7,583
    Trophy Points:
    113
    Location:
    Winnipeg
  19. Luckbad

    Luckbad Traded in a unicorn for a Corolla

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    3,409
    Trophy Points:
    113
    Location:
    Holly Springs, NC
    Nobody seems to agree on what to call it...

    Bifrost Multibit, Multibifrost, Multibitfrost, Frostybit, , Mifrost, Bifrost Multibit, MBifrost, Emby
     
  20. Sqveak

    Sqveak Friend

    Pyrate
    Joined:
    Sep 28, 2015
    Likes Received:
    934
    Trophy Points:
    93
    Location:
    NSW, Aus, Terra, Sol.
    Multifrost?

    Newcomers might be a bit perplexed by Bifrost Multibit.
     

Share This Page