A tale of 3 digital transports (shootout with Unison USB and AES)

Discussion in 'Digital: DACs, USB converters, decrapifiers' started by rlow, Jan 11, 2020.

  1. rlow

    rlow A happy woofer

    Friend
    Joined:
    Jul 18, 2017
    Likes Received:
    3,384
    Dislikes Received:
    3
    Trophy Points:
    113
    Location:
    Canada
    TL;DR: The digital transport makes a difference in the sound of your system, and Unison USB is not the “great equalizer” making even your shitty PC sound as good as a multi-kilobuck purpose-built digital audio transport devices (it may get close enough for many people however). I’m not advocating spending multi-kilobucks on a transport, but don’t expect your 3 year old Chromebook to do as well as a device that’s specifically designed to feed audio to your DAC (Unison USB or otherwise).

    Intro:

    Not all digital transports are created equal. Bits aren’t bits. We’ve seen this through the evolution of CD players as well as USB as an audio interface. The quality of the transport matters, even with modern computer-based digital audio solutions that claim to have vanishingly low jitter.

    Many are eagerly anticipating the launch of Unison USB for Schiit’s upgradable DACs, wondering “does it really beat SPDIF?”, “is it better than AES”? I’ve had a Unison USB Beta board in my Yggdrasil A2 for over a year now and have been extremely impressed with it from the start. It has handily beaten any other USB implementation I’ve heard (granted I’ve never heard the TOTL stuff like the Berkeley Alpha USB or the dCS Network Bridge).

    @Hands has said recently however that he prefers SPDIF (his DigiOne Sig) over Unison on the new Bifrost 2. How can this be? Didn’t Mike say Unison was better than SPDIF or AES? Does @Hands have better ears than Mike? He must be smoking something right? Or maybe there’s just something more at play here than just SPDIF vs USB.

    I have owned a number of USB and SPDIF transports including the Cambridge Audio Stream Magic, microRendu, UltraRendu, Sotm SMS-200, SMS-200Ultra, and recently the SMS-200Ultra Neo and the Pi 2 Designs Pi2AES (I’ve also got an Arcam CD73 player I use as a CD transport). And of course I’ve also used my laptop (Dell XPS 13) and hooked it direct through USB to my DACs at various times.

    And what I’ve come to find so far is this:

    It’s not just about how good the input is on the DAC, it’s also what’s feeding it!

    Unison and Adapticlock and all this other DAC input technology is intended to raise the level of (and potentially equalize the level of) the digital transport. But it only goes so far - the transport can still have an appreciable effect on the sound. We can see this even with the very mature SPDIF technology and the DigiOne Signature - many agree the DO Sig was an appreciable step up above other low cost SPDIF transports even through Yggdrasil’s Adapticlock input.

    Now for the “bits are bits” crowd, of course they will find this mostly bullshit - once everything is “up to spec” you will have perfection in a digital audio world. If you want to live in that world, fine. Try another thread.

    And many people have tried different transports and say they can’t hear the differences between them. Ok, but there’s a few things to remember:
    1. The transport has an even lesser effect on the overall sound than the DAC. DACs can actually do a lot to make up for a shitty transport, back to the “as long as it’s in spec” part. And I think this is what Unison shows us - it can do a pretty damn good job even with a noisy/jittery source like a Windows laptop USB.
    2. Also because of #1, you have to have a system that is resolving enough to hear the differences clearly. I’m not saying you need TOTL stuff, but gear that masks detail and doesn’t do pretty well in most areas won’t likely display much of a difference between transports.
    3. And maybe this is the most important point - the differences between digital transports is not in the tonality, but in the technicalities (mainly). People who can’t hear the differences might be looking for frequency response differences, because this is what you mainly listen for with all the other gear in your chain, and where you tend to hear the most immediate diffs - is it brighter/darker, is it warmer/colder, does it have more bass or midbass, etc? You can sometimes find a bit of this between transports, but I find the majority of the diffs between transports include:
    • cleanliness/sibilance control/digititus
    • noise floor/blackground
    • instrument/vocals separation
    • soundstage width/height
    • depth and layering
    • transients/macrodynamics
    • microdetail, plankton, atmosphere
    Ok so enough with the intro, let’s get to the meat.

    The shootout: Intel NUC (USB) vs. Pi2AES (AES) vs. Sotm SMS-200ultra Neo (USB) on a Yggdrasil A2 using Unison USB beta board input as well as the AES input for the Pi2AES.

    Rest of the chain is: Schiit Freya, 2 Schiit Vidar Monoblocks, and Graham Audio LS6 speakers.

    Also note that I made a power supply change on the Pi2AES part way through to see if I could discern a difference.

    I swapped the transports back and forth on each track at least a few times. Most tracks were rock/alternative, jazz and folk, with stuff that is at least decently recorded. Roon/RoonBridge was used with all 3 transports to keep everything as equal as possible. The Pi2AES used the Ropieee OS. Nuc had the Roon ROCK OS.

    Pictures or it didn’t happen:
    1C4C59E9-C7FD-49D3-B8CD-C315E2204C10.jpeg

    Note: This is mostly stream of consciousness thoughts and immediate impressions - I don’t have the time to pretty things up.

    Also note that the degree of differences I’m mentioning are nothing compared to the differences between transducers or even DACs from different manufacturers.

    Pi2AES (AES) with stock power supply vs. Intel Nuc (Unison USB)

    Nuc USB out (Unison):
    • Good/Pleasant but a bit boring sounding - lacking dynamics
    • Fairly flat dimensionally, lacking depth
    • Stage mostly caught between the speakers
    • Decent seperation, but a touch of greyness and a bit blobby - instruments can blend together especially in busy passages
    • A bit mushy in bass overall, not the best with slam and impact
    • cymbals and high hats a bit tizzy sounding, not clean.
    • Pretty smooth and not much in the way of digititis or fatigue, which is an improvement almost all other USB implementations I’ve heard. But a touch soft and lacking scale, separation and dimensionality
    Pi2AES (AES) with stock power supply:
    • Better blackground
    • Immediately more 3D soundstage- better depth and layering, a bit wider stage as well - less sound trapped between the speakers
    • More dynamic, less soft, less blurred
    • Tighter bass with more slam
    • Better transients and macrodynamics overall with all instruments
    • Cymbals cleaner, more shimmery, better decay
    • More extended sounding esp in the highs. The NUC almost seems dark in comparison.
    • Better instrument seperation and vocal intelligibility
    • More open and transparent sounding overall - much better atmosphere and more plankton than the NUC
    • Maybe slightly more sibilance, but this could be due to better detail and transients
    After this initial comparison, I decided to try putting a better power supply on the Pi2AES, so I tried my Sbooster 10.5V linear power supply (which I usually use with the Sotm) and compared it to the stock switching power supply.

    Pi2AES (AES) with Sbooster power supply vs. stock supply:
    • Smoother. A bit blacker background. Cleaner sounding.
    • Even more depth (if the recoding has it)
    • A bit less edge, less fatiguing, better sibilance control
    • A bit less dynamic however and slower transients than the stock switching supply
    • Cymbals a bit more clear with better decay
    • Slightly wider, taller and more open soundstage
    • Slightly better seperation of instruments
    • Maybe a touch warmer/darker sounding?
    • Overall I prefer the sound with the Sbooster, but it may lose a slight amount of dynamics and skew a touch warmer sounding. The differences are pretty slight though and the stock supply does a very good job.
    Sotm SMS-200ultra Neo (USB) vs Pi2AES (AES) - both using the Sbooster power supply:
    • WOW depth forever and huge soundstage. No boundaries in my room anymore.
    • A bit tighter sounding, a bit better transients
    • Yet even more separation of instruments and better layering between them.
    • Even more open and more transparent sounding
    • More atmosphere, more plankton. I’m at the recording session!
    • Absolutely as clean as a whistle - no edges or etchiness of any kind. ZERO digititus.
    • Less warm, more neutral sounding than the Pi2AES with the Sbooster, and more dynamic. More like the Pi2AES with the stock switching supply, but way cleaner and more open sounding.
    After having been listening to the Pi2AES for a number of weeks exclusively, I was actually shocked going back to the SOtM with Sbooster - it sounds so non-digital and has so much detail and scale. Just beautiful. It should be noted however that without a direct comparison like this, I’ve actually really been satisfied with the Pi2AES.

    As a last comparison, I decided to go back to compare the Sotm/Sbooster with the Intel Nuc. Both of course are using Unison USB and the same USB cable.

    Nuc vs Sotm

    Intel Nuc:
    • Bass bloated and soft. Softer sounding overall sounding.
    • Greyer, noisy, edgier/dirtier sounding but not terrible. Increase in sibilance but not fatiguing.
    • Way flatter dimensionally - huge loss of depth
    • Loss of atmosphere and microdynamics
    • Narrower stage and not as tall
    • More congealed, less open, muddled in busy passages
    • Sotm is far more relaxed sounding, but more dynamic and detailed at the same time. Bass more balanced. Everything cleaner and smoother and more separated. Larger scale. Less grey. Vocals much clearer and more articulate and real sounding - more “organic” and less reproduced.
    So that’s it for now.

    My conclusions at this point are:
    1. Unison makes shitty USB transports (i.e. a standard computer USB output) sound pretty decent - acceptable to a lot of people I think and a worthwhile upgrade over Gen 5 USB definitely.
    2. Pi2AES IMO is going to be better than a standard computer USB out going into Unison, and you can improve things even further with a better power supply (but the stock supply is still pretty good). You gain in dimensionality and soundstage and atmosphere/plankton, and it sounds more dynamic and less soft/rounded. Considering the price of it with the stock supply, it is absolutely an awesome value. I consider it better than the microRendu and Cambridge Stream Magic that each cost much more.
    3. The Sotm with Sbooster can take Unison USB clearly above the level of the Pi2AES. It’s better in almost every respect IMO, aside perhaps from macrodynamics (where they are pretty close). It’s also many times the price of the Pi2AES, and if other parts of your chain need improvement, you’re probably better off spending your money there, since you’ll likely gain much more obvious differences.
    I’ll probably continue to bring in more transports to try them out and report back (and mess a bit with power supllies as well, although I feel that the Sbooster probably gets me mostly as far as I need to go - I’ve owned a few others as well)

    Hope this was a useful exercise.
     
    Last edited: Jan 13, 2020
    exocer, econaut, Mike_WI and 59 others like this.
  2. Biodegraded

    Biodegraded Friend

    Friend
    Joined:
    May 28, 2017
    Likes Received:
    2,933
    Dislikes Received:
    0
    Trophy Points:
    93
    Location:
    Vancouver BC
    CD73 vs any of the others please :)
     
  3. RobS

    RobS treboR (Psalmanazar Groupie)

    Friend
    Joined:
    Feb 21, 2019
    Likes Received:
    1,523
    Dislikes Received:
    25
    Trophy Points:
    93
    Location:
    Appalachia
    Wow. Great comparision, rlow. Appreciate your effort.

    Interesting. Schiit's rumored transport with USB hopefully gets even more gains out of Unison.

    I'm disappointed by the mixed reception of Unison. rlow, you said the Yggdrasil A2 was more tipped up in the higher frequencies, is there any kind of stridency you are hearing with Unison?

    USB continues to be a shit format for music playback it seems.
     
    FlySweep likes this.
  4. Ksaurav402

    Ksaurav402 Friend

    Friend
    Joined:
    May 7, 2019
    Likes Received:
    680
    Dislikes Received:
    4
    Trophy Points:
    93
    Location:
    NY
    Would love to see where Digione Signature with good power supply stands among them. Thanks for this.
     
    skem and txroadandmtb like this.
  5. Gaspasser

    Gaspasser Flatulence Maestro

    Friend
    Joined:
    Jan 16, 2016
    Likes Received:
    3,643
    Dislikes Received:
    10
    Trophy Points:
    93
    Location:
    Suburban DC
    I think that the mixed reviews should be taken with a grain of salt. My experience with Yggdrasil A2 Unison and the old Allo USBridge with cheap SolaHD LPS yields similar results that @rlow obtained with Sotm + SBooster.
    The easy takeaway from this excellent writeup is Unison input will scale like mad with better transports & power supplies.

    Edit: Another clear trend that @rlow and @earnmyturns noted is Pi2AES with meanwell switching PS is a great value in comparison to any of the Allo products that need usually expensive add-on power supplies to sound better.
     
  6. rlow

    rlow A happy woofer

    Friend
    Joined:
    Jul 18, 2017
    Likes Received:
    3,384
    Dislikes Received:
    3
    Trophy Points:
    113
    Location:
    Canada
    @RobS, @Gaspasser nails it here. Unison gets USB from a standard computer to an “acceptable” level finally IMO. But you can still scale it upwards with a purpose built audio device that uses USB.

    At some point I’m going to try to get my hands on a DigiOne Sig and a USBridge Sig to compare as well. Anyone else who’s compared any of those or others, please chime in.

    @Biodegraded I’ve done some really quick, non-critical comparisons between the CD73 and the Pi2AES and SOtM, but will do a bit more careful listening and report back. I’ll probably try the coax on the Pi2AES as well just so it’s apples to apples.

    By the way, I should have mentioned as well that I used Roon as the player for all my comparisons, so that everything was equal. I’ll update the first post to reflect that.
     
    Last edited: Jan 11, 2020
  7. Gaspasser

    Gaspasser Flatulence Maestro

    Friend
    Joined:
    Jan 16, 2016
    Likes Received:
    3,643
    Dislikes Received:
    10
    Trophy Points:
    93
    Location:
    Suburban DC
    I think @Clemmaster may be helpful here.
     
    Biodegraded likes this.
  8. Biodegraded

    Biodegraded Friend

    Friend
    Joined:
    May 28, 2017
    Likes Received:
    2,933
    Dislikes Received:
    0
    Trophy Points:
    93
    Location:
    Vancouver BC
    Gaspasser likes this.
  9. Mindbender

    Mindbender Acquaintance

    Joined:
    Jul 15, 2019
    Likes Received:
    40
    Dislikes Received:
    2
    Trophy Points:
    13
    Location:
    Anywhere the Wind Blows
    I’d really like to see a comparison where the SoTM is being fed between a Gen V and a Unison USB build. Although the tester would likely need two Yggdrasil’s or very nimble fingers.

    But to get a side by side comparison between the two USB formats with the same source would be informative.

    @rlow - Great analysis!
     
    allegro likes this.
  10. haywood

    haywood Friend

    Friend
    Joined:
    Oct 22, 2015
    Likes Received:
    267
    Dislikes Received:
    2
    Trophy Points:
    63
    Just for like, science, if powering the hat and Pi separately wasn’t part of the upgraded power supply test it might be worth trying (there’s a jumper that by default is bridged) with the lps powering the hat only and using a decent 5v switcher to power just the Pi.

    One area the SOTM probably has an advantage vs the Pi is latency since it’s likely running a lightweight real-time operating system. If your Pi distro allows a kernel with the preempt patch enabled it’d be worth trying, the idea is this allows kernel threads to be preempted as well which hurts overall perf a bit but improves latency quite a lot.
     
    Michael Kelly and Mindbender like this.
  11. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    53,119
    Dislikes Received:
    89
    Trophy Points:
    113
    Location:
    Antarctica
    What this says that USB source matters as much as the USB receiver on the DAC. I'm sure SOtM had paid a lot of attention to their USB.

    As for the Pi2AES, it sounds like I would prefer the sound of it with the stock switcher more than with the sbooster LPS.

    How did the physical CD transport compare?

    This is a good discussion. Many of you know I care not for USB or digital transmission nervosa, but when it's in the context of "transports" which actually have a play functionality, the discussion is worthwhile. This is a new fascinating frontier for me as I cannot wait to get rid of my laptops and PCs for playback.
     
  12. rlow

    rlow A happy woofer

    Friend
    Joined:
    Jul 18, 2017
    Likes Received:
    3,384
    Dislikes Received:
    3
    Trophy Points:
    113
    Location:
    Canada
    I’ll do some more detailed comparisons on various aspects at some point soon, but the very brief and non-critical comparisons I’ve done highlighted differences in two areas initially:

    • Transients/timing/immediacy: This was the most instant and really puzzling difference between the 2 streamers and the CD transport (using coax SPDIF). I’m not really sure I’m using the right names for it, but with the CD transport everything seems “faster” and snappier - transients seem more instantaneous and have an almost elastic “snap” to each note. It’s actually quite addictive to listen to, but I can’t fully explain it. Ironically “PRAT” is probably the best term I can give it.
    • Etch and sibilance: Everything seems just slightly brighter, harder and dirtier and a bit etched, compared to the Sotm especially, and somewhat more than the Pi2AES with stock supply. In my system, it borders right on the line with fatiguing, but never quite going over the edge.
    Maybe these 2 things are related in some way, not sure. Like I said, I do find the CD player more addictive to listen to because of the timing and dynamic/rhythmic snap, but I just find streaming so much more convenient and Roon such a great interface and way to discover new music, it’s kind of going unused lately.
     
    Last edited: Jan 11, 2020
  13. msommers

    msommers High on Epipens

    Friend
    Joined:
    Nov 6, 2016
    Likes Received:
    2,491
    Dislikes Received:
    34
    Trophy Points:
    113
    Location:
    Calgary, Alberta
    Home Page:
    I'm hesitant to agree with this based off of real world observations from @MisterRogers comparing power supplies on the PI2.

    Furthermore, since @rlow was using Roon I can posit that he's using RoPieee which is a bare-bones OS and frankly does very little processing in the PI unit. I can't imagine latency is a problem at all.
     
    Michael Kelly and rlow like this.
  14. rlow

    rlow A happy woofer

    Friend
    Joined:
    Jul 18, 2017
    Likes Received:
    3,384
    Dislikes Received:
    3
    Trophy Points:
    113
    Location:
    Canada
    Yep I am using Ropieee on the Pi2AES.
     
    Poleepkwa and Michael Kelly like this.
  15. haywood

    haywood Friend

    Friend
    Joined:
    Oct 22, 2015
    Likes Received:
    267
    Dislikes Received:
    2
    Trophy Points:
    63
    Iirc @MisterRogers posted that he was using two linear power supplies in his setup, one to hat and one to pi. It’d be interesting to see if rlow also found an improvement with dual power supplies with lps to just the hat, and if that made any difference in the comparison with his sotm.

    Latency isn’t load (though it can be affected by it), it’s how quickly the system can switch contexts when needed. Even if the os is lightweight it might be worth checking with whoever maintains the RoPiee project to see if it already uses or can use a preempt-able kernel:

    https://wiki.archlinux.org/index.php/Realtime_kernel_patchset
     
    msommers likes this.
  16. Scott Kramer

    Scott Kramer Friend

    Friend
    Joined:
    May 3, 2016
    Likes Received:
    1,102
    Dislikes Received:
    3
    Trophy Points:
    93
    Thanks @rlow for the awesome comparisons!

    This quote really reminded me of going to piCore realtime/cpu affinity noticed something about transients (timing on original digiONE), then the elastic/snap w/the digiONE_SIG on top of everything -->

    ref: https://www.diyaudio.com/forums/vendor-s-bazaar/294940-fifo-buffer-rpi-sbcs-243.html#post5552272
     
    Last edited: Jan 11, 2020
    haywood likes this.
  17. rlow

    rlow A happy woofer

    Friend
    Joined:
    Jul 18, 2017
    Likes Received:
    3,384
    Dislikes Received:
    3
    Trophy Points:
    113
    Location:
    Canada
    While I never did an quick A/B comparison as such, my initial impressions going from Gen 5 to Unison was an elimination of etch and fatigue, blacker background, better separation and more speed/immediacy (not necessarily more slam or dynamics however). At one point we had to send the beta board back for a firmware update, and putting the Gen 5 board back in place confirmed those thoughts. In fact, since I only had a USB transport at the time, I found I had to put the Schiit Eitr back in my chain in order to eliminate the etch and fatigue from Gen 5 (Eitr has some other issues with tonality and a bit too much softening of edges however, but it was less offensive to my ears). Once the Unison beta board came back however, all was well again.
     
    Jinxy245, msommers and Mindbender like this.
  18. RobS

    RobS treboR (Psalmanazar Groupie)

    Friend
    Joined:
    Feb 21, 2019
    Likes Received:
    1,523
    Dislikes Received:
    25
    Trophy Points:
    93
    Location:
    Appalachia
    Weird how these transports open up the soundstage like this. One of the common criticisms I've read regarding the Yggdrasil (with either of the two analog cards) is the width of the stage is fairly constricted. I definitely notice clear boundaries with the Gungnir Multibit A1 hooked up to both my Vidars (using Eitr). I'll have to try my old CD player if I can get it to read discs. Currently warming up a Yggdrasil A2 which sounds like a congested muddy mess at the moment. These Schiit Mulitibit DACs are really good with HD650s, great with tubes that open up the headstage. My shitty home theater budget speakers need a kick in the pants DAC like a Convert-2 to wake em up. I need new speakers...
     
    allegro likes this.
  19. lehmanhill

    lehmanhill Almost "Made"

    Contributor
    Joined:
    May 3, 2018
    Likes Received:
    127
    Dislikes Received:
    0
    Trophy Points:
    43
    Location:
    Michigan US
    I agree. The choice of power supply for the Pi2AES (and possibly other transports) seems to be system/preference dependent. Using Pi2AES with i2s out to an Onyx, I ended up preferring an old laptop switch supply over a quiet DIY linear supply. Of course, my DIY supply may not be as good as the Sbooster.

    Also, I am sure that rlow and Mister Rogers' improvements with better supplies are correct for their systems. There are just different trade-offs.
     
  20. msommers

    msommers High on Epipens

    Friend
    Joined:
    Nov 6, 2016
    Likes Received:
    2,491
    Dislikes Received:
    34
    Trophy Points:
    113
    Location:
    Calgary, Alberta
    Home Page:
    When all the talk of various LPS and the PI2 were going on months ago, there sounded like there were minor improvements with megabuck LPS units. @Michael Kelly can hammer this home but iirc using two power supplies, which is encouraged by Allo, was conversely discouraged by PI2 for their units.

    Again my own interpretations and impressions is that the PI2 is less susceptible to power supply changes than the DigiOne Sig. Overall I feel PI2 presents much higher value relative to Allo's offerings while still performing, at the very least, the same if not better as AES input into the Onyx I use has marginal gains over coax.

    This is my own personal sentiment as well but the fact @Michael Kelly is very active and open to feedback from the community here, I really like supporting his operation.
     

Share This Page