DACs: Improve your USB (break out PCI USB card with Unison)

Discussion in 'Digital: DACs, USB converters, decrapifiers' started by purr1n, Jan 1, 2022.

  1. Psalmanazar

    Psalmanazar Most improved member; A+

    Slaytanic Cliff Clavin
    Joined:
    Sep 27, 2015
    Likes Received:
    4,814
    Trophy Points:
    113
    @purr1n once you go pci-e (only options are RME and Lynx) or good Thunderbolt (only options are RME, Lynx, and UAD), you don’t want to go back. Only Lynx is great sounding of those. Otherwise you still need a stereo converter.

    the apogee symphony II is tbolt but has a fan so it’s not really for 99.999% of sbaf members. Motu is wtf with support and product development. Don’t know anyone with the Steinberg tbolt and only one with the Slate pci-e interfaces.

    keep in mind that most of these audio guys are running Mac minis, old iMac Pros (slim and doesn’t throttle), or custom silent pc builds. They’re not using white girl or gamer computers. Most of them will laugh at people using cards other than stuff like the fanless 1030 or 1050 where you can safely turn off the fan.
     
  2. crenca

    crenca Friend

    Friend
    Joined:
    May 26, 2017
    Likes Received:
    2,653
    Trophy Points:
    113
    Location:
    Southern New Mexico
    Right, essentially I'm just confirming Purr1n's order - #6 vs. #2 in my case. The Pi is running RopieeeXL, so my stream is unmolested from PC to Yggdrasil (Windows OS sound libraries not involved, RAAT, etc) except when I intentionally apply DSP within Roon, which itself is mostly SoX based as I understand it or at least used to be.

    As far as PC OS, playback software (e.g. JRiver vs. Roon vs. ...), stream format, and the like having a "sound" I agree with @earnmyturns when (and only when) the end user knows his software (and hardware) well enough to set up an apples to apples comparison. However, like @Beefy points to how many folks know whether or not their OS is processing/resampling, or their playback software/streamer/endpoint, and is software (particularly low priced and "free") even always honest about it?

    When I set up an apple to apple, such as Roon vs. JRiver, or a more specific comparison between Roon's internal DSP and my JRiver/DMG Equilibrium (VST) setup and try to hear the sound of my EQ, the differences are well within placebo range so 96.58% of the time I'm listening to Roon because convenience trumps small differences, let alone placebo.
     
    • Like / Agree Like / Agree x 5
    • List
    Last edited: Jan 4, 2022
  3. Beefy

    Beefy Almost "Made"

    Joined:
    Apr 10, 2021
    Likes Received:
    860
    Trophy Points:
    93
    Location:
    Canada
    This is a good point, where I can tell the story of a silly own-goal....

    I used to pride myself on being pretty informed about all this stuff, and for the longest times was smug in the self satisfaction of my 'bit perfect' playback chain. Except, for literally years, I had replaygain enabled in foobar. Turned it on once for a wonkily mixed album, and forgot to turn it off again. That's every single track I had listened to for years, having a variable amount of attenuation or gain applied, for volume matching within and across different albums. Unequivocally not bit perfect.

    One, I didn't notice, for shame to both my ears and my smugness. Two, if I had made any comparisons between computer USB versus any other source to my DAC, it might be a significant confounding variable.
     
    • Like / Agree Like / Agree x 5
    • List
  4. earnmyturns

    earnmyturns Smartest friend

    Friend
    Joined:
    Sep 25, 2016
    Likes Received:
    2,028
    Trophy Points:
    113
    Location:
    Palo Alto
    Home Page:
    I can't help with Windows, which I left behind around year 2000 :D All my digital sources are Linux-based, and all they do is to move bits through the driver and interface without pretending to know better about sampling rates.
     
  5. AllanMarcus

    AllanMarcus Friend

    Friend
    Joined:
    Oct 23, 2015
    Likes Received:
    2,533
    Trophy Points:
    113
    Location:
    Los Alamos, NM
    Home Page:
    just make sure you are using audiophile grade ethernet cables, an audiophile grade router, and an audiophile grade switch (if needed), audiophile grade SSD drives in the audiophile optimized server, audiophile grade feet (on the server, the endpoint, and the DAC!), it all has to sit on an audiophile grade rack.

    The sad part is, all these things exist.
     
    • Like / Agree Like / Agree x 6
    • List
  6. rlow

    rlow A happy woofer

    Friend
    Joined:
    Jul 18, 2017
    Likes Received:
    5,955
    Trophy Points:
    113
    Location:
    Canada
    You forgot these.
     
    • Like / Agree Like / Agree x 5
    • List
  7. dericchan1

    dericchan1 Acquaintance

    Joined:
    Sep 14, 2021
    Likes Received:
    39
    Trophy Points:
    18
    Location:
    Canada
    The sad part is I can’t afford any of these!!! I have to settle for utp cat 6 cables with plastic connectors, a $30 Used Cisco switch from 2008.
     
    • Like / Agree Like / Agree x 1
    • List
  8. exocer

    exocer Acquaintance

    Joined:
    Jan 30, 2020
    Likes Received:
    65
    Trophy Points:
    18
    Location:
    NY, NY
    FWIW - I hear major differences between OS's i've used on the Pi2AES. I don't have any concrete arguments. I had no desire for one OS to sound better than the other as all of the OS's were freeware. No scientific blind tests to validate it either... so, probably not worth much. I completely skipped Roon so I can't speak to whatever SQ differences there are...but with a quick search you'll see many complain about how much better Audirvana sounds (or HQPlayer) than the more user friendly easier to use, more expensive Roon.
     
    • Like / Agree Like / Agree x 1
    • List
  9. exocer

    exocer Acquaintance

    Joined:
    Jan 30, 2020
    Likes Received:
    65
    Trophy Points:
    18
    Location:
    NY, NY
    You forgot to mention reference clocks :)
     
    • Like / Agree Like / Agree x 1
    • List
  10. dericchan1

    dericchan1 Acquaintance

    Joined:
    Sep 14, 2021
    Likes Received:
    39
    Trophy Points:
    18
    Location:
    Canada
    Out of all the free os, i felt moode sound the best to me, not sure I would be able to do a blind ab tho.

    Hqplayer is a notch above

    Deric
     
    • Like / Agree Like / Agree x 1
    • List
  11. haywood

    haywood Friend

    Friend
    Joined:
    Oct 22, 2015
    Likes Received:
    660
    Trophy Points:
    93
    Roon uses its own bespoke (proprietary?) networking scheme between the client and endpoint, it’s not really surprising it’d have a different sound. I didn’t do any comparisons between roon and a stand-alone distro when I tried it out but didn’t really see the point in using roon if you had tidal + connect (or Spotify lossless + connect if they’d ever get off their butts).

    HQP has had a long time to develop their audio stack and the various resampling filters in particular. People who use it tend to primarily be those who want to do extreme resampling (like to 768k) or dsd conversion. If your dac benefits from that kind of resampling then no doubt a Pi that is limited to 24/192 (or 24/384 on some systems) isn’t going to measure up.
     
  12. dericchan1

    dericchan1 Acquaintance

    Joined:
    Sep 14, 2021
    Likes Received:
    39
    Trophy Points:
    18
    Location:
    Canada
    actually I use 2 rpi4 as hqplayer network adapter to connect to my two systems. 1 system is hqplayer desktop pc to a pi2aes as a hqplayer network adapter upsample pcm to 24/192 and apply a linear phase filter to a Marantz hdamp1 dac, the improvements over moode is pretty significant!!!!

    another set up is hqplayer desktop to rpi4 to ifi neo idsd upsample to 24/768k for pcm and dsd256 for dsd, that sounds pretty nice as well
     
    • Like / Agree Like / Agree x 1
    • List
  13. earnmyturns

    earnmyturns Smartest friend

    Friend
    Joined:
    Sep 25, 2016
    Likes Received:
    2,028
    Trophy Points:
    113
    Location:
    Palo Alto
    Home Page:
    It would be surprising for correctly implemented processing. RAAT is just a scheme to move bits from the Roon server to the Roon endpoint at a rate driven by the endpoint's clock. RAAT is layered on TCP, so it has error correction and retransmit; the only way something would go wrong there would be a long-enough packet loss that RAAT gives up and skips the track, not any change in sound quality. The only sound quality change between RAAT and other protocols would be if the endpoint is poorly implemented and in assembling a synchronous bitstream for rendering it causes the downstream DAC to behave differently. My main hypothesis after over 7 years of messing around with streaming audio, 5 with Roon, is that (like all computers), most streamers are buggy in one way or another, and SQ differences between sources are caused by those hardware or software bugs. To give you a contrast, I own two Linn systems and have listened to several others, and there's zero difference between how they sound with Roon as a source vs their own DLNA implementation.
     
    • Like / Agree Like / Agree x 6
    • List
  14. haywood

    haywood Friend

    Friend
    Joined:
    Oct 22, 2015
    Likes Received:
    660
    Trophy Points:
    93
    Well theoretically all bit-perfect audio would sound the same, but people usually don’t think that’s the case. I don’t see why Roon should sound noticeably different (or worse as someone claimed) using the same source material but if there are differences then it makes sense to look at where platforms differ to try to deconstruct the why.
     
    • Like / Agree Like / Agree x 2
    • List
  15. quattr0

    quattr0 Rando

    Joined:
    Jul 2, 2021
    Likes Received:
    2
    Trophy Points:
    3
    Location:
    TX
    Would you please educate me on blackground? What does it mean? Is there a particular song you listened to that you could easily notice the difference? I cannot tell a difference.

    As for transport, I have compared so far Pi3 b+, Hifi Rose RS250, iFi Zen Stream, sotm sm200 ultra neo with SBooster, the transport from Lyngdorf tdai-3400, all sounds the same to me. I wonder it is because the RME ADI-2 FS Pro BE DAC doing such a good job of filtering all garbage, thus I cannot hear the difference?

    My speakers are Focal Aria 926.
     
  16. RestoredSparda

    RestoredSparda Friend

    Friend
    Joined:
    Jan 7, 2021
    Likes Received:
    56
    Trophy Points:
    18
    Location:
    New Richmond, Wisconsin US
    Here are explanations of the various terms, including "blackground".

    https://superbestaudiofriends.org/i...logy-subjective-terms-used-on-superbaf.11665/
     
    • Like / Agree Like / Agree x 1
    • List
  17. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Friend MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    11,990
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    There's a difference between data transfers over a networking protocol like TCP, which is bi-directional and has error detection and correction, vs. a unidirectional, realtime protocol such as SPDIF.

    Now, if your source is sending data to a DAC over USB, I'm less certain if any, some, or all implementations have native error detection and correction. Lots of conflicting and sketchy info online about that one. Though it seems at least some implementations are also unidirectional and without error correction/detection.
     
  18. purr1n

    purr1n Super Friend

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    75,432
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    To follow up, let's continue to search because maybe we are looking at the wrong measurement.

    Here I decided to so a simple 1kHz 0dbFS signal. This time I increased bandwidth to 250kHz and plotted the frequency axis in linear scale.

    1kHz 0dbFS
    250kHz bandwidth
    USB Port "A"
    upload_2022-1-7_15-3-52.png

    1kHz 0dbFS
    250kHz bandwidth
    USB Port "B"
    upload_2022-1-7_15-9-0.png

    Care to guess which USB port which is which? One port is a mobo USB port on a bus shared with mouse, keyboard, another DAC, or whatever else. Could be the MOTU or USB ethernet. I have no clue, but I do know that it's shared. The other USB port is on a dedicated bus off this Renasys controller USB PCI card.

    Of course none of these measurements should be audible. Then again, why is there this "common belief" that LPS sound better than switchers (which tend to have lots of ultrasonic junk), and likewise why do IMO, less noisy super-higher frequency switcher sound better than < 1MHz switchers?

    We don't have all the answers, but this is worth looking into, including blind tests to see if subjective differences are discernable.
     
    • Like / Agree Like / Agree x 12
    • List
    Last edited: Jan 7, 2022
  19. crenca

    crenca Friend

    Friend
    Joined:
    May 26, 2017
    Likes Received:
    2,653
    Trophy Points:
    113
    Location:
    Southern New Mexico
    So the isolated USB card is cleaner. IMD is what is affecting the audio band perception of stage, detail, and instrument isolation/delineation??

    Perhaps these sorts of in and out of band distortion measurements are just indictive of something else entirely - the measured distortion is a secondary indicator of whatever is the primary cause in causing the perceived stage/detail differences. Can't imagine what the primary would be...
     
    • Like / Agree Like / Agree x 1
    • List
  20. purr1n

    purr1n Super Friend

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    75,432
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Going out on a limb, but I suspect the primary to be in the transient type measurements. The problem with trying to visualize transient distortion on an traditional visualization such as above. However, this is not possible because the frequency spectrum FFT requires a window of 32768 samples to process, and then after than, ten of those measurements to get an average reading. We will need to wait until quantum computing.

    We can always of course try to compare a waveform, but eyeballing that doesn't exactly work.
     
    • Like / Agree Like / Agree x 4
    • List

Share This Page