Delta Sigma: An Inconvenient Truth

Discussion in 'Digital: DACs, USB converters, decrapifiers' started by k4rstar, Jun 21, 2020.

  1. k4rstar

    k4rstar Done his time

    Friend BWC
    Joined:
    Jun 11, 2016
    Likes Received:
    5,192
    Dislikes Received:
    99
    Trophy Points:
    113
    Location:
    Toronto, Canada
    thanks Dan for trying it at my insistence and keeping an open mind. I have to disagree about the stridency only, the rest is open for interpretation. could you just add your chain including transport and examples of music used to your post for posterity.

    the 6DJ8 sent along is an early Amperex Holland, as good a tube as any, and leans a little warm if anything.
     
    Jerry and Gazny like this.
  2. ChaChaRealSmooth

    ChaChaRealSmooth SBAF Gearmaster

    Staff Member Friend Gearmaster
    Joined:
    Dec 6, 2018
    Likes Received:
    6,874
    Dislikes Received:
    16
    Trophy Points:
    93
    Location:
    Avalon
    Thanks for including me!

    I will edit my post later, but I'll post the chain used here:

    Pi2AES (via Coax) -> EC Ultralinear/DNA Starlett -> various headphones.

    FWIW the Abbas seemed to synergize better with the vintage Beyer and HD250 Linear. Also, the "stridency/ringing" I mentioned is not really a big issue. It's slight; I've observed that most R2R DACs seem to exhibit this character (Schiit DACs and Soekris 2541 included, and like I mentioned the only one I heard that does not seem to exhibit this trait is Wavelight). IMHO, this is much preferable to the "digititus" present in many poorer DS implementations (even X-Sabre Pro, which is IMHO the best Sabre DAC, has a bit of this).

    I listened to a lot of Daft Punk, Jamiroquai, and some Bryan Adams. Right now that stuff is hitting the spot for me.
     
  3. 7seven

    7seven Rando

    Joined:
    Oct 28, 2018
    Likes Received:
    7
    Dislikes Received:
    3
    Trophy Points:
    8
    Location:
    UK
    This is the perfect oportunity to add oversampling with a PC/Pi to see how it influences the sound, after all it's the only way to directly compare NOS and OS.
    IME NOS only has a small influence on sound in the grand scheme of things and most of what your hearing is the DAC,
    OS most of the time just made things sound more 'digital' without improving resolution/technicalities.
    I bet NOS sound can depend a lot on how your amp responds to the ultrasonic content too.

    Also keep in mind the treble roll off that NOS exhibits :
    [​IMG]
     
    Last edited: Jan 21, 2021
  4. wormcycle

    wormcycle Friend

    Friend
    Joined:
    Aug 13, 2016
    Likes Received:
    1,356
    Dislikes Received:
    43
    Trophy Points:
    93
    Location:
    Toronto, ON, Canada
    That's interesting. So let's say I have JRiver reading files from NAS and the setup would read the entire encoded file into memory before playing it. The same can be done with Roon endpoint. So cables, switches are not a part of the playback process. I would think that if the Ethernet components are not a part of the playback process, they cannot possibly sound different?
     
  5. dubharmonic

    dubharmonic Friend

    Friend
    Joined:
    Mar 30, 2017
    Likes Received:
    1,890
    Dislikes Received:
    2
    Trophy Points:
    93
    Location:
    Evanston, IL
    @wormcycle noise from networking usually doesn't impact the digital files, but it can introduce analog noise to the rest of the chain. Even when there's no digital traffic over an ethernet cable, it's possible for noise to travel from a switch into a DAC.
     
    skem likes this.
  6. Azteca

    Azteca Friend

    Friend
    Joined:
    Sep 26, 2015
    Likes Received:
    1,556
    Dislikes Received:
    1
    Trophy Points:
    93
    Location:
    Cincinnati, Ohio
    Evidence?
     
  7. robot zombie

    robot zombie Friend

    Friend
    Joined:
    May 11, 2016
    Likes Received:
    1,432
    Dislikes Received:
    2
    Trophy Points:
    93
    Location:
    Ennui, FL
    I actually don't know if I would implicitly doubt that. Ive observed pretty terrible noise from USB dac connections before. It has to do with the USB power bus, no effect on the data or the dac function. If you put your ear up to the motherboard you can hear the same noise that gets carried to the speakers coming directly off of the power section. Only running toslink or converting the USB to something else stops it. AFAIC its down a bad combo of subpar USB implimentation and poor isolation on the board. Not all dacs even pick it up. But when they do its mostly a lot of beeping, chirping and whining that's constant, whether anything is playing or not.

    I don't know if ethernet has that same vulnerability. I dont even understand how it happens with USB.
     
  8. rlow

    rlow A happy woofer

    Friend
    Joined:
    Jul 18, 2017
    Likes Received:
    5,697
    Dislikes Received:
    5
    Trophy Points:
    113
    Location:
    Canada
    I have no skin in this game, however here’s a link to a white paper from an engineer that designed an Ethernet isolation switch, the Uptone EtherRegen. Most of this is way over my head, so don’t come back expecting me to argue for or against it. I’ve not heard the product, so nothing to offer there either. I just link it for those who may be interested in reading:
    https://cdn.shopify.com/s/files/1/0...enson_EtherREGEN_white_paper.pdf?v=1583429386

    There’s a discussion about the white paper from the owner of Uptone and occasionally the author (John Swenson) over on the former Computer Audiophile forums (I refuse to call it by its current name):
    https://audiophilestyle.com/forums/...f-the-uptone-jswenson-etherregen-white-paper/

    Apparently the measurements to back up some of the claims in the white paper is still WIP (which obviously seems a bit dicey right off the bat). Maybe @atomicbob would weigh in on his feelings of the content.

    Edit: Amir apparently “measured it” and found it no better than a regular $60 switch. But we all know how much that opinion is likely worth...
     
    Last edited: Jan 21, 2021
    robot zombie and crenca like this.
  9. wormcycle

    wormcycle Friend

    Friend
    Joined:
    Aug 13, 2016
    Likes Received:
    1,356
    Dislikes Received:
    43
    Trophy Points:
    93
    Location:
    Toronto, ON, Canada
    Right, this is Important point regarding the difference between USB and Ethernet, power line, unless of course you use POE switch. In this case I would be interested how different or similar it is.
    On the other hand Ethernet is not usually connected to you DAC.
    EDIT: yes it is possible it get your DAC over anything but Toslink
     
    robot zombie likes this.
  10. Dzerh

    Dzerh Friend

    Friend
    Joined:
    Apr 1, 2018
    Likes Received:
    403
    Dislikes Received:
    6
    Trophy Points:
    63
    Location:
    San Ramon, CA, US
    continue offtopic about ethernet, sorry.
    Keep in mind, for all cables in general:
    1. There is always analog noise.
    2. Any cable can affect performance if it is not up to spec.
     
  11. Psalmanazar

    Psalmanazar Most improved member; A+

    Slaytanic Cliff Clavin
    Joined:
    Sep 27, 2015
    Likes Received:
    4,717
    Dislikes Received:
    139
    Trophy Points:
    113
    It's not just the roll off. There's a heavy phase shift that sounds like trash. Linear phase oversampling always sounds better. The filter in modern dsp (both dacs and plugins) beyond the audible range and mostly affects what comes after it.

    Those who prefer NOS DACS do not prefer analog sound. Analog sound, not vinyl or tape, does not have these electronic or digital filters. They prefer the sound of decades old digital conversion. Converters were originally not oversampled or used minimum phase filters. Outside of masturbatory audiophile whack off gear and extremely low end prosumer gear, e.g. Behringer and Focusrite, these filters do not exist for good reason. Quite frankly, they are very audible and sound like shit. Anything with them sounds awful or would sound a great deal better with a competent linear-phase oversampled filter. This is not preference. It just sounds awful and cannot come close to adequately converting redbook audio. NOS DA conversion just makes everything sound like an awful mid 1980s CD master. Given that some audiophiles (many of the posters on the Steve Hoffman forum) prefer CD masters made from cassette back up copies converted with 1980s converters over more modern 90s and 2000s masters made from the original tapes with modern converters, audiophile preference is pretty much worthless. Granted the movie and music industries still try to sell butchered, noise-reduced products to modern consumers. However you won't hear any of that with a non-oversampled DAC. You're just moving it even further from the ur source with more phase shift and more distortion that will impact whatever equipment it is fed into.
     
    Josh83, crazychile, Xecuter and 2 others like this.
  12. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    72,274
    Dislikes Received:
    115
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    It may not matter. An NOS design is likely to have a brick-wall analog filter to get rid of the 44.1kHz pulse which sits a little bit over one octave above human hearing. Such a filter will cause phase shift in the audible spectrum. Now, I could see an NOS DAC design with a variable analog filter that filters accordingly to the sampling rate it is fed.

    You have to make up your mind here in respect to your opinion on this. You can't say NOS has a small influence and that most of it is the DAC, and then suddenly say that OS makes things more "digital" sounding (which would imply a significant influence). Personally, I feel that delta-sigma DACs, particularly implementations a generation earlier, are responsible for the high-end nasties associated with "digital" sound. NOS or OS is just that, something altogether different. I've heard NOS delta-sigma DACs, and those had high-end nasties too.

    I would hope not because a properly designed NOS DAC should still have an appropriate LPF / anti-alias / reconstruction filter.
     
    Last edited: Jan 21, 2021
    crenca and Dzerh like this.
  13. k4rstar

    k4rstar Done his time

    Friend BWC
    Joined:
    Jun 11, 2016
    Likes Received:
    5,192
    Dislikes Received:
    99
    Trophy Points:
    113
    Location:
    Toronto, Canada
    would be great if you could stop misleading people who don't know any better. when is the last time you put any of your approved gear list on loaner?
     
  14. atomicbob

    atomicbob dScope Yoda

    Friend BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    15,232
    Dislikes Received:
    4
    Trophy Points:
    113
    Location:
    On planet
    Let's look at a particular measurement, Jitter Attenuation, for two DACs.
    Both are measured with spdif input and SE output.

    Here is the first one, a Topping D30:
    01 Jitter Attenuation D30 - annotated.png
    The yellow line to which the red arrow points is the jitter attenuation for 80 nS jitter insertion with the jitter modulation swept over the audio frequency range. The blue FFT is one frequency on the sweep. At approximately 4200 Hz a strong peak indicates substantial jitter contamination. Also look around the base of the 8400 Hz stimulus. Basically the D30 has no jitter attenuation and would likely benefit from an external jitter reduction device if the digital audio source is plagued with jitter.

    Now look at this one for a Soekris dac2541:
    02 Jitter Attenuation dac2541 - annotated.png
    Here we see the jitter attenuation sweep demonstrates dac2541 has significant jitter reduction capability. One of the not so well known, great features of Søren's (@soekris ) designs. The overall noise floor is significantly lower than the previous DAC. Doubtful any external device will have any positive benefit for this great DAC. More than likely it will either do nothing or possibly detract from overall performance.

    *edit* added clarity for measurement setup input and output
     
    Last edited: Jan 21, 2021
    Cryptowolf, Azteca, blubliss and 13 others like this.
  15. Psalmanazar

    Psalmanazar Most improved member; A+

    Slaytanic Cliff Clavin
    Joined:
    Sep 27, 2015
    Likes Received:
    4,717
    Dislikes Received:
    139
    Trophy Points:
    113
    It's quite easy to program a linear phase low pass filter outside the audible range with no audible ringing in the audible range. Only the stop band is smeared unlike with linear phase equalization of music, which can sound horrible easily. Pretty much every modern converter chip has an adequate filter built in. The ones that allow you to select minimum phase filters are stupid design choices. The NOS DACs that sound pleasant sound that way in spite of all of that. They would only sound better with adequate filtering.
     
    taisserroots and crenca like this.
  16. earnmyturns

    earnmyturns Smartest friend

    Friend
    Joined:
    Sep 25, 2016
    Likes Received:
    2,697
    Dislikes Received:
    5
    Trophy Points:
    113
    Location:
    Palo Alto
    Home Page:
    Having owned several NOS DACs I'd have sort of agreed -- until I got a
    What input? If USB, why would a DAC be at all sensitive to source jitter, given that it should be able to buffer bits and reclock them in async USB class 2? Am I missing something (not unlikely)?
     
  17. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Friend MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    11,699
    Dislikes Received:
    42
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    It's about a 50/50 shot if a NOS DAC has anything "more" than gentle LPF starting at 60-100KHz. Metrum, for example, barely has any sort of LPF. I think 1st order, 70KHz? I assume this is the main reason their DACs tend to having increasing IMD as the frequency increases.

    Others will have more aggressive filtering, and I assume at the expense of phase linearity. But that's about the most you'll get out of non-oversampling, just some sort of analog LPF of varying types and strengths.

    I've done direct A/B tests on a couple such DACs recently and can't say it's anything I hear much of a difference with one way or another with headphones.
     
    Cryptowolf, Baten, nishan99 and 3 others like this.
  18. atomicbob

    atomicbob dScope Yoda

    Friend BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    15,232
    Dislikes Received:
    4
    Trophy Points:
    113
    Location:
    On planet
    I updated the two measurement annotations to clarify this question. spdif input and SE output. All DAC buffering is not created equal.
     
    Cryptowolf and Psalmanazar like this.
  19. 7seven

    7seven Rando

    Joined:
    Oct 28, 2018
    Likes Received:
    7
    Dislikes Received:
    3
    Trophy Points:
    8
    Location:
    UK
    Using a proper reconstruction filter is what was intended but are you sure most of them actually do? I think it's most of them don't, especially for this kind of boutique stuff and DIY stuff, the idea being that aliasing is less harmful than phase shift/more complex signal path, and I would agree with that . My opinion is based on NOS stuff without one.


    NOS or OS has a small influence, didnt mean to imply it wasnt.
    Digital was a bad descriptor though since it implies 'bad' (I was thinking digital vs vinyl), SS vs tube would be better, not on the same scale of course. It could be nothing more than an EQ effect from the rolloff, I think there is slightly more to it though.
     
    Last edited: Jan 22, 2021
  20. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    72,274
    Dislikes Received:
    115
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    It's a good question. What's worse or how is it different: improper reconstruction or phase shift? And do we know if or how a filter is implemented.

    For all we know, the NOS "tone density" or (even maybe the specific "R2R stridency" that @ChaChaRealSmooth mentioned) could be the result of aliasing / improper reconstruction.

    It can be argued that somewhere along the signal path, the signal ends up bandlimited, therefore could be in affect subject to an anti-alias / reconstruction filter.
     
    Cryptowolf likes this.

Share This Page