Native DSD makes DSD 512 version of the Sound Liaison DXD original, upsampling???

Discussion in 'Music and Recordings' started by Peter Apperup, May 17, 2019.

  1. Peter Apperup

    Peter Apperup Acquaintance

    Joined:
    Sep 30, 2016
    Likes Received:
    11
    Trophy Points:
    18
    Location:
    europe
    Help me understand this;
    Native DSD has just released a sampler with 5 tracks in DSD 512.
    It says; "These higher bit rate DSD 512 tracks are all pure DSD created. They are not up samplings, for there are no PCM or DXD conversions involved in their production."... ?
    I happen to own the last track of the sampler, the sublime live recording of The Feenbrothers playing Brubeck,
    and I know that the original recording has been done in DXD 352. That's what it says on the https://www.soundliaison.com/ site where the DXD version is still on special offer.
    But how can it not be an upsampling? Or is it OK because DSD is a different format from DXD?
    [​IMG] https://www.soundliaison.com/index.php/453-new-the-feen-brothers-one-mic-recording
    [​IMG]

    https://nativedsdpresents.nativedsd.com/albums/NDSD013-nativedsd-presents-5-tracks-in-dsd-512
     
  2. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    10,961
    Trophy Points:
    113
    Location:
    Philippines, The
    I stopped giving a bother about anything other than PCM but hope this is some help (DXD is basically super high-rez PCM, 352.8kHz, while I barely trust the alleged benefits of 192kHz/24bit):

    https://dsd-guide.com/faq/what-dxd-it-dsd#.XN9hb6RS9PY
    https://www.mojo-audio.com/blog/dsd-vs-pcm-myth-vs-truth/

    Also from nativeDSD themselves: http://help.nativedsd.com/high-resolution-dsd-listening/additional-reading/dxd-resolution-explained

    They're either lying, which does happen surprisingly enough, or they're being disingenuous. If the original recording was done in PCM (which is what DXD is) then there's no way short of their having had an entirely separate recording system right next to it operating simultaneously to capture in DSD.

    Brain fart: or it's an entirely new performance by the same act?
     
    Last edited: May 17, 2019
  3. drgumbybrain

    drgumbybrain Science Nut

    Pyrate
    Joined:
    Jan 9, 2017
    Likes Received:
    2,219
    Trophy Points:
    93
    Location:
    Lives in Fortaleza, Heart in Girona
    Don’t have hard disk space to those 512 dsd tracks. Life is too short
     
  4. Peter Apperup

    Peter Apperup Acquaintance

    Joined:
    Sep 30, 2016
    Likes Received:
    11
    Trophy Points:
    18
    Location:
    europe
    It's definitely the same version.
    Here is what Native DSD says; " These higher bit rate DSD 512 tracks are all pure DSD created. They are not up samplings, for there are no PCM or DXD conversions involved in their production. They are re-modulations of the original DSD 256 encoding modulation that produced the DSD 256 releases. The sonic advantage to these new Stereo DSD 512 releases, as with all higher DSD bit rate releases, is the wider frequency passband prior to the onset of modulation noise. This results in the listener’s DAC using gentler and more phase linear filters for playback of the music."
     
  5. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    10,961
    Trophy Points:
    113
    Location:
    Philippines, The
    Not upsamplings but remodulations. Gotcha. I'm guessing the argument is that the process is nondestructive or immune to the quantization errors of PCM upsampling? I used to know much too much about DSD for my liking, but have forgotten most of it and had to read up on it again.

    I will say that I'm generally not a fan of (PCM) interpolation because it just seems to haze everything up and that well-done Redbook is far preferable to the same recording upsampled to 768.4kHz/48-bit (Lord I hope this doesn't actually exist).

    The way I understand remodulation is that it's analogous to capturing a photo in RAW and then doing a finer job editing it in Lightroom? If so then I think it's bull because you simply cannot get any more data out of a photo than was captured by your sensor to begin with, and if it's different then that's a VSCO filter at work; you could maybe use interpolation to turn your 12MP image into a 48MP one, but I doubt it'd look nearly as good as a native 48MP photo, or even a 24MP one when viewed at the same distance with the same quality, unless they gimped the original DSD256 and are just now allowing its full, DSD512 glory to come through.
     
    Last edited: May 18, 2019
  6. Baten

    Baten Friend

    Pyrate
    Joined:
    Mar 18, 2018
    Likes Received:
    1,130
    Trophy Points:
    93
    Location:
    EU
    As you say it's likely just bull, DSD512 is pretty ridiculous as a redistributable format.
     
  7. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    10,961
    Trophy Points:
    113
    Location:
    Philippines, The
    Always tend to give people benefit of the doubt, but you have no idea how much I've had to tamp down my knee-jerk response of "DSD is bull!" the last little while, haha. I've not got excellent ears or very well-developed listening skills, and I'm aware there are others who have both these, so eh.

    Also edited my last post to make a point I forgot to originally.
     
  8. Peter Apperup

    Peter Apperup Acquaintance

    Joined:
    Sep 30, 2016
    Likes Received:
    11
    Trophy Points:
    18
    Location:
    europe
    btw.....there is only a few days left of the special offer on the album at the Sound Liaison site.https://www.soundliaison.com/
    The DXD recording sounds incredible. It has been done with only one very special stereo microphone so it's completely phase coherent. It is one of the best sounding albums that I own;
     
  9. Peter Apperup

    Peter Apperup Acquaintance

    Joined:
    Sep 30, 2016
    Likes Received:
    11
    Trophy Points:
    18
    Location:
    europe
    The 5 tracks from Native DSD seems to be on the new Astell & Kern portable player in DSD512 format.
    And no wonder they are very exited about the product:)

     
  10. Mike Clark

    Mike Clark New

    Joined:
    Jun 6, 2019
    Likes Received:
    1
    Trophy Points:
    1
    Location:
    europe
    Can it have something to do with the lowpass filter being used on DSD masters. Maybe the higher rate make the filter works more smoothly?
     
  11. Taverius

    Taverius Smells like sausages

    Pyrate
    Joined:
    Dec 27, 2017
    Likes Received:
    3,026
    Trophy Points:
    113
    Location:
    Rapallo, Italy
    Why the hell does anyone need a nyquist frequency greater than 24khz is still something none of these bit peddlers has been able to explain.

    Unless we're hoping to charm mosquitoes.
     
  12. Peter Apperup

    Peter Apperup Acquaintance

    Joined:
    Sep 30, 2016
    Likes Received:
    11
    Trophy Points:
    18
    Location:
    europe
    That could be it. Sounds logical.
     
  13. Peter Apperup

    Peter Apperup Acquaintance

    Joined:
    Sep 30, 2016
    Likes Received:
    11
    Trophy Points:
    18
    Location:
    europe
    The Sound Liaison label are selling the album as well a the other ''One Mic recordings" in lower resolutions down to 24/96 and DSD64; https://www.soundliaison.com/
    [​IMG] [​IMG] [​IMG] [​IMG]
     
  14. Peter Apperup

    Peter Apperup Acquaintance

    Joined:
    Sep 30, 2016
    Likes Received:
    11
    Trophy Points:
    18
    Location:
    europe
    but even a CD is 44.1khz and from what I have read digital reverbs does sound better in 24bit.
     
  15. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    10,961
    Trophy Points:
    113
    Location:
    Philippines, The
    Oh. Oh no. Ohhh no. Oh dear. If I'm reading what you're saying correctly there may be a bit of a misunderstanding.

    44.1kHz as per Redbook CD spec does not mean that the data therein includes sound all the way up to 44,100Hz. The sampling rate refers to the number of times per second datapoints are taken in order to approximate an analogue waveform for reconstruction of that same waveform down the line during the playback process. Given the nature of how this works the highest sound Redbook standards can capture and reproduce is maybe half the sampling rate, or about 22kHz, which is already generally accepted as higher up the spectrum than is really necessary.

    From what I understand the only reason people invested time into developing higher-res formats is that older ADCs sometimes generated sonic artefacts within the audible spectrum (i.e. below 20kHz), and using higher sampling rates pushed the same above the audible threshold. I think modern gear isn't necessarily susceptible to these issues, at least according to what I've read (because I've no experience at all on the production side of things).

    I do think ultrasonics and infrasonics might have an effect on human perception and overall being, but research into sonic weapons was banned by the Geneva convention last I checked, so I'm not touching that stuff.

    24 bits allow for remarkably high dynamic range, but whether or not any playback system on the planet is capable of reproducing such quiet sounds while actually-audible things are playing is debatable given few people listen to music in anechoic chambers I think.
     
  16. Psalmanazar

    Psalmanazar Most improved member; A+

    Pyrate Slaytanic Cliff Clavin
    Joined:
    Sep 27, 2015
    Likes Received:
    5,345
    Trophy Points:
    113
    Who gives a shit? CDs and 16-bit, 44.1khz is fine for playback and 24-bit, 44.1khz is fine for production. Modern gear and plugins upsample for the filtering where it actually matters.
     
  17. skem

    skem Friend

    Pyrate
    Joined:
    Nov 1, 2017
    Likes Received:
    1,911
    Trophy Points:
    93
    Location:
    Charles River
    Maybe it’s time for me to address the erroneous view that sampling at twice the limit of human hearing enables, at least a theoretically, perfect audio reconstruction. This is simply not true in the general case, nor does the Nyquist-Shannon theorem say it is.

    The Nyquist-Shannon theorem says if the original signal has a no spectral components above some cut-off band (the Nyquist limit)—then the signal can be perfectly reconstructed in theory. That said, (1) this condition isn’t necessarily met in music, and (2) DAC’s don’t do “in-theory” reconstruction.

    Let’s look at problem (1): A drum hit (delta function), the square waves of a synthesizer, a triangular ramp, all necessarily have spectral components well above their fundamental frequency. In fact, these are all spectrally unbounded.

    Well, you may say, that is true BUT those frequencies are outside of the limit of human perception. Tut tut... not so fast. The inaudible frequencies sum together in ways that cancel out lower-frequency artifacts generated by the Fourier reconstruction (here I mean the analytical reconstruction by math, not your DAC). Unless special countermeasures are taken, the sound of a Nyquist-limited reproduction of all non-sinusoidal acoustic information will be polluted by crackly artifacts and warbling. You can experiment with this here.

    That said, most real music has very little of this kind of information to contend with. Natural instruments are primarily sinusoidal (though they can have ultrasonic overtones); and square waves from synthesizers are intended to be...weird. Coloration of transients still occurs, but that isn’t something that people get in a huff over. Actually, year ear is rather insensitive to these distortions because it’s doing its own Fourier decomposition of sorts. So, all said and done, basically fine to use Redbook as a standard.

    But all the above is subject to the caveat “in theory.” This goes problem (2) and to @Lyander’s comment, above.

    In reality, your DAC doesn’t read the entire data-stream and perform a simultaneous transformation to generate the analog acoustic signal—as the mathematics does. Instead, it attempts to convert a discrete-time signal into a continuous-time signal bit-by-bit. It does that (initially) by holding the voltage from the previous sample until the next sample arrives. What results looks like a staircase going up and down in an approximation of the original acoustic wave. To smooth out the steps, we use “reconstruction filters” that are essentially attempting to fit curvy waves to our staircase. Here I am skipping a metric shit tonne of details, like aliasing—but suffice to say filtering doesn’t come for free. We threw away some information when we sampled the original signal at discrete points in time and we do not get to have it back now. That’s a law of physics. The filters used create “ringing” artifacts (swoosh sounds to transients) and phase (i.e., time) ambiguity/distortion, where the low frequency sounds may come out with a delay relative to the higher frequency sounds. Here lies the great debate about minimum-phase, linear-phase, and MegaComboBurritos. Here also is where new technology keeps making things better.

    Turns out that, generally, humans are also remarkably insensitive to time shifts. Partly this is because the room does it for you. Partly because the rest of most people’s stereos are also f'ing with the time signals. BUT... you can indeed hear these errors as off timbre and squishy transients.

    What to do? All these problems are lessened as you sample faster and faster—think of it as your discrete-time sampling is getting closer to continuous sampling. And that is the core argument for high sample rates. Indeed, some newish DACs actually operate the reconstruction internally at rates well above the original sampling rate to help fix some of these artifacts, as @Psalmanazar points out. These kinds of tricks and their equivalents are getting VERY good. Before these features existed, having more samples of the original signal was the old way to help minimize the DAC’s reconstruction artifacts—and theoretically still a better way since you’re using more real data and less fake/interpolated data. But reconstruction methods have gotten so good, I’d say it’s not obviously worth it anymore ... except for maybe in exceedingly well architected sound reproduction listened to by very discerning people. Like us, of course. ;-)
     
    Last edited: Sep 12, 2019
  18. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    10,961
    Trophy Points:
    113
    Location:
    Philippines, The
    @skem always happy to get input from someone who knows more than theory (and actually deals with the matters at hand), haha.

    So the TL;DR is, unless I'm misunderstanding again, that higher sample rates reduce errors that manifest as floofy transients (among other things), but the math on modern filters are getting to the point where these errors are now largely negligible, and seeking any further refinement is for the sake of fixing that last fraction of an iota of guesstimation?

    Yup, definitely for a more discerning crowd.


    Was about to post this before you replied but opted not to because it's noise, figure I might as well sneak it in now:

    @Psalmanazar LOL! I was actually planning to tag you for poops and giggles because I wondered what your stance on the topic was, as relevant to your experience with the same. Imagine my surprise when you replied regardless.
     
  19. skem

    skem Friend

    Pyrate
    Joined:
    Nov 1, 2017
    Likes Received:
    1,911
    Trophy Points:
    93
    Location:
    Charles River
    @Lyander yes... you got it. Obviously this is about chasing down more and more precision. Whether that precision is worthwhile is entirely subjective. I, for one, find virtually all DACs to have unacceptable reconstruction of Redbook piano. But I envy the person who can’t tell or doesn’t care.
     
  20. Peter Apperup

    Peter Apperup Acquaintance

    Joined:
    Sep 30, 2016
    Likes Received:
    11
    Trophy Points:
    18
    Location:
    europe
    this is all a bit much for me but again that is why i'm a member of this forum; to learn......anyway when is it good yo upsample? How does that benefit the audio?
     

Share This Page