1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.

    Dismiss Notice


Discussion in 'High-end Audio Forum' started by magiccabbage, May 14, 2015.
731 732 733 734 735 736 737 738 739 740
742 743 744 745 746 747 748 749 750 751
  1. rayl
    I asked a similar question several months back when deciding whether to go all-in on Chord and add a Blu2 or not. (I did.)

    My question was, to paraphrase -- a pure DSD recording... with pure DSD processing... how can anyone argue that going to PCM and then to DAC will be better than a good DSD DAC based on the merits of PCM over DSD?

    I believe there were some replies around the issue, but I would have to say the answer is -- if you are in pure DSD land, then that is a tough argument to make.

    However, other than specialty productions targeted at the SACD market, most other commercial music has been processed in PCM on a DAW already, so going back to DSD would have the detriments you quoted.

    In my research, I could only identify two DAWs that process in pure DSD -- Sony (father of DSD) Sonoma and Merging Pyramix. Every other DAW I research may have DSD capability but actually convert into PCM.

    So I concluded that unless I go out of my way to find a demo SACD or demo DSD file, none of the music I actually listen to avoids an existence in the PCM world.... Thus, a system optimized for PCM works just fine for me!
    GryphonGuy and miketlse like this.
  2. ecwl
    I think that's one way to look at it. But as I use my Chord DACs more and more, and re-read what Rob Watts said and thought about it, I would present a different argument.
    If most music is recorded in PCM, then why don't we all play music using ladder DACs because those are the values that the PCM recordings got.
    Except that is the wrong way to think about DSD or PCM... DSD or PCM in the recording space is supposed to sample the analog music. The role of the DAC is not to play the samples back but to re-construct the original analog waveform. That's why Rob Watts doesn't program his Chord DACs just to a non-oversampling filter and play the PCM samples back at 44kHz. He uses as many taps as possible to upsample the PCM signal into much higher frequencies to reconstruct the original analog waveform.
    Similarly, although it always sounds like playing DSD signals natively using a DSD/PWM DAC would be the most natural thing to do, because you're avoiding additional noise that would be generated from additional electronics, it is also suboptimal in some ways. In an ideal world, we would want to be able to take the DSD sampling signals and be able to upsample them and reconstruct the original analog waveform at higher frequencies. It is also why many DSD DACs would actually upsample standard DSD64 signals to DSD128 or DSD256 for playback. So in my mind, DAVE DSD+ mode/Blu2/Hugo 2 are doing the same thing with DSD files using the Chord DAC technology.
    onsionsi and Em2016 like this.
  3. Rob Watts
    I like this post, as I hadn't thought of the DSD filtering that I do in the same way. Of course, the WTA filter is trying to reconstruct the original analogue signal, and the same is true with my DSD filters - trying to get back to the original analogue signal without the hideous out of band RF noise and nth order harmonics distortion (and anharmonic too) that DSD introduces. The benefit the PCM has though, if it is an ideal bandwidth limited signal, then an infinite amount of processing will perfectly recover the original analogue waveform in-between the samples. Not so with DSD - the timing errors with amplitude can never be repaired, as information is permanently lost by the DSD noise shaper(s).

    A common strand with audio is to make things as simple as possible. And in analogue design, this almost always works - a simple number of components will sound more transparent than a larger number (assuming that everything else is equal, noise floor modulation and distortion of course). But the digital domain is different, in that vast amounts of processing can radically improve the sound quality (assuming of course that the data path is capable of perfectly reproducing -301dB or better than 64 bit in-band accuracy!)

    In Hugo 1's time, I got a lot of flak for admitting that DSD was filtered. Now the original DSD filter was good, in that it gave better SQ - but with the DSD+ mode I realised the original filter was not good enough. Hence the Hugo 2 DSD filter, which is even better in it's performance; so much now that I am not so vigorously against DSD as before. But I still easily prefer redbook, and you can't get away from the fundamental and huge technical problems that DSD has - problems that can't ever be eliminated.
    Chord Electronics Stay updated on Chord Electronics at their sponsor page on Head-Fi.
    https://www.facebook.com/chordelectronics https://twitter.com/chordaudio http://www.chordelectronics.co.uk/
    Aslshark, onsionsi, miketlse and 4 others like this.
  4. marcmccalmont
  5. doraymon
    Fellas, a little help to properly set up my Roon connection with Dave.
    Roon Core is running on a Mac mini connected via ethernet to the router.
    Dave is connected to my MacBook Pro (Late 2013) through a Sys.Concept Toslink to Mini cable. The MacBook is running on battery and is connected to the network via WiFi.

    I managed to "force" Roon to recognise the Dave in the Device Setup / Audio Device menu, so at list I got the right icon with Dave's silhouette!
    Please have a look below and let me know if you have better suggestions.

    Device Setup:
    - Private Zone: NO
    - Exclusive Mode: YES
    - DSD Playback Strategy: DSD over PCM v1.0 (DoP)
    - Volume Control: Fixed Volume
    - Volume Limits: Volume is Fixed
    - Resync Delay: 0ms

    Advanced Device Setup:
    Max Sample Rate (PCM): Up to 192kHz (this is the highest available on the menu, is that only because of the optical connection or I'm doing something wrong?)
    Max Bits Per Sample (PCM): 32
    Zone Grouping Delay (ms): 0.00
    Clock Master Priority: Default (what is that???)
    Enable MQA Core Decoder: YES
    Enable Integer Mode: YES
    Force Max Volume At Playback Start: YES
    Use Maximum Buffer Size: NO
    Use Power-of-2 Buffer Sizes: NO
    Multichannel Mixing: Downmix As Needed
  6. Jawed
    Isn't it decades since music ADCs changed to be based on delta-sigma (sigma-delta) or multi-bit delta-sigma?

    DSD's sample rate, when it was introduced, seems to have been too low. If it had been 4x, would we be having this discussion?
  7. marcmccalmont
    I think the point is delta sigma (dsd) is flawed in its basic concept, not that it doesn’t sound good it does but it can never be as accurate as pcm. PCM is like building a car by taking every measurement from a reference line, if your precision is .001in every sample is accurate to .001 in so your car is built to a tolerance of .002 in, not bad! If every part is measured from the last part after 1000 measurements your car is accurate to about a foot! Hmm one car is 15’ long the next is 17’ long not so good. My understanding of the development of delta sigma DACs was driven by Japanese companies that did not want to pay royalties to Burr Brown who held the patents on PCM
    Last edited: May 5, 2018
  8. rayl
    Yes to the ADC part. Hence, my earlier reservation. But the total number of pure DSD (through post-production, through the file/CD I have) songs I have based on information I can find on the jackets, etc... well, one demo SACD from PS Audio produced on Sonoma.

    There were more when SACDs were more heavily marketed, but as I am one of those primarily "modern" music listeners (how many plugins are avail on Sonoma vs on more prevalent DAWs?), I can defer further exploration and be happy with the best PCM system there is!
  9. Mython Contributor

    Not being 100% serious, here, your above comment reminded me of the so-called 'Serenity prayer' popularly adopted by AA:

    Then again, others contend:

    Personally, I'm very content with Redbook recordings, and I hope when the day comes that I can afford one of Rob's TOTL DACs, I won't have a reason to change my mind about Redbook.
  10. Em2016
    Having spent
    Everything looks perfect. Those are the settings for bit perfect playback.
  11. Jawed
    For making a recording, at say 11.3MHz sampling rate (256x 44.1KHz)? I doubt that's true.

    All modern recordings are derived from multi-bit delta-sigma ADCs as far as I can tell, though getting evidence for the actual functioning of the ADC (not the format that it outputs) seems to be quite hard.

    Do you mean that single-bit delta-sigma is flawed? Regardless of sample rate?

    Has Rob built a 24-bit 16FS PCM converter for Davina that doesn't start with delta-sigma conversion from analogue?
  12. marcmccalmont
    I’m not a EE or a digital expert but in my way of thinking the most accurate systems reference a common point whether 0 voltage or a centerline in a drawing this is basic engineering that I was made aware or in my first year as a ME student. Any system that references a previous measurement or sample is inherently less accurate. How multibit delta sigma overcomes this basic premise is beyond my math but I assume it is a technique to push noise and distortion higher in frequency but does not solve this fundamental issue?
  13. Rob Watts
    There is a fundamental difference between a 1 bit delta sigma and an N bit delta sigma. The problem is the noise shaper for DSD has levels of -1 and +1. Now when it's reproducing 0 it is fine - the output oscillates between +1 and -1. But the instance that it goes slightly positive, or slightly negative, then the natural sequence is for the noise shaper to occasionally request +2, which is not allowed. As the signal gets more positive, then the desire to go to +2 gets larger; eventually it goes unstable. So the noise shaper behaviour is constantly signal dependent. But an N bit noise shaper has no such limitation as the behaviour of the noise shaper is independent of input signal. The other issue is switching activity; a fundamental flaw of DSD is the switching activity is also signal dependent; with an N bit noise shaper, it's possible to design the activity rate to be signal independent. Again, this is a fundamental flaw, as noise in the DAC which is used in the ADC return path is signal dependent as switching activity modulates the OP - and this in turn creates noise floor modulation. Both these situations exist irrespective of OP frequency.

    Another couple of issues - timing of transients is amplitude dependent (small step changes have different times to big step changes) and it's impossible to achieve the required 350db noise shaper performance that my listening tests have shown is needed for depth perception. Both of these issues are solved by an appropriate use of a noise shaper architecture and N bit depth and noise shaper output frequency.

    Davina is a N bit delta sigma with a pulse array reference DAC for the ADC return path, and of course has none of the above issues - 400 db noise shaper performance, delay times independent of signal amplitude, and constant noise shaper activity rate, plus pulse array's signal independent switching activity. All these issues will ensure no noise floor modulation, and accurate depth reproduction... You can't do this with DSD at any OP rate.
    Chord Electronics Stay updated on Chord Electronics at their sponsor page on Head-Fi.
    https://www.facebook.com/chordelectronics https://twitter.com/chordaudio http://www.chordelectronics.co.uk/
    Siz., 514077, onsionsi and 6 others like this.
  14. Jawed
    So the pulse array in Davina has better noise-shaping than seen in DAVE? More than 20 elements? It's running at 2048FS?

    Presumably there is an opportunity with Davina to use a better WTA filter for upsampling 16FS to 256FS. But you then risk adding more latency in the return path, I suppose.

    Alternatively, I suppose there's an opportunity to upsample from 16FS to 1024FS with a WTA filter, instead of only up to 256FS?

    Now playing: King Creosote, John Hopkins - John Taylor's Month Away
  15. Em2016
    I've whipped out my Frankenstein setup for the weekend - battery powered microRendu v1.4 and battery powered optical-to-ethernet converters feeding the microRendu.

    Battery powered doesn't mean Uptone LPS-1. I mean proper battery powered.

    Although it's a pain in the a$$ to setup, this is even better than my battery powered wireless USB source.

    Dave is driving my AEON closed (thanks for the recommendation @Rob Watts ) directly, so there's no leakage loops involved.

    Both are outputting bit perfect playback. Since both USB sources fully break leakage current loops, there's no RF to worry about.

    But the battery powered microRendu still sounds smoother. It's only a difference I can pick after a couple hours of listening.

    I can only imagine it's the superior signal integrity of the microRendu versus the iPod Touch.

    I thought I knew this album back to front, inside out, but have never heard it like this:

    Screen Shot 2018-05-06 at 4.10.45 pm.png
    Last edited: May 6, 2018
    flacre likes this.
731 732 733 734 735 736 737 738 739 740
742 743 744 745 746 747 748 749 750 751

Share This Page