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

Hugo M Scaler by Chord Electronics - The Official Thread

Discussion in 'High-end Audio Forum' started by ChordElectronics, Jul 25, 2018.
First
 
Back
606 607 608 609 610 611 612 613 614 615
617 618 619 620 621 622 623 624 625 626
Next
 
Last
  1. racebit
    The sync issue only exists in FPGA. With a CPU/GPU no sync is required, we just need to give enough delay (with safe margin) to play. So when it is played it has been calculated and ready for some time, it doesn't matter if it was calculated one second before or after the other sample.
    Unlike CPU/GPU, with FPGA there is limited memory, the data has to go out for the process to continue, so sync is crucial.
     
    miksu8 likes this.
  2. miketlse
  3. delirium

    Yes i know @mlxx[/USER has made some interesting posts!!! Ceep it up
    :relaxed:
     
    Last edited: Nov 10, 2019
  4. miketlse
    https://www.head-fi.org/threads/chord-electronics-dave.766517/page-273#post-12778587
    https://www.head-fi.org/threads/cho...-official-thread.831343/page-28#post-13472763
     
  5. racebit
    I think the difference is that HQPlayer processes only the samples required to meet the frequency domain criteria, while Rob goes much beyond that as he aims for time domain accuracy for transient accuracy.

    This is what Rob said here http://www.the-ear.net/how-to/rob-watts-chord-mojo-tech :

    "it is a fact that if you had an infinite tap length FIR filter you would perfectly reconstruct the original bandwidth limited analogue signal in the ADC. It would make no difference if it was sampled at 22uS or 22pS you would have the same digital signal. But it was very clear to me that having a limited tap length would create timing errors. And I know from my studies and from my own listening tests that that would be a major subjective problem.
    Now, unfortunately, nobody else has recognised this problem for two reasons. One is electronic engineers do not study hearing, and the second problem is that they are stuck on the idea that filters are a frequency domain problem and not a time domain problem. So if you design a filter where your only concern is frequency, then a 100 taps or so is enough. But if you think from the timing perspective, it categorically is not enough. What I have done is to make no assumptions about whether something makes a difference to the sound unless I actually do a listening test. And listening to increasing tap lengths always improves the sound quality. "


    Now HQPlayer developer has this to say here https://audiophilestyle.com/forums/topic/19715-hq-player/page/638/#comments :

    Question: "I thought perfect interpolation requires an infinite length (time) sinc, so infinite processing and infinite delay. Of course that is not possible, but doesn't a higher number of samples used on the sync, increase the output resolution (real bits of each sample)? If that was the case, more processing would result in better quality. Or is it not noticeable?"

    HQPlayer:
    "It is not quite that simple. Theoretically yes, if you want to reconstruct up to exactly Nyquist (fs/2) freqency and you are also OK with infinitely bad time domain response.
    Practically, if we begin with source content, it is certainly not relevant and proper up to Nyquist. In fact, depending on ADC/mastering tools, there can be quite a bit of aliasing band at the top of the frequency band. You wouldn't want to reproduce that distortion part. That's why we have apodizing upsampling filters, for example in HQPlayer, to clean it up. In addition, even with RedBook, there is some amount of margin between wanted frequency band (let's assume 20 kHz here) and Nyquist frequency (22.05 kHz). So we have 2.05 kHz wide transition band available between wanted frequency band and Nyquist.
    So in first place you need to consider what was used to produce the content you are trying to reconstruct, because the source content is certainly far from perfect. So instead of assuming perfect content up to Nyquist, you need to consider imperfect content up to Nyquist and what you can make out of that."
     
    zenlisten, kennyb123 and ZappaMan like this.
  6. dmance
    @racebit
    IMO, the latest HQPlayer 4.2.0 'poly-sinc-long-lp' enters the realm where it is indistinguishable(*) from WTA1. It's now available (on HQPlayer website) for any HMS owner to download and test using USB for A/B comparison (A: set PCM Mode, DAC bits=24, filter=poly-sinc-long-lp, dither=none, rate=768k (HMS does pass-through) or B: set filter=none so HMS does the upsampling ). If you run the test on a x64 CPU then Jussi uses 80-bit extended precision floating point. The SQ is unlike anything else.
     
    Last edited: Nov 10, 2019
  7. racebit
  8. racebit
    Need to test it. Not able to compare to mscaler though.
    Tbh, one thing that puts me a bit off about HQPlayer is it seems to focus much more on DSD than on PCM. But that is my personal bias.
     
  9. racebit
    @dmance I never heard a mscaler, but this post makes very careful about short period A/B comparisons.
    I am mentioning this for my comparisons, as you have used mscaler for long time, so are able to judge correctly.
     
    Last edited: Nov 10, 2019
  10. miketlse
    Thanks that's interesting info.
    I have read Robs posts about 'the second problem is that they are stuck on the idea that filters are a frequency domain problem and not a time domain problem' whilst I have been trying to understand the issues with DSP, and it is a key reason why I believe that it is important to keep all the samples in sync.

    However looking at things from another point of view, it is obvious that the frequency domain problem thinking built into off the shelf dac chips proves perfectly satisfactory for 99.99xxx % of use cases for dac use in phones, TVs, radios, etc, so cannot be simply dismissed as 'wrong'.
    The remaining fraction of a percent of audiophiles who want the most realistic reproduction of music, are well served by Rob and his time domain problem approach, but delivering a commercial product to customers (especially those who are not interested in coding, or maths) is clearly not easy or cheap.

    It is just my personality to be interested in 'understanding the problem', rather than selecting a solution even if it doesn't meet my needs, so I am planning to continue trying to understand for a few months yet.
    :slight_smile:
     
    zenlisten likes this.
  11. Uncle Monty
    "The included 15V 4000mA power supply simply plugs into the Hugo M Scaler."
    from the manual
     
  12. nomad777
    i think I'll go with the designers recommendation 9v or 12V
     
  13. Uncle Monty
    I'm confused (again) - does this mean not using the power supply provided?
     
    bikutoru likes this.
  14. miketlse
    The recommendation will always be use the supplied power supply, if you need to use mains power.
    However it is clear that for alternative scenarios, especially when travelling that the options using battery packs are perfectly viable.
     
  15. nomad777
    no with batteries use 9v or 12v don't see why your confused with the mains you don't have a choice
     
First
 
Back
606 607 608 609 610 611 612 613 614 615
617 618 619 620 621 622 623 624 625 626
Next
 
Last

Share This Page