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.
434 435 436 437 438 439 440 441 442 443
445 446 447 448 449 450 451 452 453 454
  1. simorag
    Thanks for the hint. Just wanted to clarify that I was not implying in my post that I have glare issues in my current setup, only that the ISO Regen does not affect (for better or for worse) the treble smoothness / digital glare in my system at its present stage.

    Anyway, I am not saying that my front end is fully devoid of digital artifacts. Sometimes you discover that you have noise in your system only when you lower that noise :)

    I would be curious to listen to the Stream cables in my system, given the very good feedback from other HMS users, sometimes in the future.

    I am also investing in better power cables, interconnects and grounding devices … oh my, it never ends!
  2. Uncle Monty
    WAVE Storms made my set-up sound noticeably 'darker' - Rob Watts stated somewhere that a darker sound means the upgrade has succeeded, so I'm going with that. It does sound better, though, smoother and darker...
  3. STR-1
    I found the Storm to be darker than the Stream. But, bucking the trend of accepting darkest as best, I found it to be too dark (just) for my liking with the tonal balance of my system as it is at the moment. I might give the Storm cables another try at some point.
    Uncle Monty likes this.
  4. ray-dude
    I am waiting on one final piece to the puzzle to repeat my experiments end to end, taking away various upstream tweaks to see if they are still needed (soon...soon).

    For now, with my NUC end point, I hear huge SQ improvements when running Squeezelite as the end point software (with large buffers) vs Roon Bridge. SL is VERY configurable and open source, which makes it a good experimentation platform. With gratuitously large input buffers (the current and next track are loaded into memory in a second or so) and gratuitously large buffers to the Linux ALSA sound driver (buffer between SL and USB driver basically), SQ is amazing. Reduce the buffers, and SQ becomes exactly the same as Roon Bridge.

    Practically, running SL with large buffers is extremely flakey: skips, playback seizing up, etc. impractical for civilians, but a great test bed for what is possible

    With one ISO Regen the gap between SL with large buffers and Roon bridge narrows significantly. With two in series, it almost disappears. I have not yet gone back to a generic unoptimized PC to see how close that gets to an optimized NUC. I’ll be doing the full test ladders once the final downstream tweak is back in my setup

    I have tested the ISO Regen with stock SMPS to AA battery packs (no mains connection but poor voltage regulation) to laptop battery pack (no mains but ok voltage regulation) to LPS 1.2 (no mains but great voltage regulation) to see what power aspects could be most impactful.

    The LPS1.2 was clearly the winner. Based on that I have a Paul Hynes SR4 (even better voltage regulation) on order, while I wait (and wait and wait and wait) for my 3x dual regulated Paul Hynes SR7 (ultimate voltage regulation). If I can, I’m also going try to track down a txusb-ultra to test as well

    Lot more experimentation to be done. It just gets harder and harder to motivate myself to do the experiments as the music gets better and better.
    TheAttorney likes this.
  5. Uncle Monty
    So here's a question - if the Mscaler re-clocks everything and you use optical to cut out electrical noise, so long as your server / streamer / renderer whatever can handle your FLAC or WAV files, does it matter what server (from a pure sound quality perspective)?

    Will a decent portable music player or laptop sound as good as a Cambridge or Sony server sound as good as a top of the line Antipodes CX when played through the M-Scaler via optical?
    Siz. likes this.
  6. Triode User
    Jay Luong at AudioBacon ( @bacon333 ) agrees with you and he says he prefers and uses the Stream in his system instead of the Storm. I have told him he is wrong and that the darker sound is an indicator of least RF noise but it is a free world and we have agreed to differ! Haha, :stuck_out_tongue:
  7. musickid
    My belief is that if the optical source is bit perfect then the type of 'transport' becomes more or less indistinguishable. However for usb the source 'transport' may affect the sound as widely reported.
    Uncle Monty likes this.
  8. Gibson59
    I’ve come to the point that with the exception of dealers being able to provide simple explanations of products, I always take their opinions with a grain of salt and I’ve found that I often know as much about the products as they do or I trust my own ears better than theirs. It remains that no one can tell you what you will or won’t like the sound of, you have to decide from your own learned sound preferences and/or demo stuff. The dealers may know product specs but they can’t tel you what sounds “best”.

    It’s a shame for those who don’t spend time on forums doing their own research and using the various resources we are able to source to properly demo gear before buying. They’re just left to the dealer’s opinions which is a risky proposition when spending thousands of dollars unless it’s a really good dealer with a trustworthy demo/return policy.
    Last edited: May 2, 2019
    Uncle Monty likes this.
  9. dmance
    All digital transports are bit perfect. TOSLINK,SPDIF, USB all deliver the bits right into the heart of Rob's DAC completely error free.
    All we are doing with all our myriad tweaks is help the final conversion in the DAC from bits to volts to be as linear as possible. This means follow the metal and squash noise on those paths. Defy your incredulity. Use battery vs AC...and even then, regulate the supply to ultra low noise. Always use optical vs metal conductive signals. Avoid external amplification to muddy all the benefits you just made. Get better headphones/speakers to reveal more of the transparency you've uncovered ...and keep looking for more.
    You will be rewarded immensely.
  10. musickid
    I agree totally. But there must be a point for which you decide to stop and just enjoy the music. Otherwise it never ends.
    Uncle Monty and xxx1313 like this.
  11. Uncle Monty
    if using TOSLINK can any electrical AC noise or other noise travel along the optical cable into the M-Scaler?
  12. musickid
    No it's galvanically isolated. It's transmission of data by light pulses so it's impossible for electrical 'noise' to get in. By the very laws of physics.
    Uncle Monty likes this.
  13. Uncle Monty
    so would swapping out my AK music player for a dedicated server make any difference in sound quality?
  14. Rob Watts
    The FPGA is rated at 85 deg C - in that the timing tools are such that worst case it will still function perfectly well. The trip is set so that outside case temp will not burn one, so that is 65 deg C. You would need ambient temperatures of 55 deg C for this to occur; you won't want to listen to music at that temperature!

    Take a 44.1 recording against a 176.4. A 44.1 will sound better via an M scaler than the 176.4 that is not M scaled; but of course an M scaled 176.4 would sound better than an M scaled 44.1, as the decimation (the conversion to 44.1) isn't very effective, as it has lots of aliasing.

    And it's not because the M scaler creates more information - and I mean information in the mathematical sense - it's just that it recovers the timing of transients much more accurately, and it's the better accuracy that leads to less uncertainty in transients that is important.

    To show you the transient reconstruction issue, take a look at the slides below:



    So this illustrates how conventional audiophile filters will mess up the transients; indeed the peak error with these filters is huge - it's one bit accurate! With the M scaler, it is identical to an ideal sinc function to better than 16 bits - this means it will reconstruct to 16FS at better than 16 bits. If we use a sinc function filter, we will perfectly reconstruct the bandwidth limited analogue signal in every single regard - it would then make no difference if it was sampled at 22uS (CD) of 1 pS - the result would be the same, if its bandwidth limited to 22 kHz.

    So we can actually think of conventional filters as destroying transient timing information, because the interpolation filter simply isn't accurate enough - it's not that the M scaler creates new information (in the mathematical sense) it just degrade it like conventional filters.

    That post explains it well. I would clarify that the 4 or 6uS resolution applies to the interaural delay, and this is used by the brain for left right localisation. We don't have any numbers about the other aspects of transients - such as bass perception, timbre, instrument separation perception etc - and I suspect that actually very tiny uncertainties in transient recovery is important. I now know we need to be talking about nS and not uS errors.

    However I would argue with you about other designers. No one else has made the link between transient recovery accuracy in timing and interpolation filters - they just go on about lack of ringing, which of course is completely the wrong idea and will actually degrade timing recovery and create more errors. And when you get more errors - the brain can't make sense of the music, and everything sounds soft and out of focus. Which of course superficially helps with a hard bright sound from endemic DAC noise floor modulation (something which again no other designer also does not talk about either). But two wrongs don't make a right...

    Actually I only recently characterised it as this - the tempo sounding slower. I heard it as more natural flow, with less dominance effect - that is with one's attention constantly switching from the loudest instrument to the next loudest instrument. Hence tempo sounded jerky or mechanical before the M scaler; with it, it flows much more naturally. The technical explanation is that with smaller tap lengths the transients that are created are modulated by program material, so transients are subtly jumping backwards and forwards in time depending upon other instruments. Reduce this effect and dominance and tempo becomes much more natural and less mechanical sounding. Another impression that is down to this error is instrument power; when transients are being less modulated by instruments, you get a much stronger sense of individual instrument power.

    It often amuses me when the sound science anti listening faction complain about not being possible to hear these changes. Of course, it's easy to hear these things; but the real problem is not that one can hear the difference, but that people draw the wrong conclusions about whether something is better or worse - this is the real problem. In the case of optical against BNC/coax it's often stated that optical is worse because coax has lower jitter. This myth actually isn't always true - my previous AP had lower jitter measurements from optical than from BNC (my current APX555 has slightly better coax now). Anyway, the jitter from optical is tiny - about 1 nS - and of course my DPLL means that it actually doesn't matter at all, as the local clock is used. But when people hear optical against BNC the optical sounds a lot darker and warmer - and it's nothing to do with jitter, just RF noise from the electrical connection creating more noise floor modulation, which artificially gives the wrong impression of things sounding more transparent.
    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, miksu8, darkless and 8 others like this.
  15. dmance
    Photons do not carry an electrical charge ..so toslink is 100% galvanic isolation. Further, optical stops any leakage current flow due to ground loops.

    Why can toslink sound different from other sources? I don't design DACs but know enough about digital audio to comment:
    1) it's not jitter because Rob's DACs always reclock inputs to align with his super stable output clock. This process is pretty straightforward logic and likely the 'extra' effort by an FPGA to manage this does not cause more internal RF noise.
    2) Toslink is optical SPDIF that gets modulated back to electrical...and even though SPDIF has zero DC bias, this process could pump noise into the ground plane. maybe.
    3)SPDIF on coax is not galvanically isolated and skips the optical demodulation but then goes through many of the same input circuits as toslink.
    4) USB has many of its own RF noise problems and perhaps the induced brightness is false detail (as warned by Rob)
    5) switching inputs on MScaler changes signal paths through the FPGA and (perhaps) causes changes in the character (frequency, pulsations) of the internal noise. That's possible.

    Any and all the issues described above that manifest RF noise have to get to the DAC. MScaler is not a DAC, it's a digital-to-digital box that cares not a whit about RF noise for its own operations.
    Concern yourself with where the conversion to analog happens. ...at the D/A. Our ears don't hear bits. We hear moving air moved by a vibrating transducer, which is a modulated in dimension and frequency by an analog signal. We hear the analog signal.

    Try this:
    Robs amazing digital technology says the output voltage at time stamp T should be 1.000 volts,. say ...and at T+1msec it should be 1.025 volts. Add just a smidgen of RF noise to the ground plane or reference voltage and those values become 1.0001 and 1.024. ..and we can discern these variations.

    Hope this helps.
    wswbd, ray-dude and Uncle Monty like this.
434 435 436 437 438 439 440 441 442 443
445 446 447 448 449 450 451 452 453 454

Share This Page