USB Nervosa Thread Decrapifiers, pro interfaces, and bears oh my

Discussion in 'Digital: DACs, USB converters, decrapifiers' started by zerodeefex, Sep 28, 2015.

  1. Cspirou

    Cspirou They call me Sparky

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    8,200
    Trophy Points:
    113
    Location:
    Northwest France
    Reading about all these regenerators
    for Schiit DACs with adapticlock seems strange. If regeneration gives that much of an improvement then what exactly is the adapticlock doing?

    On a similar note, how much does the Gungnir or Yggdrasil level the playing field when it comes to sources? Like if you hear a major difference between a chromecast and a better source on a low end DAC. Are those differences minimized on the Yggdrasil?
     
  2. Sanlitun

    Sanlitun Acquaintance

    Joined:
    Sep 4, 2016
    Likes Received:
    47
    Trophy Points:
    18
    Location:
    Trudeauland
    I gather that's why they pulled the update off of the page.

    The firmware went without problems on my F-1 and in my case brought a little bit more treble control and smoothness. I felt it was a slight step up.
     
  3. Torq

    Torq MOT: Headphone.com

    Pyrate
    Joined:
    Jan 15, 2016
    Likes Received:
    8,193
    Trophy Points:
    113
    Location:
    Seattle, WA
    As mentioned earlier, once you're not using USB the relative differences become a lot smaller. All USB 2.0 Audio Async DACs must do some level of internal clocking/re-clocking, where as with AES and S/PDIF this is not strictly required. USB issues are much more commonly down to noise on the wires causing issues with the sensitive analog circuits in the DAC than any other factor. In general, it doesn't matter how good your USB clock is feeding the DAC, since its a packet-based delivery system into a buffer. The DAC reads from the USB buffer using it's OWN internal clock, which is COMPLETELY divorced from the source clock.

    Excepting issues with the USB receiver emitting more electrical noise with a "less accurate" source clock (.e.g. the typical "makes the receiver work harder" explanation), which is not necessarily the case anyway, the accuracy of the input clock has no bearing on the accuracy of the clocking/jitter to the actual converter ICs/boards.

    For non-USB interfaces ...

    Many DACs , including Gungnir and Yggdrasil have excellent built-in re-clocking. A DAC using AES or S/PDIF interfaces is highly dependent on the source clock in determining jitter performance, unless is performs internal re-clocking. And, once it does, it is the benefits of better external clocks diminish markedly. In most cases, with those two DACs, as long as you're not using a source that causes the "buy better gear" LED to come on, you'll be lucky to discern a difference between sources (as long as we leave USB out of it). And when you can, its certainly not as big as many seem to want to claim.

    Out of interest, the one time I tried a Chromecast with Yggdrasil it did light up the "buy better gear" light, which was as far as I bothered to test with it, as I'd literally never seen another source do that.

    I, personally, didn't find a reliably discernible difference between using a RedNet 3 interface (the primary benefit of which was having a dedicated network-addressable AES interface, rather than some USB-to-AES DDC box ... i.e. keeping USB entirely out of the audio chain), and then adding a Mutec MC-3+ USB into Yggdrasil. Other have had different experiences, but there are also other differences in our respective chains.

    I HAVE seen differences doing this with some other DACs.
     
  4. drez

    drez Acquaintance

    Joined:
    Sep 30, 2015
    Likes Received:
    51
    Trophy Points:
    18
    Don't forget PHY noise - this is the only mechanism I have come across that explains why USB signal quality and clocking could possible matter. I guess the best way to isolate that is the clock in the USB controller - a lot of people seem to find that it matters. Also say with USB cables - I'm not 100% on what causes the differences there and if that can solely be explained by noise, but that is not something I have evidence on either option. One could easily claim the differences are due to shielding or that they are due to signal integrity and not be able to provide evidence as to which of these is valid (or if both are valid). Just rationalising. I would still use the product that sounds the best, and IME those products tend to treat the USB digital pair as high frequency transmission line with maximum bandwidth. Techniques to lower noise while keeping this high bandwidth also seem to help. Again just my rationalisation, but based on experience.

    USB, with some polishing, it very good option. I don't think it's reasonable to expect to use noisy computers with high end audio gear without some work being put in. I think if more DAC used reclocking USB inputs with galvanic isolation, this might be even better. That way the isolated side uses very little power - on my F1 this is like 40mA, while USB receiver uses a heap of power (I think like 200mA). Also, in general I think computer audio is exercise in turd polishing compared to disk transports, vinyl etc.

    If I2S lines were better, we might see even more difference. I really don't like these ribbon cables.

    Adapticlock reminds me of the PLL lock range adjustment, eg like I have on my Pioneer XPA-700. It seems best if you can use a smaller lock range as long as there are no dropouts. Strangely this is similar to what I experience with computer software buffer settings.
     
  5. Torq

    Torq MOT: Headphone.com

    Pyrate
    Joined:
    Jan 15, 2016
    Likes Received:
    8,193
    Trophy Points:
    113
    Location:
    Seattle, WA
    Doesn't matter what the source of the noise is; it it's present, however, "cleaner/de-crapifier" devices that take USB in and then give USB out can only improve this so much as everything at the transceiver/PHY level will have it's own noise profile anyway. So, at best, the physical interface on the USB-out side of things is lower noise than whatever you're plugging in to the units input.

    All DACs with Async USB inputs are doing re-clocking.

    They may not all be using super-accurate clocks to do it, but they have to implement a scheme to provide a suitable replay sample-frequency clock to drive the converter ICs (or board for discrete implementations), and use that clock to synchronize the movement of sample data out of the USB receive-buffer and onto the I2S lines (generally) to feed to the converter.

    This means that, for any DAC that doesn't support an external clock-source, you can never get more accurate clocking than the DAC is using internally. You might "reduce the load" on the receiver and thereby reduce it's internal noise, but those effects are going to be minuscule.

    What tickles me is using heavy-handed multi-core USB solutions like XMOS vs. simple, super-specific, function-limited receivers. The more complex your device, particularly if we're talking high-rate signal switching (as found in any processor), the more noise it'll generate.
     
  6. drez

    drez Acquaintance

    Joined:
    Sep 30, 2015
    Likes Received:
    51
    Trophy Points:
    18
    Trues - I should have rephrased that sentence. I was referring to USB input cards that implement galvanic isolation after the USB receiver, and then re-clock the galvanic isolated signal.

    This seems to have been the prevailing logic a couple of years ago where we had all the USB/SPDIF boxes. Although I'm not sure if they needed to do that because SPDIF, rather than noise considerations. Although, even Berkely BADA and AES with isolation transformers seem to have used isolation between USB PHY and CPLD/FPGA with audio clocks. BTW does AES need pin 1? To be honest most of my previous experiments with lifting ground between components (which were designed to have ground connection on the IC) seemed to show an advantage of connecting the grounds as described in the spec. In any case it can't be too bad to separate the grounds from the USB PHY/computer to the DAC digital side? Also, IIRC Amanero implementations with galvanic isolation (and nice clocks) were generally preferred all things being equal (which they never really are).

    Actually that reminds me - apparently for audio clocks NDK are apparently better phase noise and stability than Crystek CCHD-575/950. I think quite a few newer USB cards are moving to NDK. Not my Singxer F1 unfortunately (they use Crystek 575s). Actually by the time I install it something better will probably come out...

    Actually maybe you can advise on how I intend to install the F-1 in my DAC? I was planning to power the CPLD and clocks from external very low noise shunt regulator (Salas R-D). The other alternative was to use the 3.3V from the DAC - likely higher noise, but arguably less chance of grounding issues. AFAIK even the external USB/I2S need ground connection, so I am probably overthinking. I can of course take measures to limit the ground connection to only the I2S patch cable to avoid ground loops.

    I'm not sure what is driving modern high power/functionality USB implementations. Maybe higher resolution formats and DSD? Lower latency? XMOS and Amanero both seem to use a lot of power - possibly Amanero using even more. I think the idea with new XMOS is that you can do away with the CPLD that Amanero uses. Strangely, despite the high power consumption, both of these seem to have a better reputation that other lower power USB solutions. Could also be the clocks I guess (Amanero and XMOS both use dual audio clocks).

    In any case I expect ANY high functionality I/O will produce a lot of noise. I expect NAA/ethernet cards also produce their share of noise, esp taking into account the power supplies on some pro-audio solutions are not really super low noise (nor that great clocks). Presumably the (alleged) performance advantage lies elsewhere?
     
  7. Torq

    Torq MOT: Headphone.com

    Pyrate
    Joined:
    Jan 15, 2016
    Likes Received:
    8,193
    Trophy Points:
    113
    Location:
    Seattle, WA
    Need? No.

    And if you do connect it, best-practice is to do so only on the source side.

    Preferred by whom?

    Not really - I've never used it and don't know what DAC you're running.

    Both the Amanero and XMOS solutions are SoC/SoB implementations that provide a complete, end-to-end, buzzword-compliant USB to I2S solution. They make it almost trivial to build a DAC. You can literally take one of their boards, a suitable PSU, a conventional CoTS DAC chip, and any of a number of "stock" I/V stages and have a USB driven DAC up and running in a day, maybe two (if you're familiar with I2S, the Amanero/XMOS boards, and how various DAC ICs interface and are configured).

    Even supporting specific converter ICs can be as simple a selecting a known configuration, adding a few pull-up/pull-down resistors to configure converter IC operation that isn't addressable in I2S preamble. There's a ton of functionality on those boards ... and it goes well beyond simply getting a buffer of samples via USB 2.0 Audio. They're the absolute opposite of a "simple-is-best" solution.

    It's their ease of use, and broad capabilities, that make them popular. In particular, you find them disproportionately often in "Chi-Fi" products because they dramatically reduce R&D and help satisfy the all-too-common tendency for such products to be accompanied by marketing copy of the "how many buzzwords, relevant or otherwise, can we quote and get away with" form.

    We moved DACs out of transports to isolate them from as much other electrical nonsense as possible. And then we take something like Amanero or XMOS and stuff what amounts to a full-blown computer not only right back in the chassis, but connected to the bloody I2S bus! Not that it's all bad necessarily, those solutions definitely have some merit, but they're much more complicated than the comparatively simple implementation that you need for AES3 derived interfaces (AES and S/PDIF).

    The level of functionality, complexity and switching going on in the Amanero or XMOS boards is orders of magnitude higher than in any NIC I know of. And since those are typically right inside the case with the sensitive analog electronics ... I would err towards the simplest solution if audio-quality, rather than ease of engineering, was my goal.

    NAI/AOIP solutions, pro, or otherwise, offer advantages on several levels. For one, they're fed with galvanically isolated, sometimes air-gapped, connections that have guaranteed delivery, absolute source-clock immunity (packets can arrive out of order, so there's no "clock" there in the first place), and ECC/re-transmit capability. Then they operate outside the DAC's chassis and typically output AES, S/PDIF (Optical and COAX) and/or ADAT (think TOSLINK connectors, but a different protocol), which are all, assuming they're implemented to spec (pretty much guaranteed for "pro" gear at least), galvanically isolated.

    Yes, pro devices rarely implement clocks as fussy as those in higher-end DACs, and they often have much noisier PSUs, but since they operate PURELY in the digital domain, that's not that much of an issue as long as they're properly isolated (and I don't know of any that aren't) from the DAC. And in most cases, the clocks are more than good enough since in pro-applications there will invariably be a higher quality master-clock in the setup, and they're not targeted at audiophiles. Audiophile DACs, the "seriously engineered" ones at least, implement proper re-clocking schemes anyway ... and it's that internal clock that ultimately governs the achievable jitter performance of the DAC in question.

    And then, of course, they offer greater source flexibility and the ability to locate them anywhere, not just within a few feet of the source.
     
  8. Luckbad

    Luckbad Traded in a unicorn for a Corolla

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    3,408
    Trophy Points:
    113
    Location:
    Holly Springs, NC
    While in my gear slump, I grabbed an ESI Juli@ XTe again. It's about the most expensive thing I can afford right now.

    Really solid for the price. Better than USB or TOSLINK from the PC.
     
  9. Wfojas

    Wfojas Friend

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    848
    Trophy Points:
    93
    Location:
    CA
    I also use a Berkeley Audio USB converter from a microrendu to an Yggdrasil, and it does make it sound better. I am feeding it from a microrendu, but also note that I could easily live without the BA USB if I didn't have it, as the BA effect is to make it sound more organic, where microdynamics are better heard, and the sound staging gains depth cues. If your system has prominent mid to upper bass, you may not hear the effect.
     
  10. Torq

    Torq MOT: Headphone.com

    Pyrate
    Joined:
    Jan 15, 2016
    Likes Received:
    8,193
    Trophy Points:
    113
    Location:
    Seattle, WA
    Out of pure, morbid, interest, what output are you taking from your Berkeley unit?

    It makes sense to me that microRendu -> USB to <anything else> DDC (e.g. the Berkeley), would be better than the raw output from the microRendu.
     
  11. Madra

    Madra Acquaintance

    Joined:
    Oct 18, 2015
    Likes Received:
    26
    Trophy Points:
    18
    A couple of friends compared the microrendu>usb>devialet 400 monoblocks to microrendu>usb>berkely>spdif>devialet 400 monos. They felt that the berkley lowered the noise floor, improved bass tightness and improved the sense of space. They saw it as a worthwhile upgrade.
     
  12. drez

    drez Acquaintance

    Joined:
    Sep 30, 2015
    Likes Received:
    51
    Trophy Points:
    18
    RE: XMOS - flexibility in USB is pretty important in my mind - it needs to work with range of OS, driver, computer, cabling etc.

    With AOIP you have PHY, microcontroller and FPGA? Seems not far removed from USB. Maybe this is better way to do things - I can't fully appreciate. I don't design electronics. If someone releases an AIOP I2S card I can install into my DAC I can compare.

    But, if I might turn the question back on you - which USB implementation sound better than Amanero or XMOS?
     
    Last edited: Dec 7, 2016
  13. Torq

    Torq MOT: Headphone.com

    Pyrate
    Joined:
    Jan 15, 2016
    Likes Received:
    8,193
    Trophy Points:
    113
    Location:
    Seattle, WA
    If you implement USB to spec, it'll work with any OS that supports that spec, using standard drivers and cables. There's no magic to it. Special purpose chips (ASICs), like typical USB transceivers, are much simpler than the XMOS core (which is a multi-core SoC, among other things), and are much less electrically noisy.

    However, they're also harder to implement if you've not had to built a USB interface before. Which is a roundabout way of re-iterating the XMOS solutions are less focused on "audio quality" than they are on making the engineer's life easier. Again, not to say they necessarily sound bad, but if you're just doing USB Audio there are vastly simpler ways to get the job done, that use quieter, simpler, cheaper components with less potential to inflict noise on the rest of the system.

    Depends on the AOIP in question.

    You'll always have a physical interface with any transport of this nature. Whether there's a "micro controller" (which, in the strictest terms, I would define as a general-purpose, programmable, controller) depends on the unit in question ... but you will have some kind of controller which may, or may not, be task-specific. Again, ASICs and special-purpose components tend to do a better job than general purpose ones. FPGAs are typically not needed for AOIP units, and using one would be down to the designer's preferences/needs.

    For example, the AOIP box I am building has no FPGA and no CPLD, it uses a CoTs ASIC USB implementation and a general purpose ARM CPU.

    It may or may not be ... depending on the interface in question.

    It'll generally be outside the DACs enclosure, on a separate PSU, both of which are things that bring useful benefits, even if everything else was the same. They'll generally be using a simpler protocol. And, since they operate purely in the digital domain (leaving the sensitive analog electronics inside the DAC) if they're properly isolated at their point of connection, which they typically are, you've side-stepped almost all of the sound-quality issues that plague USB Audio.

    Now, if you stick the AOIP implementation INSIDE the DACs enclosure, and run it off the same PSU, then you're going to lose some of those benefits. Like most things in engineering there are trade-offs and no free-lunches. Operating inside the DAC means its easier to interface to I2S, at the cost of all the undesirable electrical noise also moving inside the DAC.

    Implementation details matter.

    In every evaluation I've done, an external AOIP solution connected via AES or COAX has beaten the DACs internal USB implementation. And that's proven true across every DAC I've tested (which is quite a number at this point).

    No idea, since unless you have both an XMOS or Amanero USB input board and one built using ASICs or another approach that work in the same DAC, you wouldn't be comparing apples to apples. It's not something I've spent time investigating either, since what I have been able to test, i.e. XMOS or Amanero vs. other inputs, has consistently given the best results using the non-USB interface.

    So, even if XMOS was the best way to do USB Audio (and perhaps it is, but that's counter-intuitive to me as an engineer) it's still not achieving the same level of results that much simpler interfaces are on the SAME DAC.
     
  14. msommers

    msommers High on Epipens

    Pyrate
    Joined:
    Nov 6, 2016
    Likes Received:
    2,750
    Trophy Points:
    113
    Location:
    Calgary, Alberta
    Home Page:
    So I bought a little USB power meter to test the USB-output of my Squeezebox Touch. As a result, I find that it's putting out 4.63V and 0.12A. My desktop on the other hand puts out 5.04V and 0.0A through every single USB output.

    It seems like a Wyrd or Regen would be the answer here for the Squeezebox, if I'm understanding them correctly, will "boost" the USB voltage to what it should be. Does 0.37V even matter in the grand scheme of things?
     
  15. Torq

    Torq MOT: Headphone.com

    Pyrate
    Joined:
    Jan 15, 2016
    Likes Received:
    8,193
    Trophy Points:
    113
    Location:
    Seattle, WA
    I'm guessing that you used the power meter between the SqueezeBox and your DAC, but when you connected it to your PC you didn't have anything on the other side of the meter.

    If that's the case, the readings are not comparable.

    The current reading should be the same regardless of the source, since that's how much power the connected device is pulling. The voltage difference is likely to be sag due to the load.
     
  16. msommers

    msommers High on Epipens

    Pyrate
    Joined:
    Nov 6, 2016
    Likes Received:
    2,750
    Trophy Points:
    113
    Location:
    Calgary, Alberta
    Home Page:
    Thanks for the reply, Torq.

    In the first example. yes it was between the Squeezebox and the Hegel amp with the respective values previously mentioned.

    In the latter example, no the device was between my desktop motherboard USB 2.0 output and the Schiit Gungnir Multibit->Mjo2->650.

    In both examples, I was playing music while checking the readings.
     
  17. Torq

    Torq MOT: Headphone.com

    Pyrate
    Joined:
    Jan 15, 2016
    Likes Received:
    8,193
    Trophy Points:
    113
    Location:
    Seattle, WA
    That suggests the USB input in the Gungnir Multibit is not USB powered then (I've never actually checked).

    To be properly comparable, you'll want to test with the same devices in place after the power meter.
     
  18. Wfojas

    Wfojas Friend

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    848
    Trophy Points:
    93
    Location:
    CA
    Oh, AES to the Yggdrasil, and Spdif coax to the Gungnir Multibit, just out of the slight sound differences and preference. I do think the Gungnir Multibit was discernibly better with the Bada USB, than the microrendu alone. It's only when you factor in the price does it become a less rationale proposition, so like @zerodeefex posts at the top of the thread, the better Dac trumps the converter.

    Also, the USB 3.0 in the Yggdrasil is better than the one in the Gungnir Multibit, as I listen to them.
     
  19. rott

    rott Secretly hates other millenials - Friend

    Pyrate
    Joined:
    Oct 23, 2015
    Likes Received:
    1,176
    Trophy Points:
    93
    Location:
    Nation's Capital of failure
    Here's to hoping that the USB input module for Gungnir Multibit is slated for an upgrade >= the USB 3.0 currently in Yggdrasil. At this point it's the only improvement I'm wishing for.
     
  20. colorsquid

    colorsquid Friend

    Pyrate
    Joined:
    Oct 13, 2015
    Likes Received:
    30
    Trophy Points:
    13
    Location:
    Los Angeles, CA
    Home Page:
    I'm thinking bigger. Here's hoping the next Gungnir MB upgrade includes ethernet.
     

Share This Page