Asynchronous USB

A couple of years since I published this article, numerous discussions still spread over the internet: is the USB better than S/PDIF, or is it vice versa?

And, after all these years you can still find the discussants who, trying to remove the question, but in fact moving it down to the more basic level, claim something like “none is better or worse, because both ‘just’ simply carry the digital data, and it is the trivial task, especially because this data is not even the one of the highest rate there is”.

Yes, they both do carry the same digital data, however the digital audio requires also some clock for D/A conversion, because the audio playback operates by converting the data into the audio signal in the real time, using the same very clock. Hence, the quality of this clock directly influences the quality of resulting audio signal. D/A conversion needs clean and steady clock frequency. Otherwise, frequency modulated clock translates into the frequency modulated audio signal, a.k.a. jitter.

And in this way, the way they handle and recover the clock, the USB and S/PDIF are fundamentally different. These differences and their consequences on the audio performance I described in the same article from 2008.

Now, two and a half years later, what’s really new?

As opposed to the S/PDIF, USB is a bidirectional protocol that makes possible to make peripheral device (here the D /A converter) act as a master device, and PC as a slave that delivers the data only when and as requested by the DAC. This way, the overall clock performance at the DAC is not related to and limited by the quality of PLL used, or by PC. It is solely determined by the quality of free running oscillator, placed locally in the DAC.

Such a solution became lately known as an “asynchronous USB”.

As for the S/PDIF, unless the whole S/PDIF system, which includes an external D/A converter, is designed to generate the clock at the DAC side, and to route the same clock back thus slaving the source, an S/PDIF must use some sort of PLL, to lock the DAC clock onto the source clock.

The S/PDIF also went a long road since the days of poor performing early S/PDIF receivers. Today’s industrial S/PDIF receivers show incomparably better performance, achieving sub 100 ps jitter, and some being able to lock onto the S/PDIF preamble, thus mostly eliminating data related jitter, traditional Achilles’ heel of S/PDIF, though not really without its own cost. Please see the measurements I posted one year ago to the diyhifi.org:
http://www.diyhifi.org/forums/viewtopic.php?f=2&t=1879

Thus, once we go the asynchronous USB route, the question “USB vs. S/PDIF” becomes rather the question of “free running crystal oscillator vs. PLL”. Practical reasons will leave PLLs in use for quite some time, and, as pointed out, nowadays fortunatelly we can make very good PLLs indeed. The general advantages of using non PLL, free running oscillators should be however still obvious and out of question. Just as you would probably never make a question like “is it better to run with or without 10kg of load?”.

Of course, Audial will also continue making the highest quality S/PDIF DACs, as we will continue to employ the best solutions in their own areas, at the given price and at the given moment.

And, regarding the USB DAC developments, even though not a lot of my work is visible online anymore, there was actually a lot of it happening behind the scene. And most practically speaking about the upcoming releases, here is the good news for those looking for Audial to come with its own async USB device:

As a part of a cooperation with Tone Distribution, and the Gramofone brand, which they are promoting, during the upcoming month Audial will release its first async USB DAC design. At the same time this will be also up to 192 kHz compatible device with 24 bits D/A conversion, and which employs no oversampling. Targeted to the highest end market, I’m sure it will be in many ways the holly grail of the long term hi resolution digital audio search.

More comprehensive article on the associated audio topics will follow soon too.

 

4 Replies

  1. 25-10-2010

    Hi Pedja – great article.

    A couple of questions – will the new DAC have both USB and S/PDIF inputs? Or only USB? Also, are you planning to make a USB/SPDIF converter?

    I was planning to buy the D-09 but wished to have USB capability, too, of course.

    Regards

  2. 25-10-2010

    Sorry I forgot the D-09 has USB already – but presumably not asynchronous – so I guess it’s better asked, will the D-09 be retrofitted with asynchronous capability?

    Thanks

  3. 26-10-2010

    Hello Alec,

    Good question, thanks for asking. Of course, I’ve been considering this option, however the answer is no, and let me explain why.

    A real non compromise USB interface, as the one used in the Gramofone DAC, includes, apart from dedicated USB audio controller and Windows driver, also a galvanic isolation. Also, such an asynchronous USB operation requires two quality local master oscillators as well, one for 44.1 / 88.2 / 176.4 kHz, and another one for 48 / 96 / 192 kHz. All this just wouldn’t be possible to do within D-09 price range. Hence, such an ‘updated’ unit would imply either price change, or technical compromises, or possibly both.

    Besides, and even if sampling frequency is normally not associated to the bit depth, and though the D/A conversion is of course much more than nominal number of bits, I can witness that people usually expect 24 bits conversion accompanying 192 kHz compatibly. This would practically rule out a TDA1387 D/A converter chip (or any other 16 bit D/A chip for that matter), and then it definitely wouldn’t be the same device successor anymore.

    (Just to avoid possible misconceptions: in this case, 16 bits D/A chip will still work with 24 bits words, using the first 16 bits and truncating the rest.)

    So, D-09 DAC will remain as is now. Consequently, for previously sold units, no such a USB front end update will be offered, either.

    In any case, and for what its worth, let me remind you that a D-09 still can accept up to 192 kHz sampling rates at its S/PDIF inputs.

    Regarding the USB to S/PDIF converter, and I assume that you are referring to up to 192 kHz async USB here, it is not impossible that, at one point, I will come with such a device. However it will surely not happen for now, and, for the reasons stated above, such a converter, if well done, also can not be cheap device.

    As for the Gramofone D/A converter, it will not have an S/PDIF input, not for now at least, but please note that all the questions regarding Gramofone devices are best directed to the Gramofone / Tone Distribution.
    http://www.gramofone.co.uk

    BRs,
    Pedja

  4. 26-10-2010

    And, one clarification regarding galvanic isolation…

    The D-09 was designed to include relatively simple (but effective) galvanic isolation. It converts the USB to S/PDIF internally, so the ground path is cut after the USB decoder, at the S/PDIF line, by use of an S/PDIF transformer. Of course, internal conversion to S/PDIF wouldn’t be desirable in itself, but I’ve settled on it deliberately, as this was effective compromise that made users free of any care regarding ground loops, HF noise, and all the associated problems when PC was coupled directly to the audio system. As you might remember, AYA II had such a direct (I2S) connection to PC, and even if quite nice results were possible that way too, for typical user it required too much of work at PC side.

    Non compromise solution in this regard, as done in the Gramofone DAC, implies isolation with no intermediate conversions, and master oscillators located at the DAC side.

    Regards, Pedja