I read that article [re: DF96's post] on S/PDIF and USB timing/DAC conversion. Carefully.
FIRST: the entirety of the measured "problem" was packet overhead in the DAC itself, having nothing to do with USB. Like so many pieces of equipment, it has attractive specifications, yet very likely a pathetically underpowered microprocessor handling the USB interface data.
SECOND: The data frequency shift/drift was less than 10 parts per MILLION. Sorry boys 'n' girls, but the world's finest piano tuners, masters of "tone" and hypersensitive to relative changes in tone - can not detect 10 parts per million tone changes. Indeed, the Fourier transform of the whole sequence resulted mostly "parts per billion" noise.
THIRD: The only thing that moving to S/PDIF did in the article was to change the packetized data to a steady stream of serial fed bits. Yet, if the DAC with its USB connector had better design done at the USB level - there would have been no measurable difference at all. None. Identical.
THE MORAL OF THE STORY really isn't that "USB isn't good for carrying timing" (which is absolutely and undeniably true), but rather "Some DACs have crappy USB front ends." (which itself is kind of shocking ... for such potentially expensive devices.)
GoatGuy
FIRST: the entirety of the measured "problem" was packet overhead in the DAC itself, having nothing to do with USB. Like so many pieces of equipment, it has attractive specifications, yet very likely a pathetically underpowered microprocessor handling the USB interface data.
SECOND: The data frequency shift/drift was less than 10 parts per MILLION. Sorry boys 'n' girls, but the world's finest piano tuners, masters of "tone" and hypersensitive to relative changes in tone - can not detect 10 parts per million tone changes. Indeed, the Fourier transform of the whole sequence resulted mostly "parts per billion" noise.
THIRD: The only thing that moving to S/PDIF did in the article was to change the packetized data to a steady stream of serial fed bits. Yet, if the DAC with its USB connector had better design done at the USB level - there would have been no measurable difference at all. None. Identical.
THE MORAL OF THE STORY really isn't that "USB isn't good for carrying timing" (which is absolutely and undeniably true), but rather "Some DACs have crappy USB front ends." (which itself is kind of shocking ... for such potentially expensive devices.)
GoatGuy
if it's properly implemented it should be OK. we shouldn't forget that USB allows for a logical clocking scheme, not a backwards one like with S/PDIF.
IME radiated noise is a real problem. some USB DAC makers have no idea how to get it right.
just a little off-topic, but not so much, i have a stanley car battery charger Product review: Stanley 500-amp jump starter - Autos.ca that I bought at samsclub and it has a 12V 5amp output, and also a USB output, can I use it to feed my Musical Fidelity 12V 500ma device? I think that DC power is a lot cleaner than the standard DC psu that comes with the DAC.
1. The author attributes the problem to the driving PC, not the DAC.GoatGuy said:
2. You can't equate a steady frequency error of X ppm with a jitter (random or periodic) of Xppm. If this seems counter-intuitive, think how amplitude errors work: a steady amplitude error just make the sound louder or quieter; varying amplitude creates IM.
I think the Cambridge Audio DAC is not asynchronous. it may be using some sort of async resampling (which is something else and I would guess that at least some people confuse the two), but I don't think it uses async isochronous USB.
relying on the PC clock instead of slaving it to the DAC is an audio dead end, IMO.
relying on the PC clock instead of slaving it to the DAC is an audio dead end, IMO.
No necessarily. Since the DC from the battery is around 12V it must be regulated down to 5V. If this is done with a linear regulator, then it's probanly very quiet. If iy's done with a high efficient switch-mode type, is there any reason it'd outperform a plug-in switch-mode?
hi
IMO it's useless to generalize. look inside any commercial DAC and you'll see that very different "stuff" happens in the S/PDIF and USB paths.
one example is the popular Arcam rDAC. many users reported inferior sound with S/PDIF. one day some guy on a Chinese forum decided to measure jitter and the USB input was indeed superior, measurement-wise.
maybe some manufacturers pay more attention to the S/PDIF input, maybe some to USB.
also, the cheaper DACs resort to off-the-shelf chips, which are made by different manufacturers with different design strategies etc.
with the high end models, I believe most manufacturers don't care much for the USB input. imagine you have a high end company and you want to make a new CD player. what's the average user? likely not exactly in his 20s, has a large collection of CDs. of what use is USB to him?
and I'm absolutely aware that there's a "dogmatic" belief that S/PDIF is intrinsically superior, nothing new to me.
Last edited:
sure.
random DAC (not made by a known brand).
while doing some measurements and playing digital zero, I notice some rather large (~2mV high) spikes at the output, occurring some 690 times per second. they were going away when stopping playback on the computer.
what the...?
turns out the cheap interconnect was picking up USB noise. solved with a shielded one (I was not using my day to day IC for the measurements).
I eliminated all other causes (the notebook itself, the notebook supply etc). it really was the USB cable or maybe even DAC internals generating that.
I can easily imagine similar phenomena going on at not so easily observable but nevertheless audible levels.
there was a S/PDIF to USB shoout out over at computeraudiophile (not a fan of that site myself but...). looks like the winner's designers went to lengths in order to reduce noise. for instance mounting the USB connector on a plastic plate, reportedly to reduce capacitive coupling. there is screening between two sections of the converter and they even fitted a ferrite above one of the isolators used (Analog Devices iCoupler technology).
yes, it's anecdotal evidence but don't many studies begin with "there's much anecdotal evidence"? 🙂 I haven't had the chance to do comparisons myself but I know enough people who report differences.
the CA thread: 15 USB/SPDIF converters shootout
ADI iCoupler technology:
Digital Isolators | Interface & Isolation | Analog Devices
http://www.analog.com/static/imported-files/overviews/isoPower.pdf
inside pic of the thing + the Berkeley implementation:
Attachments
Yes, of course. Going back to the article, looking - again - at the graphs, there wasn't "jitter" in the classic sense, but rather there were periods every 10 seconds where the conversion rate dropped by 8 ppm (48K/16bit ... the 44.1k/16 bit was much tighter). Let's put that in perspective, yes?
Speed of sound is 343 m/s
8 ppm of that is (343 x 8 / 1,000,000) = 2.7 millimeters per second.
Or... about the amount of "head jitter" of one sitting in a comfy listening chair, merely breathing. In other words, inaudible. My point is - just because with the remarkable accuracy of digitization we can measure things, does not make those things audible, in any meaningful way. One must put the numbers, into perspective.
FURTHER... "Fourier transforms" of long tone streams are absolutely meaningless in the "intermodulation distortion" sense. Oh, sure, using Fourier's theorem, one could precisely reconstruct the original waveforms, including their periodic dips ... but that doesn't tell the story of the signal tone changes.
Just as you regularly champion "thinking practically" here, I am solidly behind the idea that the numbers are important, but they must be remembered in context for which they're being applied.
GoatGuy
If something can get its timing that wrong then there is a design error and/or an invalid assumption somewhere. I agree that numbers matter.
Our brains are used to making allowances for head movement; here we have the apparent effects of head movement without the movement - that can lead to nausea if taken too far.
Our brains are used to making allowances for head movement; here we have the apparent effects of head movement without the movement - that can lead to nausea if taken too far.
a dedicated fifo is the best way for sure, but not the only way, most modern receivers will have a small fifo.
the timing for the data transfer from the computer has nothing to do with the audio timing in all but the most ancient (and pretty clueless back then) USB implementations. there is a dedicated local USB/MCU clock in any half decent receiver now (usually 12MHz), as well as 1 or 2 audio XO's. conducted common mode noise is really the last area of some concern depending on how crap your computer and interface is.
Last edited:
I have to agree that noise could be a problem with some of the layouts I have seen, leave a lot to be desired. Also a couple of designs have rise times far greater than required, adding high frequency harmonics that aren't desired further adding to the noise and signal integrity problems.
The output is not USB, it's SPDIF- very different animals. The jitter that counts is what's at the DAC.
This is the one I was using:
Hi Fi CM6631 192kHz 24bit Asio USB to Coaxial Optical SPDIF Convertor for DAC | eBay
edit: The one I used in the HiFace comparison no longer seems to be sold. My current one (above) goes to 192, but the results at lower sample rates are the same.
This is the one I was using:
Hi Fi CM6631 192kHz 24bit Asio USB to Coaxial Optical SPDIF Convertor for DAC | eBay
edit: The one I used in the HiFace comparison no longer seems to be sold. My current one (above) goes to 192, but the results at lower sample rates are the same.
- Status
- Not open for further replies.