I’m not kidding about the fundamental flaw in S/PDIF. And indeed there are issues in USB audio too, but because the clock is not directly embedded in USB it isn’t as appallingly bad as S/PDIF. The flaws in design is well known in the professional audio design community. Of course you can embed and recover the clock for the purposes of digital signal transfer, this isn’t the flaw. The flaw is more subtle.
The problem is this, S/PDIF sends the data as a serial stream that is trivially encoded so that there is a direct correlation between bit density in the stream and the audio signal. You can actually rectify a S/PDIF stream and hear recognisable audio. That is the core problem. The DAC needs to do things with the data stream - it needs to extract the data, and it needs to reconstruct the sample clock from the same data stream. Note, this isn’t just a simple data clock to cheerfully drive the logic, it is the audio sample clock. Jitter in the sample clock is directly audible as distortion in the audio. If the jitter was stochastic noise, it would simple raise the noise floor, But as noted above, the S/PDIF stream has a low frequency compoenent that is correleated with the audio data. This component is extremely difficult to prevent from contaminating the recovered clock. The usual designs use a phase locked loop to create a local clock, but a PLL simply works as a low pass filter, and typical designs leave a huge amount of audio correlated jitter in the recovered clock. Because the jitter is so correlated the DAC tends to create very objectionable distortion products. You get products that are a fixed frequency away from the desired audio for instance. These are not harmonic distortion products, and tend not to be easily masked by the ears natural tendency to mask harmonics. Rather they add a nasty edgy and sometimes metallic edge. DAC designers have to work hard to design systems that can recover a clean clock, and cheap DACs still do a bad job of it.
Avoiding clock jitter on USB audio is also painful, and requires care. A big part of the problem with USB is that the clock much be recovered from the received packaet rate (not the USB clock itself) and this is subject to very large variations in timing due to operating system activity. Simple DAC chips like the TI 2702 and it family boast a two stage PLL system that first sues a broad capture PLL to sync up, and then switches to a much sharper peak PLL to keep the clock. These are considered one of the better one chip USB DACs. Advanced DAC designs like the Benchmark use techniques such as an asynchronous sample rate converter and a local clock to try to stabilise the clock.
You can run USB audio in isochronous mode, where the SUB sink - the DAC, sends back pacing commands to the source, and the source is required to send data at the rate demanded by the sink. This allows the DAC to provide its own master clock and avoid the need to recover the sample clock from the revived data rate. However the vast majority of USB DACs don’t support this mode, and a lot of sources don’t either.
Firewire does provide isochnonous transfers, and was designed from the outset to avoid problems in audio sample clock distribution.
I’m not talking about audio, I’m talking about how those interfaces all seem to be able to transfer data at upwards of 5 GB/s w/o issue. but, I’m far too tired of arguing against woo in the audio field, so I’ll drop it here.
You and me both, I mentioned what actually works, I don’t care what others are saying about DACs and soundcards.
Just take the outputs from your soundcards, put them through your audio amp get some hifif speakers, the result is good.
Forget jitter, forget clocks, forget all that shite and just stick with what works, I am not going to go into explaining the whys and wherefores.The properties that are often described by amateur engineers and psuedophiles are rarely significant at audio frequencies. Gaining technical knowledge and working in the field in audio takes time and huge effort, it means college and night school and the audiophools seem to think they can pick it up from a few like minded magazine that are full of repeating the same old same old without ever a double blind evaluation.
You might run across a few issues if you were involved in digital studios with various delays and timing issues, but this is not what we are talking about.
So where did I say you couldn’t transfer data with an embedded clock? Note this isn’t woo. Most crap you see in the popular press on audio is woo. This stuff isn’t, and is mostly derived from professional audio and the published papers of the engineers that designed the devices (from companies such as TI, Cypress, and even Phillips.) Clock distribution and recovery is one of the difficult and painful areas of digital audio. The idiot audiophile products jumped on the jitter bandwagon too, and mostly designed products that made the situation worse, or used insane designs, but also a lot of work was done to understand and ameliorate it in device design.
I’m not saying jitter doesn’t exist. I’m saying it’s been a solved problem for a long time and is rarely an issue worth worrying about, even with S/PDIF’s limitations. It’s also one of my hot buttons since jitter is a favorite canard of woo peddlers seeking to sell junk to schmucks.
I think on this we are in agreement. There are times when home audio systems manage an own goal and create a new problem for digital audio, and using computers as a source is one - but so long as nothing stupid is done it is a problem that is reasonably under control. Certainly you get a level of performance vastly better than anything you could have achieved with pure analog systems. I only mentioned S/PDIF as flawed in passing - and it does remain a very curious bit of history as to how a pair of companies like Sony and Philips could let past such a dumb design. It isn’t as if they didn’t know how to do it right either. Internally even right back in the 80’s DAC chips and systems used I[sup]2[/sup]S to communicate.
Personally, I use Apple Airport Express devices all over the house, and stream to them. There are a few amusing pitfalls to doing this, but they mostly work very well. The internal DAC is a TI 2702, and it will produce quite acceptable sound all on its own. A big buffer no doubt helps a lot. But you can noticeably improve it. Most of the time I don’t care, and listen to them directly. Many new HT amps support wireless streaming, and this will probably become more and more the main way we handle music.