Toslink has enormous amounts of jitter. It's an inherently shitty interface. This is a thing we can measure, it's not up for debate. Although glass Toslink cables are better than plastic, that's still just polishing the turd that is the Toslink interface. It's not because the light is "bouncing around" inside the cable. It's because S/Pdif is not an optical signal. It's electrical. So at the Tx end, the electrical signal must be converted to optical, and at the Rx end, the same process happens in reverse. It's this double conversion process that introduces so much timing error. The length of the cable is basically irrelevant. The shittier these electrical/optical converters are, the worse Toslink will sound. The one Apple uses in their Macbooks for example seems to be of particularly atrocious quality.
I2S is the native format of digital audio for the DAC chip itself, this format includes the following signals:bit clock, wordclock, masterclock, and data, plus ground return(s). SPDIF includes all of these signals on a single wire, and then requires the SPDIF receiver to strip out the I2S signals, and then re-align them with the DAC masterclock, typically using a PLL circuit. All of this embedding, stripping out, and re-aligning adds unneccessary jitter. By using async USB in the DAC itself, all these extra steps are avoided, and the I2S data stream is created with a single fixed frequency clock, resulting in jitter levels as close to that of the clock reference itself as is possible....Re-clocking is not the panacea that some suspect it is. Only truly asynchronous re-clcocking, used fixed frequency oscillators, can really address jitter issues. And listening tests even reveals that this type of re-clocking does not result is the same performance as using an approach which does not generate the jitter in the first place (well implemented async USB in the DAC). The problem is that building a true asynchronous re-clocker is very difficult, and costly to implement. Most do pseudo re-clocking, using a PLL-this does not eliminate jitter problems-it just changes them into different distortions (as does an ASRC). The technical details of these approaches are too involved to go into here, but my advice is to never believe marketing materials which claim "jitter reduction" or "jitter immunity". My experience has proven time and again that having low jitter from start sounds better than any "jitter reduction" scheme.
1. Because you are comparing apples and oranges. USB streaming works over Isochronous mode, which works in real time and has no error correction. Older adaptive mode implementations are also especially bad, because the clock is controlled by the computer, and the CPU isn't particularly interested in the level of output jitter going to your USB DAC. USB hard drives, printers, thumb drives, etc work in Block mode. The data is sent is sent in packets not in real time, and these packets are error checked and corrected. That's how you can copy your Word document to your thumb drive and get a bit-exact copy, without the need for a USB filter. Keyboards and mice work in a 3rd mode called Bulk Interrupt, and they don't need filters either. 2. USB noise has nothing to do with having a "bad" computer. Any channel motherboard from any brand is not designed for ultra low noise applications when there is no demand for it from 99.9% of buyers. Similarly, even the best 80+ Gold or 80+ Platinum ATX switching power supply creates a substantial amount of noise, as does any switching power supply. It's baked into the cake. Aside from the switching noise itself, the very best ATX power supplies can get down to around 15mV of DC ripple on their 12/5/3.3V outputs. Any decent linear supply can produce closer to 1mV of ripple. NO ATX power supply can get anywhere close to that, and the best linears will be under 1mV. The amount of DC ripple produced by the ATX power supply on its outputs are also not what you actually end up with on the VBus line from a motherboard output. That will be on the order of 250mV-700mV. A typical laptop running on its power brick will be closer to 1000mV. Compare that to <1mV produced from a 5V linear power supply.
Are there any measurements that show these atrocious levels of errors and jitter from computers?How much dc ripple is an actual audible problem?Oh and I found this: http://www.gigabyte.com/products/product-page.aspx?pid=5497#ov (scroll down a little bit to the audio part)It turns out there are some BIG manufacturers advertising some kind of audiophile usb power delivery
Three Wyrd in chain: http://www.head-fi.org/t/603219/schiit-gungnir-dac/2400#post_11897126
