There are a couple of methods for sending audio data over an usb cable. Here’s an explanation of the principles behind them and the advantages of one over another
Synchronous USB connections use a one way digital connection for music replay and are considered the worst type of connection for audio purposes.
Adaptive mode is a little smarter. It assesses the amount of data in the frame and adjusts that dac’s clock-timing to it.
Asynchronous mode is technically most advanced in that it has a feedback loop so that the amount of data in the frame can be controlled.
Conventional USB connections use a one way digital connection for music replay and use the computer’s bus frame rate as source for the clock which is less stable than a fixed one. A computer sadly cannot maintain perfect timing of the data sent via USB. For adaptive mode, the receiving chip adapts to this drifting signal by re-adjusting its own frequency every milisecond to match the incoming frequency. This is already considered better. The downside of both methods is that they tend to cause high levels of jitter. On top of this, the computer’s clock is not as stable (powerline fluctuations/RF pollution) as you would like for highend audio purposes.
Asynchronous USB (not to be confused with asynchronous samplerate conversion) uses a clock housed near the dac (usually in the external dac’s casing) and allows it to drive the converter directly, thereby not relying on the unstable computer’s clock. It is called asynchronous because the DAC’s master clock isn’t synchronized directly to any clocks within the computer. Instead, the DAC is controlled by a (potentially high-precision) fixed-frequency clock. This clock controls the datastream from the computer to a buffer near the DA converter.
Asynchronous USB data transfer, however, doesn’t automatically guarantee perfect sound. J. Gordon Ranking informed me of the following: “Many of the less expensive units do not use a fixed oscillator and instead use a frequency synthesizer and therefore have worse jitter than a dedicated fixed oscillator. Many companies don’t even know how to power fixed oscillators. If you go to say Crystek and ask for a really good oscillator and then power it with a 3 terminal regulator the results will be less than staggering. For high end you need a discrete regulator with ultra low noise to establish a really good low jitter Master Clock.”
Finally, having bit-perfect transfer doesn’t guarantee enjoyable sound. This may sound contradictory, but depending on the rest of the audio system, and taste perhaps, the very accurate transfer can sound overly tight and controlled compared to a more loose, but inaccurate, transfer that you may be used to hearing so far. For this reason I myself for some time have preferred the warmer sounding Winamp over the cleaner sounding Foobar in the past. If everything is correct though, a bit-perfect audio reproduction should indeed sound best.