Actually there is a very important problem with digital audio and sampling, and it is not about rise time itself, as this is always limited to FS/2, as sampling theory demands a bandwidth limited signal and this will define the rise-time. But the issue is when does the timing of a transient occurs; is it at the beginning of a sample, or 10% between the sample, or at halfway, or towards the next sample. Now this timing uncertainty is resolved by the DAC interpolation filter, and the maths is very simple - use an infinite tap length interpolation FIR filter with a sinc impulse response, and you will absolutely and perfectly recover the original un-sampled bandwidth limited signal. But it will still be bandwidth limited, and this will determine the signal rise time. Now I don't believe rise time per se is important, but the precision that you reconstruct the timing of the edges is radically important subjectively as the ear/brain is very sensitive to timing errors. We talk about timing errors in terms of jitter in terms of pico seconds, but the timing errors can be tens of microseconds long with conventional short tap length filters.
That's why Hugo has over 26,000 taps on the FIR interpolation filter (the longest tap length of any other production DAC at any price), as this sounds much better because the timing errors are much smaller. Indeed, I have been listening to 164,000 tap length filter against 100,000 taps, and you can still hear improvements, so we are a long way from solving this issue. I am clear in my own mind that we do not need infinite tap length filters, but we certainly need much more taps than even 164,000 before one can no longer hear the difference. If you take the view that the sinc impulse response should be 16 bit accurate, then we need about a million taps before the timing problem is resolved. this would them guarantee that the reconstruction in the time domain was 16 bit accurate.
Rob
