You all need to realize that playback is not the same as recording ..
Obviously, when recording, any 'latency' is unacceptable -
YES, the guitarist,bassist, drummer whatever - will notice a 30ms delay from doing something till hearing it ..
It's enough to make the drummer off-beat !
When playing back a digital file from a computer - There is no such thing as 'latency' !
You are NOT playing a LP-record, 'decoded' in 'real-time' .
Ultimately, you will want the entire track read from disk to RAM and played from RAM without further disk-access .
CD's are almost read like LP's - That's why Sony made a load of money on that 'anti-skip' thing on the Discmans .
They did what foobar and all other computer-playback software does : Load it to RAM and play from there .
I think you misunderstood me, I said I prefer to use a small buffer for reasons"other than input delay."
There most definitely is such a thing as latency and it is there in varying amounts at every stage of the computer setup. You want to to input a command to a music player, there is latency, to access a hard disk, there is latency, you have different bus speeds, different speeds of data flow, and on top of this data does not usually arrive in a smooth manner so you need buffers at strategic points to ensure there is a constant flow of data and no dropouts. If you have a problem where the latency is inconsistent (very typical in Windows machines) you need a larger buffer, or on the other hand you can use techniques like ram disk etc to ensure a smoother flow of data and therefore not need such a large buffer. You can also not use a ram disk at all and just use a larger size buffer, but I find that this does not sound as good to my ears. Most music software do not load the entire file into ram, they instead load a number of samples ahead of time thus use a buffer. Audio data coming out of a USB port does not have consistent timing, which is why timing correction mechanisms are needed at the USB receiver, maybe even another buffered control mechanism. IMO none of these systems are perfect at fixing timing issues, whether this is audible is another matter.
The simplistic view is that you can just use a large buffer at any stage and you will get perfect results, but in my opinion this is not the case, as in I believe the buffer size and nature affects the sound quality. You may not hear any differences from these settings and this is fine. Even if I could provide evidence for any of my opinions I do not think it would be in anyone's interest. I could for argument's sake throw a million ferrite beads on a USB cable, prove I can hear a difference with DBT and then say a million ferrite beads sounds better, but still not be contributing anything useful. What I mean to say is that audio is so inherently subjective that even if a difference can be heard, people will not agree on whether there is an improvement. IMO If you are really interested in something you need to do your own research and not rely on extrapolation and prediction from other peoples research, and not mistake one (or more) negative result/s for a universal truth in circumstances outside the scope of the test setup. If you don't worry about subtle differences in sound quality then you could just not do the testing and direct your efforts elsewhere.