it's likely the minimum buffer size for smooth playback is going to depend somewhat on the computer you're running this on as well as the API. best bet would be to do some investigation by setting up a streaming source with a large buffer (say 1024 samples) and then stepping down in size until the sound starts to exhibit popping/glitching artefacts. if all you actually want though is a sine tone that changes pitch you can create a sample that is a single period of the wave, enable looping and then adjust the pitch during playback. of course, that's less fun
...
Less fun, indeed. I already did that.
You are right about the other points
about the speed and the API, no question about it, yet I assume my computer to
be fast enough for generating a sample for the purpose at hand. This is how I
look at it; the driver grabs the ready flag from the sound hardware and routes
it into the upper tiers until we see it as an OnBufferEnd within XAudio2
introducing sort of an overhead which might be ok. But if the XAudio2's thread
awakes for every 10ms, then this will ultimate require a minimum buffer of 441
samples (assuming 44100Hz) to be buffered ahead of time for continuous play,
which is usually not a problem if one plays a common wav file. So if my
assumption (well, I've read it on a Microsoft page) about the XAudio's 10ms
is true (XBox360 ~5ms) the sound should start to crack at a buffer size <=
440 samples. If this will be he case, XAudio2 won't be a solution for me.
Anyhow, stepping down will give the lower limit, as you wrote.
