akira^8GB wrote:Do you think recording at a higher volume, even maybe clipping it a bit, would help? How about amplifying by software? That'd make the noise worse I would tend to think...
Clipping would make things worse, actually. Although, approximating the sine closer to a square wave might help a little- the closer to a square, the less volume jumps you'll have, or in other words you're decreasing the time period between voltage flips. The downside of this is of course the sound will approxiamte that of a square as well. Post-amplifying would only increase the sound (including the aliasing), as you said. The restriction is inherit, and if it's as I think it is, you're stuck with 6bit samples in one channel. A quick check with google seem to confirm this (4x8bit sample channels with 6bit volume)
I don't know.. my next impulse would be to suggest interpolation of the output signal. This doesn't work well (as the interpolation should ideally be on each individual channel) and will dull the sound, akin to a low-pass. You might be able to polish up the sound a little with post EQ.
Edit: Just did a quick test mixing square and sine waves. The square becomes very apparent wit as little as 15% mix, and the effect on the aliasing is neglectable.
Edit 2: Currently testing with sigmoid functions and results are somewhat more pleasing, (1/(1+exp(p*x)))-0.5 makes the sound more bell-like for increasing p, but the faster voltage change reduces the aliasing
Last edited by µB (Apr 6, 2012 4:12 pm)