[sdiy] digital delay, was ...

[Glen]

At 10:24 AM 9/12/03 , Cornutt, David K wrote:

>To make it even more bizarre, do an analysis on the BBD device
>to get a general idea of what its noise spectrum looks like.
>Find the frequency band in which the device has the best S/N.
>Then, divide the input signal up into bands, and pitch shift
>them all into the optimum band and run them each through their
>own channel.  On the output side, pitch shift them back to
>their original frequencies and mix them back together.

As an simplified alternative, what if we tried this without splitting the
original sound into different frequency bands? I mean, use pitch shifting
to compress the frequency range of the original signal to fit into a small
band of frequencies that the BBD handles best. (Make the original low
frequencies higher, and the high frequencies lower.) Pass that through the
BBD, and then re-expand the bandwidth of the signal to its original state.

You could also place a compander around the BBD stage. The reduced
frequency range going through the compander should greatly reduce the
amount of compander breathing and pumping artifacts. It should be much
easier to tune the response time of the compander for a narrow band signal.

I'm not saying the single band approach would be better. Indeed, I think
that a parallel approach should yield lower noise, and quite possibly less
distortion as well.

I just think the single band approach might be simple enough for someone to
actually get around to trying it.   :)

Now, if I only had some accurate pitch shifting devices...


later,
Glen