Randy Yates wrote in message ...
This is the problem with these outrageous sample rates. Why do we
need to process at a rate in which over 50 percent of the available
bandwidth is unused?

I think there are valid reasons to process at high sample rates:

1. There's some software out there now (DVD-A) that delivers content
at high sampling rates, and decimating it to lower rates seems to
defeat the purpose of having such high-sample-rate content in the
first place (assuming such high sample rate content is audibly
different due in part to their high sampling rate). For the really
optimistic, some day there may even be a standardized hi-res digital
interconnection which may pass along hi-res data unaltered.

2. There're DSP operations which require high sampling rates, such as
non-linear algorithms like compression, because non-linear algorithms
generate harmonics which may alias. One could argue that such
algorithms should upsample, do their non-linear processing, and then
downsample, but for the sake of fidelity through simplicity, it may be
easier to implement a good-sounding system if the digital processing
were kept to a minimum.

Optimized FIRs? Never heard of that phrase. One way of efficiently
computing a convolution is "frequency domain filtering."

Sorry. That's what I meant --- faster ways to do FIRs than just brute
force time-domain computations.

Yes, decimation is reasonable. 48 kHz was a fine sample rate - we
didn't need to throw it away and go to 96 kHz.

The stuff that has been implemented by Michael Gerzon and Peter Craven
in the early 90s for the B&W correction system is a multi-rate system,
where the bass, which needs the most precision and longest filters, be
decimated to 1 kHz, the mid-range from 500 Hz to 3 kHz would be
decimated to 6 kHz, and everything above that is running at 48 kHz.
The difficulty then moves to making the filters for the downsampling
and upsampling transparent and efficient. Here's a link to the Gerzon
article:

http://www.audiosignal.co.uk/Digital...alisation.html

--Andre