Paul Winkler wrote:
> 
> Hi all,
> 
> Sorry if this is a repeat, but I haven't seen the message yet on this
> end, and I'm quite eager to get nice reverb happening on my machine--
> I'm doing some volunteer sound design for a play that's opening in four
> days, and I need to make some things sound very spacious!
> 
> I've been playing around with various csound reverbs I've found on the
> list or on people's web pages. So far nothing I've found sounds as
> smooth as I would like.

Most of the reverb algorithms that I have seen are variations of
Schroeder's original algorithms from the 1960's. These were a great
innovation, but I have the feeling that most good commercial reverb
units use far different algorithms. Joe Dattorro has described a Lexicon
"plate" reverb algorithm in a recent issue of the Journal of the Audio
Engineering Society. This uses several allpass filters in a sort of
figure 8 feedback loop, with the output summed from various taps within
the allpass stages (something that the Csound alpass opcode can't do).
In addition, the delay lengths of some of the allpass stages are slowly
varying, which is also difficult to do in Csound.

You may wish to look into Feedback Delay Network (FDN) reverb
algorithms. These are essentially a variation of the comb filter stages
in the Schroeder algorithms, but with the output of each delay line
feeding back into the inputs of the other delay lines (the gain is
determined by a feedback matrix). This is apparently a way of getting
very smooth reverbs. Interestingly enough, most of these designs also
use slowly varying delay times, to avoid metallic resonances. Look for
papers on this by Julius Smith, Davide Rocchesso
(http://www.dei.unipd.it/english/csc/people/roc/), and a 1982 paper by
John Stautner and Miller Puckette, "Designing Multichannel
Reverberators," Computer Music Journal 3(2), (pp. 52-65.), that is
reprinted in The Music Machine, ed. Curtis Roads. Cambridge, The MIT
Press, 1989. (pp. 569-582.) . The Stauner/Puckette paper has Music 11
code for such a reverb.

Another option is the work with waveguides being done by Julius Smith at
CCRMA. This is somewhat related to the FDN stuff, but from more of a
physical modelling standpoint.  There are several papers available on
Julius Smith's page at http://www-ccrma.stanford.edu/~jos/

Hope this helps,