[sdiy] Filter Banks & CSound

[Ian Fritz]

Tim --

Filter bank references:

"Electronic simulation of violin resonances," JV Mathews and J Kohut, J. 
Acoust. Soc. Am., vol. 53, 1620 (1973).

"Audio engineering improvements for clavichords," RW Burhans, J. Aud. Eng. 
Soc., vol. 23, 635 (1975).

"Filter bank design," IJ Fritz, Electronotes, #107, 16 (1979).

"A 39-channel variable-Q filter bank," BA Hutchens, Electronotes, #115, 3 
and #116, 3 (1980).

It's the peak-to-valley ratio that mainly characterizes the overall 
sound.  This ratio is strongly influenced by the shape of the resonance 
curves and the number of resonators, since many long tails can add up to an 
appreciable contribution.

I used a single transistor circuit for my resonators -- just a phase-shift 
oscillator with a bit of degeneration to set the Q.  I started with nominal 
Q values around 20, but had to increase this somewhat to get 10 dB contrast 
with all the bands turned up.  My design has variable amplitude on each 
filter and variable Q in sub-banks of six frequencies.  That way I can -- 
for example -- do some formant shaping at low frequencies and timber 
enrichment at high frequencies at the same time.

The hollow, echoey sound you describe is from too large a peak-to-valley 
ratio.  Several people have remarked on this before.

   Ian


At 07:50 PM 12/10/2002, Tim Ressel wrote:
>Ian,
>
>Yes, in fact Dr. Matthews was mentioned in the
>article. I'd love to see the article if you can lay
>your paws on it.  What kind of filters are in your
>filter bank? What frequencies and Qs??
>
>By the by, the CSound filters I used did not seem to
>parallel well. I didn't get that timbre change for
>slight frequency changes. Also the bank had a hollow,
>echo like sound. If I knew the real filters used I
>might be able to get it working better.
>
>--tr