[sdiy] Craig Anderton's Multiple Identity Filter

[Olivier Gillet]

> I'm guessing you went this way. A 17-bit code gives you at least 8 resistor options for each of 5 inputs to the mixer (e.g. a 4051 for each one). Under uP control this gives fantastic flexibility and the problem becomes finding "useful" or "interesting" responses from a system which will include many duplicates and things that just aren't that good - including 30 from the 131,000+ possibilities suggests you did the required serious pruning!

Pruning indeed! I prototyped a design in which each filter stage goes
into a an attenuverter whose gain is controlled by a MCU. The goal was
to allow continuous morphing from one response to the other, with each
stage's "coefficient" being computed from 3 parameters: slope,
response (LP to BP to HP) and "polarity" (LP becomes HP, BP becomes
notch, HP becomes LP). And it's not that interesting, with many
combinations of parameters sounding mild. I think you get more
flexibility, for the same hardware budget, with two SVF with
LP/Notch/HP/BP crossfading, and crossfading between serial/parallel
routing.