[sdiy] Filters

[Grant Richter]

Bandpass filters get a little confusing. For example, a state variable or
biquad has a theoretical 6dB/Oct slope, but with undamping (resonance) the
actual slope near the resonant peak is much steeper. But the resonant peak
has a tendency to "ring" adding holdover time.

An unresonant (maximally damped) 36dB/Oct bandpass filter would require 12
integrators (six for lowpass and six for highpass) but nearly the same
effect could be achieved by a smaller parts count filter with some
resonance. So it is a balance between parts count and the amount of resonant
"ringing" that can be tolerated.

> All this talk of non-resonant filters and how to make them resonate 9ie the
> RC stuff) prompts me to ask a similar question.....for reasons to do with my
> lack of sense I'm thinking about trying to design a vocoder. Becasue I'm
> lookgin for quite high quality vocoding I'm looking to build it around 24
> channels of 36dB/Oct filters plu sa Bode hi-pass bleedthrough, and if I get
> that far also ripping off the Elektor voiced/unvoiced detector circuitry (or
> at lest its principles)....so what I'm looking for is a stable sensible 36dB
> filter, preferably built around npn transistors (I'll be using 3046-based
> VCA's so I might as well use 3046 based filters if I can) and capable of
> running a fairly high signal level (to improve signal to noise) If I can
> manage it I'm going to try to get 2 channels per Eurocard and mount the
> whole lot in a 3U rack, which I have kicking around somewhere....Any ideas
> on where to start, Horowitz and Hill weren't a lot of help in that
> regard...simple fixed fitler and hwo to change the frequencies by swapping
> out resistor adn cap values....
> 
> chrisc
> 
>