SC filters for vocoding

[Paul Schreiber]

I certainly have lots of CEM3320 VCF chips that would make dandy vocoders (hint hint).

Paul Schreiber

----------
From: 	Chris Crosskey[SMTP:chrisc at zetnet.co.uk]
Sent: 	Thursday, April 03, 1997 9:07 PM
To: 	synth-diy at horus.sara.nl
Subject: 	Re: SC filters for vocoding


Jim Clark wrote
>
>Has anybody used Switched-capacitor filter chips in vocoders? A while
>back I built a 4 formant resonator, for demonstration in a computer
>speech course I taught, using two National Semiconductor MF-10
>chips. These are dual SC filter chips. Each of the filters has 2
>poles. The data sheet claims that one can make Bessel filters of
>arbitrary order using multiple filter blocks, so it would seem that
>you could use these for vocoders. The clock frequency is 100x the
>centre frequency, so for the lower frequency filters (such as a 50 Hz
>centre) the clock noise could appear in the audio range. It would be
>simple even to construct a divider chain to produce a nice
>distribution of centre frequencies.


I had thought about it a while ago, but couldn't persuade myself to 
buy 100 UKP's worth of filter chips to give it a go....


>
>Anybody tried this? The MF-10 chips are about $4 each, so if you used
>a 4-pole Bessel filter for both analysis and synthesis, a 16 band
>vocoder would need 32 of the MF-10s, costing about $130 US.
>
You'll need more than four-poles for a 16-band, four-pole is what 
Elektor used for their ten-band, I guess you'd need 6-pole for 
16-band, unless you are really talking abouta full-audio bandwidth 
unit rather than a bias towards vocal frequencies like the 
Elektor...and what about unvoiced stuff, Bode Hi-pass bleed or a full 
UV/V detector....the Elektor UV/V detector is three boards about the 
size of a Eurocard.....I was thinking of getting my Elektor unit 
running again (been saying that for ages <g>) and then look at 
replacing a channel with the MF10 based circuits... might as well run 
a 555 on each for the clock, plus a simple lo-pass to killl the 
clock-noise will be OK, 24dB/Oct on the units will give you no signal 
at 100x filter freqency anyway, you could chuck a 24db filter tuned to 
a third the clock frequency, base it round a pair of TL07x dual 
capacitor circuits, and you won't be touching anything below five 
octaves above the vocoder band freeqency, that's a 61-note keyboard 
length if it helps to envisige like that, a lot of harmonics as 
well...that bit shouldn't be too hard...if you can be bothered to 
design a decent UV/V detector rather than the monster that Elektor 
used it's a good idea....


chrisc