New Filter Designs

[Tim Ressel]

Grant,

I'll take a stab at this. The capacitors get tied to the circuit one at a time
in a rotating fashion. This implies a certain period for each cap to be tied to
the circuit. Now a signal is flowing into the circuit which has a certain period
as well. imagine these two periods are identical. Each time capacitor N is tied
to the circuit, the incoming wave is at the same part of its cycle. So if you
were to plot the voltage across the caps, you'd see a rough approximation of the
wave. Kinda like a standing wave, no? This forms a primitive analog memory, like
a bucket brigade. If you take the average voltage across the caps and sum it
with the original signal, you will get a nice notch. 

Now if the incoming wave speeds up a bit, the caps are now seeing different
parts of the wave each time they are connected. The cancellation decreased
dramatically at even slight differences. If the wave speeds up even more, the
point is reached where there are two complete cycles of the incoming wave for
every one cycle of the cap rotation. Once again cancellation occurs. This
repeats on up the spectrum. By the way, you should be able to configure either a
bandpass or notch filter this way. A PLL is a nifty way to generate the clock
for the counters. And if you were to sync it to another wave....

Tim Ressel--Hardware DQ
Hewlett-Packard
Verifone Division
916-630-2541  
tim_r1 at verifone.com 



> ----------
> From: 	Grant Richter[SMTP:grichter at execpc.com]
> Sent: 	Monday, August 16, 1999 12:07 PM
> To: 	synth-diy at mailhost.bpa.nl
> Subject: 	New Filter Designs
> 
> I have been experimenting lately with new filter designs
> based around the state variable filter.
> 
> One design that has always intrigued me is from Jan Hall
> in EN #95 the "Bi-N-Tic" filter. This is a state variable using
> commutating capacitors for the integrators.
> 
> To elaborate, build your usual state variable, for the capacitors
> in the integrator, use a CD4051 with eight capacitors of the
> same value. Power the CMOS off of +/-7.5 volts and multiplex
> the 4051s with a counter like a CD4024. You'll need two
> 4051s and 16 caps. Since the CMOS is powered from +/-7.5
> the switch point is at ground, making it easy to clock.
> 
> The circuit produces the sharpest comb filters that I have heard.
> The sound is very similar to a flanger or phase shifter.
> While the original design used an ultrasonic clock,
> by far the most interesting sounds are produced by
> clocking it in the audio range. It sounds really good
> with a rhythm box driving a sequencer controlling the
> clock frequency.
> 
> Can anybody explain why the commutating capacitors
> create comb filters at what seem to be octave intervals
> of the clock frequency? My brain doesn't seem to be
> able to grok this.
>