Multipath Filters

[Grant Richter]

> 
> thank you very much for this article !
> This does explain something. Whished I had it before I went to the
> breadboard. Yes, I held back my "results", because they were so
unpleasant,
> but now I dare speak.
> 
> I built a circuit similar to the EN design (changes see below), and I was
> very
> disapointed with the results. Guess I was expecting miracles ...
> There were some spectacular comb filter sounds indeed, but all was
spoiled
> by heavy clock feedthru and aliasing distortion.


I think this technique is best suited to large passive L-C-R circuits
which will "soak up" a lot of switching noise. Passive circuits should
eliminate
any offsets.

In my research, I managed to uncover a design from an early
organ manufacturer. The organ included a mechanical "vibrato"
circuit (read "get around Lesley patent"). This was a string
of about 25 passive LC circuits with a motor driven commutator.
Effectively the unit rotated through a series of taps in a passive
delay line. Apparently the manufacturer felt they could get away
with this trick even using mechanical contacts

> filter are open doors for the clock to come thru, and going high
impedance 
> (1nF caps to fit the OTAs) was not improoving things either. I even
suspect
> capacitor mismatch (didn't select the caps) plays some role, because the
> clock
> became even more dominant when I touched one of the capacitors with the
> scope
> probe.
> 
I tried different sets of capacitor values. The more mis-match between
capacitors, the more feed thru of the commutating frequency.
Try using 0.1 uF capacitors and set the state variable to
maximum frequency. The action is not actually that of
a state variable and the larger the capacitors, the more
pronounced the effect is. Excite the unit with a LFO and "ring"
the filter with slow pulses. With Q set to maximum, this
should produce metallic decay sounds (apparently all the
comb points will "ring").