AW: Use of Diodes in Filters??

[Haible Juergen]

	>>(2)Location of these poles determined by unbuffered structure,
and by
	>>using
	>>several diodes in series for the last (upper) stage results in
a
	>>response that
	>>is more like 18dB/Oct than the 24dB you'd expect from a 4-pole
filter.
	>
	>Why is that?

If you look at each diode pair + capacitor as one symmetrical RC
stage, then you have several *passive* RC stages in series, without
buffering, i.e. loading each other. This will shift the poles' start
position
(with feedback=0) rather far apart from each other, so that the 
transition from passband to the final 24dB slope is not sharp at all.
(Compare this with the Moog cascade, where all poles start at one point,
or with a butterworth filter with already complex poles ...) 

	>I don't remember ever seeing diode ladder filters other than
4-pole. Do
	>they exist?

I *guess* that's just because additional stages would have less and less
effect. With four stages, you get something similar as with 3 Moog
stages
in the "interesting" region above the cutoff frequency, which seems
reasonable
somehow. Now idea how many stages you would need to get a sort of
24 dB slope.

You might try the effects of pole spread and feedback with a simulation
(or
calculation), if you de-symmetrize the circuit, replace each diode by a
resistor, 3-diodes-in-series with a resistor 3 times as large, and so
on.
Then build a little feedback loop and experiment.

JH.