AW: LM13700 VCF update no.2

[Haible Juergen]

	>Well I've built two of these low pass and high pass VCFs now.
I've used
	>TL074 buffers instead of the darlingtons and all is well
but...control
	>voltage breakthrough is terrible. One of the filters is worse
than the
	>other. The LM13700s do have different batch numbers, so it may
be case
	>of selection (yuk!!!). Oddly enough, I have a hunch that it may
be the
	>linearising diodes. When these are in use Iabc seems to have
more of an
	>effect on dc offset.

You know that you need a very different circuit for offset trimming
when you use the diodes ? The 3080-style 1Meg-Resistor + trimmpot
method is quite useless here.
The data sheet suggests splitting the resistors from the two input
pins to gnd into a fixed and a variable part. For example:
Trimpot 220 Ohm with variable pin to gnd, and both ends connected
to a 1k resistor each, with the resistors going to pins 3 and 4 of
the 13700.
I think it was the famous Serge (of Serge synthesizers) who suggested
an even better way in an old EN issue: use larger resistors, and then
connect the trimmer to the negative supply. I tried this, and it works
fine.

BUT: This (both methods) only works with AC-coupling to the inputs,
or with a constant source impedance if you want to dc-couple the
inputs, BECAUSE when you use the linearizing diodes, the inputs are 
not nearly at 0V anymore (by design) !

IMO, for most musical applications, it's a good choice to not use
the diodes at all. You don't gain much of SNR when you compare
the 3% THD levels (which is more reasonable for filters etc. than
comparing 0.1% THD levels, IMO).

If you really want to use a gilbert VCA, the 3280 is the better
choice by far. Or these Curtis 8-pin dual VCAs which Paul
Schreiber sells.
The 13700 is a great chip, nevertheless, if you want a dual
3080 or 3094.

JH.