[sdiy] SSM2164 VCAs

[David G. Dixon]

(You probably saw this one coming based on my last post...)

I've been looking at the datasheet for the SSM2164 Low-Cost Quad VCA (which
I just bought a pile of in the form of CoolAudio V2164D clones), and I'm
thinking about building a -- (wait for it) -- quad VCA!  Based on the
datasheet, this looks like a refreshingly straightforward thing to do,
except for one little thing: the gain of the VCA goes DOWN as the control
voltage goes UP, and 0V CV represents unity gain.  Also, the gain can
increase as much as +20dB or decrease as much as -100dB.  At one point, the
datasheet mentions that a 0 to 10V CV range will give gains from 0 to -100
dB.  Hence, I'm presuming that a 0 to -2V CV range will give gains from 0 to
+20dB, although this is never stated.  Also, the inputs, outputs and control
voltages can all range safely from -V to +V, and this can be as high as
+/-18V.  So, am I supposed to presume that control inputs between -2V and
the negative rail have no effect, or does the positive gain bear a severely
attenuated relationship with voltage over this range?

In any case, presuming I am correct in my original presumption (and noting
that it really doesn't matter too much anyway, since one typically doesn't
operate a VCA above 0dB gain), here is my tentative scheme for controlling
this puppy: bring the CVs in (through attenuating pots) to opamps which
shift the DC level by 10V and invert.  Hence, for example, an envelope from
0 to +10V would actually give +10 to 0V, such that the peak of the envelope
would represent 0dB gain, and ground would represent -100dB gain.  Does this
sound right?  Of course, independent gain and CV attenuation controls would
give flexible control over the true relationship between gain and control
voltage.
 
Another consideration: Does having the CV turned upsidedown complicate using
a logarithmic CV?  At first pass, it doesn't seem like it would, but if
anyone has any specific experience with this 2164 gadget, I'd appreciate
hearing any "horror stories".  However, on first reading, it seems a lot
easier to use than LM13700, with significantly lower parts count.