[sdiy] There's no such thing as a free lunch :(

[David G. Dixon]

Hello Fellow SDIYers,

I have finally figured out why the 2164 Expo VCO is not all it's cracked up
to be, tracking-wise.  (I know, it only took two years!  Don't remind me.)

First, a little background for the uninitiated.  When using a 2164 VCA as an
exponential converter, it is possible to use a second 2164 VCA to compensate
for the 3300ppm/C temperature coefficient.  One simply places the second VCA
between all CV input resistors and the CV inverting summer.  When one then
adjusts the voltage at the VC pin of the second VCA to about 270 mV (which
fixes its gain to about 39%), and adjusts the gain of the summer to tune the
VCO, the temperature coefficient of the 2164 expo converter magically
disappears.

At least, that's the theory.

However, there is a problem.  When a constant positive voltage is applied to
the VC pin of a 2164 VCA, the gain of the VCA is a weak (but measurable)
function of the current fed into it.  Over a 10V CV (10-octave) range, for
example, if the optimum gain for tempco is, say, 39.3%, the actual gain will
vary over the 10-octave range from about 38.2% to about 40.4%.  This has the
effect of causing the mV/octave at the VC pin of the exponential converter
VCA to change from one octave to the next.  If the VCO is tuned such that
the VC pin voltage is, say, 187 mV/octave in the middle of the range, then
the actual voltage will vary from about 182 mV/octave to about 192 mV/octave
over the 10-octave range.  This has the effect of making the lowest octaves
slightly flat, while adding significant negative tracking error (in addition
to any negative tracking error from the VCO core itself) at higher
frequencies.

As far as I can tell, this changing gain problem does not occur when the VC
pin of the 2164 VCA is grounded (i.e., the VCA is programmed for unity
gain).  Under those circumstances, the VCA is quite transparent in the
circuit.

The sad part of the whole thing is that, if the tempco VCA is removed from
the circuit, the 2164 expo converter is actually a pretty decent log
conformer.  Hence, using it with a conventional tempco resistor glued to it
should still make a pretty decent expo converter.  Why bother, you ask?
Well, the fact that it processes current in both directions is very
convenient in certain applications.

So, if anyone has anything constructive to add to this sorry saga, I'm all
ears.