[sdiy] Temperature Compensated Exponential Converter Using SSM2164

[David G. Dixon]

If anybody still cares...

I've been developing a full-featured VCO around this expo converter with
saw, triangle, pulse (from saw or triangle) and sine (or sigmoid, remember?,
from saw).  So far the thing is shaping up to be really nice, and has a very
low (and very cheap except for the 2164) parts count.  I'll share it all
very soon.

Now I'm playing with high-frequency compensation.  With no compensation, at
3V the simulated frequency is about 7.82kHz (should be 8kHz).  I'm presuming
this decrease is caused by the integrator reset delay, which (with a
suitably small cap across the saw comparator) is about 1us.

The way I've decided to implement the compensation is to tap the
current-sourcing VCA's control voltage, invert it with variable gain, and
feed it through a 1N914 diode and a 1M resistor to the CV summer.  It's
pretty standard, and it seems to work very well.  The diode passes a small
negative voltage (-55uV) when CV < 0V, but this can be compensated or just
ignored.  I think that 1N914 must be used here -- 1N4148 passes a much
larger negative voltage, according to my simulator.

I also came up with a very convenient way to sync the oscillator.  By
placing a small (1k or so) resistor between the 5V reference and the - input
of the saw comparator, a syncing square wave can be fed in through a small
(1nF or so) cap directly into this input.  The 5V reference is AC ground, of
course, and can therefore be used as ground for the RC network.  One can
even attenuate the sync input, provided that a suitably small (10k or less)
pot is used.

Any comments or criticisms will be much appreciated...


> >Let's face it: unless you're some kind of 'accuracy fetishist', the
> >references are only really necessary at critical points, such as the
> tempco
> >voltage for the expo we've been discussing, and perhaps for certain
> >operations like precision wave shaping, where it might ease or eliminate
> a
> >trim.
> 
> I always use subregulation for freq controls and for reset trip points in
> VCO cores (as examples). Maybe it is a fettish, but to me it doesn't make
> sense to only fix some sources of error and drift.
> 
> Ian