[sdiy] Linearized 2164 VCA Warning

[David G. Dixon]

> Or simply... don't use a cheap-ass op-amp like the TL074 for a
> critical DC application.  Its "good enough" for non-critical use
> (typical audio path) where DC accuracy is not paramount, but for
> important CV circuits you really need to choose components that
> favour DC operation (low offsets, low drift, etc).

Yes, but first one needs to understand the problem.  This is actually the
very first time in three-plus years of synth-DIY activity that the offset of
TL07x presented any discernible problem.  And, funnily enough, the dual VCA
that worked perfectly had the even cheaper-ass TL074CN in this application,
whereas the one which wouldn't shut off had the slighly less cheap-ass
TL074BCN (with its guaranteed lower offset).  As it happens, it wasn't the
offset that was the problem per se, but the fact that the offset was in the
wrong direction.

Other than that, the circuit works absolutely perfectly, and always has,
with TL074, just like Mike Irwin said it would in his article.  Given the
exact nature of the problem, I'd rather add a two-cent resistor and a couple
of traces than replace a one-dollar opamp with a five-dollar one to achieve
exactly the same end.

> Try an LT1012 instead (max 25uV offset).  Or if you don't want to
> pony up the $$$ for the op-amp then you need to provide some manual
> means of trimming out the offset, for example as you suggest with a
> large resistor to a stable bias voltage (suitably decoupled else any
> noise will amplitude-modulate the audio).

I'm using the same -5V reference (79L05) that servos the linearizing VCA,
and its decoupled with 10uF and 100nF, as are all my rails.  This works
perfectly.

> Did you also include the series resistor that George Hearn wrote
> about from his 2164 experiments?
> 
> http://search.retrosynth.com/synth-diy/search/lookit.cgi?-v0906.246
> 
> Otherwise you'll be cooking those 2164s nicely (perhaps not a bad
> thing at the moment tho' ;).

Well, I had forgotten that little tidbit -- good advice.  In fact, the
dissipation is about 140mW across the entire chip, which according to the
datasheet, should be expected to raise the temperature of the DIP package by
4 to 10 degrees C.  However, I just left mine on in a quiescent condition
for about 15 minutes, and I think I may have been able to discern that the
2164 was ever so slighly warmer than the other chips on the board, maybe.
In any case, it's a good idea.  I'll add those resistors to my layout along
with the offset bias resistors.