[sdiy] Zener clipping Was : ...that 1K resistor on the output

[David G. Dixon]

(Long reply... apologies in advance)

> Hmmm ... Little known?  It's used extensively in the IC Op Amp Cookbook
> (Jung).  :-)

Well, I'd never seen it, and I don't think it's in any other book I've got.
Danjel van Tijn pointed it out to me.  It's pretty obvious once you see it.

> I've used it, but the diode matching really can be an issue.  That's why I
> dropped the idea and used commutating CMOS switches in my Tri VCO.
> http://home.comcast.net/~ijfritz/sy_cir4.htm

I'm trying to avoid the use of digital logic in my new TZFM VCO.

> And for the Teezer I use hand-matching plus a trimming circuit.

Is this the same trimming circuit you used in the Tri VCO, where you added
or subtracted a tiny bit of current directly to the integrator from a
trimmer between the two zener references?

Although my circuit also uses two comparators in the core, it is actually
very different from your Tri VCO circuit, and is based on an oscillator core
which I haven't seen used in any other VCO.  I'm using the back-to-back
zeners as a clamp on the comparator output in such a way that they also
define the current to the integrator.  It's a pretty odd core, but
simulations suggest is should work very well.

I'm hesitant to use real comparators in my circuit for two reasons: 1) they
suck a lot of current, and 2) as you said in your Tri VCO page, they give
asymmetrical output, which makes it difficult to centre the triangle about
ground.  Hence, I've opted for TLE2072, which is the faster, better version
of TL072.  It's got a slew rate of 35V/us, which doesn't hold a candle to
LM317, of course, but it will give a dead symmetrical triangle, will suck
only a small amount of current, and won't create any other problems.  I also
considered at using LM13700 for the comparators, but it had the same issues.

I may yet change my mind about the whole thing...

In any case, to compensate for the crappy HF tracking, I've come up with an
improved version of the HF trim circuit I used in my 2164 Expo VCO.
Essentially, I've still got an inverter off of the CV summer driving a
diode-connected 2N3904 to inject a correction current back to the CV summing
node.  However, this time I took a slightly smarter approach.  I derived the
I_C vs V_BE relationship for 2N3904, then I simulated the VCO tracking over
10 octaves starting from 16Hz, calculated the required additional control
voltage to be fed into the 2164 Expo VCA at each octave to get perfect
tracking, converted this to the required correction current to the summing
node, and took this as the I_C of the transistor.  Then I plotted the
corresponding V_BE of the transistor vs. the required 2164 control voltage,
and this gave a perfect straight line.  The slope of this line is the
required gain of the HF inverter, and the intercept is the required V_BE at
a control voltage of 0V.  This V_BE is achieved by feeding a bias current
into the inverter (the -15V rail through a certain input resistor).  That
bias current is the key to the whole scheme, and accounts for the fact that
the VCO tracking may start to fall apart well below a control voltage of 0V.
With this scheme (in principle), just about any opamp could be used as the
comparator (even the lowly TL072), provided that the correct input resistor
is chosen to bias the HF inverter.

Concerning the other comment you made on your Tri VCO page regarding the use
of zener clamps, specifically their temperature effect, that should not be
an issue in my circuit.  If the peak triangle magnitude increases due to the
zener voltage increasing with temperature, then the current to the
integrator will also increase by exactly the same proportion, thus giving
exactly the same ramping time.  That just leaves the zener matching problem,
which is partially alleviated by using the diode bridge (or by
hand-matching, although I would need four zeners in my circuit without the
bridges).