Frequency multiplication

[Martin Czech]

> I want to make a high frequency VCO, 32 times the normal frequency, to use
> as a driver osc for a walsh bank. There are several options:
> 1. Use a HF VCO... like a CEM3340 with a smaller than usual cap. I've tried
> this before, it worked sort of...

And it will have a flat response at the top, because of nonideal 
comparators etc. You could try to compensate like in 3340.
Use a dual ota as current switch for the tri integrator.
One is used for the real vco, the other mirrors the current into
some network, which is in turn fed back into the vco input.
The 3340 simply converts the curent into voltage and then
feeds back into the expo converter. Maybe  a more elaborated
PWL network coulkd be used. Of course temp compensated.

> 2. Use an ordinary pitched VCO's triangle output and FWR and level shift it
> 5 times over. Never tried this but it could work... but will it track fast
> enough?

As the frequencys get higher, FWR gets unprecise. This means that parts
of the incomming frequency remain. Maybe this doesn't nurt for
your purpose, since at the end you want to come up with 1:1 frequency
anyway.

> 3. Use an ordinary VCO's sawtooth output and have a load of comparators to
> generate a pulse train as the sawtooth drops in voltage over one time
> period.

This seems reasonable. At the fast edge care has to be taken because
of possible glitches. A lot of comparators means a lot of different
delays. This is likely t give a hell of toggling. The best would be to
use the discharge spike to shut off all comparators, or outputs.

> 4. I'm open to any other ideas.
 What about pll? Yes, there is lag and overshot. But only because
a very simple linear osc. is used, so the loop has to be strong fedback.
Or in other words: You have the information where the frequency
should be (incomming control coltage) but you don't use
it and rely instead on the control circuit.

How can we make better use of our knowledge of desired frequency?
Now , what about using a normal precision vco as master and the same 
design with lower cap (flat track compensation? tempco?) as slave?
Since the slave will follow the master anyway within some tolerance,
it is not necessary to have very strong feedback. Maybe the error
is only four semitones, i.e. the pll will have  to pull only 4*83mV.
If the master frequency changes, the slave will immideately follow,
without any regulation, simply because of control voltage. The pll
will only fine tune then.


m.c.