Pitch -> CV conversion

[Martin Czech]

>      The loop filter was likewise designed 
>      by sheer fumbling around with various parts. My findings were that the 
>      best tracking and response happens when there is *no* loop filter at 
>      all (!) but the CV is un-useable for controlling an external synth. `

Yes the loop filter is the problem in pll designs, because  the vco integrates
the phase comparator information back to frequency information, that makes
already 90 degrees phase shift. Now add a improper lowpass filter to the loop
and oooops, 180 deg. : instable ! 
So this finding makes sense. 
What about averaging the CV OUTSIDE the loop ?
Could work , maybe.
What about using a lead-lag circuit for filtering ?
Could help against the phase problem. Never tried that.

The 4046 has two comparators : 

A simple exor, crude multiplier,
that will also lock on harmonics (easy to fool).

And a flip-flop state machine thing, described in some old RCA databook
with state diagramm. If I only could remember...
One could interprete the function as follows:
It counts the pos edges at input a and b. If a has more edges than b per
time interval T, it says : lower (nmos on). 
If the number of edges is equal : don't move (tristate).
If a has less edges : higher (pmos on).
This can be done with a few flip-flops, you don't really need counters.
a and b provide the clock edges.
As long as a und b have sqa shape, you can't fool this comparator, it will
always try to lock on the right freq.

m.c.