pll and phase

[Martin Czech]

In my mail

>From owner-synth-diy at horus.sara.nl Thu Mar 13 14:40 MET 1997
Subject: Re: Re[2]: Pitch -> CV conversion

there's a big mistake. I was in a hurry yesterday andoso I
mixed things up.

Ok. Reset. Throw the mail away.

So again : what's the problem with pll ?
It's a feedback loop, so one has to avoid phase rotations >=180 deg
given that the feedback is strong enough (loop amplification > 1).
If not, one gets a instable loop (how would that sound? maybe interesting ?)
You have a vco in the loop which gives frequency information.
Then you have a phase comparator. Phase and frequency are related
via this simple equation (unit circle) :

d/dt phi = ommega or phi = integral(ommega) dt.

This says : frequency is variation of phase in time or the other way:
phase is the time integral of frequency.
 
So, no matter how we realize that phase comparator, it'll always work
like an integrator thus giving already 90 deg phase shift from it's input
to the output.

Next we want some kind of lowpass filtering in the loop to avoid
switching noise of some comparators (like 4046).
If the filter is good enough, it will add more phase shift and so we get
near 180 deg and the loop tends to get instable.

What can we do ?

1. Use other types of phase comparators with less switching noise.
   Maybe use tow comparators : one wide for initial coarse tune , second
   narrow for final control.
2. Use PI filter with large P amount (that's what 4046 data sheet says)
3. Use good old PID filter with phase enhancement at higher frequencys.

Any experiences with pll and PID out there ?

m.c.