PLL input filter vs. loop filter

[Tom Becker]

... feed a small sawtooth, made from the VCO, into the comparator to cause it to PWM...

This was super-easy to do, Ken (just three resistors and two caps on an
existing comparator), and it instantly tripled, in my case, the ripple
frequency. I also used an active notch and highpass ahead of the PWM.

Very nice - and, by directly tying the top of the triangle-generating RC
cap to a processor pin, I can turn the PWM on and off by floating or
grounding the pin in code; if the cap isn't permitted to charge, the
comparator switches at Vcc/2 as it normally would. I can do the same
with two loop filter caps, too, so I think I've got the wide capture and
narrow tracking modes I sought, very easily, under program control.

It's interesting how a long learning path can bring one close to where
one started, but with insight. I think I grok it. Thanks again, folks.


Tom

 

[Ken Smith]

comparator switches at Vcc/2 as it normally would. I can do the same
with two loop filter caps, too, so I think I've got the wide capture and
narrow tracking modes I sought, very easily, under program control.

I assume you are doing this to drop out part or all of the loop filter:


----/\/\/--+-----+---
! !
--- ---
--- ---
! !
Pin-------- GND



There can still be a bit of an issue with the capacitor trapping an error
when you switch the bandwidth. Too bad the CA3080 is going away, there is
a cute circuit to make the bandwidth smoothly sweep from one to the other
using it. You may be able to do this however:



----/\/\/--+-----+---
! !
! ---
! ---
! !
! GND
!
! Gain=1
+--!>---
! !
! \
! /
! \
! ! 1/3
O O CD4053
^ ............. CPU pin
!
---
---
!
GND

You start off with the amplifier charging the capacitor. If you just
randomly switched to having the capacitor filtering, the ripple voltage
would get trapped and cause a phase offset.

You can solve this by making the switch at just the point where the ripple
crosses through the center line. You will still trap noise but not
ripple.

If you really want to have some fun, you could try PWMing the switch
slowly from 100% the amplifier to 100% filtering. I haven't really
thought this through but I throw it out there.

 

[Tom Becker]

... There can still be a bit of an issue with the capacitor trapping an error when you switch the bandwidth...

Yes. Your buffered VCO input voltage idea is very good, and PWMing a
cap between it and the VCO input is intriguing; a PWM-variable cap might
be a flexible design component.

I could, I suppose, avoid the issue by switching a parallel R (instead
of switching caps of the loop filter RC) to change the loop bandwidth.
Any surprises likely there?


Tom

 

[Ken Smith]

an error when you switch the bandwidth...

Yes. Your buffered VCO input voltage idea is very good, and PWMing a
cap between it and the VCO input is intriguing; a PWM-variable cap might
be a flexible design component.

I could, I suppose, avoid the issue by switching a parallel R (instead
of switching caps of the loop filter RC) to change the loop bandwidth.
Any surprises likely there?

You don't really avoid the issue that way. The capacitor will still trap
the ripple voltage at the point of switching. You could use a JFET and
slowly bias it off to ease the resistor out of the circuit.

 

[Tom Becker]

... you could try PWMing the switch slowly from 100% the amplifier to 100% filtering...

Someone on another group says he's done this; with a sufficiently high
PWM rate, he says it works well to effectively produce a
continuously-variable capacitance. Very nice.


Tom