PLL locking problem

My PLL can not lock. I am using an XOR as the PD, cross coupled VCO.
The control voltage is in the tuning voltage range of VCO. However, it
is an oscillation the same frequency of the output of the PD. When the
PLL is locked, the voltage should be a straight line. What can be the
problem? Thanks!

 

The output from an xor phase detector will be a 50% square wave when in
lock, i.e. the RC network will be charged and discharged at the same rate
maintaining a fixed frequency. It sounds like it's working fine. Maybe
you're getting confused with an edge detecting PD?

Mark.

 

Check your phase. The control voltage may be driving your VCO away from
lock.

 

Sounds maybe like a loop stability problem. I've built PLLs that were
so unstable that they did all they could to run away from locking. If
you don't know this stuff already, get a good book on PLL dynamics.

Are you using a lowpass filter after the PD? Got any more details?

John

 

I think an XOR pd will just lock on the opposite edge if the loop
phase is reversed.

John

 

I can't really remember because it was sooo long ago, but I ran into a
similar problem and it was driving me nuts. I flipped the phase of something
(although I can't rember what or how) and it worked just fine after that.

 

John, You *are* correct (except it's not *edge*, but on the opposite
90° phase). I think the OP's problem is *lack* of an appropriate
filter.

...Jim Thompson

 

Indeed, it sounds like there's no filter at all if the control voltage is
oscillating at the same frequency as the PD.

Mark.

 

Should be at *twice* frequency... maybe he has a mis-wire.

...Jim Thompson

 

I think the OP says it's the same frequency as the output of the PD, which
as you say should be twice the frequency of the VCO in lock.

Mark.

 

Bingo!

 

probably talkin' bout differential type phase detector, yeah i get the polarity
reversed too.


Remove "HeadFromButt", before replying by email.

 

Yes, I am using a RC low pass filter, followed by a source follower.

 

Try adding a small R in series with the C... that will damp an
otherwise very ringy loop dynamic.

John

 

Hmmm.

Excuse me but........ didn't you suggest that DM, or some other, was in the
realms of suck it and see coz you thought you were so ded clev at doing the
job proper?

And then you suggest adding a small R in series with the C to deal with a
ringy loop dynamic.

That is a big fucking schnort......

No offence meant but....... c'mon.

DNA

 

Heavens, you're surely not...gasp...sputter...comparing....

Well, geez, I can't deliver a 2-semister course in control theory in a
newsgroup. I did suggest, somewheres north of here, that the OP get a
good book on PLLs. The resistor is the typical bottom line here, so
let the guy give it a try; the notion *is* theoretically sound, ain't
it? Maybe?

John

 

Could be three or four easy steps, or perhaps one or two..... mind you it'll
tax my small one.

Step 1)

Wot's the gain of the EXOR gate?

NAND

 

They tax your small one? I guess there's a VAT on everything over
there.

Step 2)

Wot's the transfer function of the VCO?

(You do #3; it gets harder from here on.)

John

 

So.... don't be an old fart. If you give me 1 and 2 then number 3 should
drop out of my bum.

Caveat is you tell me and the rest of the planet what 1 and 2 is.

DNA

 

Soumds like classical instability problem. Measure the frequency of
oscillation and you can get back to the transfer function - it will be
something like

K/s*(1/(1+sT)

where K is your unknown gain. To find K you will need to draw a Bode plot
or do a calculation which takes the magnitude of the above and sets it to
unity using the measured oscillating frequency - then work back to K.
Once you have K you can do a proper design.
I expect the phase margin is pretty much non-existant. What you really
need is a lag-lead compensator. A series resistor (sombody mentioned that)
and a parallel resistor across the capacitor to give a steep roll-off at
high frequencies (for noise attenuation).You could try fiddling but these
things are best calculated otehrwise it is a shot in the dark. Trouble is
nobody does control courses anymore - they all want to do digital!

Tom