Can PLL Freq Error be zero?

pretty much. Like I said, its pure pedanticism, viz.:

#define PLL_hunting phase_noise

IOW define any signal we dont want as "noise".

If the VCO was *perfect* then once the error amp (PD + loop filter) had settled on the voltage required to give the exact frequency (and phase), the frequency/phase error would be zero and stay that way forevermore (assuming the tracked signal is likewise constant, and the error amp contains an integrator). This needs a perfect EA too. Of course there is no such thing as a perfect VCO or EA (or anything else, if we look close enough) so the output F/phi drifts, the error amp wobbles appropriately and all becomes well again. Said wobbling = phase noise (brutally abridged description eh wot?)

A s***ty PLL will have metric truckloads of phase noise - in the worst case it will oscillate at its closed-loop bandwidth (a-la nyquist, thats a whole lotta phase noise). So to build a good PLL, firstly make the damn thing closed-loop stable. Then try and inject as little s*it into the loop as possible to reduce phase noise (good VCO, stable ref, good PD, low noise EA etc). I've never built a stupendous PLL (only nice easy ones), but odds on it gets pretty tricky as the phase noise spec gets smaller (well duh).

Just like any other control loop, I betcha there are quite a few PLLs in existence that *do* have nyquist oscillations.

Cheers Terry

OK, closed-loop instability. whose criteria? depends how many times you want to spin around -1 if you like drawing Nyquist plots. Or take the easy way out and draw a Bode plot :)

More than one? yeah, sure, all it takes is more unstable poles (I think

- I haven't actually proved it though).

Hell, its probably not that hard to build a chaotic PLL that does some really crazy stuff.

Cheers Terry

to spin around -1 if you like drawing Nyquist plots.

just me, being cryptic :}

Google looks like a great way to ensure (*NOT* Insure. Aargh, that particular mangling of English drives me bonkers /rant) conformity (aka mediocrity).

I think it would take a Ouija board to find out :)

But I *have* seen the term "Nyquist oscillation" used to describe closed-loop instabilities. just not frequently (har har).

haven't actually proved it though).

Its actually a pretty good question. I'll put my maths hat on, and do some thinking. For example can a pair of not-too-closely spaced unstable poles cause oscillation at the beat frequency?

crazy stuff.

That doesnt surprise me (nothing ever works), but please do elaborate

Regards, Terry