Phase Noise vs. Jitter

There's a math operation that maps one into the other, at least for wideband noise and RMS jitter. There are several calculators online.

One of the guys here did this:

In the telecom business, stuff below 0.1 Hz is called "wander" and faster wiggles are "jitter."

What you seem to be talking about is like the old turntables which has spec s for "wow" and "flutter". One was fast and the other slow. Same but diff erent. Or not so different really, they just produced different effects on the sound and usually had different causes, but both are just variations i n the speed of the platter.

Likewise your jitter and phase noise are the same thing, just different par ts of the frequency spectrum. Same but different.

Why do you care exactly? How are you using this?

so

y,

is

ecs for "wow" and "flutter". One was fast and the other slow. Same but di fferent. Or not so different really, they just produced different effects on the sound and usually had different causes, but both are just variations in the speed of the platter.

arts of the frequency spectrum. Same but different.

Actually "jitter" is variation in period, and is measured once per period, and phase noise is variation from an ideal - perfectly regular - waveform a nd is integrated over the whole period.

It's not always a particularly useful distinction, but Cursitor Doom doesn' t think about what he's doing.

A recent acquaintance of mine has built some fancy super-fast (to me) oscillator which he wants to hook up to one of my SAs to check for stability and purity. I don't want to look an ass if he asks me something technical that's above my pay-grade and as a mere hobbyist, my-pay grade is pretty low.

I don't want a mathematical explanation, though; just an intuitive one.

In that case I've discovered a new phenomenon I've dubbed "twitch". :-)

"Exe files can't be previewed" it says on your link.

Let's try re-phrasing it thus: phase noise produces sidebands. Does jitter?

Something dumbed down enough for Cursitor Doom to understand.

The kind of deviations from the ideal perfectly repetitive waveform that constitute phase noise, and can be characterised as jitter, show up as low level sidebands on the Fourier transform of a noisy and jittery waveform.

Bill Sloman wrote in news: snipped-for-privacy@googlegroups.com:

clock jitter.

Bill Sloman wrote in news: snipped-for-privacy@googlegroups.com:

A really good piece on it.

Download it and run it. It's cool and perfectly safe.

Sure. If the period of a signal isn't exactly constant, then it's frequency is being modulated. And FM makes spectral components in addition to the main frequency spike.

If the jitter is fast and random, the frequency-domain noise floor is wideband, so a spectrum analyzer might not show classic sidebands, just an elevated noise floor. But some kind of systematic jitter, like time wobble caused by power supply ripple, will make classic looking sidebands around the main frequency line; FM.

I often measure jitter as a function of time. The shortest time is a single cycle of a waveform, which is what you usually see on a scope. But you can measure one rising edge relative to the 10th edge away, or the millionth one. A graph of jitter vs time correlates directly with the phase noise graph of the oscillator. Low frequency phase noise makes long-timebase jitter. You can, in theory, measure the jitter of a 1-year time delay; it will be huge.

A cheap crystal oscillator will have maybe 10 nanoseconds of RMS jitter when it's used to time out one second of delay. A really good OCXO will have a few picoseconds. The phase noise plots are indicators of the jitter behavior.

Allan variance is one formal way of expressing the complex phase noise/jitter of an oscillator.

What kind of resonator is he using? An LC or equivalent will probably show lots of sideband junk on a decent spectrum analyzer. Some really good ones could wind up displaying the phase noise of the SA itself.

Check for harmonics too, just for fun. Even some high-end, supposedly low phase noise, signal generators have horrible harmonics.

Look for line-frequency sidebands too. The moral equivalent of roaches in the soup.

Post a spectrum pic here, if you can.

Thanks, John. That's exactly the clear explanation I was looking for. Now watch some awkward bastard come along and find fault with it. :-)

-- This message may be freely reproduced without limit or charge only via the Usenet protocol. Reproduction in whole or part through other protocols, whether for profit or not, is conditional upon a charge of GBP10.00 per reproduction. Publication in this manner via non-Usenet protocols constitutes acceptance of this condition.

No idea as yet.

I'd be surprised. The guy is a heavy-duty scientist in his own field and retired from various blue chip pharmaceutical gigs, but just a hobbyist (albeit advanced) like me in electronics.

I have no SA that goes beyond 22Ghz without deploying some external source which would take it/them up to just above 40, and as yet I have no idea what frequency his fundamental is even. I'm sure 40Ghz is practically DC for a bloke of your calibre and worthy of haughty derision. :)

From the mains power you mean? Not much chance of that. Min sweep starts at 10Mhz.

Might take a couple of weeks before we can meet up and perform the relevant tests but I'll post back with pics if I can.

Not really. We're time domain, not RF. I did recently do some low-phase-noise VCOs with coaxial ceramic resonators, around 600 MHz. And a triggered LC oscillator at 40 MHz.

Could have close-in sidebands, +- 60 or 120 Hz from the main line.

One fun thing is spread-spectrum switching regulators. They jitter hard in time domain, to spread out the RF spectrum and pass FCC limits. It is sort of cheating.

It surprised me that the DC output is very quiet. They must shape the timing of the jitter, like a 2nd order delta-sigma modulator, to push the noise spectrum up where the output filter kills it.

Oh, ricky and sloman are both wrong. Jitter is a time-domain measurement, and phase noise is frequency domain. Either can be mapped into the other mathematically, but the units and measurement are different.

And jitter is not just a single-period measurement. Low frequency phase noise causes long-interval jitter.

I like when people ask me questions above my pay grade. I ask them to explain what they are talking about. That's when things get interesting.

And where did I say anything else?

"Jitter" is obviously not a single period measurement. You got to take at least two period measurements before you've got a base period to jitter from.

John Larkin tries to do analytical comment, but can't get to first base.

Well, Sloman of all people I would have expected to know that, given it's the only subject he's professed any qualification in. Perhaps he'll be along in a minute or two to clarify what he meant. Pity I won't see it if he does.

John Larkin clearly doesn't know what he's talking about. I've never suggested that jitter was a single period measurement. The idea is absurd.

Cursitor Doom is similarly inane. I've never "professed" to have any qualifications in electronics - I've never followed any academic course that might have lead to any such qualification.

This hasn't stopped me from working as a high level electronic engineer, or becoming a life senior member of the IEEE.

Cursitor Doom does have his blind spots - his habit of plonking people who point out when he's talking nonsense does mean that he's not reminded that he's talking nonsense quite as often as he might be - but this is sillier than usual.