XOSC jitter

I meant to say mounted on 1.6mm FR4 i.e. quite flexible. I've just checked a second board and it's doing exactly the same thing. Looking at the (AD9901-style) PFD output, the edge due to the external standard is rock solid; but the edge due to the XOSC is all over the place. Phase noise at high offsets looks fine on the spec an; but ultra close-in looks quite bad when viewed as jitter in the time domain.

Must be, since it's happening! I'd guess that these small smt oscillators would be fairly stress and vibration sensitive.

Analog ICs, opamps and I guess PLL chips, can also be stress sensitive. It will take very little voltage change at the Vc input to bash the edge picoseconds per second.

Most VCXOs have pretty good jitter up into the low milliseconds, then get bad. A cheapie Mouser-type VCXO can have inherent jitter in the ballpark 20 ps plus a few ns per second RMS.

A 1 KHz loop is about right to discipline a cheap VCXO. A TCXO or better yet an OCXO can work well with a narrower loop bandwidth.

We had to mount one DIP14 canned VCXO, a Vectron 155.52 MHz thing, on specially fabbed springs to kill lethal microphonics. It looks like one of those bobble-head dolls some people put in their cars. The loop uses a d-flop as a bang-bang detector, around 3 KHz in acquisition mode, 500 Hz or so after lock.

John

r

ds

it

The crystal isn't all that different from an accelerometer. It isn't optimized to be one, but a crystal is effected by motion.

I've never done the experiment, but perhaps you could measure a frequency change of a crystal if it is changed from vertical, which the gravitational force would be symmetric, to horizontal, where gravity would aid movement in one direction.

I bet the engineers at Swiss Federal Institute know all the nuances.

r

ds

it

According to John Larkin, it is.

You may also need to worry about the 1/f noise spectrum of the amplifier in your phase-locked loop. Any noise it contributes gets to integrate as a phase excursion over the period of up to one second defined by the bandwidth of your phase-locked loop.

White noise averages to zero (eventually) but 1/f noise builds up as the square root of the period over which it is averaged.

It might pay to take a carefull look at the noise specification of the amplifier that you are using

-- Bill Sloman, Nijmegen

That's a common spec for good OCXOs, delta-F for a 90 or 180 degree rotation. 1 part in 1e10 per G is good, 1 in 1e7 is not so good.

John

If the crystal itself is deforming, absolutely - by the piezoelectric effect if nothing else. A crystal should also be an extremely hi-Z output, so no telling what you're doing by pressing around it.

-- Les Cargill

ee

C or

dreds

ase it

ed

Very interesting. Many times I've used test gear in vertical mode, i.e. sitting on the feet in the back. I suppose I should always use test gear positioned horizontally, which is how I assume the gear is designed.

r

ds

it

Thanks to all for the responses.

Can someone point me at any references on flicker noise? I don't understand why it builds up.

It sounds like I should widen the loop bandwidth somewhat to reduce microphonics and flicker noise; however, the oscillator is actually

0.5 ppm TCVCXO and my client says his external standard has poor short term stability, so I don't want to go too far.

As previously mentioned, apart from the microphonics, I can see other jitter sources present all the time. Originally, I was locking to the EXT STD OUT of a 30-year-old sig gen. I see less jitter using the STD OUT of my counter instead. I still don't know how much of the jitter I'm seeing is due to my ancient test equipment!

and if you sync your scope trigger to the XOSC you should se the same jitter.

In cases like this you are comparing two signal sources, both with some (usually unknow/unspecified) jitter or short term behaviour. The difficulty is to decide which is the chicken and which the egg, and how much of the jitter is contributed by the loop filter.

There's an interesting (free) paper on this at

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
hobbs at electrooptical dot net
http://electrooptical.net

r

ds

Not a surprise. First: your oscillator is HCMOS, and isn't wholly jitter-free. Secondly, ALL crystal oscillators are audio transducers, audio signals will naturally be picked up. The trick, is to mount the crystal so the important stress directions aren't coupled acoustically through the mount. The smallest-possible-package oscillators may not be optimized in this fashion.

If picoseconds of jitter is a problem, usually one goes to oscillators that don't have logic-level outputs.

can see

XOSC or

hundreds

increase it

shocked

Depends on the test gear. Some operate within specification in any orientation given reasonable ventilation, and some has a seriously inviolable up.

see

XOSC or

hundreds

increase it

shocked

Maybe it is rental time? Rent something good for a month, about a tenth of the price of buying it (maybe a little less).