PLL Jitter, Frequency, and Linewidth (0.25um, 0.18 um, 90 nm, etc).

[john_cambridge]

Hi,

I'm am new to design, and had a basic question - for a PLL with a set output frequency, e.g. 500 MHz or 1 GHz, to what degree will jitter change as linewidth decreases? or is jitter more related to the frequency?

I have looked at a couple specs (0.13 um 90 nm), and it the jitter values seem about the same. Does this seem right?

Thanks,
John

 

[vfone]

RMS jitter have a kind of square-root variation vs gate width

https://www.ece.ucsb.edu/~long/ece594a/Weigandt_ROjitter94.pdf

 

[biff44]

I don't know about line width, but for digital signals, the voltage headroom is a big factor. If random noise stays the same, a 5V logic family is going to have a much lower phase noise than a 1.3 v logic family! Simple signal to noise issues.

Also, the type of stucture has a big effect. CMOS, with interstitial trap states, is going to have more shot noise than a junciton device.

 

[dick_freebird]

Voltage headroom is not what matters. It's edge rate and
the input-voltage-width of the linear region of the phase
detector. Edge rate is much much better in the finer geometry
processes. What is not better, is how much stuff people pack
together and how tightly, making on-chip interferers; rail
bounce from all the faster transitions against not-much-
different supply trace and bondwire inductance, and so on.
It's not a simple line fit. Get used to that if you're chasing
a performance improvement in anything analog besides
raw frequency or raw voltage.

You should be so lucky, that individual device noise is
the main contributor.

If you get serious about measuring jitter you will find it
very difficult to get instruments reading reliably below
10pS. Very expensive 'scopes and a lot of work to clean
up your bench of every noise source, etc. You may be
seeing the limit of measurement more than reality.
Especially if it's a production limit on ATE that has to
be guaranteed, it's probably way sandbagged and all
about the ATE platform limitations.

 

[vfone]

Yes, is the same when somebody try to design the best sensitive with lowest noise floor receiver, when actually the receive frequency band, is few times noisier than the RX noise floor (due to multiple factors as: power lines, catv, ignition, industrial, cellular, radars, weather, sky/solar).
Soon, due to so much noise, we have to design crappy receivers, and use kW of TX power