[time-nuts] Phase noise confusion II

[Chris Caudle]

On Thu, April 9, 2020 11:20 pm, Hal Murray wrote:
> Suppose I measure the edge to edge times and make a histogram.

How precisely can you measure the period.  You would be using the rising
edge as start time and stop time, with no dead time, i.e. measure the time
of every period?

> Can I get jitter out of that?

I think that is typically referrred to as period jitter in the literature
I have seen. I think you could in theory derive any other view of the
timing error if you had enough measurements, and if the measurements had
enough precision and accuracy.  I think in practice that is difficult to
achieve, otherwise there would not be need for DMTD, PLL based mix down
for phase noise measurement, Miles's fancy quad ADC instruments, etc., you
could just use a time interval counter for everything.

Most of the jitter terminology has come from the digital communications
industries, because that is typically where you care about edge position,
vs. phase noise in an application where you care about modulation, noise
in an RF transmitter or receiver, etc.

This app note has an overview of the definitions of the various ways to
measure jitter:
https://www.sitime.com/sites/default/files/gated/AN10007-Jitter-and-measurement.pdf

I have some small quibbles, like in section 2.1.2 it begins with the
statement "Because the period jitter from a clock is random in nature with
Gaussian distribution...."
I would rather phrase it "If the period jitter from a clock is random in
nature...." because there are a lot of ways that systematic noise can get
into a clock and make the jitter behavior not random, or at least
non-Gaussian.

> Where is the clock recovery loop?

I think that comes from a lot of the measurement techniques where you
don't care if the clock you are measuring is slightly off of nominally
perfect frequency, because the receiver will have a PLL which will track
the average value of the clock, what you care about is short term
variation around that average value, so the measurement techniques utilize
a PLL (either physically implemented, or simulated in software) to mimic
the behavior of the receiver PLL so that you effectively ignored slow
variation in the average period time.

On Fri, April 10, 2020 5:05 am, Dana Whitlow wrote:
> Question about definition of jitter:   Is it the variation in
> pulse-to-pulse spacing, or is it
> the variation in pulse positions with respect to a jitter-free waveform?

See the document linked above, there are various distinctions made to what
and how you measure the period based on how the clock is being used or
what particular behavior you care about.

-- 
Chris Caudle