[time-nuts] The Nature of Noise

[Attila Kinali]

On Thu, 30 Jun 2016 18:30:35 +0000 (UTC)
Bob Stewart <bob at evoria.net> wrote:

> With my recent Cs problems, I've been wondering about the subject of noise
> generation and measurement.  Specifically, my question is this:  Let's say
> that I have 3 disciplined oscillators: A, B, and C.  So, I use the same 5370
> to create a 1000 second ADEV and discover that the 1s noise value between A
> and B is low, while the 1s noise between A and C, as well as B and C is
> high.  Can I then say that both A and B are low noise devices?  Or is it
> possible that even though I'm measuring 1000 points, both A and B are high
> noise devices, but somehow are noisy in the same exact way, and it's
> actually C that's the low noise oscillator?

Short answer: yes
Long answer: Yes, but unlikely

:-)

You can see the output of the oscillators as random processes and model
them as such (MIT opencoursware has some good lectures on that topic).
If the processes are all independent, then you can assume that A and B
will have the lowest noise and estimate the ADEV values of C and the
second nosiest oscillator with good confidence, and of the lowest noise
oscillator with some confidence (keyword: "three cornered hat").
In reality you will never have totally independent oscillators, there
will be always some coupling through power supplies (even mains power)
or through RF coupling. Heck, even your measurement instrument might
couple the oscillators if you are not very carefull. So it really might
be that A and B are just bad oscillators, that injection locked to
each other, and thus seemed more stable/less noisy than A+C and B+C.
But generally, high stability/low noise oscillators are usually made
such, that the output and the power supply are decoupled from the
oscillator itself as much as possible, to make the effect of coupling
(which is always present) so small that it get swamped out by all
other noise sources.

			Attila Kinali
-- 
Malek's Law:
        Any simple idea will be worded in the most complicated way.