[time-nuts] How to measure 1 sec tau
warrensjmail-one at yahoo.com
Sat Sep 12 22:05:40 UTC 2009
tau expert's advice sought
Basically my question is that is the correct bandwidth to use when taking 1 sec and above tau data?
What I want to know is how to record raw 1 second phase data to use for calculating 1 sec and above tau.
Should I including in the data the results of every high freq noise source in the system?
My limited understanding is that none of the high freq noise has any true effect on the 1 sec freq stability of the Osc.
Would it not be better, when all I want to know is how much the Osc freq is changing from second to second
is to take the 'average' phase error over the previous 1 second period to use as each raw data point.
i.e oversample to get rid of the high freq noise.
When I want to know how the Phase is changing between two high freq Osc,
what I now use to get the highest accuracy and resolution possible,
is a 1 sec time interval average running at from 1 KHz to 10 MHz.
By averaging 1000 to 10 million answers over a 1 second period the
phase shift noise is reduced and the resolution increased by 10 to 1000 times
over what it would be if I just took a single sample each second.
With this technique I'm able to get phase noise resolution down in the low fs. (< 0.01ps)
Some have told me that this will give wrong false tau answer.
Maybe so, I do not know enough about what noise sources should be includes in tau.
My though is, I am only interested in 1 second and above stability of my osc
and not the high freq noise of my Osc or test system, so this is the best way to get the correct answer
My test have shown that this 1 sec averaging does not change the 1 sec
and above tau answers, it just lowers the noise floor.
More information about the time-nuts