[time-nuts] Disciplining Rubidium
magnus at rubidium.dyndns.org
Thu Apr 24 00:16:29 EDT 2008
From: "Tom Duckworth" <tomduck at comcast.net>
Subject: Re: [time-nuts] Disciplining Rubidium
Date: Wed, 23 Apr 2008 20:57:40 -0700
Message-ID: <E1Josb2-0007LX-EK at febo.com>
> tau, the Greek symbol for a defined measured time (1 sec., 10 sec., etc.)
> when using it in a timing application measurement. In nuclear physics it has
> a very different meaning (weakly interacting subatomic particle).
To clarify a bit further, when doing Allan deviation measurements (and friends)
it is the gate time between measurements. To properly collect data,
measurements needs to be done back-to-back.
Back-to-back data also has the nice property that you can build integer
multiple length gate-times from the same data, which allows multiple tau-
lengths to be analyzed from the same dataset.
The use of Allan deviation is needed to handle noise sources of f^-1, f^-2 etc.
i.e. not flat white noise. These will dominate at lower frequencies, and normal
deviation and variance measures will not converge.
Allan deviation is sensible to drift, but phase and frequency errors is
canceled since it uses the second derivate of the time errors.
For shorter taus, phase differences is normally dominated by white noise and
shorter noise forms, so noise is analyzed in the frequency domain, referred to
as phase-noise plots.
More information about the time-nuts