[time-nuts] Disciplining a TCXO

[Attila Kinali]

On Fri, 26 Oct 2012 11:46:35 +1100
Tom Harris <celephicus at gmail.com> wrote:

> I have been asked the viability of using a vanilla TCXO, with an
> accuracy of +/- 0.5ppm (+/- 15 secs per year) that is disciplined
> occasionally (perhaps only once a month) with a GPS module. The
> application is for an analogue clock, which powers up a GPS module
> every so often to learn the drift characteristics of the TCXO, which
> it then compensates to generate indicated time. The TCXO's that I have
> played with have a very predictable aging characteristic over time, at
> least in a normal home/office environment.

Hmm... The feasability of this depends a lot on your temperature
variations and the TCXO. If you have a nearly constant temperature,
then the aging will dominate. If you don't have constant temperature,
then the semi-random variations due to the temperature correction
errors will make it hard to really predict what's going on.

Yes, you can filter out a long term average, but nothing says that
this will be the same for the next measurement period if it's dependend
on the temperature variations. Modeling those errors is difficult and
not realy for the faint of heart, even if you know what type of frequency
correction is used. For a general TCXO it gets very tedious to calculate.

Of course, if you need only a "crude" accuracy, then a simple
average error correction might be enough.

			Attila Kinali

-- 
The trouble with you, Shev, is you don't say anything until you've saved
up a whole truckload of damned heavy brick arguments and then you dump
them all out and never look at the bleeding body mangled beneath the heap
		-- Tirin, The Dispossessed, U. Le Guin