[time-nuts] GPSDO disciplining algorithms

Attila Kinali attila at kinali.ch
Mon Apr 13 13:29:12 EDT 2015

On Mon, 13 Apr 2015 13:03:36 +0000
Alan Ambrose <alan.ambrose at anagram.net> wrote:

> I'm interested in GPSDO disciplining algorithms - presumably the good ones 
> are really well thought out stochastic control algorithms.
> Is it possible to point me to the seminal references / papers / datasheets 
> that describe typical algorithms and the advantages and disadvantages of the 
> various approaches?
> I can see, for example:
> http://tycho.usno.navy.mil/papers/ts-2000/gpstime.steering.kmm.ion.sep00.pdf

The paper you refer to is a about how to generate UTC(GPS), ie how to
stear the clocks of the satellites in order to get a correct time
on the ground. It has nothing to do with stearing a GPSDO.

You are right that GPSDOs are nothing but control loops with
added noise. You can approach the whole thing by standard control
theory and ignore noise in a first approximation. The book by
Franklin et al.[1] is a good start, IMHO.

The next step is to model the system with all non-idealities that
affect the oscillator frequency (temperature, aging, retrace,..).
The field you want to have a look for this is called "System Identification".
For that you will also need to know how a quartz crystall oscillator
behaves under different conditions. A good place to start for
this is Vig's Crystall Oscillator Tutorial[2], as it contains
many references.

If you have done that, you can apply addaptive control, ie measure
the systems inputs and outputs and guess from that the internal
parameters for better control. The generic technique here is
called Kalman Filters. But fellow time-nut Marek Peca would tell
you that a Wiener filter is better suited[3]. Of course, the field
of adaptive control is vast and there are many more methods to have
a look at. "Model Predictive Control" might be one of the key words
you'd like to search for.

If you are asking yourself why I am not refering you to specifc
literature on GPSDOs, then I have to tell you that there is hardly
any. The best fit for you that I am aware of is [4], but this is
nothing other than what i have written above. And you will need
to understand these things to understand the paper.

			Attila Kinali

[1] "Feedback Control of Dynamic Systems", by Franklin, DaPowell, Emami-Naeini

[2] "Quartz Crystal Resonators and Oscillators", by Vig, multiple versions

[3] "Clock composition by wiener filtering illustrated on two atomic clocks",
by Peca, Michalek, Vacek, 2013

[4] "Adaptive OCXO Drift Correction Algorithm", by Nicholls, Carleton, 2004

< _av500_> phd is easy
< _av500_> getting dsl is hard

More information about the time-nuts mailing list