[time-nuts] Orbiting crystals

Magnus Danielson magnus at rubidium.dyndns.org
Sat Jun 27 12:44:44 EDT 2009

Tom and Antonio,

Tom Van Baak skrev:
> Hi Antonio,
> A couple of thoughts...
> First, remember the quartz crystals on the GPS SV are not
> free-running; they are just the LO of Rb or Cs atomic clocks.

In the Block II series, the crystal oscillator responsible for 
transmitter circuit is a separate oscillator from any of the atomic 
references and their respective fly-wheel oscillator. That is how SA 
noise is introduced, into the loop of the transmitter oscillator.

> So you can partly answer the question yourself with a simple
> experiment at home. Have you ever seen frequency jumps in
> any of your cheap Rb or Cs clocks at home? Even one? If so
> how big was the jump?

It takes a good reference to find it. The quality of oscillators have 
reduced the rate of jumps.

> Second, can you better define what you mean by "frequency
> jump"? This is a familiar but vague term. In some respects one
> can say that every crystal, that every oscillator has frequency
> jumps.

I think looking at this reference could aid, as it covers many of the 
other terms, so "frequency jump" becomes distinct from that of 1/f 
variants of noise sources.


In general, the UFFC educational material at:
can be a good start.

For aging, have a look at this survey:
However, the FCS 1991 PDF is better formated variant, but password 
protected on the UFFC site.

However, the more I read about aging effects, the less clear connection 
I see between material effects and frequency jumps.

> The frequency averaged over one tau is always different than
> than the frequency over the next tau, and the next, etc. If not,
> the ADEV would be zero. But as you know ADEV is never
> zero. The purpose of clock statistics is to show how close to
> zero it is. So it would be good for you to define your threshold
> of "jump" before we talk too much more about detecting them.

A jump should be considered that or those frequency samples which 
significantly deviates from what the various noise sources would allow 
for. Random walk noise produces frequency jump like features, but 
through propper ADEV and MDEV analysis the levels of various 1/f noises 
can be established and random walk deviations can be limited. So, those 
level shifts which is significantly higher than expected from these 
noise sources is "true" frequency jumps.

However, there are several systematic effects which can produce a 
frequency jump observation, which includes environmental effects. Let's 
assume that environmental effects is maintained within limits for a 

Then, crystal physics and mecanics, alongside with that of the 
electronics comes into play. However, it is from these sources we see 
much of the noise we already said we elimintated.


More information about the time-nuts mailing list