[time-nuts] GPS clock stabilitiy, Rb vs Cs

[Magnus Danielson]

On 05/05/2013 10:05 AM, Attila Kinali wrote:
> On Sat, 4 May 2013 12:36:20 -0700
> "Tom Van Baak (lab)"<tvb at leapsecond.com>  wrote:
>
>> Rule of thumb: quartz is best short term, Rb or H-maser mid-term,
>> and Cs by far the best long-term.
>
> Ah.. so it's a fundamental limitation. And i was looking for something
> GPS specific.
>
> Any references i could read on those limitations? A quick google
> did not produce any good results.

There is a handful of references but picking up a book like "Quantum 
Leap" is a good start.

Quartz is a bit of (syntetic) rock, cut at some angle(s), cleaned, 
mounted in some hermetic sealed chamber with residue dirt, and mounting 
that over time shifts the stress of the crystal. The various processes 
create a long-term frequency drift (plots over 5 years shows the same 
systematic drift, non-linear). Oh, and if you shift temperature of the 
crystal, it has to re-align. The parameters of this systematic behaviour 
is individual, and by itself it is not able to handle these.

For rubidium gas-cell, there is a bunch of systematics, including 
darkening of the pumping Rb lamp, which causes shift in light intensity 
pull, temperature miss-alignments changes, causing the lamp and 
filtering cell to shift, causing de-pumping to change over time as well 
as light intensity. The rubidium reference cell has buffer gas, which 
leaks, and the buffer gas is there to reduce the speed of the rubidium 
atoms, such that they don't hit the glas wall with inevidable shift in 
frequency that has. Oh, the wall shift changes with temperature, the 
buffer gas causes a shift, which also shifts with temperature... the 
smart guys make them more or less cancel, but this cancellation shifts 
over time as some of the buffer gas escapes, especially Helium. Oh, and 
the resonator around the rubidium reference cell pulls the frequency, 
and that changes with temperature as well. No wonder that things are 
being temperature stabilized using ovens, which in itself is the major 
power consumption source of a rubidium gas-cell.

The caesium atomic beam does not have wall-shifts, but rather it has
much lower systematics. One of the major onces being magnetic field. 
Assuming that the tube is degaussed, the C-field adjustment is 
troublesome. Old caesiums had them rather arbitrary set, but you could 
adjust them to national references, so first generations where really 
secondary sources. By looking at side-bands, this can be servoed and 
C-field steered and hence the shift understood, and by that basis for a 
primary reference can be found. Then, there is wear mechanisms that 
kicks in on the physical package.

The above is a summary of things collected from a variety of sources, 
but I think this coarse walk-through of issues gives some insight as to 
what issues pops up where and the milage vary a lot within each group. 
Modern high-performance rubidium gas-cells outperform the early 
caesiums, high-performance crystals outperform several rubidiums.
The HP5065A is an example of an old clock with really good performance, 
so modern is not everything, and the modern compact telecom rubidiums 
and for that mater CSAC is more space/power oriented than ultimate 
performance of the technology as such.

Cheers,
Magnus