[time-nuts] Best Rubidium Frequency Standard

Charles Steinmetz csteinmetz at yandex.com
Sun Mar 13 16:21:07 EDT 2016


Poul-Henning wrote:

>At least some of the firmware versions do not allow you to change the
>PPSIN offset second by second, so you cannot compensate for the
>"negative sawtooth", and the resulting "hanging bridges" means that
>you have to us a very unoptimal PLL time constant.

The cure for that is to use a GPS that has a clock synchronous with 
PPS -- like a TBolt.  (Why anyone does it any other way baffles me 
completely, but they do.)

And the long time constant is not "unoptimal" -- it is optimized for 
best stability.  GPS stability (not including any uncorrected 
sawtooth) is about 5x10e-9 at tau = one second, improving by 10x per 
decade at longer averaging times.  Most oscillators a time-nut would 
be interested in start at around e-11 at one second, improving for 2 
to 5 decades (OCXO and Rb, respectively -- generally improving as a 
root per decade), then worsening at longer tau.  At some point, these 
stability curves cross.  The whole point of GPS discipline is, at 
each tau, to follow whichever curve is more stable by crossing over 
from one to the other.  This results in a TC of typically 100-1000 
seconds for disciplined OCXOs, or 1000-100k seconds for disciplined Rbs.

So, the TC that is optimum for a disciplined OCXO is NOT optimum for 
a disciplined Rb, assuming what one wants is the best possible 
stability, at all tau, that the particular GPS and oscillator can 
provide.  The long TC needed to average out an uncorrected sawtooth 
in the GPSDRb would be chosen as the optimum TC for stability anyway.

Best regards,

Charles




More information about the time-nuts mailing list