[time-nuts] The 5MHz Sweet Spot

[Magnus Danielson]

Hi,

On 11/03/2013 05:22 PM, Tom Knox wrote:
> I am not the right person to explain, So anyone with more knowledge please feel free to jump in. I think basically the reasoning is as a single clock the system at some point it would need mantainance or repair. So time is maintained with an algorithm that monitors all the clocks and oscillators that make up the system in which each part contributes and if any part deviates or fails it can be removed. This system is then characterized periodically by F1 as required to keep the system at F! accuracy. 
Basically, the caesiums and hydrogen masers operate as fly-wheel
oscillators and they are sufficiently stable, such that the F1 only
needs to operate every once in a while since the clocks have sufficient
stability and what F1 provides is accurateness beyond what the
individual clocks in the ensemble provides. It also helps to make the
ensemble estimate accurate rather than the average of the continuous
clocks of the ensemble. NIST then produce a couple of time-scales as
phase and frequency adjusted from one of the clocks, which is steered to
be what the ensemble estimate is judging it to be. It is the output of
these which is then compared to other clocks, and these differences
along with the individual differences of the clocks which is reported to
BIPM. This is to show the overall principles rather than accurate detail.

So, F1 could be run more often, but having the flywheel oscillators that
it has, it doesn't have to. Besides, it's a lab standard, so it's
physical buildup is altered in the breaks, such that it's performance
have been improved through a long set of stepwise refinements. That's
why F1 has had increased performance during it's operational period.

Cheers,
Magnus