[time-nuts] Disciplining Rubidium

Bruce Griffiths bruce.griffiths at xtra.co.nz
Fri Apr 25 04:36:38 EDT 2008

SAIDJACK at aol.com wrote:
> Hi Bruce,
> the 1ns performance per second was mentioned by someone else in the  previous 
> threads.
> When you lock a BVA OCXO to carrier phase, I would still expect the PLL  loop 
> time constant to be >>20s, thus the 8E-014 / 2s you mentioned is very  likely 
> the performance of the BVA Crystal, not likely the carrier phase  real-time 
> system output.
More complete specs are:
8E-14 @ 1 sec
8E-14 @ 10 sec
2E-13 @ 100 sec
5E-13 @ 1000 sec
4E-14 @ 10,000 sec
1E-14 @ 1 day

Loop time constant is probably somewhere around 1000 sec or so.

> For example, our Fury with double oven has an ADEV of a couple  of parts per 
> E-012 1s to 20s, but that performance is entirely  generated by the OCXO, not 
> the M12+.
> I would expect the carrier phase to improve things at the point where the  
> BVA starts having a rising ADEV. Maybe around a couple 100s? If for  example the 
> phase comparator has 1ps resolution (that's really quite a  high resolution) 
> to compare the OCXO and carrier phases (1E-012) then it would  take >12s 
> averaging intervals in the PLL just to prevent the  measurement errors from 
> affecting the system performance above 8E-014.
> It's late, and I may just be wrong about all this.
> bye,
> Said
Yes the carrier phase measurement resolution is very high (typically 1/4000 of a ~635ps L1 carrier period) however the measurement noise is around 20ps the oscillator performance is so good in comparison a very large time constant is used.

However if one had an OCXO whose performance ADEV was around 1E-11 for tau from 1 to 100 sec or so then a much shorter time constant would be appropriate. 

When the GPS receiver LO is phase locked to the OCXO, the GPS receiver has all the hardware necessary to do the phase comparisons using either via code phase or carrier phase methods.

The following Chinese implementation appears to use a commercial receiver, they just replaced its crystal oscillator:


They also use a neural net for modeling prediction of atmospheric delays.

Perhaps the ultimate is to implement a custom GPS receiver (apart from the RF front end and ADC) in a large FPGA.
One could then use L1, L2C, L5 etc plus the Galileo signals when they become available.


More information about the time-nuts mailing list