[time-nuts] Another "atomic" clock question
Tom Van Baak
tvb at LeapSecond.com
Thu Mar 6 10:45:47 EST 2014
> Tom and Bob,
> It is not obvious to me that it is "easier" to simply apply a correction in nS increments
> with a range as wide as 100nS. How is this done? Using switched delay lines or delay gates?
If you intend to measure the 1PPS there is no need to correct or adjust it prior to measurement. Each second you simply apply the numerical sawtooth correction value to the numerical 1PPS measurement value. This is the pure software solution. Most people who use GPS for timing do it this way -- since they are already employing a sub-ns TIC to compare their standalone lab reference against the GPS tick.
The software solution introduces no additional errors. This method also applies to any GPSDO which incorporates a digital TIC.
On the other hand, if you intend to improve the accuracy of the 1PPS without measurement, you need a hardware solution instead. The classic approach is to delay, each second, the hardware 1PPS by N + sawtooth correction. You choose N (depends on the GPS receiver) so that the delay is never less than or too near zero. The one-chip solution I found was the Dallas DS1020 and that's what Rick used in his CNS-II product. Maxim (bought Dallas) now has alternative silicon delay lines that do the equivalent.
Note the hardware solution is never quite as good as the software solution since there are offset, gain, linearity, and tempco issues with programmable delay lines. But it's usually close enough. Rick measured the difference between the two methods: 0.7 ns rms.
See page 27-31 of http://www.cnssys.com/files/tow-time2009.pdf for details and his wonderful graphs.
More information about the time-nuts