[time-nuts] LPRO-101 with Brooks Shera's GPS locking circuit
SAIDJACK at aol.com
SAIDJACK at aol.com
Sat Dec 16 23:20:40 EST 2006
good details on how to set the time constant for best GPSDO performance etc!
Some issues you did not mention are but that are essential to get a good
* Aging compensation
* Temperature compensation
* fault recovery, such as mechanical shock to crystal, power surge,
acceleration, thermal shock, atomic particle impact (pop noise) etc.
=> Aging is mostly constant over periods of several days, except after first
cold power-on. I just received a plot from an OCXO factory showing 1E-07
aging the first day, asymptotically slowing down to <1E-010 after 10 days! The
oscillator is rated at 1E-010 per day (which is quite good) but clearly does
not meet this spec in the first couple of days.
Aging will show up as a constant offset as the GPS and the DAC is constantly
"chasing" the crystal. The slope of the change in DAC voltage over time can
be measured (but only after it has been temperature corrected!!) and then the
software can compensate for aging so the TIC/GPS/Loop filter doesen't have
to do it. This helps tremendously for maintaining stability during hold-over
as well. I have seen some cheaper OCXO's age more than 1E-013 per second after
being on for weeks and that's not a bad spec either!
=> Temperature compensation is helpful even on a near perfect oscillator (Rb
or even Cesium!) because as long as the unit is being corrected by a voltage
from an external DAC, the DAC, and Dac reference will have temperature
dependencies that can be corrected. The problem becomes measuring temperature
changes on the order of 0.001 Degrees Celsius resolution...
Temperature compensation should be done for Aging prediction/compensation to
work well. Otherwise one would not know if the drift is caused by aging or
Electronic temp compensation works great to make a good OCXO out of a
mediocre one, but the temperature sensor's thermal response must match the OCXO's
(and DAC/Reference) response as well as possible, otherwise the unit will be
=> Some non-linear processing can be done to speed up fault recovery. The
problem then is: is it a GPS induced error, or an OCXO induced problem? Fuzzy
logic helps to get the OCXO locked quickly when a big jump happens by for
example momentarily changing the loop time constant for fast locking, then slowly
going back to the intercept point on the ADEV curve you mentioned.
More information about the time-nuts