[time-nuts] GPS DO Alternatives
Tom Van Baak (Lab)
tvb at leapsecond.com
Sun Dec 9 03:31:56 UTC 2012
Bob, Charles, Chris,
If the goal is a GPS locked frequency counter then there is no need for disciplining the LO at all. Just let the XO or OCXO or Rb free-run. Use a two channel counter: one capture input to measure the frequency of the DUT and another capture input to leisurely and independently monitor the 1PPS.
Over time the counter will know the frequency error (and drift) rate of the LO relative to GPS, which it then applies as a *math* correction to all DUT frequency readings before the user sees them. I do this all the time -- an XO-based picPET measuring a GPS 1PPS is really measuring the XO using GPS, not the other way around.
It takes one line of code to apply a *(1+e) correction to a reading. The averaging interval over which e is calculated replaces the GPSDO time constant. Effectively you are disciplining the readings rather than the oscillator. No DAC used. This also allows you to use oscillators that lack EFC. The LO doesn't need to be a nice number like 10 MHz either. Saving e in EEPROM is effectively holdover mode.
This GPSDFC is simpler than a GPSDO-frequency counter but delivers the same accuracy. Note any stable 1PPS source will do; doesn't have to be GPS. The quality of the LO determines the short gate time stability. The accuracy of the 1PPS determines the long gate time accuracy.
More information about the time-nuts