A long time ago, I did something very similar with an early release Magellan 5000 board (and a couple of other models of GPS cards) that did not implement the 1PPS signal. I think the best time sync that I ever got was around 1 mS (most boards were around 10 mS). True 1PPS signals are implemented and synchronized in hardware. The NMEA messages are generated in software. Due to the vagaries of things like software interrupt latencies, serial port bit synchronization, etc I doubt that you could ever get better than 1 serial bit time of accuracy out of any board and 10-100 bit times is more likely. At 4800 bits/sec the jitter/offset of even 1 bit time would be huge at best. Perfectly OK for things like clock displays, but totally inadequate for things like oscillator locking. _________________________________________________________________ Send e-mail faster without improving your typing skills. http://windowslive.com/online/hotmail?ocid=TXT_TAGLM_WL_hotmail_acq_speed_122008

Mark, 2008/12/30 Mark Sims <holrum@hotmail.com>: > > A long time ago, I did something very similar with an early release Magellan 5000 board (and a couple of other models of GPS cards) that did not implement the 1PPS signal. I think the best time sync that I ever got was around 1 mS (most boards were around 10 mS). I'm not actually looking at time syncing, only making a GPSDO. > True 1PPS signals are implemented and synchronized in hardware. The NMEA messages are generated in software. Due to the vagaries of things like software interrupt latencies, serial port bit synchronization, etc I doubt that you could ever get better than 1 serial bit time of accuracy out of any board and 10-100 bit times is more likely. At 4800 bits/sec the jitter/offset of even 1 bit time would be huge at best. Perfectly OK for things like clock displays, but totally inadequate for things like oscillator locking. Ok, so I can see what you mean about jitter but would that not be taken care of by a low pass filter with a reasonable time constant following phase-frequency detection between the PPS and a divided down ocxo? Agreed the short term stability (ADEV for small tao?) would be poor but would it not improve significantly in the longer term (ADEV for large tao?)? As for offset against an absolute PPS synced to GPS time, well this is not important to me. 73, Steve -- Steve Rooke - ZL3TUV & G8KVD & JAKDTTNW Omnium finis imminet

Steve Rooke wrote: > Mark, > > 2008/12/30 Mark Sims <holrum@hotmail.com>: > >> A long time ago, I did something very similar with an early release Magellan 5000 board (and a couple of other models of GPS cards) that did not implement the 1PPS signal. I think the best time sync that I ever got was around 1 mS (most boards were around 10 mS). >> > > I'm not actually looking at time syncing, only making a GPSDO. > > >> True 1PPS signals are implemented and synchronized in hardware. The NMEA messages are generated in software. Due to the vagaries of things like software interrupt latencies, serial port bit synchronization, etc I doubt that you could ever get better than 1 serial bit time of accuracy out of any board and 10-100 bit times is more likely. At 4800 bits/sec the jitter/offset of even 1 bit time would be huge at best. Perfectly OK for things like clock displays, but totally inadequate for things like oscillator locking. >> > > Ok, so I can see what you mean about jitter but would that not be > taken care of by a low pass filter with a reasonable time constant > following phase-frequency detection between the PPS and a divided down > ocxo? Agreed the short term stability (ADEV for small tao?) would be > poor but would it not improve significantly in the longer term (ADEV > for large tao?)? As for offset against an absolute PPS synced to GPS > time, well this is not important to me. > > 73, Steve > Steve Instead of waxing lyrically try doing some estimates of the averaging time required for a given accuracy: You need to measure the ADEV of the reference source (ie the relevant NMEA sentence). As a crude model say the jitter is 10 millisec of white phase noise and you want the GPSDO to be within 1ppm for averaging times from 1 sec to a time equal to the loop time constant then the loop time constant will have to be about 3 hours. If you hope to achieve 1E-9 averaging over something like 3000 hours will be required. Whilst one can achieve lower noise by averaging the GPSDO over a day or so, this may not be what you want. Aging and thermal drift of the OCXO over 3000 hours will make it impractical to achieve a stability of 1E-9. Bruce