Hi Tani, > Q: Why is there an timing offset of 20ns average between identical > commercial off-the-shelf GPS modules?
Without testing I can't tell if 20 ns is normal for your receiver or not. It doesn't seem totally out of the ordinary but some investigation would be in order in case part of the problem is on your end. The paper I mentioned is: Critical Evaluation of the Motorola M12+ GPS Timing Receiver vs. the Master Clock at the United States Naval Observatory, Washington DC http://www.gpstime.com/files/PTTI/PTTI_2002_CNS_Testbed.pdf Here you can see even with a popular, vintage, high-end GPS timing receiver there are variations in time offset. A similar thing happened with they replaced the M12 with a ublox 6T: Low cost GPS-based time and frequency products, an update http://www.gpstime.com/files/PTTI/Low_cost_GPS-based_time_and_frequency_products.pdf One thing we time-nuts can do is to measure N of your favorite GPS timing receivers and get a feel for the unit-to-unit variations. It would not surprise me if some are consistent at the 5 ns level while others at the 20 ns or 50 ns level. To further complicate matters, it sometimes also depends on the f/w version. So precise, repeatable, absolute time is much harder to know for commodity "timing" receivers. Fortunately a GPSDO never actually needs to worry about this -- since it's all based on relative time, not absolute. > Q: When it comes to using a GPSDO to measure timing performance (Sigma > -Tau) any hiccups on GPS could make this calculation fail and you would In general with a GPSDO, if you find a gradual departure in phase between GPS and your OCXO you can't be sure if the error is due to the GPS receiver or with the OCXO. This is all part of how a GPSDO works -- it makes an informed guess, and this is where filtering algorithms comes in. The *assumption* is that if there is a gradual, sustained change in phase that it's the OCXO's fault and so it is corrected to "close the loop". All this to maintain the 10 MHz output as close as possible in real-time. The feature I mentioned is applicable to situations where one is making time/frequency measurements over the long-term. In this case there's no point in having a GPSDO make assumptions in real-time. Instead it's easier to just collect 1PPS phase differences and make retroactive corrections to the numerical phase or frequency counter readings after-the-fact. For example, glitches are easier to detect and remove if you have a full before/after history. And "disciplining" with post-facto numerical corrections to frequency readings are more accurate than having to correct an OCXO using a DAC and EFC h/w. For example, in my lab, I tend not to use a GPSDO to compare a high-end clocks (Rb, Cs, H-Maser). Instead I just use a plain GPS/1PPS receiver. For these sorts of standards, a GPSDO can add nothing but noise to a measurement. But this "trick" only works if you are content to measure after-the-fact, not in real-time. Does that help? Or further confuse things... /tvb ----- Original Message ----- From: "Estanislao Aguayo" <[email protected]> To: <[email protected]> Sent: Tuesday, January 26, 2016 9:58 AM Subject: [time-nuts] Tom Van Baak: a conversation about GPS timing > Hello Time-nuts mailing list, > > Let me introduce myself, my name is Tani Aguayo, I am a physicist > working on precision clocks for Stanford Research Systems (SRS). We had > the pleasure of having Tom Van Baak visit our lab yesterday, and I want > to write down some of the topics we talked about, and I thought this > list would be a good place to keep it, and share it. And also have Tom > go over it, and correct me if I got something wrong. > > We talked GPS, and here is a couple of thought inducing comments from Tom: > > Q: Why is there an timing offset of 20ns average between identical > commercial off-the-shelf GPS modules? > TVB: A low-cost (~$25-$50) GPS receiver gathers timing information via a > data lock to the local oscillator in the receiver, meaning the the > timing information is based on data passed to the receiver and then the > receiver uses that to produce a signal based on its own oscillator. This > is one part of what can explain the differences between units, the > stability of the receiver local clock will mask the precision timing of > the GPS 1 pps. The other part is that this timing data is based on the > GPS antenna parameters, and so any disturbing effects (i.e. cable > tempcos, impedance mismatches, active filtering in the antenna that > could introduce group delays... ) on the transmission path from antenna > to receiver, will affect the timing precision of the GPS 1pps at the > receiver with respect to UTC. This effect could change from unit to > unit, as different antennas would have different effects on the signal. > > Q: When it comes to using a GPSDO to measure timing performance (Sigma > -Tau) any hiccups on GPS could make this calculation fail and you would > not know until after the fact, is there a way around this dependence? > TVB: It is hard to get a beautiful 10MHz and see how in a matter of > days, the timing quality is lost due to aging, tempcos, GPS hiccups... > you name it. A better approach is to keep a log of your clock against > GPS (without locking to it) and use an offline approach to compute your > timing performance (Sigma-Tau) and this way you can look at your clock > performance without mixing the GPS locking loop parameters. > > Thank Tom for sharing this insight with us! > > - Tani Aguayo > > > > > > > _______________________________________________ > time-nuts mailing list -- [email protected] > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
