Unruh wrote: > All I was saying was that just because the time routines are capable of > recording time with nsec precision does not mean that that accuracy of > the clock is anywhere near that, and I have great difficulty believing > that the accuracy is any better than usec.
You are somewhat correct: Those ntp servers that do run at better than us precision/accuracy need some additional effort, i.e. phk's replacing the motherboard crystal with OCXO/TCXO/Rb oscillators, and then using an FPGA to measure the offset between the arrival of the PPS signal and the dispatch of the interrupt handler. Even without the FPGA it is possible to do significantly better than 1us, if you let the interrupt handler twiddle an output pin so that you can use an oscilloscope to measure the average latency, and then fudge that away. NTPD with a PPS signal does not need all interrupts to be handled very quickly, only that a large fraction of them have short & relatively stable service time. Statistics will do the rest. Terje -- - <Terje.Mathisen at tmsw.no> "almost all programming can be viewed as an exercise in caching" _______________________________________________ questions mailing list questions@lists.ntp.org https://lists.ntp.org/mailman/listinfo/questions
