[time-nuts] NTP linked PC clock is slightly ahead of Lady Heather GPS time
Tom Van Baak
tvb at LeapSecond.com
Thu Oct 12 12:04:51 EDT 2017
>> These days the speaker is almost the last usable interface for this.
> ….. and the mic input is one of the few inputs left to feed “events” into.
Yes, and by connecting the speaker to the mic and issuing a pulse through the speaker you can estimate the latency of the OS. It should act like an echo chamber and the number of echoes per second is your inverse latency.
The other trick is to use a TIC to compare the output pulse(s) against your house GPS/1PPS. You then compare the timestamp that the OS thought it was with the real UTC+TIC timestamp. Similarly, put a GPS/1PPS into the mic and compare the time the OS thinks it saw the pulse.
The results of these should give a nice pair of UTC offset and jitter plots. If at the same time you also measure the XO directly you can then subtract and pinpoint how much of the error is due to the XO and how much is due to OS design. For extra credit put NTP into the mix and see how badly NTP s/w performs compared to what a h/w solution could so.
More information about the time-nuts