[ntp:questions] Adjusting PPS offset

unruh unruh at wormhole.physics.ubc.ca
Sun Sep 4 17:11:09 UTC 2011


On 2011-09-04, A C <agcarver+ntp at acarver.net> wrote:
> On 9/3/2011 22:21, unruh wrote:
>> On 2011-09-03, A C<agcarver+ntp at acarver.net>  wrote:
>>> I'm still digging through code for PPS and NMEA plus the routines in the
>>> OS so that part is going to take a long time for me to figure out.  So
>>> for now I'm just going to stick with PPS only and the pool servers.  As
>>> for PPS itself, what is the proper (or at least most reasonable way) to
>>> adjust the time1 offset so that PPS agrees with the tick of UTC?
>>
>> And how would you decide what that offset was? Since the PPS is almost
>> certainly the most accurate time you have already.
>> (actually it is off by about 1us due to the interrupt handling.)
>
> Well, after leaving ntpd running for about a week straight, six network 
> servers have settled down to report an average 2 ms offset (according to 
> ntpq).  Each of the offsets is within about 0.5ms of each other.  The 
> offset shown on PPS was 0.000 with a small amount of jitter.
>
> I figured that if the entire collection of servers had settled into a 
> 2ms offset then the tick of PPS was probably off by that amount.
>
> If that's not a reasonable assumption then what would be?

It is a very unreasonable assumption. If your PPS is off by 2ms, throw
it away immediately. It is completely useless. In general PPS is off by
less than 1 us and probably 100ns or so. But the computer cannot run its
interrupt routine fast enough to make use of better than 1us or so
without veryspecial hardware. 
And 2ms for a bunch of network servers is bad. It seems that there is
something along the route out of your system which is delaying the
packets packets really badly but consistantly. 

What is the source of your PPS? 




More information about the questions mailing list