[ntp:questions] GPS-UTC Leap Corrections
"terje.mathisen at tmsw.no" at ntp.org
Mon May 4 20:35:14 UTC 2009
> We have some time servers with GPS clocks. One of their uses is in
> operations centers where satellites are being flown. The satellites
> GPS (as opposed to UTC) internally, so we want the computers on the
> ground also to use GPS (rather than UTC). The time servers speak NTP
> and hand out GPS time as if it were UTC. The ground computers think
> they are on UTC, but it's really GPS.
> But we'd like some of the computers to really be on UTC. I really
> been able to think of a good way to have computers use the GPS time
> given out by the time servers as a source, but then to actually offset
> their own time by the GPS-UTC difference. Some quick Googling hasn't
> turned anything up. Anyone have some suggestions on how to manage
> something like this? Or are we just going to have to live 15 seconds
> ahead of the rest of the world?
This is one more vote for my long-standing request for the ability to
use the 'fudge' command on real (network) servers, and not just refclocks.
It is relatively easy to do, it just means that you have to add/subtract
the required offset when handling all times from the relevant network
I.e. when server utc1 asks gps1 for the time, it needs to do the following:
1) Do a completely normal t1/t2/t3/t4 NTP request and calculate the
2) Adjust the offset from (1) by adding/subtracting the fudge time.
3) Use the resulting value as in the current code.
I.e. only step 2 is new.
You do have to make very sure that none of your GPS-based servers will
ever serve time to normal ntp clients, that would really mess them up!
PS. The easiest (and also ugliest!) hack would be to take the ntp source
code and add a single line to the server clock calculation code where
you add the current offset. You _must_ recompile/fix this value on every
upcoming leap second though!
- <Terje.Mathisen at tmsw.no>
"almost all programming can be viewed as an exercise in caching"
More information about the questions