[ntp:questions] question about logs of PPS

Richard B. Gilbert rgilbert88 at comcast.net
Tue May 15 11:45:23 UTC 2007

Christian Pinedo Zamalloa wrote:
> On 2007-05-09, Christian Pinedo Zamalloa 
> <christian.pinedo.zamalloaNOSPAM at NOKKgmail.com> wrote:
>>hi all,
>>I'm a newbiew on NTP services, but anyway I'm trying to setup a Stratum1
>>server with the following hardware and software:
>>- GPS Motorola Oncore M12
>>- Server:
>>    - CentOS 3.8
>>    - Linux kernel
>>    - ntp 4.2.2p4
>>    - PPSKit 2.1.7
> After reading emails of the mailing list and searching the web i decided
> to migrate the ntp server to FreeBSD 6.1. I enabled PPS_SYNC in the
> kernel, compiled the kernel and the world, and configured again the ntp
> daemon. The result is ok, the server is running better and it has
> been announced as a stratum one public server in ntp.isc.org wiki.
> Anyway, i would like to ask to the users of a Motorola M12 receiver what
> should be the accuracy of the time server. Reading the logs and using
> ntpq tool i see that is from several useg to hundred useg. I though that
> it could reach nanoseg. Was I wrong?

The Motorola M12+T is specified to deliver its Pulse Per Second output 
with the leading edge within 50 nanoseconds of the top of the second. 
Getting that into your computer is where the problem lies.  Things like 
cable length can be compensated for.  Interrupt latency is more 
difficult because it's difficult to determine exactly what value you 
need to compensate for.  Depending on interrupt priorities and what else 
is going on, the latency may not even be constant.

I'd say that you are probably within one microsecond of the correct time 
if you are using the Motorola M12+T as a reference clock.

More information about the questions mailing list