[time-nuts] Need Time Help

Chris Albertson albertson.chris at gmail.com
Wed Oct 5 19:20:29 EDT 2016

First off you are right, NTP does not run well as it can with only a single
reference.   But that is a general statement and assumes there is some
error in a reference clock.   Certainly if using Internet pool servers as
reference clocks you want to have 5 to 7 of them.  But GPS is so darn good
that NTP will be unable to find the difference between any two GPS
receivers.   GPS is a special case.    Remember the NTP can use about a
dozen different kinds of reference clocks, even if today only a few are
used in practice.  (really who uses WWV as a standard for their computer

Of course there are other ways of getting time into a Windows PC then using
NTP.  The most primitive thing possible is to simply "jam set" the PC clock
from a NEMA sentence.  This can be as much as a full second "off" and has a
50% chance of making your PC clock run backward (if it was fast rather then
slow)  but LOADS of software offer to do just this, Lady Heather included.

But if you want something that avoids all those problems and makes the
clock perform reasonable at all times then there really are two options NTP
and PTP.  If you are using a GPS as a reference Either if OK,

The problem is harder then most people think. To avoid jumps in time either
forward or backwards the software must be something that runs continuously
and monitors your clock vs. one or more reference clocks.   Logically there
is no other way.  The correction process must be continuously.   PTP and
NTP do this well enough that I doubt anyone would bother to write something
to compete with them.

Between the two PTP performs best in the special case that you own all of
the hardware on both ends and the communications network.

Most casual users can live with a system that "jam-sets" the clock as they
are just reading the display and don't care if  (say) 14:45:30.345 happens
AFTER 14:45:30.355    But if you are doing something like trying to aim a
telescope such a thing will cause it to jump and ruin a time exposure.
But I doubt an EME antenna needs to be pointed with the same arc second
level precision.

The BEST way to get GPS time into a computer, and this is how, I think the
world record accuracy was done is to use a GPSDO in place of the cheap
crystal on the motherboard and then spend many hours comparing the
computer's clock to GPS, sorry I lost the link to this  But this method
gets you fractions of microseconds.  The worst way is to parse NEMA
sentences and jam-set them into the clock.  This gets you to roughly one
second, more or less.

On Wed, Oct 5, 2016 at 8:19 AM, Chris Caudle <chris at chriscaudle.org> wrote:

> On Wed, October 5, 2016 6:14 am, Bob Camp wrote:
> > If you buy a GPS receiver and get it set up for timing, just use it.
> > Then there is no need for NTP at all.
> Is there another way to get computer system time set from a GPS receiver
> other than NTP?
> In the case that the system clock is controlled by GPSDO and the seconds
> delineated by PPS, there should be no need for the NTP clock discipline
> code, but I am not aware of any way to inform the NTP daemon that no
> disciplining is needed.  Presumably the code should determine that
> eventually.
> In the case that the system clock is free running, the clock discipline is
> still needed, but I found a note in one of the NTP documents (that I can
> no longer locate at the moment) that stated something to the effect that
> NTP did not run well as it could with a single reference, which would seem
> to directly affect the case where a GPS receiver is the only reference.
> That document had only that short note, no details on why or specifics of
> behavior.
> --
> Chris Caudle
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/
> mailman/listinfo/time-nuts
> and follow the instructions there.


Chris Albertson
Redondo Beach, California

More information about the time-nuts mailing list