[ntp:questions] NTP architecture recommendation

Hal Murray hal-usenet at ip-64-139-1-69.sjc.megapath.net
Sun Aug 19 22:14:10 UTC 2007

>accuracies are about 100m with SA enabled and 20m without.  See also:

That number only holds for inexpensive commercial units.

You can do much better.

Surveyors get to cm or mm accuracy.  I'm not familiar with the
details, but there are several tricks that I know about.

Listening to L2 as well as L1 gives you a much better
correction for the delay in the ionosphere.  Using fancy
antennas (multipath rejection) in good locations helps too.

Another is differential GPS.  You put another unit on a known
location.  Nearby units will have the same error pattern.
If your known unit says it is off 100 ft to the north,
you can correct nearby units by subtracting that from
their reported position.

Another way is to simply average lots of data.  I think
they go for several hours, maybe up to 24.  There may be
some complicated post processing they can do.  Mumble.
I don't understand that area.

For timekeeping, GPS is used to compare atomic clocks around
the world.  Here is a fun read:
  Time Too Good to Be True
  Daniel Kleppner
  Physics Today, March 2006, page 10

These are my opinions, not necessarily my employer's.  I hate spam.

More information about the questions mailing list