[ntp:questions] Frequency Offset
snews at lordynet.org
Fri Feb 24 08:33:52 UTC 2012
> On 2012-02-23, Alby VA <albyva at empire.org> wrote:
>> On Feb 23, 4:29?pm, unruh <un... at invalid.ca> wrote:
>>> No, that is a very typical figure for the frequency offset. Remember
>>> that the crystals used to control the timing in computers are not
>>> supposed to be terribly accurate. (They are chosen to be cheap, not
>>> It is because such frequency offsets exist that ntp was invented to
>> When you say "crystals used to control the timing in computers",
>> are you referring to the parts that make up my actual FreeBSD
>> Server (ie: Motherboard)? Or Parts in the GPS Device (ie: Sure Elec.
> On your motherboard.
>> Or in the Satellites receiving the GPS signal?
>> I'm just trying to gauge, what hardware should be looking to obtain
>> improve that frequency offset. Or is it a moot point unless I'm going
>> invest in some high dollar atomic clock?
> An atomic clock.
>> Or better yet, can I buy quality crystals from DigiKey and do some
>> soldering in whatever devices that need crystal upgrading?
> And why do you care? As I said this is waht ntp was invented for. It is
> doing its job.
When I started using ntpd it didn't work with some of my
systems as the clocks were too far out, >> 50ppm, and/or
too temperature sensitive. Some of my motherboards had a
3-pin jumper already fitted so that an external clock
source could be connected.
One of my desktop systems, p4-2667, has just taken two
days to get to an offset of under 2 ms after a kernel
change. I suspect because that coincided with the hottest
day in February since 1998. All my pcs were affected
except the one connected to PPS from a Sure GPS. The one
with an MSF source, no PPS, was also fighting to keep sync.
> Note that even all of the time standard labs around the world use
> programs like ntp to track their frequency offsets.
> And all atomic clocks have them.
More information about the questions