On Sat 2000-09-30T17:09:33 -0400, matsakis.demetrios_at_usno.navy.mil hath writ:
> I believe the cheapest clocks that can keep 1 second over a year are
> oven-controlled crystal oscillators. I don't know how much they cost, but I
> can refer you to John Vig's excellent tutorial
> (http://www.ieee-uffc.org/fc). Of course, commercial atomic standards
> (rubidium, cesium, maser) can also do this, and I have purchased clocks with
> rubidiums in them for a few K.
Without actually calling up sales reps to find out the cost of the
units, it looks to me like the best cheap clocks most likely to be
built into some other hardware are either the oven-controlled crystal
oscillators (OCXO) and the microcomputer-compensated crystal
oscillators (MCXO). Both come in PC board-mountable packages and the
tradeoffs are warm-up vs. power consumption. But in either case they
appear to be able to keep time to about 1s/year at best. And my
impression of the cost is that they are still too expensive to be
included in any device which does not have time keeping as one of its
principal purposes.
> It can also be pointed out that any electric clock does this because the
> line frequency is adjusted for this purpose.
I would agree if my local power grid had a reliability of 0.99999997,
but as I actually experience something more like 0.999, this is not
really useful for keeping a consumer-grade timepiece accurate to
1s/year.
> Also, any computer's clock
> will do this if Network Time Protocol (NTP,
> http://tycho.usno.navy.mil/ntp.html) is used
Except that the common underlying operating systems don't really
acknowledge the distinction between UTC and TAI either, thus getting
into trouble when leap seconds happen.
> and any GPS receiver will do this as well.
Presumably because GPS is the only widely-available time source whose
protocol clearly provides the distinction between TAI and UTC.
Anyway, it appears that no consumer-oriented device can truthfully be
said to be keeping either UTC or TAI unless it is reset from an
outside source by about a full second at least once per year.
(Clearly the choice of whether to reset to UTC or TAI should be made
wisely.) In such a system leap seconds can't really make a
difference. It is either effectively experiencing some whether or not
IERS inserts any, or it really doesn't care what time it is to that
accuracy.
--
Steve Allen UCO/Lick Observatory Santa Cruz, CA 95064
sla_at_ucolick.org Voice: +1 831 459 3046 http://www.ucolick.org/~sla
PGP: 1024/E46978C5 F6 78 D1 10 62 94 8F 2E 49 89 0E FE 26 B4 14 93
Received on Sun Oct 01 2000 - 22:27:14 PDT