Re: [LEAPSECS] The real problem with leap seconds

From: Tom Van Baak <>
Date: Sun, 8 Jan 2006 13:23:42 -0800

> : As I understood it, it was mainly that TAI is a post-factum "postal"
> : timescale.
> How is it that UTC can be realized in realtime, but TAI isn't. I
> thought the difference between the two was an integral number of
> seconds, by definition. Is that understanding flawed?
> Wanrer

Not flawed. But recognize that the words UTC or TAI
have more than one meaning.

Strictly speaking UTC isn't real-time either. However,
individual laboratories produce a real-time realization
of UTC, in the form of UTC(k) where k = NIST, USNO,
NPL, etc. which is then steered, over time, toward or
around the mean. Your own lab or home may have its
own realization of UTC, such as Doug's UTC(DWH)
or my UTC(TVB).

Note there has been strong resistance to the use of
the phrase UTC(GPS). True, there exists UTC(USNO)
but what a GPS navigation or timing receiver delivers
is many layers of uncalibrated phase and frequency
offset and uncertainty from UTC(USNO) which it itself
just an good estimate of UTC. This is all at the
sub-microseconds level.

Up at the seconds level the generic terms "UTC time",
"TAI time", "GPS time", or, for that matter, "local time",
are simply a matter of integer seconds offset; they all
nominally follow the same synchronized seconds
epoch (1PPS).

So when you see "UTC" or "TAI" you have to infer the
context. Is the timescale; is it seconds; is it nanoseconds?

The danger is mixing the generic use of the acronym
(meaning seconds offset) with a more technical use
of the word (meaning the realization of the timescale).

I can easily set a cesium clock or wristwatch to display
in UTC, TAI, GPS, EST, or local time. The ticks occur
at the same time. The hands or digits differ. But does
this mean my UTC(TVB) clock can produce UTC? or
TAI? Yes, it can produce a display that matches TAI to
some finite accuracy. No, it does not produce TAI itself.

If I had to guess, the resistance to the use of the word
TAI by the BIPM or others is that in the past and still
today the word TAI represents *first* the notion of a
well-organized international ensemble of atomic clocks.
Only secondarily does the TAI convey something as
mundane as an N second offset on your wristwatch.

Mixing the two meanings can lead to trouble. For
example, occasionally you hear that Galileo is
somehow better than GPS because it uses TAI time
instead of GPS time. That's bogus. To the second,
any existing WWVB or DCF77 or GPS or Glonass
or Galileo receiver can display time as TAI or GPS
or UTC, or EST, or sidereal. Clever ones can even
handle DST. But at the nanosecond level Galileo
can no more deliver TAI than pizza to your receiver.

Received on Sun Jan 08 2006 - 13:25:41 PST

This archive was generated by hypermail 2.3.0 : Sat Sep 04 2010 - 09:44:55 PDT