Re: [LEAPSECS] what time is it, legally?

From: Rob Seaman <>
Date: Tue, 12 Dec 2006 17:14:33 -0700

I'm given to wonder how much of the friction on this mailing list is
simply due to the shortcomings in the technology that implements it.
I've appended a message I sent in August with four plots attached.
Can someone tell me whether it is readable now or was successfully
delivered back then? I rummaged around on the list archive and on
archives accessibly via google and find no copy of this message that
survived the communications medium.

On Dec 12, 2006, at 2:17 PM, Tom Van Baak wrote:

> Is there a technical definition of the "mean" in "mean solar
> time" that would help guide the discussion?

See the appended message. There appears to be a natural excursion of
several minutes – even in the absence of first order lunar effects
– in accumulated "leap" offset over the course of several
centuries. Undoubtedly an expert could wax poetic on this subject
should one care to speak up. Perhaps this natural variability could
be used to start to wrestle with the issue.

> One could argue that adding 50 or 100 leap milliseconds a
> few times a year (as was done in the 60's) to preserve the
> mean is just as valid as adding a couple of leap seconds
> every few years (as is done now) is just as valid as adding
> a couple leap hours every few thousand years (as has been
> proposed).

I'm with you for the first two, but not the third. An approximation
that is as large as the width of a timezone is equivalent to
eliminating timezones.

> I'm not arguing for one over the other but it seems to me
> all three models achieve a mean.

See my previous message. (Assuming it was delivered.)

> None of them prevent secular drift.

Secular means you never zero it out. Pretending that we can get away
with that is where the ALHP fails.


Begin forwarded message:

> From: Rob Seaman <>
> Date: August 5, 2006 6:47:29 PM MST
> To: Leap Seconds Issues <LEAPSECS_at_ROM.USNO.NAVY.MIL>
> Subject: Re: trading amplitude for scheduling
> John Cowan wrote:
>> Rob Seaman scripsit:
>>> Third result - even in the absence of lunar braking, leap jumps
>>> (or equivalent clock adjustments) would remain necessary.
>> Why is that?
>> If the SI second were properly tuned to the mean solar day, and the
>> secular slowing were eliminated, there would be no need to mess
>> about with
>> the civil time scale, because the random accelerations and
>> decelerations
>> would cancel out in the long run. Of course, we'd have to
>> tolerate larger
>> differences between clock time and terrestrial time, but we'd
>> expect that.
> Excellent discussion. The answer depends on how much larger the
> clock differences are, and on the meaning of the word "tolerate".
> As Tom Van Baak said:
>> My understanding is that, in addition to astronomical
>> effects (lunar/solar tides), no small number of geological
>> and climatological phenomena also contribute to the
>> instability of the mean solar day. That all the random
>> accelerations exactly cancel all the random decelerations
>> in any finite time, short- or long-term, is very unlikely.
> Which is to say that "proper tuning" may not even have a
> definition. It certainly is non-trivial.
> Consider the historical trend (from
> leapsecs/dutc.html):

> Detrend the data by removing the 1.7 ms/cy secular effect:

> (I read the LOD from the plot every century - should repeat with
> the original data, but results should be acceptably accurate. I'm
> sure somebody would be happy as a clam to point out any errors I
> may have made :-)
> There are positive and negative excursions from "normal" that
> persist for centuries. For the purpose of civil timekeeping, we
> don't care what geophysics causes these excursions, or even whether
> the rather evident sinusoid is real or not, but just that the
> residual ~ +/- 5 ms length-of-day variations exist.
> Leap seconds represent the accumulation of these daily residuals:

> A very small daily residual becomes +/- 9 minute descrepancy
> between TAI and UTC over millennial time periods. So even in the
> absence of the secular trend, the natural geophysical irascibility
> of the planet is very evident. Leap seconds - both positive and
> negative, of course - would be needed to resync the clocks. I
> count about 2200 over 2500 years (for instance, about 500 leap
> seconds between the battle of Hastings in 1066 and the most recent
> eruption of Fujiyama in 1707). That amounts to about one leap
> second per year even in the absence of the secular lunar effect.
> Since this entire range of phase space sits within an hour's extent
> from the "mean", the ALHP is really a proposal to completely ignore
> the chore of providing time-of-day to the world. Of course, we're
> back to the same question we've been debating with our well known
> positions all staked out – but the point is – the Moon ain't in
> it anymore. The whipping boy for hurrying a decision to emasculate
> UTC has been the quadratic lunar term – but large numbers of leap
> seconds are seen to be needed every century, whether or not the
> lunar term is included.
> Of course, on top of this long term trend are superimposed the
> decadal and seasonal effects (also from Steve Allen's page):

> This +/- 1-2 ms jitter can either magnify or quiet the millennial
> effects – for relatively short periods – or equivalently can
> accelerate or retard the scheduling of leap seconds. On the other
> hand, there is no reason to suppose that the periodicities that
> have been revealed since the time of Pericles have completely
> characterized our planet's wobbles at the long end of the
> spectrum. In particular, The LOD offset plot appears to show a
> baseline error of 0.5 or 1.0 ms. Also, the choice of slope for
> detrending the historical data was purely a result of a
> phenomenological fits to the data in hand. The fact that the lunar
> estimate, and the estimate born from the high precision near term
> data, both differ from the phenomenological slope could also be
> taken to suggest that long term periodicities are lurking in the
> data, and that even greater excursions from the mean are not only
> possible but are inevitable.
> Which is all to say that not only does the secular baseline
> guarantee that we cannot tune SI seconds to match a mean solar time
> whose rate is changing, but the geophysics of the planet lead one
> to question what this even means exactly.
> What to do? What to do?
> Challenge the premise of the question.
> Interval time and time-of-day are simply, truly, two entirely
> different things. Some pragmatic arrangement is going to have to
> be struck between the two. I'd suggest that Babylonian notation be
> reserved for time-of-day (UTC) and that interval time (TAI) be
> expressed as what it is – a simple count of seconds. But the
> details would certainly be open for discussion – if we didn't have
> to expend so much effort fending off the Absurd Leap Hour Proposal,
> again and again.
> But also, we should be investing in the infrastructure needed to
> convey both quantities to the vast array of users who need one or
> the other at whatever level of precision. Striking a "compromise"
> that eviscerates time-of-day for the imagined benefit of some
> interval time special interests is not only uncalled for - it
> doesn't even begin to address the issues. Time-of-day is solar
> time and should always remain such. If there are to be policy
> changes in the identification of civil time with mean solar time
> with Coordinated Universal Time - these should be driven by issues
> related to time-of-day, not force fit to match the pathologically
> even cadence of atomic clocks.
> Similarly, if precision time users are having troubles dealing with
> leap seconds – we've said it before – they shouldn't select time-
> of-day as a time standard! I mean - duh! Why would any sane
> engineer who needs a clock regular to one in ten billion specify a
> requirement to synchronize their precious system to a clock that
> only keeps time precise to one in ten million? Obscuring this
> simple fact by trashing UTC for the rest of us is rather – impolite.
> Time-of-day is Earth orientation. Atomic time is interval time.
> One doesn't have anything to do with the other.
> Rob Seaman
Received on Tue Dec 12 2006 - 16:15:31 PST

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