Avoiding the Leap Second

On 31 mei, 09:04, Quadibloc wrote:
Paul Schlyter wrote:

Since I am proposing lengthening the second *only in those years with
leap seconds*, obviously I would not seriously propose that the ohm,
volt, et cetera be different in those years too. That would be
insanity, and rest assured I suggest no such thing.

John Savard

But you'd have no choice, what does a frequency counter (piece of
electronic
equipment show? Pulses per second (Hertz).
And these may wel be calibrated against NIST, and some measure in pico
seconds, quite a bit of error would show!
Now endless arguments would be created if it was your new second or
the real
second or whatever,
The wavelength scale on your radio, basically all other physics
constants,
you car's km/hour or miles/hour... what not!
I say: Forget it:-)

On 2007-05-31, Quadibloc wrote:

There are also some hidden assumptions: these are -

Except for TV stations, that use cesium and rubidium clocks to
maintain their frequency on track (and they wouldn't need to do
anything, because frequency would still be measured in terms of the SI
second, not the second of civil time) nobody else in the private
sector uses anything better than a quartz crystal timer. These are
accurate to about five seconds a year, so changing the length of a
second with an impact of *one* second a year isn't going to impact the
real-time clocks in computers or in Internet routers.

A couple of points here. TV stations are not by any means the only
people with requirements for precision timekeeping, and in any case
TV transmitters don't use atomic clocks in the manner you suggest.

Secondly, quartz isn't nearly as accurate as you suggest. A
32.768kHz crystal typically has a quoted accuracy of 10ppm.
According to my HP-21 that works out at 315 s a year. That's over
five _minutes_. Mechanical clocks are capable of far better
accuracy, although of course that depends on precision engineering
that comes at a price.

Maybe not, but I am working from the following assumptions:

First, since it has been proposed that we just go on straight atomic
time, and just use "leap hours" eventually, I take it that it is at
least believed by some that:

ISTR that according to international agreements UTC must remain
within one second of GMT. This is important for navigational
purposes - not everyone uses GPS and in any case many wouldn't wish
to depend on it as their sole system (remember GPS can be turned
off for civilian purposes at any time). From memory even a one
second difference between UTC and GMT equates to a quarter mile on
the ground. At 30 minute difference would make traditional navigation
impossible.

--
Andrew Smallshaw

On 30 May 2007 17:51:43 -0700, Quadibloc wrote:

Since I'm proposing changing the length of a second, though, by an
approximation, rather than the *exact* proportion that adding an extra
second would make, this would not lead to TAI minus civil time being
an integer number of seconds at least at the start of each new year.
There are two possible cures: use the exact proportion instead
(adjusted in leap years!), or switch from a longer second to a regular
one before the end of the year (for example, splitting the leap second
up among all the seconds of the first 360 days of the year would lead
to an "even" lengthening of the second in some senses).


Do I understand correctly: you are effectively suggesting a redefinition
of the civil second from an interval to a count? That is, instead of a
civil second being a certain number of atomic vibrations, it's simply
determined by having a fixed, integral number of them in some longer
time, like a year?

That doesn't seem like a bad idea- leap seconds are rather ugly things
for ordinary applications to have to deal with. Seems like we should
really just have a formal second, and a formal time that is a count of
these seconds from some origin. That's the second used for scientific
purposes, and the time used for critical applications (power grid sync,
phone service, etc). Civil time could be trivially converted to official
time (and vise versa).

_________________________________________________

Chris L Peterson
Cloudbait Observatory
http://www.cloudbait.com

On May 31, 8:49 am, Chris L Peterson wrote:
That doesn't seem like a bad idea

Except that it's basically what broadcast radio signals were doing
during all of the 1960s up to the end of 1971 (which was called both
UTC and GMT, but was really trying to track UT2). It was unworkable
for a number of different reasons.

For broadcast time signals it only makes sense to transmit based on
the most stable time and frequency interval available to your
civilization. The current scheme for UTC makes a lot more sense for
broadcast signals. In order to have a civil second approximating the
mean solar second it would be necessary to modify the infrastructure
for distributing civil time such that it was able to handle both kinds
of time. This is a major restructuring of the systems that we use on
a daily basis. In the economic, practical, engineering sense it is
much simpler to abandon leap seconds (and mean solar time) and
implicitly say that any problems caused by this can be handled by
folks 20 generations from now. (Of course, they said the same thing
about pollution, but my bias is showing.)

leap seconds are rather ugly things for ordinary applications to have to deal with.

No. Leap seconds are a form of resetting a clock. They are an
acknowlegement of a horological reality which was understood very well
by the navigators of the British Navy 200 years ago, but which we
somehow seem to have forgotten in the interim.

The reality of running a clock is that no two clocks can ever agree,
not even if they are co-located in the same reference frame, in the
same laboratory (and there is no agreement between clocks in different
reference frames). If one clock is deemed more authoritative then the
other has to admit a mechanism by which it can change its offset, its
rate, or both. To deny that is to live in a fantasy world and risk
the consequences of believing in a system that does not correspond to
reality.

So the issue is a question of who pays how much, and when, with what
sorts of short-term and long-term inconveniences.

In article .com,
Quadibloc wrote:

Paul Schlyter wrote:
Changing the length of the second is out of the question!!! Why?
Because a change in the length of the second would affect so many
other units which all depend on the second.

I'm not proposing that we change the length of a second for purposes
of measuring speed and acceleration, so as to change the value of the
newton relative to the meter, or to change the volt, the ohm, the
joule, or so on.

I am merely proposing that we change the second of *civil time* so
that it is longer than the 1900 second of Ephemeris Time, also the SI
second. That is simply *going back* to the situation we had before
atomic time, with leap seconds, was adopted in 1972.

Before 1972, UTC time was adjusted in steps of fractions of a second,
much more frequently than the frequency of leap seconds. It was to
get rid of that mess that leap seconds were introduced.

The difference is simply that to compromise with our world in which
things are measured with greater precision, I propose that instead of
having a second that gradually increases in length in a somewhat messy
and indeterminate way, let us have a time scale that is still tied
closely to TAI, but by changing the length of the civil second in a
controlled manner, so as to produce an effect essentially equivalent
to having leap seconds.

Since I am proposing lengthening the second *only in those years with
leap seconds*, obviously I would not seriously propose that the ohm,
volt, et cetera be different in those years too. That would be
insanity, and rest assured I suggest no such thing.

And how would you determine which years are "those years with a leap second"?
Yep, that process too is as unpredictable as the leap seconds. So there
would be a need to transform the information about how long a second will
last this year ..... it would be just as awkward as the leap seconds
themselves.

I think applications which really need a time accuracy better than
a few seconds over long time interval should use TT, TAI, GPS time, or
some other suitable *uniform* time scale *without* *leap* *seconds*!
All other applications, where it doesn't matter whether the time is
off by a few seconds, should adjust for leap seconds whenever they
occur, but ignore leap seconds when computing the time interval
between two instants - much like the UNIX time() family of funtions do
today.

John Savard
--
----------------------------------------------------------------
Paul Schlyter, Grev Turegatan 40, SE-114 38 Stockholm, SWEDEN
e-mail: pausch at stockholm dot bostream dot se
WWW: http://stjarnhimlen.se/

In article ,
Chris L Peterson wrote:

On 30 May 2007 17:51:43 -0700, Quadibloc wrote:

Since I'm proposing changing the length of a second, though, by an
approximation, rather than the *exact* proportion that adding an extra
second would make, this would not lead to TAI minus civil time being
an integer number of seconds at least at the start of each new year.
There are two possible cures: use the exact proportion instead
(adjusted in leap years!), or switch from a longer second to a regular
one before the end of the year (for example, splitting the leap second
up among all the seconds of the first 360 days of the year would lead
to an "even" lengthening of the second in some senses).

Do I understand correctly: you are effectively suggesting a redefinition
of the civil second from an interval to a count? That is, instead of a
civil second being a certain number of atomic vibrations, it's simply
determined by having a fixed, integral number of them in some longer
time, like a year?

That doesn't seem like a bad idea- leap seconds are rather ugly things
for ordinary applications to have to deal with. Seems like we should
really just have a formal second, and a formal time that is a count of
these seconds from some origin. That's the second used for scientific
purposes, and the time used for critical applications (power grid sync,
phone service, etc). Civil time could be trivially converted to official
time (and vise versa).

Having the civil second change length slightly every year would
probably be more awkward than the current practice of leap seconds.

Chris L Peterson
Cloudbait Observatory
http://www.cloudbait.com
--
----------------------------------------------------------------
Paul Schlyter, Grev Turegatan 40, SE-114 38 Stockholm, SWEDEN
e-mail: pausch at stockholm dot bostream dot se
WWW: http://stjarnhimlen.se/

In article .com,
Quadibloc wrote:

Paul Schlyter wrote:
But hey! You wanted to go decimal, right? Why should one quadrant be
90 degrees? Why not, say, 100 degrees instead? There is actually
such a measure for angles - it's called gons, or (sometimes) "new
degrees": one quadrant is 100 degrees and one revolution is 400
degrees. Sometimes this has been used for surveying, and there it
fits neatly: while one degree of latitude on the Earth's surface is
some 111 km, one "new degree" is very close to 100 km and one "new
minute" (equal to 1/100 "new degree" of course) is very close to 1 km.
.
So the decimal angular units are already there - you just have to start
using them! :-)
.
I thought they were called "grads".

They are -- that's another name for them.

But, in any case, since you are apparently a stickler for SI units,
you should of course know that the official unit for angular measure,
the one that fits with everything else in the metric system, is the
*radian*.

If astronomy would go completely metric, it would have to abandon units
like the AU, the solar mass, the parsec, ..... g

Of course, it is a little awkward that the number of radians in a
right angle, or indeed any aliquot part of the circle, is an
irrational, even transcendental, number. One could include pi
explicitly in descriptions of angles to get out of this; then, an
angle would be called "N pi radians". Since the circle contains 2 pi
radians, a protractor with 100 big marks on it to cover 180 degrees,
instead of 200 marks, would make sense, covering the expanse from 0 to
1 pi radians.

John Savard
--
----------------------------------------------------------------
Paul Schlyter, Grev Turegatan 40, SE-114 38 Stockholm, SWEDEN
e-mail: pausch at stockholm dot bostream dot se
WWW: http://stjarnhimlen.se/

Quadibloc (John Savard) wrote:

Except for TV stations, that use cesium and rubidium clocks to
maintain their frequency on track (and they wouldn't need to do
anything, because frequency would still be measured in terms of the SI
second, not the second of civil time) nobody else in the private
sector uses anything better than a quartz crystal timer. These are
accurate to about five seconds a year, so changing the length of a
second with an impact of *one* second a year isn't going to impact the
real-time clocks in computers or in Internet routers.

Instead, only the national time standards people will need to modify
their equipment, so they can generate time signals based on longer
seconds during those years that require them. So the routers on the
Internet would just adjust themselves to WWV from time to time, to
avoid being out of sync, presumably as they do now, but because of the
changed second (in years requiring leap seconds) this adjustment also
automatically takes care of what an explicit recognition of the leap
second would have had to do.

Many networks have far better accuracy than a quartz crystal
timer does. This is from one of the many available network
time server boxes, specifying holdover accuracy during GPS
outages:

== start of quote ==
|
| "NetClock NTP servers are designed to maintain accuracy in the
| event of loss of the GPS signal due to severe weather (lightning
| strikes, high winds, etc.), physical damage to the antenna, GPS
| signal jamming and electromagnetic pulse (EMP), and even if the
| federal government disables the GPS signal. Internal oscillators
| ensure seamless operation if the GPS signal is lost by maintaining
| synchronization accuracy until the GPS signal is restored.
|
| "A choice of 3 oscillators are available depending on the needs
| of the application. A temperature controlled crystal oscillator
| (TCXO) is standard. Optional oven-controlled crystal oscillators
| (OCXO) and Rubidium-stabilized (Rb) oscillators offer extended
| 'holdover' accuracy.
|
| Oscillator Drift Rate Holdover Accuracy Recommended
| (nanosec/sec) (millsec/day) Holdover
|
| TCXO 2,000 172.8 days
| OCXO 20 1.728 months
| Rb 0.05 0.1296 years
|
== end of quote ==

In many network applications, having two computers be off by up
to a second (as in one using SI time and the other using Civil
time can be a Very Bad Thing.

Consider a computer that falsely concludes that another computer
has a newer version of some data and thus overwrites new data
with old data. Or a computer that falsely concludes that another
computer has an older version of the data and thus refuses to
update it's copy.



Guy Macon
http://www.guymacon.com/

On May 31, 8:38 am, Andrew Smallshaw wrote:
ISTR that according to international agreements UTC must remain
within one second of GMT.

UTC is under the control of the ITU-R. There is no reason to believe
that the ITU-R currently believes this, and lots of reason to believe
that they do not. In part this is due to the fact that there really
is no such thing as GMT anymore, at least not as any kind of precision
entity. The IAU 2000 reforms for earth rotation pretty much abolished
the concept. The reformulations which allow re-creation of the
previous entities for earth rotation are now given the name
"classical" in order to distinguish them from the currently official
conventions.

From memory even a one
second difference between UTC and GMT equates to a quarter mile on
the ground. At 30 minute difference would make traditional navigation
impossible.

If "traditional navigation" means using some sort of almanac, then
this is not the case. It is straightforward to predict earth rotation
to within one second with a five year lookahead, and the pre-
publication lead time even for printed almanacs is less than this.
Therefore, with or without leap seconds, the almanacs used for
"traditional navigation" can easily change their tabulations such that
the users of such traditional methods won't notice the change.

Andrew Smallshaw wrote:

Secondly, quartz isn't nearly as accurate as you suggest. A
32.768kHz crystal typically has a quoted accuracy of 10ppm.
According to my HP-21 that works out at 315 s a year. That's over
five _minutes_.

That's over a wider temperature than a watch sees. A typical
low-cost Casio watch spec is +/- 15 seconds per month, and
they often do better than +/- 5 seconds per month in use.