[time-nuts] Solving the UTC drift problem

Tom Van Baak tvb at leapsecond.com
Thu Jul 14 03:48:08 EDT 2005

> Instead of adding randomly-placed leap seconds to UTC or allowing UTC 
> to drift from UT1 etc, the timing community should just change the 
> second's definition from time to time as needed. That is, dither the 

Bad idea. Here's the math. An average of one
leap second a year is equivalent to a frequency
offset of 1 / (365 * 86400) =  3.17e-8. This is
how much slower your timebase has to be in
order to match time as kept by the Earth. It
is the equivalent to about 300 Hz at 9192 MHz.

3e-8 is a huge offset; a factor of a million less
accurate than what a modern Cs will put out.

It means that all atomic clocks (masers, cesium,
and rubidium) and most OCXO on the planet
would have to be re-calibrated every time you
decide to change the length of a second [*]. That
puts the number of affected time and frequency
sources in the millions or tens of millions.

But it would be very good business for cal labs.
Busy, seasonal work, mind you, since every
instrument in the world would need to be
re-calibrated only on the same day).

Wrist-watches and sundials would be ok.

I think it also means that every scientific paper
reporting time, frequency, wavelength, distance
or energy to more than 8 or 9 digits in the past
40 years has to be rewritten. But the scientists
could do all that while all their equipment was
out to the cal labs so it might work out.

> The beauty of this method is that there are only a few hundred Cs 
> clocks in the world, and they are phase locked to 10.000 MHz or 

I bet just us time-nuts together have a over a
hundred Cs! I think the total number of Cs in the
world is closer to ten thousand. Rubidium over
a hundred thousand. OCXO in the millions. XO
in the billions.


[*] For several years in the 60's this is just what was
done. HP distributed synthesizer settings or gears
to change frequency by multiples of 50 or 100 e-10.

More information about the time-nuts mailing list