[ntp:questions] A Suggestion For Abolishing the Leap Second

David L. Mills mills at udel.edu
Sun Jun 3 21:34:04 UTC 2007


Picking a small nit: WWVB and other Time and Frequency Stations happily 
transmit a variety of frequencies for carrier and modulation. That's how 
I calibrate my radios.

Last we talked you said the goal of Ultimate Timekeepers of the World 
was to the nanosecond using Two-Way Satellite Transfer. The NIST method 
used to wrangle an unruly herd of cesium clocks is described in my book. 
The method provides nominal time and frequency offsets between all 
clocks in the herd and establish a nominal laboratory timescale. I call 
this NTP distributed mode and have threatened to implement it.

A review of the IERS data from 1961 to 1972 shows periodic ephemeris 
time (ET) offsets declared in both time and frequency. That must have 
driven the microsteppers bonkers. If I could reinvent the world, I would 
run the master clocks in barycentric time and distribute offsets via the 
web. This is TAI at the mass center of the solar system where the 
gravitational potential is zero. Earthbound laboratories would need to 
provide velocity and gravitational corrections in the order of 15 ms, 
depending on the planet. Even now the GPS satellites are initially 
calibrated a few milliseconds early to account for the gravitational 
potential on orbit.

You and I discussed whether NTP should retail TAI (as in POSIX) or UTC 
and for a number of reasons decided on UTC. Of course, you can get TAI 
from a GPS receiver. It would even be possible to map UTC as received 
along with the current leapseconds table and set the system clock to 
TAI. That's how IBM mainframes map ACTS UTC to TAI. IBM mainframes 
provide programmable offsets and initiation times for UTC, leapsecond 
insertion and local timezone. See the IBM 9037 on the NTP project page 
for elaboration.


jlevine wrote:

> Hello,
>>>  In the first place, the unit of frequency plays a central role in
>>>measurements and fundamental constants -- much more fundamental than
>>>the unit of time. If the time services distributed a non-SI frequency
>>>all frequency calibrations would become much more difficult and
>>I can understand that, and I do not advocate that any different
>>frequency be distributed.
>     The difficulty is that timing laboratories do not distribute
> frequency as distinct from time.
> Precision frequency comparisons are done by measuring the evolution of
> the time difference
> (which is often expressed as a fraction of a cycle for the highest-
> precision comparisons)
> between a device under test and a calibration source. Thus, changing
> the length of the second
> effectively changes the frequency that is being transmitted. (Pretty
> much all time distribution
> systems use a single oscillator to generate both the carrier and the
> time code.) As I mentioned in my previous note, this was one of the
> reasons that a method similar to what you are proposing
> was abandoned in the early 1970s. For what it is worth, my personal
> opinion is that changing the length of a second in any manner would be
> completely unacceptable to the frequency community and it simply won't
> be accepted.
>>This means that the discrepancy between the modified
>>UTC and mean solar time would be allowed to be somewhat larger than
>>0.9 seconds.
>     This would not be a trivial change. Broadcast services (NIST radio
> station WWVB,
> for example) transmit the current dUT1 parameter as specified by the
> International
> Telecommunications Union (ITU). The format does not allow a value
> greater than
> 0.9 seconds. If this difference is allowed to exceed +/- 0.9 s then
> something has to
> change. Either the station would stop transmitting the dUT1 parameter,
> or it would be
> transmitted modulo 1 second or something else. None of these changes
> is impossible, but they could not be implemented without discussons by
> the appropriate
> group in the ITU. In fact, the leap second question is already being
> discussed there.
> Although no formal decision has been reached, my guess is that nothing
> will change and
> the current system will continue as is. However, this is just my
> private guess.
> Judah Levine
> Time and Frequency Division
> NIST Boulder

More information about the questions mailing list