[time-nuts] The TAI zero epoch of 1958

[Steve Allen]

On Thu 2019-10-24T15:37:38-0400 Scott Stuart hath writ:
> The context is that someone is going to be communicating timestamps to me
> in TAI formatted as the number of elapsed seconds since 1958 Jan 1
> 00:00:00.  To make sure that I understand how this is being calculated, I
> reviewed the history of the establishment of TAI and UTC in the late 60s
> and early 70s.  I understand why 1972 Jan 1 00:00:00 UTC == 1972 Jan 1
> 00:00:10 TAI and why TAI is now 37 seconds ahead of UTC.  I can also see
> that propagating backwards from the 1961 time conversion formulas to 1958
> Jan 1 gives TAI - UTC very close to 0 (within about 2 milliseconds).  So,
> it makes sense to use 1958 Jan 1 00:00:00 UTC == 1958 Jan 1 00:00:00 TAI as
> the 0 epoch for TAI in seconds.

On 1958-01-01 the USNO time scale systematically differed from the
international average by 0.035 seconds because USNO had adopted a
value of its longitude which turned out not to be globally consistent.
Other sources of time had similarly large offsets plus random errors.
https://www.ucolick.org/~sla/leapsecs/BHsHn05p142.html
Try not to demand exactness for time scale before 1972-01-01.

Nothing that might be called UTC existed before 1959-08 when H.M.
Smith hosted folks involved with the USNO and UK time broadcasts for
tea in his living room.  That was when the US and UK informally agreed
to use cesium chronometers to coordinate their time broadcasts by
choosing a frequency offset from cesium such that the broadcast
seconds would attempt to match UT2 seconds.  The formal agreement to
do that coordination, and the actual coordination, did not occur until
1960.  So there was no UTC before 1960, and that name was not used in
print until 1965.

In 1960-09 the URSI General Assembly recommended that everyone should
use this scheme and that BIH should choose the frequency offset.  BIH
provided a value for frequency offset starting with 1961, and in
1961-08 at the IAU General Assembly BIH was officially tasked with
providing the frequency offsets for radio broadcast time signals.

TAI did not exist in 1958.  In 1958 several labs were using cesium to
construct unpublished time scales.  During 1961 BIH took on the task
of combining all the available atomic time scales.  By 1964 the BIH
realized that some atomic chronometers were better than others, and
using the three best they created A3.  The epoch at which A3
was set 1961-01-01T20:00:00 UT2.
https://www.ucolick.org/~sla/leapsecs/taiepoch.html
The calculation of A3 evolved and on 1971-10-04 the 14th CGPM
authorized the CIPM to define TAI and work with the BIH to realize the
definition.

For practical purposes the LORAN broadcasts in the 1960s used the time
scale as the broadcast time signals, so LORAN was using frequency
offsets to match UT2 with seconds longer than what became the SI
second.  Therefore by 1972-01-01 the LORAN time scale was 10 seconds
behind what had then become TAI.

These gory details are not neatly summarized because of the
contentious climate during those decades.  The 4th meeting of the CCDS
first happened during 1966, but the results of that meeting were
voided.  The second 4th meeting of CCDS happened in 1967-07, and its
proceedings do not include detailed minutes of the discussions.  The
proceedings of the 6th meeting of CCDS best reveal why the summaries
are not complete.  In annex S15 of CCDS6
https://www.bipm.org/utils/common/pdf/CC/CCTF/CCDS6.pdf
G.  Winkler of USNO wrote "we should limit discussion of this question
to people aware of the implied problems".  Only by scouring
contemporary publications for quotes from many folks who were present
does it begin to become clear how much that attitude prevented clear
description of the history.

--
Steve Allen                    <sla at ucolick.org>              WGS-84 (GPS)
UCO/Lick Observatory--ISB 260  Natural Sciences II, Room 165  Lat  +36.99855
1156 High Street               Voice: +1 831 459 3046         Lng -122.06015
Santa Cruz, CA 95064           https://www.ucolick.org/~sla/  Hgt +250 m