[time-nuts] Absolute time accuracy pre-Cesium?

[Tom Van Baak]

> Does anyone have a pointer to information about the absolute time 
> accuracy (not stability) that was available via PZT or other techniques 
> prior to the Cesium definition?  I'm doing a presentation and want to 
> show the evolution of accuracy.  My Google-fu has failed me in finding 
> anything pre-Atomic.
> 
> Thanks!
> John

A nice example of how good astronomical timing was is how they calibrated cesium atomic time against astronomical time. The original 1958 paper is here:

http://leapsecond.com/history/1958-PhysRev-v1-n3-Markowitz-Hall-Essen-Parry.pdf

What you see there is that they spent 4(!) years and took 4(!) data points to precisely compare the best astronomical clock with the first cesium clock. It appears they got millisecond accuracy in their timings. Compared against the existing astronomical clock standard, the four measurements of cesium frequency were:

9 192 631 761
9 192 631 767
9 192 631 772
9 192 631 780

Do the math: the mean is 9 192 631 770 +/- 8 Hz. That, literally, is where the magic 9192.631770 MHz cesium number and definition of the SI second comes from. That suggests the precision was 8 Hz / 9192631770 Hz, which is 8.7e-10, the equivalent of 75 us/day, or 2 ms/month, or 27 ms/year.

As a practical matter a more accurate value of 9192631770 would have been useless because the earth is less stable than 8e-10 anyway. Here, for example, is how different UTC and UT1 would be depending on how the cesium SI second had been defined:

http://www.leapsecond.com/pages/ut/cs9192-ut1-ani.gif
http://www.leapsecond.com/pages/ut/

In retrospect we would have had fewer leap seconds if they had chosen 9192631950 Hz instead of 9192631770 Hz. But at the time it wasn't a choice; it was just a measurement.

/tvb