--------------------------- Nov. 25, 1995 --------------------------

I have just posted two .gif files on aleph for thos of you who would
like to see how the "Totally Accurate Clock" can perform. The files
are named oso-clk.gif and oso-clk2.gif. They differ only in that the
first one has some color, the second is all black-and-white.

At the Onsala Space Observatory in Sweden, a TAC is used to monitor
the performance of their Hydrogen Maser. I was there in September
and on Day#270 I set up their "TAC" for best performance.

The 1 PPS signals from their Maser and the TAC are compared in a
good quality HP time-interval counter with an HPIB for data acquisition
(sorry, but I didn't note the model number). The counter makes 30
second time-interval averages of the Maser-to-TAC time interval and
logs the data on a computer every 30 seconds.

I took their raw logs for the Day#270 to Day#312 interval. I took
their 30 second raw numbers and averaged then in 30 minute bins.
I also computed the RMS deviation of the 30 second numbers in each 30
minute bin, and the max/min values in the bin. The only "sanity"
tests were to screen out occasional bad readings (logged time-of-day
didn't increment properly, or the reading differed from a running mean
by more than 1 usec) [except for day #306 where a number of really
strange reading were in the logs and were discarded en masse].

The oso-clk.gif files are the graphical results. The heavy line (an
inverted parabola going between ~7950 and ~8150 nsec) is the 30 minute
averages of the 30 second counter readings. The small tick marks
extending ~+/- 100 to 150 nsec around this line are the individual
max/min differences from the mean.

At the bottom of the plot (use the scale on the right-hand side) are
the RMS values for each 30 second measurement in each 30 minute bin.
The overall average RMS value of these points is about 32 nsec.

The Onsala TAC is running in zero-D position constrained mode using
some average coordinates I measured before I set it up.

Thanks to Jan Johansson for making these data available. Jan might tell
us what is the HP counter model they are using. Also, can you provide
a clue as to what happened during Day#306? [And where is the data after
Day#312? -- I didn't find it on gere in the normal directory!].

If anyone is interested, we can post the raw Onsala clock logs for
you to peruse the data.

Regards, Tom Clark



    --------------------------- Nov. 26, 1995 --------------------------

I just added two more .gif plots -- oso-hist.gif (color) and oso-hst2.gif
(B&W). These show the histogram of 30-second Onsala data for the Day#291
thru Day#293 period (when the OSO clock had essentially zero drift with
respect to GPS). Superimposed on the plot is a Gaussian curve for the
measured ~34 nsec RMS  (mathematically 1000*exp((t-8151)/49)^2 ).

Tom


