A vintage rack-mount Racal (Sulzer) double ovenized laboratory quartz frequency standard was slowly warming up. I was monitoring it with a hp 53132A frequency counter. Gate time was set to 10 s and statistics were enabled (N = 30). So one measurement per 10 seconds with statistics every 300 s (5 minutes). The external reference for the counter was a 10 MHz atomic standard.
The measurements were captured via the talk-only RS232 serial port, MJD timestamped, and logged by a PC. Timestamping the readings is optional but very useful when you look back at data years later. So as a habit I timestamp any data from any counter or sensor that I use. Not only does it tell you when the data was collected, but also the sample rate. Any timestamping method works (local time, UTC, unix ctime or time_t, JD) but I always use MJD.
Sample raw data looks like this:
... 53530.612558 5,000,000.062649 Hz 53530.612674 5,000,000.062590 Hz 53530.612789 5,000,000.062598 Hz 53530.612905 5,000,000.062664 Hz 53530.613021 5,000,000.062640 Hz 53530.613137 5,000,000.062576 Hz 53530.613252 5,000,000.062607 Hz 53530.613368 5,000,000.062659 Hz 53530.613484 5,000,000.062637 Hz 53530.613600 5,000,000.062576 Hz 53530.613715 5,000,000.062613 Hz 53530.613843 5,000,000.062672 Hz 53530.613958 5,000,000.062631 Hz 53530.614074 5,000,000.062571 Hz 53530.614190 5,000,000.062639 Hz 53530.614306 5,000,000.062673 Hz 53530.614421 5,000,000.062614 Hz 53530.614537 5,000,000.062580 Hz 53530.614653 5,000,000.062661 Hz 53530.614769 5,000,000.062655 Hz 53530.614884 5,000,000.062599 Hz 53530.615000 5,000,000.062595 Hz 53530.615116 5,000,000.062651 Hz 53530.615243 5,000,000.062654 Hz 53530.615359 5,000,000.062585 Hz 53530.615475 5,000,000.062582 Hz 53530.615590 5,000,000.062659 Hz 53530.615706 5,000,000.062653 Hz 53530.615822 5,000,000.062576 Hz 53530.615937 5,000,000.062607 Hz 53530.615937 N : 30 53530.615937 STD DEV: 34.2 uHz 53530.615937 MEAN : 5,000,000.0626221 Hz 53530.615937 MAX : 5,000,000.062673 Hz 53530.615937 MIN : 5,000,000.062571 Hz 53530.616053 5,000,000.062672 Hz 53530.616169 5,000,000.062632 Hz 53530.616285 5,000,000.062570 Hz ...
This is a very stable frequency standard and the frequency drift is very low. So it accidentally makes a fine slow sweep test of the precision and accuracy of a frequency counter.
Excel plots: 2005-racal-sdev.xls
Raw data: log19816.txt.gz
Experiment dates:
Comments/questions to: tvb
Here the plot is normalized to show 2.5 ppb drift over 6 days (= 4e-10 / day)
The corresponding phase plot is a nice parabola, indicating roughly linear frequency drift
Here's a plot of stdev from the hp statistics every 5 minute
Something unusual near day 1 and day 5; zooming to the 5th day
Switching to Excel to plot both frequency and stdev in the same chart
Same y-scale but changing x-scale from 24 hour span to 8 hour span
Now changing y-scale for a better look
C:\tvb\2022-racal$ gzip -d < log19816.txt.gz | grep -v -e ": " | mjdrate
53530.609873 ... 53536.529815 5.92 days (511483.0 s = 142.079 h), 50777 lines, 10.073322 seconds/line, 0.099272 Hz
C:\tvb\2022-racal$ mjd 53530.609873
2005-06-09 14:38:13 UTC (DOY = 160, Thu, STOD = 52693) = JD 2453531.10987 = MJD 53530.609873
Thu Jun 09 07:38:13 2005 (UTC-7 Pacific Daylight Time)
C:\tvb\2022-racal$ mjd 53536.529815
2005-06-15 12:42:56 UTC (DOY = 166, Wed, STOD = 45776) = JD 2453537.02982 = MJD 53536.529815
Wed Jun 15 05:42:56 2005 (UTC-7 Pacific Daylight Time)