[time-nuts] Surplus Rubidium Oscillator Frequency Spread
skip.withrow at gmail.com
Thu Aug 27 22:37:52 EDT 2015
As an experiment I took 10 FEI FE-5650A rubidium oscillators and measured
their frequency error (units were chosen randomly). These were 15MHz units
taken directly off the telecom cards with no modifications. These units
never went to China, and are pulls from working systems. Each unit was
allowed to warmup for a couple of hours before the measurement was made.
Unit # Date Code Error
301302 0716 -2.87 x 10E-10
C312570 0839 +4.16 x 10E-11
79119 0701 -1.01 x 10E-10
C312324 0812 -2.54 x 10E-10
69008 0450 -2.52 x 10E-10
60495 0352 -6.05 x 10E-10
78894 0651 -4.30 x 10E-11
79287 0701 -2.77 x 10E-10
71379 0523 -1.80 x 10E-10
78871 0651 -8.58 x 10E-10
Date code is YYWW. The units with a 'C' in front are after production was
moved to China.
All but 1 of the units ran slow (my reference was a Trimble GPSTM, locked
to GPS, so should be within a couple parts in 10E-12). I doubt that FEI's
reference was off by a couple parts in 10E-10, so my conclusion is that
they tend to age slower while in service (they are disciplined to GPS in
the application anyway).
These oscillators can be stepped by 6.8 x 10E-13 via digital control. When
modified by us at RDR we set them +/-2x10E-11 (for a 00 00 00 00 control
word). It would be interesting to know how many hours are on each of the
test units, but I know of no way to obtain that information (unless it is
part of the secret information that be read out of the unit).
More information about the time-nuts