[time-nuts] Know thy components
James Flynn
james.flynn at csun.edu
Thu Aug 6 12:09:37 EDT 2015
In the world of time nuts, little things matter. Nanoseconds are an
eternity. We rip our hair out (if we have any left) over 1E-11 errors,
weep over phase noise and shriek about Allen deviations, modified or
otherwise.
The question is: just where are the little things?
If you are new at this, I hope my experience helps you. If you are an
old hand, you need not read on. Or you can just smile at someone else's
foibles.
I have been working on a disciplined, double oven crystal standard. I
put the resistors controlling the gain of the VCXO in the outer oven
along with the DAC that generates the EFT voltage so that I could tweak
the loop gain by selecting the values during calibration without
disturbing the main oscillator. The trade was to minimize any noise out
of the DAC while ensuring the embedded processor could steer the VCXO
with the necessary precision.
I made the measurements, pulled some resistors out of my collection and
was on my way to months of insanity while I tried to figure out why my
lovely oscillator wandered like a over-sugared kid a petting zoo.
I checked everything. The power supplies, the precision DACs; everything
had been selected to be wonderfully stable. And it was in an oven,
after all.
After eliminating all the other suspects, I tested resistors of the same
type I had installed in the EFT circuit. Using a precision, 8 digit
ohmmeter, I put the resistors through a temperature excursion of about
20 deg C by external heating. The results are summarized below. I had
used resistors of the first two types in the list.
carbon composition 5% -550 ppm/deg C
cheap 5 % metal film -240 ppm/deg C
1% metal film axial lead <-1 ppm/deg C
1% thick film surface mount <-1 ppm/deg C
My resistors were mounted near the walls of the outer oven, about 2 cm
from one of the heaters. A change of about 0.1 deg C would more than
explain what I had been seeing. Also the carbon composition resistor
drifted with time at elevated temperature. This may explain some of the
long term trends I had seen.
The moral of this tale of woe is to know thy components. Know them well.
More information about the time-nuts
mailing list