Hello,
This behavior smacks of a genuine hardware failure, not "drift"
in the usual analog sense. I'd be looking for bad solder joints,
loose or defective cable connectors, that sort of thing. If you have
series SMT capacitor(s) in your signal path, that's another
possibility. These parts are made of very brittle dielectric
materials and can develop cracks and other failure modes due
to board flexing associated with tightening nearby SMA
connectors, etc. If you can arrange to monitor the levels in
near-real-time, try wiggling the signal cable, prodding parts
on the acquisition board, and that kind of thing.
Or perhaps there's a failure in your signal generator itself.
Ordinary component drift is of the character of one (or a few)
percent change in value occurring slowly and fairly smoothly
over a span of several years.
Dana Whitlow
Arecibo Observatory
On 3/22/2013 6:00 PM, [email protected] wrote:
If other people see the drift, then I
suspect the gain of the input stages
of the ADC boards drift. If so, then
switches will be needed in some of my
spectrometer designs which currently
lack them. What do you see or think?
I had reason to return to using the old
spectrometer of CASPER Tutorial 3. This
spectrometer is old enough to be useful
for diagnostics with new designs, but I
noticed a problem when viewing the FFT
spectra. With an input of a 1Vpp sine
wave at 5, 10, 15, 20, 25 or 30 MHz,
the amplitude of the peak in the FFT
spectrum changes in two ways: during a
single run and between runs. I was
trying to determine the source of this
change because I am trying to calibrate
the vertical (amplitude) axis of the
spectrum obtained in the tutorial.
For example, during a run the amplitude
either rises to almost four times its
original value or drops to a quarter of
its original value. If I wait for a
long time, the amplitude seems to rise
slowly. If I shut down the spectrometer
and all the instruments except the
ROACH board, and then repeat the run at
a later time, the readings may be
different by a factor of five or even
ten. This is making calibration very
difficult. If this is normal behaviour
for the spectrometer of Tutorial 3,
then I may be forced to give up an easy
calibration and use a switch for all
measurements so I can monitor the drift
during a run.
The apparatus consists of the ROACH 1
board, an old Bee2 ADC board 1GHz
oscillator for the ADC clock, an old
Wavetek oscillator for the input sine
wave and a blocking capacitor at the
input of the ADC board. The oscillator
does not give a perfect sine wave. Is
the drift normal behaviour for the FFT
spectra?
