The top window shows the first 40 milliseconds of the file. The
bottom window shows the first 60 milliseconds.
With each window the top 3 plots have the 1st cal off 1
millisecond short. The bottom plots are from 17nov08 when the cal was
correct.
The color coding is:
black: total power polA
red: total power polB
green: The pdev cal status bit that comes with each 1
millisecond spectra. It will be 1 if the ttl cal input was high for any
portion of the 1 millisecond integration.
blue dashed lines: the expected cal transition
red dashed lines: the cal transition that is 1 millisecond
early.
Some properties of the problem are:
The problem with the starting phase (it is 1 too large). The
cycle length remains correct.
Once the scan starts, the phase remains constant for the entire
scan.
All spectrometers show the same cal phase.
The oscilloscope was also used to measure the time from the 1 second
tick of the start to the first cal on transition.
I took the cal multiplexer output, the 1 second tick, and the
detected rf signal and put them into the scope.
I started the spectrometer 35 times:
7 scans had a 1 millisecond offset, 28 scans were correct so it
is off about 20% of the time.
The cal on transitions after the 1 second tick start are:
Cal correct: 19.75 milliseconds
Cal off: 18.75 milliseconds.
We expect the cal transition at 19.5 milliseconds. The extra
time is probably measurement error and propagation.
Conclusions:
The pdev spectrometer starts the winking call off by 1
accumulation 20 % of the time.
The offset remains constant for a scan.
The error is coming from the spectrometer. It is probably a
timing problem in the fpga setup of the start of the calcyle.