A2130 (galfacts) adc,pfb overflows
26jun11
Intro:
The galfacts project (a2130) takes data using the mock spectrometers.
The mock spectrometer keeps track of overflows that occur in the
processing of the spectra (in the fpga's). The overflows can be:
- ADC overflows: when an adc sample is at the limit of the adc
range (-2048 or +2047).
- PFB overflows: When there is a register overflow in the
butterflies used to compute the fft.
- This normally occurs when a strong tone occurs (eg a radar).
The error count is stored as a log(base2) of the number of overflows:
- 0 - 0 errors
- 1 - 1 error
- 2 - 2 to 3 errors
- 3 - 4 to 7 errors
- ..
- 12 2048-4096 errors
- For this processing i used the max value: 2^n
The galfacts users reported a large number of overflows for the run
06-07jun11. To follow this up the following processing was done
for (06jun11,21-25jun11):
- for each day, read the compressed data files recording the number
of overflows and the time and za where the overflows occurred.
- For the daily run, bin the overflow counts to .5 deg za
steps and then plot vs za. Do this separately for the adc and pfb
errors.
- The amount of radar that gets into alfa depends on:
- the phase of the radar rotation period.
- When the radar points at AO once every 12 seconds,the
signal get much stronger.
- the telescope az,za, and alfa's rotation angle.
The plots show the number of error
cnts vs za by day (.ps) (.pdf):
- this data was taken with the azimuth at 0 deg.
- top plot: pfb overflow errors vs za
- bottom: adc overflow errors vs za.
- Each beam is a different color.
- Each page is data from a separate day (the date is the start of
the daily run).
- the vertical scale is constant for the N days.
- Vertical dotted lines are are drawn for the adc error peaks of
06jun11.
- These peaks are constant for each day.
- Page 1: 06jun11:
- lots of pfb errors at za=10.5 to 11
- Page 2-6 data for 21,22,23,24,25jun11
- pfb much less for these days. No peak at za=11
- adc errors are similar to 06jun11
Images of the 1325 MHz band from
the hilltop monitor was made for 01jun11 thru 26jun11 (.pdf)
- page 1: 01 - 08jun11
- page 2: 09 - 16jun11
- page 3: 17 - 24jun11
- page 4: 25 - 26jun11
- The thick lines 1240,1260 are the aerostat radar.
- 06jun11 doesn't have the aerostat on while 21-25jun11 does
have it present.
SUMMARY:
- The pfb errors of 06jun11 did not repeat on 21-25jun11.
- The aerostat was absent from 06-07jun11 while it was present
for parts of 21-25jun11
- the adc errors have similar counts and peaks for all the days:
- za peaks at za = [2.5,5.5,6.5,14,17] degrees.
- So i'm not sure why the pfb errors were so high for 06jun11..
- I doubt that it was the phase of the radar always lining up
with za=11 since i don't think the galfacts updown period is a fixed
multiple of 12 seconds.
processing: usr/a2130/chkovrflow/
doa2130day.sc, doa2130day.pro, makeimg1325.pro
home_~phil