The
galfacts
(a2130)
takes
data using the mock spectrometers and a winking cal. This page uses
the winking cal to monitor the gain stability of the system.
The setup, processing, and plotting of the data is:
Setup:
Dome moves 2 to 20 degrees then back down. Az is at the
meridian.
mock spectrometer setup:
two 172 MHz bandwidths centered at 1300 and 1450. the
plots below use the 1450 MHz band.
sample spectra at 1 millisecond integrations.
winking cal with 20 milliseconds on and then 20
milliseconds off
Processing:
Average calOn, calOff spectra to .1 second each (throwing
out the 1 millisecond for each cal transition).
Compute total power for each .1 second calOn, calOff
spectra.
compute calDeflctionTp= CalOnTp - calOffTp giving .2 second
resolution.
Fit a linear ramp to calDeflectionTp for the entire
nights data. Use this to convert from spectrometer counts to
calDeflection units.
Plotting:
Page 1: calOn,caloff buffers
and calDeflection using spectrometer counts:
The data for the plots has been median filtered
to 1 second.
1st : PolA calOnTotal power (black) calOffTotal
power (red).
2nd: PolB calOnTotal power (black) calOffTotal
power (red).
3rd: PolA calOnTp - calOffTp. the green
line is a linear fit to the cal Deflection
4th : PolB calOnTp - calOffTp
Page 2: CalDeflection/LinearfitToCal.
Dividing by the linear fit to the cal changes the units
from spectrometer counts to cal units.
1st: polA, 2nd polB
the black trace has been median filtered to 1 second, the
red trace has been median filtered to 12.2 seconds.
Outliers (< .6cal, > 2.0 cal) were not included in
the filtering.
Each title contains the rms of the cal deflection for the
entire nite (before any filtering).
CalOn-calOff in spectrometer units
3rd polA
4th polB
A linear fit (green line) is done for the entire
dataset. This is used in plot 2 to normalize to
calDeflection.
Cal deflection vs Za: rightmost column:
Plot cal deflection vs za for each beam. black is pola,
red is polB. The data has beem smoothed to 12.2
seconds with a median filter.
The rightmost column in the table above plots the cal
deflection amplitude vs za.
Each plot holds the data for 1 day , all 6 beams.
Black is polA, red is polB.
Beam 4b shows a za
dependence of the cal deflection
that looks a lot like the za dependence of the telescope gain
and Tsys.
This was first seen on 21apr11.
The cal value increases by about 3% at 19 deg za.
This variation is not caused by the gradient in za at high za.
Suppose calOn comes before cal Off.
Going up calOff is taken at a larger za so the cal
deflection is smaller because of a za Tsys gradient.
Going down calOff is taken at a smaller za that calOn so
the calDeflection would be larger because of a Tsys
gradient.
The plots show that the cal deflection only increases with
za. For a za Tsys gradient it should decease and increase
because of the change in za direction.
beam 2b has shown ripples
that depend on za.
30apr11: The ripple was first seen
23may11: A low za increase in caldeflection was seen
05jun11: increase around 14-15 deg za seen.
The increase was only seen when going down in
za.
05jun11 started observing with the dome at 360 degs (looking
south). The previous dataset was dome looking north.
beam 2b has a ripple around za=5. This started on 20apr11.
What could cause this?
The amount of cal that gets into the receiver increases with
za above 15 (but only in polB)
the bm 4 polB probe is moving with za?
Some of the cal signal goes out the horn and then get
reflected back. More of this reflection occurs at high za?
The electronic gain increases with za above 15 degrees.
Some cables are moving with za? but only for beam 4a,2b?
Higher za does have higher Tsys. (but 2b does the opposite at
low za sometimes?)
the gain increases as the signal get stronger?
Somehow the spectrometer is increasing the output gain as
the input level increases?
input level too low?
It is suspicious that the increase in cal deflection has the
same functional dependence as the tsys increase (but only for
4a)
processing:usr/a2130/caldefl/caldeflvsza.pro
Measurements by date
01jul11: Beam 2a caldeflection has 4%
jumps.
The drain
Current for beam 2a amp1 had large oscillations since the end
of march (which also happened to be when the Tsys for Beam 2A jumped
up).
The bias card for beam2 was replaced on 23jun11.
The beam2a amp1 drain current oscillations went away.
The Tsys did not get any better.
On 28 and 29 jun11 the caldeflection measured by a2130
jumped by about 4%. The jump lasted for about 40 minutes before
returning.
see 28jun11 beam2 (col 4) in the table above.
The jump was probably not related to the bias card change
since the jump started 5 days after the card was changed (data
from 24-27 jun11 was ok.. as well as 30jun11).
23may11:
Adjusting beam1 bias voltages fixes jumps in beam1b cal
deflection
Bias voltages for beam 1 were adjusted
on 16feb11 when an open loop bias card was installed for testing.
This card was in use for a only a few days. When a closed loop card
was put back in, the bias voltages were never put back to the old
values. The lower bias voltages set for the open loop card made the
amp leds more sensitive to power supply voltage changes.
On 23may11 ganesh readjusted the bias voltages
(stage 3 drain voltage) for beam 1 back to the levels pre 16feb11.
These were the values that made the led's less sensitive to power
supply variation. A2130 cal deflection data taken before and after
the bias change show that the beam1b jumps are gone:
changing the stage 3 drain voltage level to a lower value on
16feb11 caused bm1b output to jump around. The leds are more
sensitive to power supply jumps at the lower voltage.
restoring the drain voltage to the pre 16feb11 values fixed
the problem.
beam 5b has jumps in the cal deflection (up to 2%). This beam
also has a lower stage 3 gate voltage. Maybe if should be
adjusted upward (as long is it doesn't cause the oscillations
coupling in from dead beam 6 to increase).