A3121 430 MHz mapping feb17
image of what's been done (jpeg)
field with sources removed (jpeg)
directory holding fits files:
- Naming conventions:
- X=1..4 individual fields
- X=All .. 4 fields mosaiced together.
- 3C234 - beam map
- G11 -> 11arcmin gaussian used for gridding of individual
- _AVG.FITS . Pol A, polB averaged
- _PolAAVG.FITS .. made with only polA.
Links to Sections and Plots:
The steps in processing the data
Map positions raw data
positions for each field (.ps) (.pdf).
Plot: where all 4
fields fit into the final combined map (.ps) (.pdf).
The individual maps by field
The average maps for each field
Mosaicing the 4 fields together
Notes on the fields:
The final mosaiced maps
A3121 mapped 4 overlapping fields at 430 MHz
using the gregorian. Data was sampled at 1 hz. Each field was
covered in the ra and then the dec direction for basket weaving. The
12.5 MHz bands with 2048 channels were used to computed the total
Data was taken feb,mar 2017
The steps in processing the data were:
The steps in
processing the data (top)
- see a2125 steps in making
- PolB showed some jumps in Tsys (probably the jennings
- maps were made of the PolA + polB and just polA.
- The beam map was processed separately
The plots shows the sampled
map positions for each field take for a3121 (.ps) (.pdf).
The next plot shows where all 4 fields
fit into the final combined map (.ps) (.pdf).
- Pages 1shows the az,za positions for each field
- Page 2 has the ra/dec sampled positions for each field (labels
are not very helpful).
- each field is plotted in a different color.
- an index (0..3) is plotted on top of each field for
- The list at the bottom of the plot, maps index number of field
name, and average ra, dec for each field
Removing Tsys (top)
We need to remove Tsys (excluding the sky
contribution since that is what we want to measure). This is
explained in a2125.
Plots of Tsys for fields 1 thru 4 .
- Create a model of the za dependence of Tsys using the x102
data (tsysX102M). Normalize this so tsysX102M(za=10) is 1.
- For each map (ra,dec strips processed together):
- fit for the scale factor TsysData(za)=tsysX102M(za)*scale
- TsysRemoved= TsysData(za) - tsysX102M(za)*scale .. this
is done separately for pola, polb.
- If the actual Tsys changes over time ,this should help correct
- If the cal Values change (without remeasuring them), then the
0 of tsysRemoved should be ok, but the scaleFactor of
tsysRemoved will still be off (since we didn't scale it by the
change in the cal values).
- Plot (TsysA/TsysB -1) before
fitting or corrections (.ps) (.pdf)
- The data has been converted from correlator counts
to kelvins using the cals (after cal fitting).
- Top: TsysA/TsysB-1 vs za
- 2nd: TsysA/TsysB -1 vs az
- bottom:TsysA/TsysB -1 by field
- dashed green lines separate fields.
- The large spread is from polA varying.
- It looked to be stable for a days run.
- My guess is the jennings relay (load,antenna
switch) is probably making different contact whenever the
430 was switched to load (during testing).
- Cal scale factors by field
- These values are computed from the cals on every
strip. they convert from correlator counts to Kelvins.
- black (polA) ,red(polB) after cal fitting
- green(polA), blue(polB) before cal fitting
- Cal fitting takes adjacent cals (with no
attenuation changes) and does a linear fit to the measured
scaling factor ( throwing out outliers). The fit is then
used to convert correlator counts to kelvins rather than the
- If there are no outliers than the blue,green
overlay the black, red traces.
Some of these things will be
fixed by the basket weaving.
- Global fitting of all the
- Fitting Tsys by Field.
- When the images are made,
Tsys is removed from the ra and dec scans simultaneously
- We left them separate here to see if the ra, or
dec scans had any outliers.
residuals by field (rascans and decscans together)(.ps)
- Same as above plots but the scaling fit is done
with ra and dec strips together. This is how the images
- Black is polA, red is polB.
- The scale factor and errors are included in the
title for each plot.
After removing Tsys and scaling to janskies, each
field was gridded. The process was:
- grid the ra driven and dec driven maps separately to the same
basketweave the two maps. Create an average map and a
difference map after basketweaving.
- Grid spacing was 3 arcminutes.
- The gridding function used a gaussian with a fwhm of 11
- A sin projection was used to project onto the 2-d
The average maps for each field (.gif) show the results.
- The basket weaving will try an remove offsets in strips.
The 4 fields from A31215 were combined into
a single image. The processing to do this was:
The plots below show the pairwise difference of the overlap regions
of the maps.
- take each pair of overlapping maps.
- Interpolate the first map of each pair to the grid
position of the second (within the overlap region).
- compute mapRaDif=(map2 - interpolated Map1) in the
- Fit a linear polynomial to the differences as a function of
the ra position
- Correct the two maps by subtracting half the difference from
map2 and adding half the difference to map 1
- Repeat steps 1 5 with the corrected fields, using the dec
overlap of the pairwise maps.
- 21feb11: changed so that 1-6 only corrects 20% of the
difference. It then iterates 1-6 5 times. This gets rid of jumps
caused by the correction.
- the units are Janskies (multiply by 10 to get kelvins).
- Top 2 frames: Overlap in ra direction with
map1-map2Interpolated plotted vs ra.
- The red line is the fit vs ra.
- Bottom 2 frames: Overlap in dec direction with map1 -
map2Interpolated plotted vs dec.
- Each set is the start of iteration i (i=1 to 5). the last one
is when everything is done. The set shows how the correction
progresses with a correction factor of 20% per iteration.
The final mosaiced maps (.gif):
The final mosaiced maps with positions (.gif)
- The x,y positions are just grid points along the ra,dec
- This used a 5.5 arcmin gridding function. The final resolution
is about 14 arcminutes.
Notes on the
- PolB tsys was changing run to run. My guess is that the load
switch (jennings relay) was being moved between runs, and
the switch was occasionally making poor contact