Intro
The steps in processing the data
Map positions raw data
    Plot: sampled map positions for each field (.ps) (.pdf).
Tsys for the 4 fields.
    Plot: Measured Tsys for the 4 fields vs za and epoch (.ps) (.pdf):
    Plot: Ratio TsysPolA/TsysPolB (.ps) (.pdf)
Fitting to Tsys
    Making the model
        Plot: The x102 calibration data and tsys model fit (.ps) (.pdf):
    Fitting all the data at once
        Plot: Tsys vs za with fit overplotted (.ps) (.pdf).
        Plot: Fit residuals vs za  by field (.ps) (.pdf).
        Plot: Tsys fit residuals vs za. separate out by ra, dec scans and by field(.ps) (.pdf):
        Plot:Tsys fit residuals vs za. separate by epoch (.ps) (.pdf):
    Fitting by field and epoch.
        Plot:Tsys fit residuals vs za with fit overplotted (.ps) (.pdf):
        Plot: Fit tsys by epoch. Tsys Fit residuals vs za (.ps) (.pdf):

The individual maps by field
    Plot:The  average maps for each field (.gif)
    Plot:The difference maps for each field (.gif)

Mosaicing the 4 maps together
    Plot:The ra,dec differences of the overlap regions before the corrections (.ps)  (.pdf)
    Plot:The ra,dec differences of the overlap regions after the corrections (.ps)  (.pdf):
    Plot:The final mosaiced maps (.pdf) (.ps):

Intro: (top)

    Data was taken in nov/dec 2005 and a few strips in oct 2006.
 

mapname	map center
Field1	00:50:38 , 23:31:00
Field2	00:34:42  , 28:31:00
Field3	00:50:55 , 32:01:00
Field4	00:34:25 , 32:01:00
Field1234(combined)	00:42:25,  30:13:46  504'        by 486'

 

---

The steps in processing the data  (top)

• Input the data, remove rfi, compute total power (inputprocdat, inputmap,mkmask,testrfi).
• The routine inpprocdat was called twice:
1. The first call with firsttime=1 determines which channels are good in a strip.
•  inputmap() inputs the entire map and calls mkmask().
• mkmask()  creates a channel mask for each spectra of each strip. The masks are created by fitting an 8th order harmonic to each bandpass and flagging any fit residuals above 3 sigma.  A value less than 3 sigma is assigned 1 while residuals greater than 3 sigma get 0. An average mask for each strip is created by averaging the masks for each strip.
• The average mask for each strip is stored in an idl save file.
2. The second call uses the average masks for each strip to compute  the total power.
• The mask processing is done separately for each cal on/off for each strip.
• For a channel to be included in the total power it must be ok for 95% of the spectra in a strip and it must also be ok in the cal on/off (since the cal will affect the entire strip).
• The data is over sampled in the driven direction (ra for maps, dec for decmaps). The routine testrfi() subtracts a smoothed version of the total power along a strip and looks for points that stick up. These are usually rfi that lasts for a short time (less than the time for a source to transit the beam). A weight array (assoc()) is constructed with these impulsive points given a zero weight.
• The weight arrays are stored in an idl save file along with the total power.
• Make maps for each field. (doallmaps(), mktpimg(), getmaps(), tsysremove(), cmgaincor(), aogridzilla(), basketweave().)
• getmaps() inputs the total power data, cals, and weight arrays from the idl save files. If any duplicate strips are present, an algorithm is used to select the best 1. The cals that were taken contiguously in time are fit to a linear polynomial (to find problems where the cal was taken while a source was passing through the beam).
• tsysremove() removes the system temperature from the data (see more about this below).
• cmgaincor() converts the data to janskies using the gain curves. For 430 they are a function of just za.
• aogridzilla() grids the ra driven maps and the dec driven maps (which cover the same area).
• basketweave() will basketweave the two maps correcting for dc offsets in a strip.
• The average the ra,dec maps after basketweaving are written out as fits files. The difference between the two maps (after basketweaving) are also output for diagnostic purposes.
• Mosaic the 4 fields together.
• Compute offsets between maps for mosaicing (combinemaps_scl()).
• The ra overlap (fields 1/3, fields 2.4) and the dec overlap (fields 1,2 and fields 3,4) are used to compute offsets vs ra and dec for each map.
• The overlap region of the first map is interpolated to the positions of the second map and then the differences map1Iinterp - map2 are computed.
• The difference is then split between the two maps.
• This is done first for the ra overlaps and then the dec overlap fields.
• The 4 corrected maps are then combined using aogridzilla().

---

Map positions   (top)

• Figure 1 shows the az,za tracks for each field. Black are the ra driven scans and red are the dec driven scans.
• Figure 2: This has the ra/dec positions sampled for each field.
    processing: chkalldata_pltpos.pro

---

Tsys for the 4 fields.  (top)

• Measured Tsys for the 4 field vs za and epoch (.ps) (.pdf):

• Page 1 Tsys vs za by epoch. Each plot is a separate field. Black and red are polA,polB before jan06. Green, blue are polA, polB after jan06.
• In field3 PolB  (blue) looks higher than polA(green). This may also be true in field2 and field 4 (although it is not as obvious).
• Page 2 has the median value for each za by field and epoch. Black is data before jan06, red is data after Jan06. The different fields are plotted  with different fields.
• PolB red looks higher after jan06 than before.
processing: plottsys_all.pro
• Ratio TsysPolA/TsysPolB (.ps) (.pdf) . Black points are before jan06 while red points are after jan06.
• Top TsysPolRatio vs za.
• Middle TsysPolRatio vs az. There is a drift to higher values at lower azimuth values.
• Bottom TsysPolRatio by map. The data is laid out by map. Each map starting at first strip taken. The dashed green lines are where the maps change.
• The red points (after jan06) seem to fit in with the variation within the maps. The offsets with polB larger than polA after jan06 may be related to the sky and not the diodes.
processing: chkalldata_plttsys.pro

---

Fitting to tsys  (top)

1. Compute a single scaling value by dividing the Tsys data by the model and computing the median value. The fit value is then the tsysModel*Scale
2. Compute a robust mean and sigma  of (data - fit).
3. Throw out all points whose residuals are greater than 2.5*sigma. This tries to get rid of points that are positive because of sources.
4. Iterate the process. After about 3 passes, nothing changed.

---

Using the calibration data to make a model for Tsys.   (top)

• Use the 1.25 Mhz band at 431.
• Find the separate source tracks across the dish that had at least 10 points.
• For each set of tracks add an offset so that Tsys(za=17) = 54. This tried to subtract off any background sky temperature.
• If the cals had jumped, this would not do the correct thing (you'd need a scaling).
• If the source track did not get to za=17, it was ignored.
• Fit a linear polynomial plus  a 3nd order polynomial above za=10 deg.
• Normalize the fit so TsysM(za=10)= 1.

• Page 1 input calibration data:
• Top az, za tracks for the sources.
• Center Tsys vs za. Each track is a different color
• Bottom Tsys vs az.
• Page 2 Correct Tsys data and fit.
• Top: Corrected Tsys data vs za and the fit.
• Bottom: Correct Tsys data vs az.

---

Fitting Tsys to all the data at once.   (top)

• Tsys vs za with fit overplotted (.ps) (.pdf).
• Plot Fit residuals vs za  by field (.ps) (.pdf). The residuals for each field are plotted separately. Black is polA and red is polb. The median offsets for each field are:
 

Fit residual offsets by Field

field	polA K	polB K
1	-.25	-.81
2	-.35	-.29
3	.65	.39
4	.38	.76

• Tsys fit residuals vs za. separate out by ra, dec scans and by field(.ps) (.pdf): Each page is a separate field. The black dots are the ra driven scans. The red dots are the dec driven scans.
• For fields 1,2, and 3 the dec driven scans residuals fall off at high za. For field 4 both the ra and dec strips fall off at high za. The za dependence of the model is not fitting the data very well.
• Tsys fit residuals vs za. separate by epoch (.ps) (.pdf): The epoch break is data before jan06 and data after jan06. The red solid line is at 0 (for reference).  Tsys after Jan06 is larger than Tsys before jan06 (especially for polB).

---

Fitting Tsys by field, epoch  (top)

• Tsys fit residuals vs za with fit overplotted (.ps) (.pdf): Black is polA, red is polB. The green line is zero for reference. The shapes remain the same. The offset for field 4 polB is now closer to zero.
• Fit tsys by epoch. Tsys Fit residuals vs za (.ps) (.pdf): Tsys was fit for all the data before jan06 and then for all data after jan06. The  epoch2 data looks to have a flatter za dependence.

    The map data processing fits tsys by field. The basket weaving will attempt to correct for any offsets. Cal changes will be partially corrected for by this.
 

---

The individual maps.  (top)

1. grid the ra driven and dec driven maps separately to the same grid.
• Grid spacing was 3 arcminutes.
• The gridding function used a gaussian with a fwhm of 5.5 (it was repeated with a gaussian of 11 aminutes).
• The gridding function extended out to about 30 arcminutes.
• A sin projection was used to project onto the  2-d grid.
2. basketweave the two  maps. Create an average map and a difference map after basketweaving.
• The basket weaving will try an remove offsets in strips.

---

Mosaicing the fields.  (top)

1. Compute the ra intersections between field1,2   and fields 3,4.
2. Interpolate the first map of each  pair to the grid position of the second  (within the overlap region).
3. compute  mapRaDif=(map2 - interpolated Map1) in the overlap region
4. Fit a linear polynomial to the differences as a function of the ra position
5. Correct the two maps by subtracting half the difference from map2 and adding half the difference to map 1
6. Repeat steps 1 5 with the corrected fields, using the dec overlap of maps 1,3 and maps 2,4.

• Page 1 Ra axis (Overlap in dec).
• Fields1,2  and Fields 3,4 have  short overlaps in ra (since they are at the same dec).
• Fields 1,3 (leftmost two fields) show a slight rise in power as ra decreases (left to right).
• Fields 2,4 (rightmost two fields) show a larger increase in power as ra decreases.
• Page 2 Dec axis (overlap in ra)
• Field 1,2 have a constant offset versus dec.
• Field  3,4 have a slight increase for the first have of the maps (moving up in dec). They then level off.
• Field  1,3 and 2,4 have short overlaps in dec.
• Page 3 shows the diagonal map overlaps: field1-4, field 3-2  plotted versus ra  and then versus dec.
• Page 4 plots the data before subtractions for the long axis.
• Top field 2,4 overlap. The are the same at the beginning then field 2 rises faster than field 4.
• Bottom field 3,4 overlap. They look to have a pretty constant offset.
processing: cmpmapoffsets.pro

• These plots show the difference of the overlap  maps after correction.
• The large jumps are at sources where the interpolation of map1 to positions of map2 did not do a good job.
processing:combinemaps_scl.pro

The final mosaiced maps (.pdf) (.ps):

• The maps have been clipped to 1 Jy.
• The ra scale is for the center of the map
• This used a 5.5 arcmin gridding function. The final resolution is about 14 arcminutes.
Processing: combinemaps.pro

---

To be worked on.   (top)

• The tsys removal is not perfect. It might have been better to try fitting all the fields at once to Tsys rather than 1 field at a time. The linear fitting of the cals, the basketweaving, and the scaling of the individual maps from the overlap region try to correct for this.
• There are pointing errors at the beginning of some strips because the telescope is trying to catch up. The data points should be interpolated here.