Sky noise removal (b)- calcsky

Quite often, especially when we observe galactic sources embedded in dark or molecular cloud, it is impossible to find any bolometers that do not pick up source emission. As long as the source emission is relatively smooth, this will only affect the mean level in the map, and any base level that gets removed with remsky can be added back into the image using $\rm ADD=TRUE$. But, if there is structure in the source emission over our sky bolometers, these variations will be interpreted as sky noise and therefore affect the morphology of our map. For extended sources we should therefore use calcsky. The task calcsky computes a model of the source, which it then subtracts from each bolometer to give an estimate of the sky variation, which is put into the file extension .more.reds.sky, which can be examined with e.g. linplot, e.g.,

% linplot i86_lon_cal.more.reds.sky device=xwindows 

shows the computed sky noise variations for the scan 86, that we will re-analyze below (Fig. ).

Figure: This plot shows the mean sky variations in scan 86 as computed by calcsky as a function of time. Therefore x = 65 corresponds to start of integration 2 in Fig . We can see that it has clearly picked up the strong sky variations that occur in the beginning of integration 2 and continued during part of integration 2. Since this is a calibrated image, we can see that the sky variations are extremely strong, and would easily mask a faint source.

We can improve the model by adding data sets to calcsky the same way as we use rebin for coadding data. If one has already produced a final map of the source, one can use this map as the input model for calcsky. For multiple data sets we should make an input file with weights and offsets as we do for rebin, see Section . We now choose the default model, which will be the median of all the observed maps. Once calcsky is done, we go back and run remsky on each data file which was included in the model computed by calcsky.

calcsky works extremely well even for extended sources, if one has kept the chop position fixed. For extended sources it is therefore an advantage to chop in a fixed ra/dec frame. One can also use calcsky for a spherically symmetric source like IRC$+$10216 even for a chop throw of 60'', but then one should use az option for the $\rm OUT\_COORDS$. calcsky also includes the option to account for the chop throw and direction, and it should therefore work even when we chop differently on extended emission from one map to the next.

For scan maps calcsky is our only option for sky noise removal, because in scan maps each bolometer will normally include both source and sky (see later).

Below we show how to use calcsky on the same file we already reduced with remsky. When we now run remsky it will not ask for sky bolometers, but will use the sky extension stored in the header to remove sky noise variations.

% calcsky
OUT_COORDS - Coordinate sys for sky determination /'RJ'/ > 
SURF: output coordinates are FK5 J2000.0
REF - Name of first data file to be processed /'i86_lon_sky'/ > 
i86_lon_cal
SURF: run 86 was a MAP observation of IRC+10216 with JIGGLE sampling
SURF: file contains data for 4 exposure(s) in 3 integrations(s) in 1
measurement(s)

WEIGHT - Weight to be assigned to input dataset /1/ > 
SHIFT_DX - X shift to be applied to input dataset on output map 
(arcsec) /0/ > 
SHIFT_DY - Y shift to be applied to input dataset on output map 
(arcsec) /0/ > 
IN - Name of next input file to be processed /!/ > 
SURF Input data: (name, weight, dx, dy)
   -- 1: i86_lon_cal (1, 0, 0)

BOXSZ - Size of smoothing box (seconds) /2/ > 
MODEL - File containing source model /!/ > 
% remsky
IN - Name of input file containing demodulated map data 
/@i86_lon_ext/ > 
SURF: run 86 was a MAP observation with JIGGLE sampling of object 
IRC+10216
OUT - Name of output file /'i86_lon_sky'/ > i86_lon_sky
REMSKY: Using SKY extension to determine sky contribution

The sky noise we see in our jiggle maps is due to changes in the sky emission between nods of the telescope. For unstable sky conditions this results in a tile-pattern in your map. This type of structure will not be removed by calcsky if we apply it to a single data set. Indeed in this particular example, remsky with selected sky bolometers gave a much better result than using calcsky. However, for really extended sources and scan maps we may not have a choice. We will have to use calcsky followed by remsky.

Another advantage of calcsky which is worth mentioning is that we can run calcsky on the 450$\mu$m array and copy the calculated sky noise variations to the 850$\mu$m array after appropriate scaling. This enables us to remove sky noise variations on a very faint extended 850$\mu$m-source, without subtracting out source emission.