Noise Arcs in SCUBA-2 Maps

Some reductions of SCUBA-2 data exhibit arc-like structures. This seems most often to be noticed in “blank field” observations taken in good weather and at high elevation. When this happens, it appears that noise from some bolometers is entering the map but never being cleaned up during the iterative map-making process. In order to address this issue, two approaches were investigated:

Applying the “flt” model during iteration.
Adjusting the filtering parameters.

It was found that adjustment of the “largescale” filtering parameter was able to suppress the noise arcs. Changing from filtering only at the start (as is normally done in the blank field dimmconfig) to during the iterative process (using the “flt” model as in other dimmconfigs) did not provide much of an improvement and seemed to adversely affect noise, especially around the edges of the map.

The following table illustrates the problem with the default parameters, and shows improved results with a blank field-style and flt-based configuration.

blank field default parameters noiseclip 4.00 largescale 200	blank field modified parameters noiseclip 4.00 largescale 125	with flt model modified parameters noiseclip 4.00 largescale 125

Config Files

The config files given below were tested, varying the noisecliphigh and filt_edge_largescale parameters.

Modified blank field

^$STARLINK_DIR/share/smurf/dimmconfig_blank_field.lis
noisecliphigh = 4.0
filt_edge_largescale = 125

With flt model

modelorder = (com,ext,flt,ast,noi)
numiter = -20
com.corr_abstol = 0.8
com.perarray = 1
noi.box_size = 0
noi.calcfirst = 1
spikethresh = 10.0
noisecliphigh = 4.0
flt.filt_edge_largescale = 125

Test Reduction Results

Please use the links in the table below to view the results of tests with a variety of noiseclip and filt_edge_largescale parameters.

	Blank field	With flt model
Maps	Maps, blank field	Maps, with flt model
Exposure time	Exposure, blank field	Exposure, with flt model
SNR	SNR, blank field	SNR, with flt model

Conclusions

It seems possible to suppress the noise arcs by reducing the largescale filtering parameter — with the observations tested they were mostly removed at 150″ and more fully removed at 125″.
Reducing the noicecliphigh parameter initially seemed promising but systematic testing of the two parameters did not confirm this.
Switching from a blank field style dimmconfig to one which is flt-based does not appear to be indicated by the test results. It is possible that some of the other parameters added to the flt-based config under testing have adversely affected the noise.