| biaseq | stsdas.hst_calib.nicmos | biaseq |
biaseq -- adjust inter-readout bias levels for NICMOS MultiAccum images
biaseq input output skysamps
NICMOS bias levels can drift in an unpredictable way during the course of a MultiAccum exposure, producing undesirable residual bias offsets. Within a given readout quadrant of the NICMOS array, the bias level can change from sample to sample of a MultiAccum sequence, sometimes drifting slowly, and sometimes jumping suddenly. The result is that the net counts accumulated in a given pixel as a function of time can increase in a non-linear fashion. This can cause a variety of data processing problems, especially when the pipeline calnica task evaluates the count rate for each pixel by fitting a slope to the counts vs. time data, and attempts to reject deviant pixels as outliers in the CRIDCALC step.
The biaseq task adjusts the DC bias levels in each quadrant of each readout so that the net counts in that quadrant increase linearly with time. Note that this "bias equalization" is not sufficient to completely remove a constant bias offset in NICMOS data. Rather, it removes any non-linear components of the bias offset, but leaves an unknown linear DC bias which must be removed by some other method. The nicmos package task "pedsky" is one procedure for doing this, applicable to images which are only sparsely filled with astronomical sources. The biaseq task, in principle, will work on any NICMOS MultiAccum image, however, regardless of the nature of the astronomical sources present, provided that there are enough MultiAccum samples available. Caveats for using biaseq are discussed below.
The biaseq task should be run on an intermediate MultiAccum image file (*_ima.fits) which is produced by partially processing a raw NICMOS data set through only the first few processing steps of calnica. The pipeline processing steps ZSIGCORR, ZOFFCORR, MASKCORR, BIASCORR, NOISCALC, DARKCORR, and NLINCORR should be performed before running biaseq, but not FLATCORR, UNITCORR or CRIDCALC. In other words, the image should not be flatfielded and should be in units of counts (not counts per second). The nicmos task nicpipe provides a convenient way to carry out the partial calnica processing needed as preparation for biaseq.
The task works by assuming that the astronomical signal in each pixel (sources plus background) increases linearly with time, and that (modulo intra-quadrant bias jumps - see below) the bias offsets are uniform across each readout quadrant of the array. A model for the linearly increasing component (sources plus sky plus a linear bias term) is created and subtracted from each readout, and the bias offsets for each readout are measured and subtracted away to produce a corrected image. Here we describe the process in somewhat greater detail with attention given to the various parameters which the user may set.
The task first converts the MultiAccum data set to first differences, meaning that for each sample the preceeding sample is subtracted from it. This creates a set of short, independent exposures of the field, each with their own potentially different bias levels. For each readout quadrant, these first differences are then combined, after first applying a multiplicative scaling by DELTATIME exposure interval, and an additive zeropoint offset which accounts for the potentially drifting bias levels. The image combination is done with minmax rejection (set via the "nlow" and "nhigh" task parameters) to eliminate cosmic rays and other transient defects. This combined image provides a cosmetically clean "model" for the astronomical sources and background in the data, plus an average bias offset level.
The "skysamps" task parameter is used to specify which samples (IMSETs) of the MultiAccum image to combine to produce the model sky image. Although in principle all IMSETs may be used, in practise there may be some advantage to using only a subset. For example, the first few readouts of most MultiAccum sequences, and the last few for the MIF sequences, are very short and do not contribute much signal to the total image. Therefore it may be advantageous to exclude these readouts. The last readout (IMSET number 1 in the MultiAccum file) often contains intra-quadrant bias jumps (see below) and it may be a good idea to exclude it (although any such problems should also be handled by the minmax rejection during the combine process used to create the sky image). Finally, most MultiAccum sequences have readouts which are non-linearly spaced in time. The NICMOS detector signal phenomenon known as "shading" (see the NICMOS WWW pages and/or the Data Handbook) depends on the readout interval DELTATIME, and is not always perfectly removed by subtraction of standard dark frames in the pipeline process. However, first differences created from readout samples which are linearly spaced in time should exactly remove all shading, which is an advantage for the biaseq procedure. Therefore, if the readout sequence of the data set being processed has some number of readouts which are linearly spaced in time, then it may be best to specify the use of only those readouts with the "skysamps" parameter. The readout times of any MultiAccum sequence are listed in the NICMOS Instrument Handbook. The nicmos task "sampinfo" can also be used to conveniently list the relevant exposure time information, including DELTATIME, for all IMSETs of a MultiAccum image.
This combined sky image is then scaled to the DELTATIME interval for each first difference sample and subtracted from it. This effectively subtracts away all the constant astronomical signal and the mean bias level from each readout, leaving only the bias offset (the difference between the bias for that sample and the mean bias present in the combined sky image) plus any cosmic rays, etc., that are present in that particular readout.
The task then loops through the readout residual images, quadrant by quadrant, and determines the median level, which it uses as the bias offset. It subtracts this median level from each quadrant of the first difference images, and finally reaccumulates the first differences to produce a new, "bias equalized" MultiAccum image. This image is now suitable for continued pipeline processing. Most commonly, the user will want to use the nicpipe task again to apply the remaining calnica processing, including FLATCORR, UNITCORR, and CRIDCALC steps, in preparation for removal of the residual constant bias with pedsky or other algorithms.
Another disruptive phenomenon in NICMOS data is the occasional presence of intra-quadrant bias jumps or "bands." Occasionally the bias level within a quadrant will have one or more discontinuities along one direction, rather than being uniform. This often occurs during the final readout of a MultiAccum sequence, but can also happen during intermediate samples as well. The biaseq task includes an option ("fitjumps") to search for these jumps within quadrants and remove them as well during the course of the bias fitting procedure. This option should be regarded as experimental for the time being: the jump finding algorithm is quite crude and can no doubt be improved. Occasionally the jump finder can cause more harm than good and in these cases should be turned off ("fitjumps=no"). This may happen if, for example, the data has substantial unremoved shading signal, which may then trigger the jump finder in undesired ways.
1. Equalize the bias for the image n3z324g9m_ima.fits, which is a NICMOS MultiAccum exposure taken using the readout sequence SPARS64 with NREAD=18 (1023.964 seconds total exposure). The FITS file has 19 IMSETs including the "zeroth" readout, and has been partially processed through calnica using the nicpipe script. Consulting the NICMOS Instrument Handbook or the nicmos task "sampinfo," we find that the 4th through 18th readouts are linearly spaced in time with equal DELTATIMEs of 63.998 seconds, and that these correspond to IMSETs 1 through 15 of the image. We decide to exclude the last readout (IMSET 1) from the sky image computation, because it may have bias jumps. We do not fit for intra-quadrant jumps or keep any intermediate products. Default minmax rejection settings are used.
cl> biaseq n3z324g9m_ima.fits n3z324g9m_beq.fits 2-15
2. Do the same, but also correct for intra-quadrant bias jumps and keep a log of the jump regions. The jump log file will be named "n3z324g9m_ima_jump.log."
cl> biaseq n3z324g9m_ima.fits n3z324g9m_beq.fits 2-15 fitjumps+ \ >>> keepjump+
3. Equalize bias levels for n4ls01i1q_ima.fits, a MIF512 MultiAccum image, keeping the various possible intermediate data products for later inspection. Consulting the Instrument Handbook, or the "sampinfo" task, we see that IMSETs 9 through 16 have equal DELTATIMEs. We explicitly set nlow=nhigh=3 in order to reject all but the middlemost readouts when constructing the model sky image.
cl> biaseq n4ls01i1q_ima.fits n4ls01i1q_beq.fits 9-16 nlow=3 \ >>> nhigh=3 keepsky+ keepbias+
4. Process several MIF512 images which are listed in the file "ima.lis", creating the corrected images with names listed in "beq.lis". Use IMSETs 9 through 16 to produce the sky image for all datasets, and use default settings for all other parameters.
cl> biaseq @ima.lis @beq.lis 9-16
Authors: H. Bushouse, M. Dickinson, STScI
calnica, nicpipe, pedsky, pedsub, sampinfo
