17.2 Dark Current Subtraction Errors

17.2.1 Dark Current Pedestal

NICMOS data show what is known as the dark current pedestal. This is an additive signal that appears whenever the detector's amplifiers are switched on, which happens at the beginning of an observing sequence, after a filter wheel motion, spacecraft motion, or Earth occultation. The level of the pedestal is quadrant-dependent and random, with typical values below 40-50 DN.

At the time of writing (August 1997) a flight software modification is being developed to remove the pedestal. The basic idea is to leave the amplifiers on at all times (instead of keeping them off, and switching them on only when exposures start). The degree to which this strategy has succeeded will not be known for some time after the change is made.

Being an additive signal, the most noticeable effect of the pedestal is to leave flatfield residuals in the calibrated images; the pedestal is a uniform offset without any flatfield variations, and so the flatfield correction step in calnica will impose an inverted flatfield response pattern in the final calibrated image. In addition, the absolute photometry of extended objects (i.e., those for which a reference sky level cannot be obtained within the frame) will be altered. Finally, the presence of the pedestal makes the calnica automatic cosmic ray rejection processing of MULTIACCUM images less effective. Since currently the pedestal effect lacks full characterization, it cannot be calibrated out by the pipeline.

For observations of compact sources, where most of the frame is occupied by blank sky, an iterative procedure can be used to remove most or all of the pedestal. The procedure exploits the large scale non-uniformity of the NICMOS flatfields, which will produce a large scale modulation of the blank sky in a calibrated NICMOS image with a considerable pedestal signal. The goal of the procedure is to remove the pedestal by minimizing the flatfield non-uniformities. This technique was thought out and developed by Mark Dickinson of the Johns Hopkins University, and is outlined below.

The procedure is iterative and identifies the minimum of the residual flatfield non-uniformities on the calibrated images (run through calnica, only). It should be applied on a quadrant-by-quadrant basis, becuase the pedestal level is different in different quadrants. For MULTIACCUM data, the procedure should be applied on individual imsets; however, it is not useful for the very first few readouts (where there is very little signal), and as a rule of thumb should be applied to those imsets with exposure times greater than about ten seconds. Increasing values of the pedestal level are subtracted from the each quadrant in the raw images, and the latter are run through the calibration software calnica, up to the flatfielding step only. The calibrated images are then smoothed (e.g., using the IRAF tool rmedian) to remove sources and pixel-to-pixel variations of the flatfield residuals. Finally, the variance of the residual large-scale modulation in each image is computed relative to the appropriate constant sky level. The pedestal level which yields the minimum for the variance is selected as the value for that particular quadrant. Because the reference value for the image is a constant sky level, it is clear why this technique can be applied only to those frames containing a large fraction of blank sky.

17.2.2 Synthetic MULTIACCUM Darks

NICMOS dark images are highly dependent on the readout history of the array since it was last reset, and, therefore, cannot be simply rescaled to the exposure time of the science data (as is done with most conventional CCD data). Each science file must be calibrated with a dark frame of equal exposure time and number of readouts.

A NICMOS dark frame can be decomposed into three basic components:

Dark current proper.
Amplifier glow.
Shading (a bias change). The three components are highly reproducible and can be easily calibrated. On-orbit darks obtained during SMOV have been used to characterize the dependence of the three components on the pixel position and on time for each of the three NICMOS detectors. This information has been used to construct synthetic dark current calibration reference files for all MULTIACCUM readout sequences, using as basic data the on-orbit darks obtained during the first part of the Cycle 7 calibration program. The synthetic darks have been used to populate the calibration database for those sequences for which on-orbit data do not exist yet or are currently contaminated by the dark current pedestal. These darks are routinely used to calibrate NICMOS data in the pipeline. Here we describe in more detail the three components of a synthetic dark.

Dark Current

The dark current component is the detector current when no external signal is present. This component is a function of time only:

Where D is the observed signal in a given readout, t is time since reset, and dc is the dark current (e^-/s). The NICMOS dark current is of the order of 0.05-0.06 e^-/sec for Camera 2, and < 0.03 e^-/sec for Cameras 1 and 3.

Amplifier Glow

The characteristics of the amplifier glow are described in the section "Instrument Artifacts" on page 17-11. For the purpose of dark frame modelling, it is enough to know that the amplifier glow is a signal: infrared radiation emitted by the amplifiers at the four corners of each NICMOS array, which is detected by the pixels in the array. It is like having a small light bulb in each corner, which produces a pattern of light that is highest in the corners and decreases towards the center of the array. Because the signal is injected in the array every time this is read, the amplifier glow depends on the number of readouts performed in a particular exposure. In particular, it is proportional to the number of readouts since the last reset: