Space Telescope Science Institute
Table of Contents Previous Next Index Print

NICMOS Data Handbook > Chapter 4: Anomalies and Error Sources > 4.6 Flat Fielding

4.6 Flat Fielding
NICMOS flat fields show response variation on both large and small scales. Figure 4.9 compares some of the flat field exposures used for Cycle 11 (post-NCS) calibration to those used in Cycle 7/7N (pre-NCS). It is evident that the morphology and amplitude of the variations of all three cameras is reduced in Cycle 11. This behavior is due to the fact that “cold” pixels (i.e. pixels with a lower than average responsivity) show a higher than average Detector Quantum Efficiency (DQE) increase with temperature. The flat-field variations are due also to a strong function of wavelength. At 0.8 μm, there is variation by a factor of ~5 (minimum to maximum) in the relative response across the array. This declines to a factor of ~3 at a wavelength of 2.2 μm, and at 2.5 μm the array is almost flat. Naturally, both the spatial distribution and amplitudes of these variations are different for each camera. NIC3 has a large contiguous region with low sensitivity.
An inevitable consequence of these large DQE variations is that science data will have large spatial variations in their pixel-to-pixel noise level, and hence in signal-to-noise for sources. For images where statistics are limited by photon shot noise (either from the sources themselves or from the sky background) the S/N variations scale as the square root of the flat field amplitude (e.g., a factor of ~2 at F160W). Many NICMOS observations, however, especially with NIC1 and NIC2, are largely readout noise limited and for these the signal-to-noise modulation will scale linearly with the flat field, e.g., a factor of ~4 at F160W.
Figure 4.9: Flat-field responses for NIC1, NIC2 and NIC3.
Normalized pre- and post-NCS flat-field responses for NIC1 (left) through NIC3 (right) for F110W, F187W, and F113N, respectively. The images are inverted to better display the grot; therefor the dark regions have higher QE. The color stretch is the same for both temperatures in each camera. The histograms show the “flattening” of the arrays at the higher temperature (narrower distribution). The decrease in the dynamic range between bright and faint targets is a direct result of the decreased well depth at the higher temperature.
As with the dark current and bias, the quantum efficiency of a NICMOS detector is a function of the detector temperature. Because the instrument temperature varied somewhat throughout Cycle 7 and 7N (when operating temperature was ~62K) as well as in Cycle 11 and beyond (operating temperature ~77K), the flat-field structure also varied slightly due to this effect. For the temperature range experienced, about ΔT~2K at the most, this has a relatively small effect, but does affect both the overall DQE as well as the relative structure of the flat-fields (after these have been normalized to unity). Calnica now (version 4.4 and later) uses a temperature dependent correction factor to account for the overall DQE change (described in the PHOTCALC section in Section 3.3.2). The effect of this correction is up to a few percent depending on the detector temperature at the time of the observation. To correct also for the temperature dependent changes in the structure of the flat-fields, calnica now uses a set of temperature dependent flat-fields (version 4.4 and later). The effect of this on the photometry is at most ~1% and depends on the position on the detectors. Users that are concerned about their photometry at these levels and use data retrieved earlier than 2008 should consider re-calibration using the latest reference files and calnica version 4.4 (or later), or retrieve the data again from the Archive using the OTFR (see PHOTCALC and FLATCOR sections in Section 3.3.2 for more info and detailed dates). For further updates please refer to the NICMOS Web page at:
The strong wavelength dependence of the NICMOS response may affect the quality of flat fielding for objects with extreme colors. The color of the internal flat-field lamps may not match that of the sky background, or indeed of many astronomical sources. In most cases this should only affect photometry by a few percent, but concerned users may wish to experiment with constructing customized, spectrally weighted flat fields. These can be constructed from linear combinations of (and interpolations between) existing NICMOS narrow band flat fields, as described in Storrs, Bergeron and Holfeltz 1999 (NICMOS ISR 99-002).
One situation where the color dependence of the flat fields may have a visibly noticeable effect is on the flat fielding of the sky background. The background has a significantly different spectrum than the flat-field lamps, particularly at λ > 1.8μm, where thermal emission from the telescope and instrument become important. After dividing by ordinary internal lamp flats, flat-field residuals may still be present in the sky background. Unfortunately, in practice it is difficult to separate these from the characteristic variations induced by the pedestal effect discussed above in Section 4.2.2. Pedestal removal routines like pedsky (see Section 4.2.4) key off these residual flat-field variations, and depend on having an accurate model of the spatial structure of the sky. If actual, observed sky frames are not available, then synthetic color-dependent flats may be useful when running routines like pedsky.
A discussion of flat field color dependence is presented in Storrs, Bergeron and Holfeltz 1999 (NICMOS ISR 99-002). That paper describes methods for constructing color-dependent flat-field reference files, and provides IRAF scripts for doing so (currently applicable to Cycle 7 and 7N data, but see the NICMOS Web page for updates on this script).
In general when using NICMOS data, you should keep in mind the large spatial variations in sensitivity and signal to noise that result from the flat field structure. This can have important consequences when you are analyzing NICMOS data, requiring, e.g., careful handling when setting source detection thresholds in automatic routines like DAOPHOT or SExtractor, or for error analysis on photometry.

Table of Contents Previous Next Index Print