WFC3/IR channel Flat Fields, Alpha Release
L and LP Flats
UPDATED: December 8, 2010
Variations in detector properties, such as the pixel thickness and non-uniform doping can cause differences in the pixel-to-pixel response of the detector that affect the accuracy of astronomical data. Additional wavelength-dependent low-order structures are introduced by the system illumination pattern and variations in the filter response. These variations are usually removed by normalizing the astronomical data to a uniform illumination of the detector with light passing through the entire optical path of the telescope. This process is commonly called flatfielding, and as a result the collected flux of a source should not depend on its position on the detector.
During the spring of 2008, in the third and last thermal vacuum test (TV3), the Wide Field Camera 3 (WFC3) team carried out an intense ground-based campaign to, among other purposes, create flatfields that will be used for the reduction of all WFC3 on-orbit data. Ground-based flatfields have been obtained by simulating the sky illumination of the UVIS CCDs (Sabbi et al., ISR 2008-46) a nd the IR array (Bushouse, ISR 2008-28) using the optical stimulus (CASTLE). These flat-fields should include both the high frequency pixel-to-pixel (P-flat) and low frequency (L-flat) structures. After WFC3 was installed on Hubble in Servicing Mission 4 (SM4), tests performed during the Servicing Mission Observatory Verification (SMOV) in proposals CAL-11452 and CAL-11453 indicate that the ground-based flatfields do not fully remove the low-frequency structures. These results were confirmed during in the Cycle 17 calibration programs CAL-11911 and CAL-11928.
Here we present an alpha release of the new IR flatfields obtained from a combination of the existing ground-based flat-fields (LP-flats) and a new low-frequency correction (L-flat) derived from combining a large number of long exposure science images after masking out objects, creating an equivalent of sky image. As a first step, sky images were produced in multiple filters. Investigating the wave-length dependence of the sky images showed no significant deviation between filters, with a variation of less than 1%. Therefore, a single gray L-flat correction was calculated and applied to all filters. In total over 2000 images were combined, all with exposure times >300s.
Independent measurements of the star cluster Omega Centauri in different broad-band filters confirms the results from the sky residual both when it comes to the spatial features and the amplitude of the L-flat corrections, as well as the lack of a significant dependence on wave-length.
The uncertainty in photometry due to the flat-field correction has an rms of 0.7% over the whole detector, with a maximum peak-to-peak range of -2.0/+1.9%. The uncertainties are typically larger at the edges of the detector. For the central part of the detector [129:896,129:896] (i.e., excluding a region around the detector edges with thickness 1/8 of the detector size), the uncertainty has an rms of 0.5%, with peak to peak range of -1.5/+1.6%. For the edge region of the detector only (i.e., a frame with 128 pixel thickness), which includes the "wagon-wheel" feature, the rms is 0.8%, with peak-to-peak -2.0/+1.9%. The introduced uncertainty in photometry due to the flatfielding therefore typically has an rms of less than 0.01 mag, at the edges of the detector the errors may reach 0.02 mag. In addition, the uncertainty in the wave-length dependence of the skyimage correction should also be less than 0.01 mag.
The new LP-flatfields are:
These LP-flats replaces the previously released alpha-version from May 25, 2010.
The new flat-fields were ingested into CDBS in December 7, 10.31pm. Data retrieved after this date are processed using the new reference files. (Note that the file names in CDBS differs from the names used on this webpage.) WFC3/IR users interested in applying these new flat-fields to older data can either re-retrieve the data from the archive, or re-run the standard calibration pipeline CALWF3 on the IR raw data. Before running CALWF3, the header keyword PFLTFILE must be updated in the raw images to the new flat-field name, while DFLTFILE and LFLTFILE should remain unchanged (= N/A):
>hedit my_raw.fits PFLTFILE flat_filter_lpflt.fits >calwf3 my_raw.fits
Alternately, users who do not wish to reprocess their data with CALWF3 can apply the L-flat
correction only to their calibrated data products in the following way:
>copy my_flt.fits my_copy_flt.fits >imcalc my_flt.fits,wfc3_IR_lflt.fits my_flt_corr.fits “im1/im2” >imcopy my_flt_corr.fits my_flt.fits[1,overwrite+]
The new L-flatfield is:
UPDATED: April 17, 2015
The seven images described here are "blob flats" generated for the IR channel of WFC3. They contain images that are unity everywhere except within the IR blobs cataloged by McCullough et al. (2014). These "blob flats" were generated by combining many images taken of the dark side of the Earth using WFC3 IR using filter F153M.
Caveat emptor! The correction of blobs by dividing by one of the blob_maskNN.fits file is experimental and may produce poor results, potentially even worse than making no correction, depending on the result desired. First, we do not recommend dividing by these "blob flats" if your scientific goal is stellar photometry. On the other hand, if your goal is either an aesthetically pleasing image, or a scientifically more-accurate surface brightness, then dividing by one of the "blob flats" may improve results. Also, there is some color dependency of detailed shape of the blobs with different filters. User feedback, either positive or negative, is welcome and may help us improve the methods, data products, or documentation.
Some users may wish to divide each WFC3 IR "ima.fits" image by an appropriate blob_mask file, in order to correct for the absorption of blobs that creates dips in the measured surface brightness of sky or other approximately uniform, extended IR emission at the locations of the blobs. (The user MUST use the ima.fits file, not the raw.fits file, to be consistent with the method by which the blob_mask files were constructed. Also, the blob_mask files are 1014 pixels by 1014 pixels, consistent with ima.fits image sizes and inconsistent with raw.fits image sizes, which are 1024x1024.)
Nominally, the appropriate blob_mask file will be "blob_mask00.fits" because that one was created from images taken when the CSM was in its nominal position. Because of the slight play in the CSM mechanism itself, the most appropriate blob_mask file may be different than the nominal. The user can experiment by dividing by any of the other files with other values of NN.
The filenames take the form "blob_maskNN.fits.gz" where NN = -04, 00, 04, 08, 12, 16, or 20. The value of NN corresponds to the rotation of the channel select mechanism (CSM) about its nominal position (NN=00) in units of millidegrees. For example, for NN=12, the blob_mask file was generated from a set of images in which the CSM position was 0.012 degrees from nominal. The blobs shift in unison to the upper right (larger values of both X and Y in pixels on the IR detector) with larger values of NN. (A CSM rotation of 0.024 degrees corresponds to a shift of 1 pixel at the center of the IR detector.)
The files names "blob_avgNN.fits.gz" in the second link contain the flat fields from which the "blob_maskNN.fits.gz" files were derived. The only difference is that for the latter the pixels outside the blobs were set to unity. Therefore, dividing by the "mask" has limited risk of doing harm, because it has absolutely no effect on pixels outside the blobs. Dividing by the "avg" will adjust all of the pixels, typically by a very small amount; whether doing so may improve upon the pipeline flat field has not yet been determined and is outside the scope of this analysis.
Download all blob masks (530 K tar file)