STScI Logo

Hubble Space Telescope
IR alpha Flats

WFC3/UVIS Channel Flat-Fields
(Alpha Release)

April 29, 2011: NEW F850LP Flat now available

A subset of new flat-fields for the filters: F336W, F390W, F438W, F555W, F606W, F775W, F814W, and F850LP is now available and can be downloaded from this webpage, with other filters soon to follow. The maximum error from the flat-field calibration is now reduced from ~5% to ~1%. Once a more complete set of flat-fields has been produced further validation of the solutions has been achieved, these new reference files will be delivered to CDBS for use in the calibration pipeline. News updates and additional flat-fields will be posted to this page as they become available.

WFC3/UVIS users interested in applying the new flat-fields can manually recalibrate their data using CALWF3. Before reprocessing, the header keyword PFTFILE should be updated in the raw images to reflect the name of the new LP-flat (PFLTFILE), while the DFLTFILE and the LFLTFILE values should remain unchanged (=N/A).

> hedit my_raw.fits[0] PFLTFILE filter_lpflat_alpha2011.fits

>calwf3 my_raw.fits

The flat-fields currently used in the calibration pipeline were produced in spring 2008 at the Goddard Space Flight Center (see WFC ISR 2008-46) during the third and final thermal vacuum campaign (TV3) using the optical stimulus CASTLE. The ground-based flat-fields were taken with the detector at its nominal temperature of -82C using a Xenon lamp.

Soon after launch, inflight tests using aperture photometry of stars in the galactic globular clusters 47 Tucanae and ω-Centauri showed that the apparent magnitude of the same star changes by ~5%, depending on detector position (see WFC3 ISR 2009-19).

In August 2010, a preliminary set of alpha-release flats were posted to the WFC3 webpage for off-line reprocessing by users. These flats were computed from the stellar photometry and were intended to correct low-frequency residuals in detector sensitivity due to differences in the in-flight light path. Comparison between ground and in-flight internal lamp flats suggested that the pixel-to-pixel sensitivity was well constrained by the ground flats, and that only low-frequency modulations would need to be corrected. While the ACS residuals were well-constrained using low-order polynomial solutions, the WFC3 UVIS flats proved to be more complex. Further inflight calibration observations have given a more clear picture of the interplay between three separate corrections which were required to improve the ground-based flat-fields:

  1. A wedge-shaped flare extending diagonally from quadrant A to quadrant D.

    Observations of the moonlit Earth limb show a bright wedge-shaped feature similar to the ground flats. This is not a true variation in sensitivity, but a ghost, or flare, caused by light reflected multiple times between the detector and the chamber window. The flare is a superposition of four specific ghost reflections varying in intensity from ~1% in the center of amp D to ~2% in the upper left corner of amp A, and this feature has been removed from the flat-fields using a geometric model of the tilted UVIS focal plane and the projection of four defocused ellipses created by the interfaces of the two detector windows. (See McCullough, WFC3 ISR 2011-xx, in press, and WFC3 ISR 2001-17).

    some_text

  2. Amp-dependent gain variations.

    Each of the UVIS quadrants has a unique gain value at the 2% level. When CALWF3 applies the gain correction to convert detector counts to electrons an average value is used, with the expectation that the flats will include the needed changes. The quadrant-dependent offsets are clearly visible in the F606W LP-flat shown below.

    Prior to Feb 2010 (and the delivery of CALWF3 v1.8.1 and STSDAS v3.11), the amp-dependent gain values stored in the CCDTAB were applied to each quadrant of the science image, in the same way as done for ACS. Since the WFC3 ground flats already contained offsets between amplifiers, the resulting science images received a double correction for the gain between quadrants. The solution for both UVIS and IR channels was to modify CALWF3 to apply the mean gain value to all four amplifiers. All data obtained from the archive prior to this date will be impacted by this error.

    Recent tests of stellar photometry from in-flight observations indicate that adjustments of a few tenths of a percent are required to fully correct the quadrant-dependent gain differences. The residual correction to the gain values will be made with respect to quadrant A and will therefore not affect the photometric zero points. However, since the intensity of the flare is highest in the upper left corner of amplifier A, where photometric zero points were measured, the improved flat-fields are likely to change the computed zero points by ~1-2%.

    some_text

  3. Low-frequency residuals in detector sensitivity.

    Once the flare and the quadrant-dependent gain corrections have been applied, stellar photometry still shows low-frequency residuals in detector response at the ~1-2% level, which are most likely due to differences in the ground-based and in-flight optical paths. These residuals have been derived using the same software and methodology for the ACS L-flat corrections (see ACS ISRs 2003-10 and 2002-08) via stellar photometry from images obtained with large dithers over the field of view and at various roll angles which place the stars in different regions of the detector. The L-flat solutions are shown in the panel below for the 5 alpha release filters (top= F438W, F555W, F606W, bottom= F775W, F814W), where the detector has been divided into a 32x32 grid (and each grid point is a unique solution based on the stars falling within those 128 detector pixels in multiple observations of the cluster over time.) The residuals are shown with a stretch of +/- 2.5%, where black indicates that the photometry obtained using the ground flats will be too faint. The final L-flat correction is obtained by interpolating over grid points which deviate by more than 3-sigma from their neighbors, and then smoothing each chip separately with a gaussian function with sigma equal to one grid point. In contrast to the IR channel, the residuals for the UVIS channel show a moderate dependence with wavelength.

    some_text

The ratio of each of the 5 ground flat-fields with respect to the new alpha-release flat-fields are shown in the figure below, and these show a combination of the flare correction, the residual gain, and the smoothed low-frequency variations. The largest correction is found in F438W, which has a min to max ratio of –2% to +3%.

some_text

The figure below is a ratio of the previous alpha-release (August 2010) with the current alpha-release (March 2011) for each filter. The primary discrepancies are due to fitting the residuals in the previous alpha-release with low-order polynomials, following the same methodology as ACS. These were unable to adequately model the sharp features of the four ghosts, which were only recently understood to be an internal reflection, and not a true feature of the flat-fields. The images below are shown with a stretch of +/- 1.5%. Note that the largest residuals are at the vertex of the flare which was not well modeled using a polynomial solution.

some_text

The plot below shows the difference in apparent magnitude for all the stars in the range -15 < mF814W < –12 (with photometric errors less than 0.5%) in two images of ω-Centauri acquired with a 185 degree roll angle. The difference in magnitude is plotted as a function of the x-position for four separate regions selected by the y-position. The plots to the left were created using the ground-based flat, and those to the right using the new alpha-release flat. The red line represents the average magnitude difference in that region. Note the large deviation from zero for x < 2000 pix and y > 3000 pix, where the flare is strongest in the ground flats. Also note the deviation from zero at y < 1000 pix and x ~3000 pix where the silicon detector layer thickness is lowest.

some_text

The new flat-fields have been independently verified by computing new aperture photometry on reprocessed images in both drizzled images and pipeline 'FLT products' multiplied by the pixel-area map to account for varying pixel areas. This photometry has been used to compute residual L-flat solutions to verify that the deviations are less than 1% and show no spatially dependent signature.



New LP-flats released 03/25/2011 by the JET team
(J. Mack, E. Sabbi, T. Dahlen)

Created 03/25/2011 MJD Modified 04/11/2011 MJD