STScI Logo

Hubble Space Telescope

S T A N / W F P C 2 - Number 3, January 1995



Each cycle 5 PI has at their disposal two people, their Program Coordinator (PC) and Liaison Scientist (LS), dedicated to helping them develop and implement their phase II HST program.

By now, all cycle 5 PIs should have been contacted by their PC/LS team, and received their preliminary phase II template, and a submission timeframe for their phase II program. Detailed instructions on the phase II submission were sent out to cycle 5 PIs in December.

PIs should feel free to contact their PC or LS for help at any time up through the execution of their cycle 5 HST program. The PCs will maintain a file of advisories on RPS2, for example, so that known problems and recommended workarounds can be accessed easily. Look for this advisory list (and other Phase II documentation) in the World Wide Web page for Phase II Proposal Development/


by John Biretta and Bill Sparks

"Strategies for Polarization Observations", a guide outlining strategies and recommendations for WFPC2 polarization observations is being compiled and will appear shortly on the WFPC2 Web page.

This guide also updates various diagrams and charts in the WFPC2 Instrument Handbook.

In planning your observations, note that regions of the field within 14 arcseconds of segment boundaries in the polarizer quad should be avoided due to cross-talk between the filter segments.


Cosmic Ray Removal Schemes: An Introduction

by Andy Fruchter

The WFPC2 is a sensitive cosmic ray (CR) detector. In a single ten minute exposure one can expect that between one and two percent of the pixels will be corrupted by CRs at the several sigma level. While substantial variation in event rate is often seen through an orbit and between orbits, on average the number of affected pixels grows linearly with dark time. Because the PC and WF chips are physically identical CCD's, the cosmic ray rates in all four chips are similar. As a result, in a 2000s exposure more than 4% of the pixels may be hit by CRs.

It is therefore recommended that except for the very shortest integration times, observers take at least two successive images of a field to allow efficient cosmic ray removal. It has been standard practice to maintain the same position while performing the second ``CRSPLIT'' image; however, GO's may wish to consider shifting the telescope pointing by a number of pixels between the two CRSPLIT images to allow removal of chip defects along with cosmic rays. The WF pixels are about 2.2 times larger than the PC pixels on the sky. As a result, a shift of 11 PC pixels corresponds to a shift in WF of 5+e pixels, where e is smaller than the usual random position shifts between successive images due to telescope pointing error. Because the readout directions are rotated between chips, the user should shift by 11 PC pixels (or 5 WF pixels if using a WF chip as the aperture) in both x and y to make sure that column defects can be eliminated in all chips. In the next articles, Rick White and Wolfram Freudling describe algorithms for detecting and removing cosmic rays. For more information on dithering, see the article by John Biretta.

For some projects, however, two successive images will be insufficient to remove cosmic rays to an acceptable level. In the case mentioned above, where the observer has 2x2000s exposures, about 1000 pixels per chip will be unrecoverable, because they have been hit in both images. Furthermore, because CR events typically affect ~ 7 pixels per event, these pixels will not be independently placed, but rather will frequently be adjacent to other unrecoverable pixels.

When taking more than three images of a field, the observer will have to decide between two general approaches to cosmic ray removal before scheduling the observations. The simplest and least CPU intensive approach is to use integer pixel shifts between images. Cosmic ray removal with three or more images is then quite straightforward: in particular, with three images, a number of observers have used the technique of dividing each image by the minimum of all three. This allows one to distinguish between cosmic rays and the large changes in flux that can occur on the edge of a star's PSF due to random pointing motion between images.

However, where CPU time is not a restriction, or where the field of interest is small, the user may wish to consider the method discussed by Wolfram Freudling in this newsletter. In this approach, sub-pixel shifts (added on to a non-zero integer shifts to eliminate bad pixels) are used both to remove cosmic rays and to deconvolve the image -- the need to deconvolve WFPC2 images arises primarily from the undersampling of the PSF by the instrument's large pixels. At present, we have little experience with the effects of real-world problems, such as errors in the estimated PSF and non-optimal sub-pixel shifts, on this method. However, Freudling's program not only returns the deconvolved image, but also the CR masks, so the user has the option of linearly combining the images using the CR masks if there is any doubt about the fidelity of the deconvolution.

Finally where the user has received an abundance of riches from the TAC, and has numerous orbits on a field, one can combine the two approaches and take several images in each of a number of sub-pixel shifted positions. We should all have to schedule such observations!

Cosmic Ray Rejection in Long-Exposure Image Pairs

by Rick White

The existing CR rejection/image combination algorithms in STSDAS and other packages implicitly assume that CRs are rare events, so that it is unlikely that a particular pixel will be hit by CRs in all the images being combined. This is a good assumption when at least 3 images are available or when the exposures are short; however, it is not a good assumption when a single pair of long exposure images is being combined. In that case, the standard tasks produce images with many remaining CRs due to their overlapping.

Brad Whitmore has used a method where he follows the 'gcombine' task with a pass by the IRAF single-image cosmicray task. This removes many of the remaining CRs and also removes most hot pixels; it is probably the easiest solution using existing packages.

A better algorithm can be developed if one explicitly assumes that some pixels will be affected by CRs in both images. Then if a CR is identified in one image, one must carefully examine the corresponding pixels on the other image to see whether it also was hit. I have developed an IDL routine for WFPC2 images that checks for overlapping CRs using proximity tests (are there CRs nearby in the image?) and shape tests (is this object too sharp to be a star?); the current method is certainly not yet optimal, but it has been used on a variety of images and appears to give better results than other approaches.

The programs generate both a combined image with interpolated values for pixels that were hit by CRs in both frames and two mask images showing where CRs were found in the two frames. The mask images can be used to compute the noise in the combined image and to determine which pixels have interpolated data (and so should not be used for quantitative analysis.)

These IDL routines are available to interested users, though be warned that they are a bit tricky to use and may still have problems for certain types of images. The algorithm is currently being implemented in an STSDAS task.

Removal of Cosmic Ray Hits from WFPC2 Images which are NOT registered

Wolfram Freudling, ST-ECF (

The severe undersampling of the PSF by the WF CCDs of WFPC2 degrades the resolution of images taken with those chips. Some of it can be recovered by dithering the exposures on a subpixel level, i.e. shift the camera by an amount less than the size of the pixels between exposures. One difficulty with the reduction of such dithered images is the identification and removal of CR hits. The task crrej in the wfpc package removes very effectively CR hits from stacks of registered images. CR removal programs which act on single images usually fail for WFPC2 images because the undersampling prevents a simple distinction between CR hits and stars. An effective removal of CR hits from dithered images has to make use of the information contained in all the images, but also take into account the sub-pixel shifts and the PSFs.

Recently, we have implemented such a method to remove CR hits from dithered images (Freudling, 1995, PASP, 107, 85) using a simultaneous Richardson-Lucy deconvolution. In each iteration, it identifies CR hits by comparing the current model with each individual image and dynamically creating masks. The output of the procedure is a combined and deconvolved image and a list of CR hits for each of the images. The combined image is identical to the combination presented by Hook & Lucy (1993, ST-ECF Newsletter 19, 6) except that it is cleaned of CR hits. The information locally destroyed by a CR hit is automatically filled in by the images not affected; no interpolation is used. The procedure also adequately removes hot pixels IF the individual images are shifted by an integer number of pixels in addition to the sub-pixel shift.

Both the IRAF and MIDAS implementations of the method are available via anonymous FTP from in the pub/swlib/crcoadd directory.

Dithering WFPC2 Observations: Image Noise and POS TARGs

by John Biretta

As Andy Fruchter mentions in the above article, dithering by integer pixels effectively eliminates the bad ones. It also has additional benefits in terms of reducing the noise contributed by warm pixels and pixel-to-pixel variations in the dark current. In a simulation of three 1800 sec. exposures of a field of faint galaxies, a non-dithered combined image was found to have an RMS error of 1.05 DN (compared to the "true" image), while dithering all three images during the observation, and later aligning and simply combining with CRREJ reduced the noise to 0.58 DN. Eliminating residual warm pixels "by-hand" in both simulations reduced the errors to 0.59 DN and 0.54 DN, respectively, for the non-dithered and dithered observations. Hence, dithering appears to have no "cost" in terms cosmic ray removal, produces a truer image, and reduces the final noise by about 10%.

The report "Simulation of Dithered Exposures" contains more details about this effect. It will be posted to the WFPC2 Documents page.

Observers interested in dithering sometimes ask what the exact relationship is between the POS TARGs and the CCD rows and columns. The POS TARG axes run exactly along the CCD rows and columns on the specified aperture. For example, if you specify aperture WF3 (or WFALL), the POS TARG axes will run *exactly* along the rows and columns on WF3. Due to small rotations between the CCDs (<0.5 degrees), the POS TARGs will run only approximately along rows and columns on other CCDs. Similarly, specification of aperture PC1 places the POSTARGs *exactly* along rows and columns on PC1, etc. For more information, please see the report "Dithering: Relationship Between POS TARGs and CCD Rows/Columns" which will be posted in the WFPC2 Documents page.


by Sylvia Baggett

In a few rare cases observers have used the optional parameter CLOCKS=ON in order to minimize the effects of severely saturated pixels.

Users should be aware that taking WFPC2 exposures with clocks=ON results in a slight decrease in exposure time (either 0.125 sec or 0.250 sec, depending on the shutterblade). For this reason, we recommend that short exposures not be done with clocks=ON.

The decrease in exposure time is due to the manner in which the application processor (AP17) in the spacecraft computer operates the shutterblades. When clocks are OFF, the WFPC2 microprocessor opens the shutter, the AP17 closes the shutter, and the exposure time is as requested. However, with clocks ON, the AP17 opens the shutter, first blade A, then blade B. When blade A is closed at the start of the exposure, the actual exposure begins 0.125 seconds after the AP17 issues the blade command. When blade B is closed at the exposure start, the exposure starts 0.250 seconds later (after the AP17 sends the open-A command followed by open-B).

Please note that the EXPTIME keyword in the science data headers will not reflect the true, shortened exposure times. It can be corrected based upon the shutter in place at exposure start (given by the SHUTTER keyword).

Cycle 4 observers affected by this problem have been notified directly.


Paper on "The Photometric Performance and Calibration of WFPC2"

The WFPC2 Investigation Definition Team has written a paper for publication in PASP that describes the photometric performance of the WFPC2. This is a followup to their paper on "The Performance and Calibration of the WFPC2". The new paper is again by Holtzman et al., and is not yet in final form so it is still subject to revision. Some of the highlights are:

The WFPC2 "Dark Glow"

by Rick White

We have recently discovered that a component of the WFPC2 dark current is very likely due to low-level light emitted by the field flattener lenses in front of the CCDs rather than to the usual detector and electronic sources. At the current operating temperature (-88 C), there is a very noticeable drop (30-50%) in the dark rate within ~100 pixels of the edges of the chip. The roll-off is seen at all 4 chip edges on all 4 CCDs.

The simplest explanation for this unexpected structure is that we are detecting light emitted within the WFPC2 and that vignetting by the mask just in front of the CCD causes the roll-off. A simple analytical model that assumes the light comes from the MgF2 field flatteners matches the observed dark distribution very well.

The glow is probably due to irradiation of the MgF2 by energetic particles, which may result in both Cerenkov radiation and fluorescense. Since the energetic particle flux varies greatly with position in the HST orbit, we expect that this "dark glow" also varies with time. We are pursuing more detailed studies of the effect in the hopes that we will be able either to predict the amount of light based on the orbital path of HST during the exposure or to derive the amount of light from the number of cosmic rays detected during the observation.

According to our model, at T = -88 C about 0.5 -- 1 x 10**-4 DN/sec of the measured dark count rate is due to the usual dark current, and the remaining 1 -- 8 x 10**-4 DN/sec (depending on the CCD) comes from the glowing field flattener. The dark count rates measured in ground tests at -90 C are about 1.5 x 10**-4 DN/sec; this is consistent with our model, since the particle rate is low on the ground so the glow contribution should be greatly reduced.

For the great majority of WFPC2 observations, this effect is negligible. In fact, it is noticeable mainly because the true dark rate is very low at the -88C operating temperature. However, if you have made or are planning observations for which the dark current is an important limiting factor (typically deep, narrow-band images), you will want to consider how a variable dark rate would affect your data. For example, the dark rate will generally not be the same in CR split exposures, which may make combining the various images more difficult. The roll-off in the dark rate at the edges of the chip will lead to a similar roll-off in the sky if the dark is not well subtracted.

This problem is still being studied; we encourage questions, comments, or reports of any problems that might be due to this effect.

Minor Movement of the Mirror Focus

The telescope is still undergoing desorption which causes a change in the mirror focus at the rate of 0.8 microns/month. Ongoing monitoring of the focus showed that by mid January we would be about -3 microns from optimal focus, hence an adjustment of +5 microns was made on January 15. Our goal is to stay within 2 - 3 microns of the optimal focus, at which point the effect on the image quality is very minor. As a reminder, a "breathing" effect of order +/- 2 microns occurs on orbital timescales. This imposes the primary limitation to maintaining good focus.

Decontamination Dates

	20 October  1994
	19 November 1994
	18 December 1995 
	13 January  1995

	week of:February 6
	Mar 6
	April 1
	May 1


We draw your attention to these papers, based on WFPC2 data, that will appear in the next few months. This list includes all preprints received by the STScI Library not yet published in the journals. Please remember to include our Library in your preprint distribution list.

"HST images of nearby luminous quasars. II. Results for eight 
quasars and tests of the detection sensitivity"  
Bahcall, J.N.; Kirhakos, S.; Schneider, D.P., ApJ in press.

"HST observations of the SN1987A triple ring nebula" 
Burrows, C.J.; Krist, J.; Hester, J.J.; Sahai, R.; Trauger, 
J.T.; Stepelfeldt, K.R.; Gallagher, J.S. III; Ballester, G.E.; 
Casertano, S.; et al.; ApJ accepted

"Hubble Space Telescope observations of young star clusters in 
NGC 4038/3039, 'the Antennae' galaxies" Whitmore, B.M.; 
Schweizer, F.; AJ 3-95


Any questions about the scheduling of your observations should be addressed to your PRESTO contact. If you do not know who this person is, PRESTO's page ( contains that information.

The Space Telescope Science Institute is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5-26555.