WFPC2 Phase II Observation Strategies

Since many observers will be new to WFPC2, or will not have used it in some years, we have assembled various resources below to assist people in writing and optimizing their phase II proposals for Cycle 16.

Recent Changes in WFPC2 Performance, Calibration, and Strategies:

Very little has changed on WFPC2 in recent years, at least that will affect Cycle 16 proposers. The largest change is the WF4 CCD anomaly, which is discussed below in some detail. Its primary effect will be a slight reduction in the accuracy of photometry on the WF4 CCD (~2% uncertainty added). Charge Transfer Efficiency (CTE) problems continue to slowly increase, and are also described below. However the net increase in CTE losses during recent years is merely a small increase over what was already present a few years ago. The dark current has remained essentially unchanged in the past few years.

Observation strategies are only slightly changed from several years ago. We are now recommending that most observers use position dithers instead of simpler CR-SPLITs. These will give better rejection of detector artifacts, and the software for combining dithered images has become more mature.

In some situations, CTE losses can be minimized by careful placement of the target on the CCD. Observers with small, single targets may wish to place the target closer to the readout amplifier rather than at the default apertures near the CCD centers; this will reduce CTE effects. Pre-flashing or post-flashing the CCD is another means to reduce CTE effects, but is generally not recommended due to the added noise and overhead times.

These and other topics are discussed in more detail below.

Phase II Resources:

A wide range of documents are already available to assist in writing Phase II proposals. Here we provide links to some key documents:

WFPC2 Instrument Handbook: The Cycle 14 version is the last full version, and the brief Cycle 16 Update describes recent calibration plans and the WF4 anomaly. Chapter 7 on Observation Strategies will be especially useful for new observers.

WFPC2 dither and mosaic patterns.

WFPC2 Exposure Time Calculators.

HST Phase II Proposal Instructions (large PDF file).

Dither Strategies:

We are now recommending that nearly all observers use position dithers instead of CR-SPLITs. This will give the best rejection of detector artifacts, and can also improve sampling of the image by the relatively coarse WFPC2 pixel grid. Dithering came into routine use on ACS, and the software for combining dithered images is now relatively mature. The exact dither strategy and optimal dither pattern will still depend on the individual program goals.

Some factors to consider when choosing a dither pattern are:

-- Removal of detector artifacts will be best accomplished with motions of two or more pixels on both the CCD axes.

-- Fractional pixel shift patterns can be used to improve sampling of the PSF by the relatively coarse pixel grid.

-- Differing pixel scales on the PC1 CCD (0.0455"/pixel) and WFC CCDs (0.0996"/pixel) can be a concern if the entire field-of-view is needed. Only carefully selected patterns will provide integer or fractional shifts simultaneously on the PC1 and WFC CCDs. These typically involve shifts which are a multiples of 0.249" (half-pixel shifts) or 0.498" (integer pixel shifts).

-- Overhead times to move the telescope. A single motion of the telescope requires about 2 minutes for WFPC2, hence complex patterns with many offsets can use large amounts of telescope time. (The overhead times for WFPC2 are many times larger than those for ACS, due to different command protocols hard-wired into WFPC2.) The overhead times become especially significant if there is one or less orbit for the observation. For observations spanning several orbits, the dithering overheads are usually hidden in target re-acquisitions.

Situations where dithering should be avoided: There are rare situations where dithering is not advised. If extremely uniform pixel sampling is required across the entire field, dithering may cause issues at the edge of the CCDs where the non-linear distortion becomes large. In effect, patterns which give, say, half-pixel offsets at the CCD centers will give some other offset at the CCD edges. Also for high-precision time-series photometry dithering can introduce additional sources or noise. There are small variations in the sensitivity across the face of each pixel, and moving the target around the pixel grid will cause spurious brightness variations.

The WFPC2 data archive does not automatically provide combined images from CR-SPLIT or dithered data, as was done for ACS. Observers will need to combine the images themselves. This should be relatively straight-forward, and work is underway to provide some new help guides for drizzling WFPC2 data.

The links below show some comparisons of possible dither strategies. These are by no means exhaustive, but serve to illustrate some of the issues and strategies that can be used to optimize the observation.

Half-orbit WFPC2 dither comparisons.

One-orbit WFPC2 dither comparisons.

Additional information on dithering can be found at these places:

WFPC2 Dither and Mosaic Patterns (with pattern examples).

WFPC2 Dither Handbook. Chapters 1 and 3 will be especially useful for observation planning.

WFPC2 Dithering FAQ (older but still useful information).

Charge Transfer Efficiency Issues:

Charge Transfer Efficiency (CTE) problems are important for nearly all WFPC2 observations. Due to radiation damage from its ~14 years on-orbit, the WFPC2 CCDs contain significant numbers charge trapping sites. After an exposure, the image is in the form of charge packets, which are then moved or "clocked" across the detector to the CCD amplifier where the signal is finally measured and digitized. During the readout process, charge will be temporarily held in the trapping sites, while the rest of the image proceeds onwards towards the readout amplifier. This trapped charge is hence removed from stars and other objects, causing the them to appear too faint. Sufficiently bright targets can also show vertical tails, typically 100 to 200 pixels long, where the trapped charge has been released and can be directly seen.

The amount of the CTE photometric loss depends on a number of factors:

- Charge loss increases linearly with the Y coordinate on the CCD (i.e. distance from the readout amplifier). There is also a similar but weaker X dependency.

- Faint objects lose a greater proportion of charge. Bright objects lose a smaller fraction.

- Images with high background illumination lose less charge, as the background fills some of the charge traps.

- Extended targets have less CTE, as the leading edge of the image tends to pre-fill the traps for the remaining pixels of the image.

The figure below illustrates the situation near the middle of Cycle 16. This is based on simulations for stellar targets at the WF CCD centers. As can be seen from the plot, a 600s exposure in a broad visible filter (e.g. F606W) leads to ~4% CTE loss for bright targets (10000 e-) while a faint target (100 e-) suffers ~17% loss. CTE losses are much larger for narrow-band or UV filters where the sky background is very low. On the plot we also indicate the background level for the F656N narrow band filter; for this background level the bright target would now lose ~12%, and the faint target ~43%, of its counts.

CTE losses can be minimized in several ways:

Moving the target closer to the readout amplifier. If you have a small single target, a few arcseconds in size or smaller, the CTE can be greatly reduced by placing the target near the readout amplifier. A factor of ~3 reduction in CTE effects is easily possible using this strategy. In practice, this means placing the target closer to the corner where the 4 CCDs meet, at low x and y coordinates.

For the WF CCDs, this could be done, for example, by specifying the WFALL aperture. The WFALL aperture is located about 10 arcseconds from the Pyramid Apex where the 4 CCDs meet, or more precisely at pixel (x,y)=(133,149) on the WF3 CCD. For the PC CCD, one could use POS TARG -7,-8 to move the target closer to the readout amplifier. (The previously recommended POS TARG -8,-8 was 1 arcsec away from a weak flat field feature, so we moved it over a little bit.) Of course, one must be careful not to move the target too close to the CCD corner, lest it move into the vignetted region along the CCD edges. Details of the vignetted CCD regions can be found in Table 2.5 of the WFPC2 Instrument Handbook; observers should keep images within the 100% illuminated region indicated in the right-most table column. Information on the aperture locations can be found here and here in the WFPC2 Instrument Handbook.

Using the WF CCDs instead of the PC1 CCD. Due to the larger pixel scale of the WF CCDs, they have about five times higher sky background than PC1, which will help to reduce CTE losses. Though, of course, the WF CCDs will have lower effective resolution.

Pre-flash or post-flash using internal lamps. These raise the background level in the image, which will serve to fill some of the traps. However, in general, we do not recommend pre-flashing as there are a number of negative side effects.

While pre-flashing will reduce CTE, it will also increase the noise due to the sky background, and in most cases the increased noise cancels any improvement due to the recovered target counts; there is no net improvement in the signal-to-noise ratio. The figure below illustrates signal-to-noise ratio as a function of background at several target intensities. As can be seen, pre-flashing does not significantly improve the signal-to-noise ratio.

Pre-flashing will significantly increase the overhead time per-exposure due to filter changes for the pre-flash exposure, time needed to command lamps on and off, and the time of the pre-flash exposure itself. A single preflash will generally require 3 minutes of overhead. During this time, cosmic rays continue to collect, which further reduces image quality.

Pre-flashing will also increase the calibration complexity, as it will be necessary to subtract the non-uniform pre-flash from the images prior to science analyses. It may also be necessary to obtain on-orbit images of the pre-flash (i.e. calibration images), so as to facilitate its subtraction from the science images.

For these various reasons we are discouraging observers from pre-flashing their images.

There may be a small range of programs on bright targets where improved absolute photometric accuracy is desired, in spite of the other issues, and pre-flashing may have some benefit for these observers. If you feel you would benefit from preflashing, please contact help@stsci.edu for additional details.

Post-observation corrections for CTE: Corrections are available for point sources as a function of epoch, brightness, sky background, and location on the detector. Several web tools for estimating CTE effects are available. The accuracy of the corrections is not well established, but we feel they are probably good to 1 part in 5 or better. The information for these resources generally pre-dates 2002, and work is underway to update the information. We are continuing to take CTE monitoring data on-orbit, and we intend to have results from that data available before the end of Cycle 16. We are also starting new calibration projects to study CTE for extended targets during Cycle 16 (e.g. observations of an Abell cluster and the HDF field).

Estimating CTE effects: The CTE web tools can also be used to estimate CTE effects for point sources during observation planning. Of the three tools, CTE tool #1 contains the most recent results. It will require the following inputs: MJD -- use 54500 for mid-Cycle 16; the source counts -- use the value of "object_counts" near the bottom of the WFPC2 ETC output page; the CCD x and y positions -- use (400, 400) for the CCD center, or (800, 800) if you need the worst case for an object anywhere in the field-of-view; the background counts -- use the value of "sky_counts" again from the bottom of the ETC output page; and finally the stellar magnitude of the target -- this only needs to be very approximate. It will then output XCTE and YCTE, which are the magnitude changes due to two types of CTE loss, or equivalently the fraction of counts lost to CTE. For example, XCTE=0.01 and YCTE=0.17 imply 0.18 magnitudes loss or ~18% loss in counts.

For extended targets, the leading edge of the target will tend to "pre-flash" the detector, and hence CTE effects will be reduced. We do not yet have a good quantatative model for extended target CTE losses. But we believe the total lost charge (in electrons) for each CCD column through the target will be roughly equivalent to the losses expected for a point source with the same brightness as the brightest target pixel in that CCD column. Hence a rough estimate for the percent losses can be made using a method similar to that described above for point sources. One would estimate the peak electrons/pixel for the target (use one of the ETCs), then use this value in CTE tool #1, and finally then divide the resulting percent losses by some measure of the target's size in the y-direction in pixels (we rcommend using the target HWHM in pixels). For example, a 1 arcsecond diameter galaxy on a WFC CCD where CTE tool #1 predicts 18% CTE losses for the brightest pixel, would have about 18%/4.3 = 4% CTE losses for the entire galaxy. This is of course only a rough estimate. Plans are underway to obtain a better quantitative calibration of extended target CTE losses during cycle 16.

Further details on CTE effects can be found at:

WFPC2 WF4 CCD Anomaly:

The WF4 CCD, one of the 3 wide field camera CCDs, suffers an anomaly in the readout electronics whereby the CCD amplifier gain and bias levels are unstable. This results in low photometric counts for targets and faint horizontal stripes (<1 DN) in the background.

For images taken at A-to-D converter gain 7, the photometry can be up to ~40% too low. At A-to-D gain 15 the photometry can be up to ~70% too low. The errors are fairly well characterized, and the photometry can be corrected to a few percent accuracy during post-processing. Work is underway to incorporate these corrections in the WFPC2 calibration pipeline, so that they will be automatically performed. For most observers, the net effect will be slightly larger uncertainties (+/-2%) in the WF4 photometry, which is related to imperfections in the correction equations.

Faint (~1 DN) horizontal stripes are also seen in the background of the WF4 images. In fields with few objects, the stripes are easily removed by spatial filtering. However, in crowded fields or in the presence of extended objects, simple filtering will be difficult. A second possible method for removing the stripes is to use information from the overscan and vignetted regions of the CCD to model and remove the stripes; work on this strategy is underway. Effort will be made to also include stripe removal in the WFPC2 calibration pipeline.

The WF4 anomaly is having an indirect effect on the relative positions of the four CCDs. In an effort to mitigate efects of the on-going WF4 amplifier failure, we have been lowering the operating temperature of WFPC2 approximately 1 deg C per six months. (This effects the optical bench and electronics boxes, but not the CCDs themselves.) This causes small changes in the optical alignment, and moves the relative positions of the four CCDs approximately 0.01" for each 1 deg C change. Hence the relative positions of the CCDs will not be as stable in Cycle 16 as in the past. We will take calibration data to address this issue, but observers requiring a high degree of stability in the CCD positions may wish to make their own astrometric calibrations. The last temperature reduction occurred on 27 March 2007, and additional reductions are planned for August 2007 and February or March 2008. At this time, we feel there is a very good probability that we can keep the WF4 CCD operational throughout Cycle 16.

Details on the WF4 anomaly are available in the WFPC2 Instrument Handbook Update and in WFPC2 ISR_2005-02.

Other Image Anomalies:

We occasionaly receive questions from observers who are concerned about image anomalies when planning their observations. In particular, people are sometimes concerned about ghosts and reflections near very bright stars. These are rare, but more information can be found in the Field Guide to WFPC2 Image Anomalies.

Choosing Exposure Times and Minimizing Overheads:

The choice of exposure times is a complex issue for WFPC2 observations. Often a non-standard choice can lead to decreased overhead times and increased integration on target. Some key factors are:

- WFPC2 has a fixed menu of available exposure times. These are listed here. A non-allowed exposure time will be rounded down to the nearest allowed value.

- There are significant overhead times which vary in complex and unexpected ways. A table giving the total time to execute various exposure times is given here. Note for example, that 350s and 400s exposures take the same total time. But a 160s exposure takes two minutes less than a 180s exposures (for CLOCKS=NO which is the default).

- For exposures 180s and longer, specifying CLOCKS=YES will reduce the overhead time by 1 minute per exposure, but will also increase the dark current near the top and bottom of the CCDs. The increased dark current may be acceptable if the target is near the CCD center, or if the sky background greatly exceeds the dark current.

- Often using different exposure times for the two halves of a CR-SPLIT or different points on a dither pattern will lead to slightly increased total integration time. For example, if one can fit an 800s exposure with the default CR-SPLIT (i.e. 2x400s) into the orbit, but not a 1000s exposure, sometimes a 500s + 400s can be made to fit (implemented as e.g. 902s exposure with CR-SPLIT=0.555).

- Filter changes take 1 minute per filter.

- Dither patterns will generally take 2 minutes per movement. Sometimes the number of movements for a given pattern can be reduced by specifying the pattern manually with POS TARGs instead of PATTERN commands. This allows, e.g., starting a new filter at the current position rather than returning to the home position.

Many of these details are discussed further in Section 7.6 of the WFPC2 Instrument Handbook.

Version 1.21 4/30/2007 JB