HST Data Handbook for WFPC2

4.3 Dark Current Subtraction Errors

---

4.3.1 Electronic Dark Current

At the operating temperature of -88 C, maintained after April 23, 1994, the WFPC2 CCDs have a low dark background, ranging between 0.002 and 0.01 e^-/s/pixel. A relatively small number of pixels have dark currents many times this value. These warm pixels, and their correction, are discussed in great detail in section 3.5.1. To remove the dark current, the standard pipeline procedure takes a dark reference file (which contains the average dark background in DN/s), multiplies it by the dark time (determined by the header keyword DARKTIME), and subtracts this from the bias subtracted image. Prior to April 23, 1994, the CCDs were operated at -76 C. The correction procedure is the same for these early data, but the average dark current was about an order of magnitude larger due to the higher temperature. Hence, the dark current correction is both more important and less accurate for these images than for later data.

The dark time is usually close to the exposure time, but it can exceed the latter substantially if the exposure was interrupted and the shutter closed temporarily, as in the case of a loss of lock. Such instances are rare and should be identified in the header keyword, EXPFLAG, and in the data quality comments for each observation; it will also be indicated by a difference between the predicted exposure start (PSTRTIME) and end times (PSTPTIME), which is greater than the total exposure time (EXPTIME); that is, PSTPTIME - PSTRTIME will be larger than EXPTIME. The true dark time differs slightly from pixel to pixel because of the time elapsed between reset and readout (even after the first pixel is read out, the last pixel is still accumulating dark counts). To the extent that dark current is constant with time, this small differential is present both in the bias image and in the observation itself, and therefore is automatically corrected by the bias subtraction.

New dark reference files are delivered on a weekly basis. However, due to the time necessary for processing, they usually aren't available until two to four weeks after the date of their observation. The primary difference between successive darks is in the location and value of warm pixels. This difference will be most noticeable if a decontamination occurred between the images used to create the dark and the observation itself. However, because direct treatment of the warm pixels themselves can be performed via warmpix, many users will find that they do not need to reprocess with the most up-to-date dark file. For more details, see section 3.5.1: Warm Pixels. An alternative is to create a custom dark reference file; this procedure is described in more detail in:Generating a custom dark reference file using "daily darks".

In order to track variable warm pixels, the weekly standard darks, prior to August 1996, were based on a relatively small number (10) of exposures and taken over a period of two weeks. However, these darks can be a significant component of the total noise in deep images. Observers whose (pre-August 1996) images are formed from exposures totalling more than five orbits may therefore wish to recalibrate their data using one of the superdarks, which have been generated by combining over 100 individual exposures (see the WWW Reference File memo on the WFPC2 Web pages), or generate their own custom superdark.

Since August 1996, the weekly standard darks have been produced by combining the relevant superdark with the warm pixel information in the dark frames taken that week. The combined file is obtained by using the superdark value for all pixels that appear normal in the weekly dark, namely, for which the dark current value in the weekly dark does not differ from the superdark value by more than 3; for pixels that do deviate more than 3, the weekly dark value is used. This compromise allows a timely tracking of warm pixels while maintaining the low noise properties of the superdark for stable pixels. Recalibration may still be appropriate since the weekly standard dark is not yet available when the image is processed and archived.

4.3.2 Dark Glow

While the electronic dark current is relatively stable between observations, a variable component has also been seen. The intensity of this dark glow is correlated with the observed cosmic ray rate, and is believed to be due to luminescence in the MgF₂ CCD windows from cosmic ray bombardment. As a result of the geometry of the windows, the dark glow is not constant across the chip, but rather shows a characteristic edge drop of about 50%. The dark glow is significantly stronger in the PC, where it dominates the total dark background, and weakest in WF2. The average total signal at the center of each camera is 0.006 e^-/s in the PC, 0.004 e^-/s in WF3 and WF4, and 0.0025 e^-/s in WF2; of this, the true dark current is approximately 0.0015 e^-/s. For more details, see the WFPC2 Instrument Handbook, Version 6.0, pages 87-92.

Because of the variability in the dark glow contribution, the standard dark correction may leave a slight curvature in the background. For the vast majority of observations, this is not a significant problem since the level of the error is very low (worst-case center-to-edge difference of 2 e^-/pixel) and because it varies slowly across the chips. Some programs, however, may require a careful determination of the absolute background level; observers may wish to consider employing techniques developed by other groups (e.g., Bernstein et al., 2001) or contact the Help Desk (help@stsci.edu).

4.3.3 Pointing Information in the Image Headers

Updating the pointing information

Improved knowledge of the detector plate scales and chip rotations, as well as changes in reference pixel locations, have resulted in periodic changes to the pointing parameters, especially early in the instrument's lifetime. These header parameters, which define the mapping between the pixel and world coordinate systems, can be updated using the STSDAS task uchcoord. The keywords affected include the reference pixel locations (CRPIX*), the values of the world coordinate system at the reference location (CRVAL*), the partial derivatives of the world coordinate system with respect to the pixel coordinates (CD*), and the orientation of the chip (ORIENTAT).

Prior to OTFR (released to the public May 16, 2001), observers requiring the most up-to-date pointing information in their science image headers ran uchcoord on their calibrated images. The new OTFR system, however, automatically calculates the best values for these parameters at the time the data are requested so there is no need to run uchcoord on freshly-processed OTFR data. OTFR data that have not been recently retrieved (i.e. have been sitting on disk or tape for some time) or pre-OTFR data may benefit from an update; any version of uchcoord may be used to update pre-OTFR data, but note that only the June 2001 or later version of uchcoord should be run on data processed through OTFR.

Only the June 2001 (or later) version of uchcoord should be used on OTFR data, as older versions of the task may over-correct the images. The version of STSDAS can be verified by typing =stsdas.version at the IRAF prompt.

Possible small errors in orientation header information in some images taken prior to Sept. 1997.

All HST data, including WFPC2 images, taken prior to Sept. 15, 1997, may have slightly incorrect orientation information in the headers (keywords PA_V3 and ORIENTAT). These very small errors are not correctable in OTFR.

For most images the errors are very small (<0.1 degrees), but for images taken during long visits, the errors can accumulate and reach many tenths of a degree. These errors affect only the rotation of the field around the proposed aperture location and otherwise have no impact on the target position in the image. The problem manifests itself in data taken during a long, multi-orbit visit containing one or more POS TARGs, slews, or other small telescope motions; the position angle information reported in the header changes slightly after each motion when in fact there was no change in orientation during the visit. The first image in a long visit contains correct header information; however, with each slew the orientation keywords in subsequent observations begin to deviate from the correct value at a rate of, at most, 1 degree per day (~7e-04 deg/min). The error also depends upon the target's position in the sky (sine of the declination); thus, keywords in images of targets near the equator will have almost no error while images taken near the poles will show larger errors. Note that moving target images appear unaffected by the problem, though further investigation is required to confirm this.

More details, including lists of possible images and visits affected, can be found in the WWW Orientation memo. Observers using the orientation information from any WFPC2 images on those lists should check the jitter data in addition to the science header. For advice on how to perform the check, please see the online FAQ "How do I best determine an observation's actual orientation?". If you have any additional questions or require further assistance, email help@stsci.edu.