Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.stsci.edu/documents/dhb/webvol2/c21_wfpc1errors.fm10.html
Дата изменения: Tue Feb 10 20:06:09 1998
Дата индексирования: Sat Dec 22 19:08:56 2007
Кодировка:
Поисковые слова: http astrokuban.info astrokuban

[Top] [Prev] [Next] [Bottom]

46.10 Data Accuracies and Problem Solving

In this section we characterize the accuracy obtainable with WF/PC-1 data and suggest steps to take to achieve the best accuracy from the data. The table below summarizes the estimated calibration accuracies and the following sections offer some final suggestions on improving the calibration.*

**Estimated WF/PC-1 Calibration Accuracies**
Attribute	Estimated Accuracy	Comments
Bias	0.2-0.3 DN	Files `d8c08261w`, `d8c0826nw`
Dark	~0.0006 DN/sec	Files `dac14274w`, `dae14273w`; worse on hot pixels
Preflash	~0.5 DN	Assuming 8 sec preftime for PC, 30 sec for WF (files `d8c0*`)
Flatfields, in general	~3% ~25-30% ~5-10% ~1-2%	Small scales (<<1") scales > 1" if flat contains F122M or F8ND (includes most broad band filters) scales > 1" if flat contains no ND filter (includes most narrow band filters) scales > 1" if high-fidelity flat used
Flatfields, additional effects	up to 5-7% ~1% rms (4-5% peak) few percent	Short exposure reciprocity problem, if flat and data have very different exposure times. To minimize, use flat with same exptime as data. Measle effects, partially correctible (see text) changes caused by decon; reduced with deltaflats
Relative photometry	5-30%	Depending on flat used
Absolute photometry	5-30%	Strongly dependent on flat used (see "Flatfield Anomalies" on page 46-7 and "Improving the Flatfield Correction" on page 46-15)
Plate Scale	WFC: 0.1016" +/- ~0.0001" PC: 0.0439" +/- ~0.0001"

46.10.1 Improving Bias, Preflash, and Dark Calibration

Near the end of the WF/PC-1 mission, high quality bias and preflash files were generated and installed in the archive; these are listed in the WWW Reference File memo. We suggest verifying that these were used in the calibration (e.g., BIASFILE, PREFFILE keywords) and if not, recalibrating.
Check that the CTE correction was correctly applied for observations that were not preflashed. That is, the header keywords should be set as follows: PREFCORR=YES, PREFTIME=0, and PREFFILE=FILENAME. Note that the CTE correction was sometimes not applied where it should have been due to a bug in the PODPS pipeline. Use the preflash files that contain the updated versions of the CTE correction.
Review the HISTORY records in the dark file used for calibration: Was it generated from individual dark frames taken close in time to the science image? If not, recalibrate with a more appropriate dark reference file from the Archive. If necessary, a new dark reference file can be generated for recalibration, by identifying dark frames that were taken close in time to the science data (e.g., with StarView), retrieving them from the Archive, then combining and normalizing them into a reference file for use in calwfp.
If hot pixels are a concern, check the WWW memo for the locations of hot pixels or see the WF/PC-1 Instrument Science Report 93-02 for a discussion of the evolution and treatment of WF/PC-1 hot pixels.
Verify the DARKTIME keyword value; for more accurate dark calibration, darktime value can be recomputed (WF/PC-1 Instrument Science Report 93-01). Also note that for interrupted observations (perhaps due to loss of lock), the dark current continues to accumulate even when the shutter was closed. A bug in the PODPS pipeline caused only the shutter open time to be used to compute the darktime.

46.10.2 Improving the Flatfield Correction

Apply a Delta Flat

A delta flat correction may be necessary if there was a decontamination procedure between the epoch of the flatfield and the epoch of the science data, particularly for PC8 data. One of the archived delta flats could be used or a new one created from individual internal flatfield exposures (see "Choosing and Generating Delta Flats" on page 45-18).

Experiment with Flatfields

After the end of the WF/PC-1 mission, Closure and high-fidelity flatfields, as well as deltaflats, were generated and archived; these flats cover ~85% of the nearly 8000 external WF/PC-1 science images in the archive. Any of these flats may be retrieved from the Archive and used for recalibrations; the reference file memo on WWW lists all of the WF/PC-1 reference files available from the archive. Small scale errors in the flatfields are normally around 3% (the high-fidelity flatfields are generally good to nearly 1%) while 10% errors can be expected on larger scales. In addition, those flatfields which were generated from Earth calibration observations taken with the neutral density filter F122M are known to have 25% gradients across the four WF CCDs. The following suggestions may improve the flatfielding.

Select an alternate version of the flatfield used; many filter / camera combinations have more than one possible flatfield. Look for the -closure or high fidelity flatfields (Cycle 2 and 3) which have been generated and installed in the Archive in 1994 or try the "Super-Sky Flatfields" from the MDS (Medium Deep Survey) group. The WWW memos provide pointers to all these files. Closure flatfields can be back-corrected with a delta flat and applied to data taken before August 1993. WWW memos are available at:

http://www.stsci.edu/ftp/instrument_news/WFPC/

wfpc1_memos.html

Try a flatfield close in wavelength to the science observation; for example, use narrow band flatfields to flatfield images of emission line targets taken with broadband filters.
Remove the neutral density filter pattern if the flatfield was taken with F122M or F8ND (or use high-fidelity flats). Alternatively, if a high signal to noise target is small and close to the center of particularly WF2 or PC6, the neutral density filter gradient may not be significant.
Generate high-fidelity flatfields for those observation modes lacking such a flatfield; these would include WF in F439W, F702W, F547M, and others, as well as PC in F889N, F850LP, and F502N (the prospects for improving the UV flats are not good, since there are few or no on-orbit earthflats). For example, a F555W, WF high fidelity flat (accuracy better than 2%) can be computed by correcting the Closure F569W flatfield reference file for wavelength dependent effects; the correction is obtained from F122M filter Earth flats, since the F122M pattern varies slowly with wavelength. That is:
newflat (F555W) = flat(F569W+F8ND) *
[flat(F555W+F122M) / flat(F569W+F122M)

In this case the flats are Cycle 3 flats taken through the indicated filters. The ratio of the flats crossed with F122M corrects for any wavelength dependent effects between 555nm and 569nm, and is applied to the F569W+F8ND Earth flat, which is known to be quite good. In practice, this can be done by identifying the appropriate filenames using the WWW Closure Flatfield memo, retrieving them from the archive, and using imcalc to perform the computation. For example:

imcalc input=e751348dw.r6h,e6d10291w.r6h,e6d1028sw.r6h output=newflat.r6h
equals="im1*im2/im3" pixtype="real"

The flatfield data quality file may be updated as well:

imcalc input=e751348dw.b6h,e6d10291w.b6h,e6d1028sw.b6h output=newflat.b6h
equals="max(max(im1,im2),max(im2,im3))" pixtype="short"
Account for exposure time dependencies in the flatfields, which can be a signficant effect; for example, we have found that flatfields based on Earth flats with exposures shorter than 1 sec, when applied to science exposures much longer than 1 sec, will result in 5-7% errors. As noted in their header HISTORY records, the input flats used in the example above (e75*, e6d*) were generated from Earth flat exposures which were ~8 sec long and so, the resulting flat is optimized for long exposures. The flatfield can, however, be optimized for short exposures by including a wavelength-independent correction factor; in this case, the factor can be determined from the F588N+F122M Closure flatfield (generated from 400 sec Earth flats) and the F588N Closure flatfield (generated from 0.18 sec F588N Earth flats). That is: newflat(F555W,short) = newflat(F555W)*[flat(F588N)*flat(F569W+F122M)] /
[flat(F588N+F122M)*flat(F569W+F8ND]

In this case, newflat(F555W,short) is the new flatfield appropriate for short exposures, newflat(F555W) is the flatfield computed in the previous step, and the other flatfields are Closure flatfields taken through the noted filters. This final high-fidelity flatfield can then be used directly in calwfp. In a similar manner, Closure flatfields based on short exposure times could be optimized for longer exposure times.
Also be aware that many of the raw earth observations (from which the flatfield reference files are generated) and the INTFLATs (from which the delta flats are generated) contain small numbers of cosmic rays.
Check for features in the flatfield which could degrade the science data; the delta flats may be used to identify the positions of measles.

46.10.3 Improving the Photometric Calibration

The resultant relative photometry at visible and far-red wavelengths should be good to 5-15% in the PC and WF2, and 10-30% for the full field of the WFC. Somewhat better accuracy is possible for narrow and medium band filters if the flats were not observed through neutral density (ND) filters; in these cases, the accuracy will be around 5-10% for all cameras. If core photometry is used in conjunction with a linear magnitude correction (WF/PC-1 ISR 92-02), and corrections are made for throughput variations and flatfield problems (ND filter gradients, edge droop, measles, etc.), an accuracy around 3-5% should be possible. The accuracy for absolute photometry will ultimately be limited by the accuracy of the absolute calibration, which is around 5% (see WF/PC-1 ISR 92-09).

The uniform contamination layer has been shown to decrease the throughput in F555W by 10-15% about 6 months after a decontamination procedure; the degradation is worse toward blue. The effects can be corrected one of two ways:
- Read the correction from the plots of the stellar monitoring results. (See Figure 46.3 and WF/PC-1 ISR 93-02).
- Use the CONTAM keyword in synphot, which uses the date of the science observation to interpolate between data points in the CDBS stellar photometry table to arrive at the correction (see WFPC ISR 96-02).
The uniform layer of contamination is also known to scatter and absorb light. This scattered light effect is particularly evident in the Earth observations (from which the flatfield reference files are generated) and in the internal flats (from which the delta flats are generated) as an edge-droop. Due to the different illumination pattern, the edge-droop is not likely to be present to the same extent in the science observation.
The non-uniform contamination (persistent measles) will have small scale effects (2-5%) which may be partially corrected with the application of a delta flat.
Note that the WF/PC-1 photometric calibration, in the form of the STSDAS Synphot tables, is based on the Cycle 1 (SV) flats and standard star observations made during Cycle 1 (the contamination correction tables have been updated through the end of the mission and may be used to estimate the effect of the contamination layer). If the data has been recalibrated with the improved reference files discussed in previous sections, and the photometric information in the header keywords (e.g., PHOTFLAM) will be used, the synphot tables should be updated in order to make them consistent with the closure flatfields. The updates would require measuring statistics on all the final flatfields and adjusting the tables accordingly, before running -calwfp or synphot. An ideal set of calibration data with which to verify any changes made would be the photometric sweep (proposal #4785), which was taken near the end of the WF/PC-1 mission.

[Top] [Prev] [Next] [Bottom]

stevens@stsci.edu

46.10 Data Accuracies and Problem Solving

Attribute

Estimated Accuracy

Comments

46.10.1 Improving Bias, Preflash, and Dark Calibration

46.10.2 Improving the Flatfield Correction

Apply a Delta Flat

Experiment with Flatfields

46.10.3 Improving the Photometric Calibration