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

32.9 Other Data Problems

In this section we describe how to recognize and correct other problems that might affect re-calibrated FOS data. The problems addressed here are:

32.9.1 Dead Diodes

During calibration processing, calfos uses the dead diode reference file (the list of all disabled diodes) to determine how many diodes contributed towards the counts for each pixel. This information is needed to calculate the exposure time per pixel and convert counts to count rates. If an incorrect or incomplete dead diode reference file is used, calfos does not have an accurate reporting of the diodes that were used for the onboard integration. This can lead to serious errors in the count rates and fluxes for affected pixels. The effect of an incomplete dead diode correction (see Figure 32.5) has a very distinct signature, which looks like an absorption or emission feature with sharp edges, extending over a fixed number (NXSTEPS x OVERSCAN) of pixels (usually 20). Further, the dead diode absorption feature typically does not extend to zero counts because more than one diode contributed to the counts in any given pixel. The depth of the absorption feature for a pixel affected by a single missed disabled diode is 1 - [(OVERSCAN - 1) / (OVERSCAN)], or usually 20%. In Figure 32.5, panel (a) shows the raw counts and the dead diodes labeled 1-20. Panel (b) shows the count rate data from the pipeline processing from the .c4h file. Some of the dead diodes were correctly removed in this pipeline calibration while others were not. If there is a dead diode in your data that is not included in your dead diode reference file, you will see absorption features similar to those in this panel. In this particular case, a dead diode reference file for the wrong epoch was used in the processing of the data. Note also the 20% "emission" feature where a correction was applied to a region in which no dead diode existed. Panel (c) shows the .c4h file after the correct dead diode reference file was used in the calibration.


If you notice a feature in your data similar to the absorption feature described above, you should suspect a spurious dead diode in your observations. Tables 31.4 and 31.5 list all diodes that were disabled during the FOS lifetime. Normally, a diode was disabled and a new dead diode reference file created after the third report of anomalous behavior by that diode. Although the USEAFTER of the new reference file was set to the date of the first reported anomaly, your data may contain an earlier, unreported occurrence. If your suspected diode has been subsequently disabled (see Tables 31.4 and 31.5) you may use a later dead diode reference file to correct your data as long as the alternate reference file does not also correct diodes that are fully functional in your data. You may also contact the STScI Help Desk (help@stsci.edu) for further assistance in producing a special dead diode correction table for re-calibrating your observation.

32.9.2 Noisy Diodes

The effect of a noisy (or hot) diode was typically to produce an emission feature extending over a fixed number (NXSTEPS x OVERSCAN) of pixels (typically 20). Figure 32.25 shows an observation where pixels 400 to 420 are affected by a noisy diode. Unlike a dead diode, the profile of the feature need not be particularly flat since the degree of spurious signal generation by the diode may have varied from one 300 millisecond internal readout cycle to another. The effect of a noisy diode cannot be removed by recalibrating the data.


Noisy diodes sometimes appear for only very short intervals within an exposure. Look at individual groups of your data in the group counts paper products plot as an initial diagnostic of the possible presence of a noisy diode in your data. You might see that one or a few groups have substantially greater signal than the rest. Similarly, for ACCUM mode you can use task deaccum (see section 33.4) to plot the whole spectrum accumulated in each individual readout interval to isolate a noisy event to only a few groups.

Cycle 6 Observations: A particularly strong noisy diode appeared around diode 250 (pixel 1000 for quarter-stepped data) in FOS/RD spectra on several occasions after July 1, 1996.

If you notice a feature in your data similar to the emission features described above, you can manually edit the data to cosmetically smooth over or blank out the affected pixels. IRAF or STSDAS tasks to do this include fixpix or splot in its etch-a-sketch mode.

32.9.3 Detector Background (Dark)

The FOS was subject to two types of background effects caused by high energy particles:

Figure 32.27: Background - Reference Files: Dotted Line: FOS/RD, Solid Line: FOS/BL

Further, there were some indications that the geomagnetic model used to scale the reference background file in the calfos pipeline1 may have underestimated the background counts in science data by approximately 10% at low geomagnetic latitudes (< 20 degrees) and by about 20-30% at high geomagnetic latitudes. At present no refinement of these uncertainties is possible without a detailed study of the dark as a function of the ambient geomagnetic field measured by the onboard magnetometer. These uncertainties are not significant for strong sources (you can verify this by comparing the counts in the .c5h and .c7h files), but could cause substantial errors in the derived flux and spectral shape of weaker sources. For example, a V~19 magnitude star with an effective temperature of 10000 K, produced the same count rate at 2700 . as the low latitude mean background of the FOS/RD detector.

On the other hand, FOS ISR 146 examined all dark measures made through July 1, 1996 and essentially confirmed the SV mean levels as a function of geomagnetic latitude, gm. No correlation of the dark rate with geomagnetic longitude (apart from the vicinity of the SAA), solar angle, or solar cycle was found. The mean background from this analysis is:

FOS/RD:

FOS/BL:

Very few FOS observations obtained simultaneous background data due to the commanding requirement that forced the allocation of half the observing time to this measurement. In these rare cases signal from an unilluminated portion of the photocathode was used for background subtraction. The error in this background subtraction was simply related to the statistical error in the background data.

The FOS grating-scatter correction (See "Scattered Light" on page 32-58.) is effectively a scattered light plus residual mean background correction. After the standard mean dark correction has been applied, the scattered light algorithm calculates the mean count level in a range of unilluminated pixels (if one exists - see Table 31.6 for a listing of unilluminated pixel ranges) and this mean correction is subtracted from all pixels in the spectral region. This procedure does serve to correct underestimates of the model dark background, but for the same statistical reasons mentioned earlier, the corrected fluxes of faint sources can often vary around or drop below zero. Nonetheless, examination of any unilluminated region in the .c4h data can help in understanding the degree to which the background of your spectrum is dominated by individual particle events.

Accuracies

Mean rates may contain 10-30% uncertainties depending upon geomagnetic latitude. In nearly all cases, these uncertainties in the dark model are dominated by the statistical uncertainties in the estimated mean dark signal. That is, individual particle events dominate the dark count distribution for shorter exposures. The distribution of dark count per pixel begins to even out and resemble the scaled model mean values for exposures >2000 seconds


You can refer to the .c7h file and the FOS paper products for information about the background subtraction made for your data. The mean dark level is reported in the exposure summary section of the FOS paper products. When it is a significant contributor to the total signal, the mean dark count per group is visible in the group counts plot, as well. Additionally, as described above, you can examine any unilluminated pixels in the .c4h file as an indicator of the statistical nature of the actual dark signal in your observation.

32.9.4 Scattered Light

As a single pass spectrometer with blazed, ruled gratings and detectors that were sensitive over large spectral ranges, the FOS was subject to scattered light which originated primarily in the diffraction patterns of the gratings and the entrance apertures, as well as in the micro-roughness of the grating rulings due to their ruled surfaces.

Pre-flight laboratory data showed that the scattered component increased with increasing wavelength. The G130H, G190H, G270H, G160L, and PRISM spectra (below 2500 Å) were substantially affected by scattered light. A comparison between spectra taken with the solar blind (scatter-free) GHRS and with the FOS (see Figure 32.28) illustrates the conclusions of FOS ISRs 101 and 114 that FOS grating-scatter contamination dominated ultraviolet observations of late type stars. The scattered light appeared to be independent of wavelength in the regions observed and depended only on the magnitude and color of the target.

FOS ISR 103 described the calfos grating-scatter correction procedure, first implemented in March, 1994. The correction algorithm determines the mean detected signal for those diodes that are insensitive (or not illuminated) in a given dispersed spectrum and uses this mean as a measure of the wavelength-independent scattered light for the entire spectral range of the grating. This mean scattered signal is subtracted as a constant from all pixels in the spectrum. Only those gratings that have insensitive or unilluminated pixels (see Table 31.6) can be corrected in this fashion. If the wavelength range readout was restricted (e.g., in RAPID mode) it was possible that no data were recorded by the insensitive pixels. No scattered light correction was made in these cases. The algorithm was altered in February, 1996 to use the median, rather than the mean, with the additional proviso that all deviations from the median greater than 4 are eliminated in order to remove the impact of strong signals due to particle hits from the determination.

The calfos scattered light correction is effectively a residual mean background correction. For those gratings for which only a small number of pixels are used to form the mean scattered light correction (e.g., FOS/RD G190H), poor results may occur. Often at low count rates the quality of scattered light correction is obviated by poor photon statistics in the target spectrum. The corrected fluxes often vary about zero or are negative for faint sources.

The scattered light modeling task, bspec, which uses optical principles to describe the contribution of scattered light to the observed spectrum, is available for use as a post-observation parametric analysis tool to estimate the amount of scattered light affecting a given observation (see FOS ISR 127). The amount of scattered light depends on the spectral energy distribution of the object across the whole wavelength range to which the detector is sensitive and on the detector sensitivity as a function of wavelength over that range. As we noted, for red objects the number of scattered light photons can dominate the dispersed spectrum in the UV. For an atlas of predicted scattered light as a function of object type and FOS disperser and additional guidelines for modeling FOS grating scatter with bspec, see FOS ISR 151.

Data contaminated by grating scatter with a ratio of scattered counts to intrinsic counts of up to 5 can likely be corrected with bspec modeling provided the intrinsic target spectrum is accurately known for longer wavelengths (that is, over the entire range of detector sensitivity), and provided the exposure times were such that sufficient S/N exists from the weak signal counts component in the total of (signal+scattered) counts. It is also advisable to have available a spectrum of the target at longer wavelengths, at least within the adjacent FOS spectral range.


You can refer to the SCT_VAL group parameter of the .c1h file for information about the scattered light correction made to your data. The mean correction level is also reported in the exposure summary section of the FOS paper products. Comparison of this value with the signal level in the .c4h file gives an indication of the severity of the scattered light contamination in the spectrum. If you have an accurate spectral energy distribution of your object at longer wavelengths, you can model the grating scatter with bspec.

Figure 32.28: Scattered Light -Comparison of GHRS and FOS



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

1 FOS ISRs 099 and 114.

stevens@stsci.edu
Copyright © 1997, Association of Universities for Research in Astronomy. All rights reserved. Last updated: 01/14/98 14:55:10