FOC Calibration: Status and Accuracies

Calibration Accuracies

The accuracies specified in Table 11.1 are requirements , in that we expect that a single observation, provided that sufficient photons are detected that Poisson noise can be neglected, will be good to the accuracy stated at the 1s level. In practice, it may be possible to use more careful analysis to determine quantities of interest to higher accuracy (calibration goals).

Absolute Sensitivity

The F/96 Absolute Sensitivity is calibrated between 1200Å and 5500Å to an accuracy of 10% (1s). This is achieved by taking high S/N images of UV spectrophotometric standards through a number of filters, and by comparing the observations with the simulations. Observers can expect to achieve similar accuracy in the absolute flux measurement of a point source as long as the aperture used for photometry is large enough so that the PSF structure is not affected significantly by orbital variations (typically radius > 0.2 arcsec). Several causes contribute to this uncertainty: some can be removed, others cannot:

Errors in the spectrophotometric standards: the published error estimates for the UV spectrophotometric standards range from 3% at longer wavelengths to 5% at shorter wavelengths.

Errors in the assumed filter transmission curves: The filter transmission curves were carefully measured on the ground, but some subsequent degradation or change in performance might have occurred since launch.

Format dependence: A variation of sensitivity with video format has been noted. In particular, differences have been found in the relative response of the more common F/96 formats with respect to the 512 512 imaging format, see Section 6.3.1. We believe that the correction factors are accurate within 5%.

Variability of F/96 DQE: The UV throughput of the FOC has been regularly monitored as part of an observatory program. There appears to be a 5% degradation in sensitivity since launch, with no evidence for any wavelength dependence.

Source spectrum:. The value of PHOTFLAM provides an exact average Fl across the bandpass, irrespective of the source spectrum's shape. Large errors can arise in situations where redleak results in a significant fraction of the light observed, especially if the source has most of its emission outside the desired bandpass. If there is any doubt as to whether there are significant color effects, the observer is advised to use SYNPHOT or FOCSIM to check their absolute fluxes.

Pattern Noise: The small-scale sensitivity variation described in "Uniformity of Response (Flat Fielding)" on page 80 modulates the measured flux from a star. The effect on the measured flux depends on the aperture size used; for apertures larger than 3 pixels radius the effect is believed to be less than 2% rms.
Other flatfield effects: Large scale and pixel-to-pixel flatfielding errors can occur to affect the accuracy of PSF photometry. Taking multiple dithered observations can reduce the effect of the small-scale contribution.

Geometric Distortion

The plate scale has been determined for the F/96 camera to an accuracy of ~0.3%. The following formats (all centered) are geometrically calibrated and maintained: 512 x 1024 zoomed, 512 x 512zoomed, 512 x 1024, 512 x 512, 256 x 256, 128 x 128. Since March 19, 1995, these formats are calibrated using the new model (described in Section 6.11). The new geometric calibration has an accuracy of better than 0.5 pixel rms. The F/48 plate scale has never been measured after COSTAR deployment. Therefore, the estimated value must be used (see "Plate Scale" on page 90).

Long Term Stability: F/96. The F/96 relay has shown a reasonable level of stability over the period since launch. This relay has experienced a very slow, steady change over this time however this appeared primarily as a simple shift (~3 pixels) and rotation (~0.2 degrees), neither of which significantly affect astrometric measurements.
Short Term Stability: F/96. As fully described in "Geometric Distortion and Stability" on page 87 and "Plate Scale" on page 90, most of the variations in distortion in the F/96 camera occur during the warm-up time. However, residual variations (~0.25%) are noticed over the next few hours. We encourage users who require the best possible astrometric conditions for their observations to write a comment in the proposal requiring the observations not to start immediately after HV switch-on. They may also want to talk to their Contact Scientist.
Stability: F/48. In January 1992, the geometric distortion behavior of the F/48 relay, (which had never been quite as stable as the F/96), began suddenly to show large, erratic variations in its distortion characteristics. Due to the long inactivity period, it is not clear whether this effect persists. The data taken in early 1995 seem to indicate overall stability.

Overall this means that, for an average observation, the separation of two stars can be measured to an accuracy of 0.5 pixels(~7 mas) or 0.3%, whichever is larger.

Flat Fields

Smoothed full flat fields have been determined at 1300, 4800, 5600, and 6600Å to an accuracy of ~3%. Pixel-to-pixel errors may be much larger due to blemishes and detector defects, as it can be seen from Figures 12.1 and 12.2. Unsmoothed flat fields are available on-line through the FOC Web pages, but we also list here all the major features, so the reader can learn to recognize them in the FOC images:

Border Effects: can be external (F/96 occulting fingers, upper and lower left corners of the F/96 image, lower right corner for the F/48 image) or internal (bad rows at the top and bottom of the raw images, leftmost columns of the raw images, and a number of columns at the beginning of the scan line). These effects are not correctable.
Video and Digitizing Defects:
- fly back of the read beam of the television camera, especially noticeable in small formats.
- noise glitches on the scan coil driver caused by changes in the most significant bits of the line counter, show up as horizontal features at lines 256, 512 and 768. This is much more noticeable in the F/48 relay.
- Both relays show the center 512 512 outlined by a sharp change in sensitivity. The effect is due to burn-in of the heavily used standard camera format.
- These effects are not correctable, with the possible exception of the F/96 fly-back.
Reseau Marks, Scratches and Blemishes: as the pipeline flat field correction is heavily smoothed, none of these effects will be flat fielded out.
Large Scale Variations: it is estimated that the F/96 large scale response is accurate to better than 3% over most the photocathode, at the wavelength it was obtained, excluding edges and corners.
Time Variability: a minor amount of time variability has been observed in the flat field response. It is largest just after HV turn-on, where changes of 1-2% are seen when compared to the response seen after an hour observation. The changes for the F/48 are twice as large.
Pattern Noise: The response of individual pixels varies with a sinusoidal pattern by up to 5% or so due to pattern noise; this is described in more detail in "Uniformity of Response (Flat Fielding)" on page 80.

These factors limit the achievable accuracy for photometry of both point and extended sources.

Pointing Accuracy

F/96. The center of the F/96 aperture has been determined to within 0.2", and is maintained at this accuracy level through routine monitoring during each Cycle. All the centered formats have the same accuracy.
F/48. The center of the F/48 acquisition format has not been calibrated after COSTAR deployment. The acquisition of an external non-astrometric target has shown that the center appears to be offset of approximately 2" from its real position.
F/48 Slit. The position of the spectrographic slit relative to the image center will be determined. Present calibration plans expect to determine this offset to within 0.1".

The astrometric accuracy for an isolated star in the FOC field is determined by the Guide Star coordinate accuracies, which have an rms uncertainty of approximately 0.5 arcsec.

Objective Prisms

The FUV has been wavelength calibrated to within ~3 pixels, which corresponds to ~15Å at 1500Å and much more towards the red. The photometric response is determined to no better than xb1 50%. The NUV prisms are wavelength calibrated to within ~5 pixels (~20Å at 2500Å), with the photometric response only being known to within 20%.

Polarization

Accuracies of polarization measurements are similar to those achievable for relative photometry, or approximately 5% (1s). The angles made by the polarizers to the V2V3 coordinates are believed to be known to approximately 3xb0 , although will possibly refine this number after analysis of the Cycle 4 polarization calibration data. Calibration of the instrumental polarization involves measurements of the flux of unpolarized targets through each of the polarizers; again, at this time we can only say that the instrumental polarization is less than 5% based on our currently-available data, although we expect that we will be able to improve on this somewhat after more Cycle 4 and Cycle 5 calibration data are analyzed.

Users should also be aware that some care is needed in analyzing polarization data since the PSF is noticeably different through each of the three polarizers. This makes estimating the relative throughputs some difficult for small apertures, and analysis of complex fields at resolutions of <0.1" will require some care in matching the PSF profiles in the observations through the three polarizers. Current calibration plans for Cycle 5 call for the acquisition of PSFs through each of the polarizer filters in combination with the F342W filter. This should aid users in understanding the smaller features provided high enough S/N PSFs can actually be obtained.

F/48 Long-slit Calibrations

In order to support usage of the F/48 long-slit facility in Cycle 6, only the most basic calibrations will be accomplished. These calibrations will include:

Wavelength calibration. The wavelength dispersion of the gratings will be verified to within 3Å.
Long-slit Throughput. The narrow width of the long-slit (0.06") and the small size of the COSTAR-corrected PSF makes it difficult to know how much of the source light is entering the FOC. This throughput will be determined using dwell scans of a target across the slit to provide a measure of the throughput for well-centered observations. These values will only be good to within 20% in the visible and 50% in the UV.

Absolute Sensitivity
Geometric Distortion
Flat Fields
Pointing Accuracy
Objective Prisms
Polarization
F/48 Long-slit Calibrations

Generated with CERN WebMaker