Frame Size and Pixel Type

For many applications, a serious restriction on FOC imaging is the amount of sky coverage available. The platescale associated with the F/96 relay (0 0225 per pixel), in principle gives us a maximum imaging area (for a 10241024 image) of 23'', but in reality this area is somewhat smaller because of geometric distortion. A further consideration relates to the aberrated PSF which has about 85%of its power in a halo which extends out to a radius of 100 pixels from its center. This means that a square perimeter with at least this width will be influenced, to a greater or lesser degree, by edge effects, further reducing the scientifically useful area of the image, at least as far as image restoration is concerned. The bottom line is that many attractive science targets, such as most Solar System planets, just won't fit. Even if you have determined that your target will fit, you should also remember that, because the scientifically useful area of any given format can be much smaller, it is almost an imperative that an interactive acquisition be included in your proposal (see later in this section).

As previously implied, successful use of the FOC usually requiresd careful planning of the exposures and filter combinations; this is especially true for those desiring to perform image restoration. These are not however the only considerations and so, as an example, let us review the factors which an observer will have to consider when deciding on an imaging format. In the first instance let us assume that the observer feels constrained, because of the target size or maybe just for maximum sky coverage, to use the largest format available, the F/96 5121024 (zoomed) format, in which case

• Because of the large area to be scanned by the read beam, the large format has the poorest linearity characteristics, i.e., significant non-linearity is to be expected in almost all data obtained with this format since the maximum reasonable count rate allowable (for non-linearity 10%) is about 0.05 counts/pixel/sec.
• FOC pattern noise (see below), appears to be proportional in some way to the level of non-linearity present. Choosing a format with poor linearity characteristics magnifies the problem of pattern noise.
• This format only has 8-bit pixels, i.e., the maximum number of photons which can be registered by a single pixel, regardless of exposure time, is 255, at which point the pixel ``wraps over'' and continues counting from zero. An individual pixel can be wrapped over several times, and therefore this problem cannot be safely corrected without a priori knowledge of the intrinsic structure within the affected area.
• For point sources, the maximum reasonable count rate (i.e., peak) is somewhat higher, (about 0.3 counts/pixel/sec), however, if the proposed target is a crowded star field, high levels of non-linearity will almost certainly be present in the cores of the brightest stars, even if wrap-over is avoided.
• The 8-bit count limit, combined with the count rate considerations, implies that for all but the most diffuse, extended objects, high signal-to-noise is difficult to achieve without prohibitively long exposure times.

A second option which users should consider is the use of a smaller format. The 8-bit count limit only applies to two formats, the 5121024 (zoomed) and the 512512 (normal) formats, so one solution would be to use a smaller format and take multiple images with positional offsets, i.e., 2512512 (zoomed) images (with 16-bit pixels) instead of one 5121024 (zoomed) image. There are other advantages to this solution in that, although your image has only half the area and all of the other problems still remain, half the area implies approximately twice the scan frequency. If the image is being scanned twice as often, the linearity characteristics are improved by about a factor of two, i.e., images are only half as non-linear. You can also expect an accompanying reduction in the pattern noise since this seems to scale with non-linearity. If you opt for the 512512 (zoomed) format you can either

a.

Obtain the same S/N with no wrap-over and twice the linearity, in the same exposure time,

b.

Obtain the same S/N with no wrap-over but with the same non-linearity in half the exposure time, or

c.

Obtain better S/N () with no wrap-over but with the same non-linearity in the same exposure time.

There are two main disadvantages involved in moving from a large format to a smaller one, particularly if image restoration is intended. First there is the reduction of the effective image area, i.e. the area free of edge effects. With the present (pre-COSTAR) PSF this is a very serious consideration since edge effects are present within a 100 pixel perimeter of every image format. The second disadvantage is that the smaller format may necessitate an acquisition image to ensure that the science target is positioned at the center of the image. Remember that the accuracy with which the telescope can be pointed depends almost totally on the accuracy of the Target and Guide Star coordinates (which can accumulate to several arcseconds of error), so when the imaging aperture is only 11'' square, an observer take a great risk by not including an interactive acquisition.

Following the installation and calibration of COSTAR we have a good news/bad news situation. The bad news is that the focal ratio of the FOC cameras will be changed from F/96 to F/151 (F/48 F/75), resulting in a reduction of the format sizes (in arcseconds) by a factor of 2/3 (F/96 plate scale is reduced to 0 014/pixel). With these smaller format sizes, the need for an acquisition image should be even more apparent. The good news is that, if you still feel a need to restore images after COSTAR, the PSF halo should be fairly negligible (only about 20%of the total light), and therefore edge effects should only be present in a relatively narrow perimeter, leaving a larger effective science imaging area (relative to pre-COSTAR). In fact, the absolute size of the effective imaging area (in arcseconds) is not going to be very much smaller than the pre-COSTAR situation. It should also be noted that the smaller size of the pixels give a slightly better sampling of the PSF. The 8-bit counting limit will, of course, be unaffected, and so wrap-over will still occur, which in turn will force shorter exposure times because of the 4-5 fold increase in the peak intensity of the PSF.

The effects of COSTAR on non-linearity and distortion, etc., are discussed in the relevent sections.