Previous: INTRODUCTION
Up: A GUIDE TO OBSERVING
Next: EXPOSURE TIME?
Previous Page: Limitations, Extra Observations, and the First Stages of Analysis
Next Page: EXPOSURE TIME?

WHICH CHIP? WHICH FOCUS?

The GEC chip or the RCA? The following are the main considerations.

1. Fig. 1.3 shows that the uncoated GEC chip offers little below 450 nm and nothing below 400 nm. Broadband U and B should not be considered with it. However, the coated GEC chip is probably the best bet for U and B, producing less severe colour terms than the RCA.

2. Fig 1.3 also shows that at the long-wavelength end there is little to choose between the chips. All are very insensitive by 1050 nm, and dead at 1100 nm.

3. If speed is the only consideration, the aims of the experiment must be well-defined and the sums done (according to section 3) in order to choose. The GEC chips have the lower readout noise by a factor of 6; the RCA chip has the higher efficiency by a factor of 2 to 3 (plus fringes). In general, then, sky-limited observations (long integrations on faint things) will suggest the RCA chip, while noise-limited observations (short integrations, bright things) suggest the GEC chip. (But remember that overheads (setting, filter changes, readout, etc) tend to make short observations (say < 60 sec) all the same effective length.) Some illustrative curves appear in Fig. 1.7.

4. Large pixels mean high speed - at least in the noise-limited case. For this reason, the INT prime focus will generally be preferred to the Cassegrain. But sampling must be considered. At this focus, the 30 m pixel size of the RCA chip corresponds to 0.74 arcsec, while the 22 m pixel size of the GEC chip corresponds to 0.54 arcsec. According to sampling theory, both chips undersample the image unless it exceeds 2 arcsec FWHM, infrequent on La Palma. With regard to resolution, the median seeing at the INT is 1.2 arcsec FWHM. The 0.74 arcsec pixels of the RCA chip broaden this by 9% to 1.3 arcsec (Recall that the Gaussian equivalent FWHM of a box filter is . The 0.54 arcsec pixels of the GEC chip broaden the median seeing FWHM by 6%. Fig. 1.6 shows the effective FWHM achieved as a function of real seeing FWHM. The difference is < 0.1 arcsec in the FWHM down to real seeing of 0.5 arcsec. If the success of an observation depends on both this difference and on better-than-median seeing, perhaps the experiment should be reconsidered. However, despite the relatively small difference in resolution offered by the two chips, they do differ significantly in terms of sampling. If subsequent analysis will involve rebinning, interpolation of pixels, reorientation, etc., the improved sampling of the image offered by the GEC CCDs may be essential. The Cassegrain focus is essential for high resolution experiments.

5. The RCA chip is notorious for `fringing', as explained in section 1.1.1. For broadband images, this is no particular problem; fringe patterns obtained during observations (see section 1.4) or available from previous observations can be used with iterative procedures to remove the fringes entirely. Observing continuum sources with narrow passbands (< 20 nm) or emission line objects is another story. In this case, the RCA chip can make its own white-light fringes, and if night-sky (and/or object) lines are within the passband, two or more fringe patterns may be present. It may be possible to remove them with an iterative process, but experience to date is limited. The GEC chip is thus recommended if passbands < 20 nm are to be used.

6. All three chips are cosmetically clean, and can be flat-fielded - with care - to better than 1% on large scales, at least in standard broad bands. The GEC chips suffer more cosmic-ray events because they are relatively thick; and this provides a somewhat shorter limiting exposure time (section1.3).In practice, such a limit is unlikely to be a consideration.

7. Photometry with CCD images reveals that undersampling is not a fundamental limitation. Profile-fitting routines may break down so that rather more basic aperture-photometry routines are required; but accuracy is not impaired.