Previous: Signal-to-noise
Up: EXPOSURE TIME?
Previous Page: Signal-to-noise
Next Page: WHAT EXTRA OBSERVATIONS ARE REQUIRED?

Time limits

In any case there are real limits to the length of exposure.

The usual limit is saturation, as described in section 1.1.1. The saturation limit is similar for GEC and RCA chips (Table 1.1), but the lower efficiency of the GEC chip means generally longer exposure times before saturation takes place.

The time to saturation can be calculated from the limits of Table 1.1, and the expression for signal, equation (2). In practice, some non-linearity begins to set in at about 0.7 saturation level; a rule of thumb is to keep counts/pixel below 40,000. Saturation is of course the principal factor in determining exposure time for bright stars, and experimentally determined curves are shown in Figs. 1.7a and 1.7b for permissible length of exposure times, RCA and GEC chips. At dark of moon, the sky background takes 55 minutes and 170 minutes to produce signals of 40,000 counts in I for the RCA and GEC chips respectively for the INT prime focus.

Cosmic-ray events may provide a limit to long integration times, if saturation doesn't. Event rates for the chips (Table 1.1) differ by a factor of 10 (section 1.1.1); and are of course filter-independent. The events are readily recognised, and can be removed in the laundry - but too great a density will seriously damage any kind of photometry. Moreover if the experiment is to detect something at a given position, cosmic-ray events may confuse the issue to say the least. In detection observations, if not in others, it is highly advisable to add frames together to reach the required integration-time. This guards against the cosmic-ray at the position in question, and is equally efficient in terms of observing time as the single long integration, provided that the sub-integrations are are long enough to be sky limited and long compared with the readout time. It is advisable to take several frames and from the median to eliminate cosmic ray events. If you offset the telescope by a few pixels between exposures you can eliminate chip defects as well.

With regard to bright stars, the shortest possible exposure times are set by shutter speed. P R Jorden has supplied the following notes about the operation of the Compur focal plane shutter which is used in both the INT and JKT CCD cameras. The sequence of events for an exposure (of N secs) is shown below:

Thus a 1.00 second exposure could give a 0.4% change in exposure from edge to centre; the status should indicate the average time. Exposures of less than 1 second are not recommended. The effect is most severe on the INT Prime Focus since at f/3 the edge of the CCD ``sees'' the edge of the shutter; whereas at the JKT f/15 focus the CCD ``sees'' only the more central part of the shutter.

The non-linearity of exposure takes the form of a linear gradient of increasing exposure towards the optical centre of the prime focus corrector - i.e. the charge-transfer direction - i.e. left on the LEXIDATA display and to the East if the rotator position angle is 180. It can be checked with a very short exposure flat field on the twilit sky.

The shutter timing has been checked and appears accurate and repeatable; in particular the shutter status accurately reflects the actual duration of shutter timing. The true exposure time for a CCD exposure appears in the packet headers on each CCD frame (except in multiple exposure frames). The value given does not exactly correspond to that requested (it is usually several milliseconds more) but it does reflect accurately the true exposure time. BUG: Occasional soft failures of the camera controller do however produce a slower shutter response than normal (0.5s): re-initialising the camera controller clears this. The exposure time displayed at the end of the exposure will still be correct, so that the bug (now rare) may be recognised from it.