Next: How to measure CCD
Up: No Title
Previous: Dekkers
Early polarimetric tests showed the presence of a considerable
scattered component with some structure. We have conducted experiments
to identify the source of the scattered light, with a view to eliminating
it. We identified several distinct modes of scattering, each of which
can be quantified to a large extent from our data. They include:
-
a very symmetrical halo round each point source on the chip, which
does not show any preference for dispersion or spatial direction and
could be due to out-of-focus reflections on optical surfaces or
scattering by dust on optics; it is of a reassuringly low (but not
necessarily negligible) level. About 50 pixels from the centre, the
intensity has dropped to less than 0.15 %. Since this corresponds to
17 arcsec in the spatial direction (with the o and e spectra separated
by about 10 arcsec) and we are aiming at polarimetry of 0.1 % or
better, it is clear that this component can influence our polarimetry.
Since astronomical spectra do not normally consist of just isolated
emission lines, the scattered light is usually much more; at any one
point in one of the spectra it is an integral over a certain length of
both polarized spectra. Light scattered within a spectrum just
affects the spectral purity, but light scattered from the other
spectrum causes errors of polarimetry. When more than one of the
Dekker apertures is filled, as in lamp or blue sky exposures, the
effect is more complicated still. Such scattered light can be
extremely troublesome in a blue spectrum, since CCDs are much more
sensitive in the red and even small amounts of scattered red light can
yield appreciable signal. A possible cure will be to use red-cutoff
post-slit filters (nature has not provided suitable glasses, but
thin-film technology can help). Whatever remains after such filtering
will have to be modeled from known point-spread functions and
observed spectra, and removed from the data. We shall conduct
experiments towards such modeling and hope to advise in future on what
auxiliary data to collect when utmost accuracy is required; please
keep us informed of any successful modeling on your part.
-
local scattering in the dispersion direction only; it is difficult
to see how this could be due to anything else than imperfections
of the grating (e.g. long-period irregularities in groove spacing).
It is of no particular significance except for a minor effect on
spectral resolution; it does not affect polarimetry.
-
scattering in the spatial direction only, with `the slit' being visible
where it is ostensibly blocked by the Dekker; these slit images show
slight flaring or defocusing at the ends of the slit. The considered
opinion of RGO detector staff is that this is probably a CCD effect
due to the inherent gross overexposure of the test frames
and that it should be of no significance in normal use.
-
a highly structured ghost image generated when there is a lot of light
just outside the spectral range that is being recorded. It is not
clear whether the light is scattered on camera components or on the
immediate environment of the chip, within the cryostat. The ghost is
well focused and shows a dark rectangular feature and an ostensible
`spider'. RGO optical staff are investigating and hope to eliminate it
once the cause has been found; a first guess from its complex
structure is that both a pupil-plane and an image-plane ghost are
involved. Under most astronomical conditions, much of its structure
will be smoothed by spectral integration. To attempt to detect it in
your application, the best possibility is to set the wavelength of one
of the available blue-cutoff filters roughly on the (red) edge of the
CCD and take a long exposure of your object with the filter in
position; such a wavelength setting may not be ideal for your
programme, but at least you can estimate one of the potentially
harmful scattered-light components.
-
a narrow long-range dispersed component, slightly skew; it shows up
also in a single-slot-Dekker exposure, making it look like a
calcite-slab polarization spectrum; in polarimetry this would be a
particularly vicious kind of error if it turns out to be of
significant strength (it would seem that usually it is not).
-
when a dichroic is used to split the light for simultaneous use
of the red and blue arms of ISIS, light reflected from the back
of the dichroic will be displaced along the slit and will corrupt
polarization data. It may be possible to cure this by an
anti-reflection coating on the back surface of the dichroic, but for
the time being we advise against using dichroics. Pressure from the
community might help.
Since CCDs are preferentially red-sensitive, one is more likely
to run into scattered-light problems when observing in the blue,
particularly if the source is strong in the red. The evidence
indicates that light just longward of where the recorded spectrum
stops is the worst culprit. At present this is all the guidance we can
give, but the question is being actively pursued. For quantitative
estimates and detailed structure of the scattered light, we obtained
(in February 1991) over 400 CCD frames under various conditions of
illumination and spectrograph settings. This database is slowly being
analysed and is available to users via the LPO archive at RGO; it
includes an ASCII observing log.
Next: How to measure CCD
Up: No Title
Previous: Dekkers
manuals store
Tue Oct 7 17:34:45 BST 1997