Introduction

The first consideration in choosing a camera is how seriously you want to pursue astrophotography, and a second consideration is your budget - how much you are prepared to spend on a camera. An occasional astrophotographer will most likely opt for a digital SLR that can be used for a general-purpose camera, or a modified digital SLR that can be used as a general-purpose camera with an appropriate white balance setting when used during daytime. If you want the maximum control over image processing in order to achieve optimum results, you will almost certainly choose a cooled CCD camera - an instrument that can be used only for astrophotography.

The cost of digital SLRs range from about $1000 up to several thousands. A camera purchased by most amateur astrophotographers is likely to be in the $1000 to $2000 range. Costs of cooled CCD cameras range from about $1500 to $10,000 and more. Whatever you decide to purchase, it is a case of "you get what you pay for" in terms of quality, features, robustness, etc.

Telescope to Pixel Size Relationship

A third consideration involves the relationship between the telescope you are intending to use, the subject of the image and the size of the pixels in the camera. Although stars are point sources of light, theoretically falling on just one pixel, the Earth's atmosphere and guiding irregularities cause the light to spread out. In a long exposure image, a faint star may have an image width of about 3.5 arc seconds in average seeing. Seeing may be defined as the full width half maximum (FWHM) of the star image. (The FWHM defines the central 50% of the Airy disk for a star image. This 50% contains most of the light from the star.) If this light falls on too many pixels, the image will be very faint as each pixel will receive only a small share of the light (this is termed over-sampling). However, if the light falls on too few pixels (that is, the pixels are large) the star image will appear "blocky" ( this is termed under-sampling). For optimal sampling, the star image should be recorded by two pixels in each direction, resulting in a pixel size corresponding to about 1.75 arc seconds in average seeing conditions. This is true for deep-sky photography where the long exposures are significantly affected by atmospheric effects and telescope tracking errors. For images of bright solar system objects, exposures will be very short so that atmospheric effects will be much less and tracking errors negligible. In this case, the same amount of light will be much more tightly focused and we can usefully use a pixel size of one third or less that for deep-sky imaging or more likely, increase the focal length of the telescope by the use of an extender to concentrate the image on fewer pixels. The moon and planets are objects that benefit from over sampling.

A table of suggested pixel sizes for various telescope focal lengths is given in appendix 1, and the basis of calculation is given in appendix 2.

Digital SLR Cameras

A digital SLR attached to a telescope is an ideal camera for imaging solar and lunar eclipses, transits and comets. For conjunctions, wide-angle images of constellations and the infrequent but spectacular long-tailed comets, a DSLR with a normal camera lens of appropriate focal length is unbeatable.

DSLRs are one-shot-colour cameras; that is, they have a colour filter array (a Bayer matrix) in front of and integrated with the detector (CCD or CMOS) such that each pixel "sees" either red, green or blue. This necessarily reduces the resolution of the camera (ie. the finest detail it can portray) but the pixels are very small, ranging from about 4.5 microns square up to around 8 microns square (there are 1000 microns in a millimetre). Resolution is therefore usually quite acceptable, particularly for the smaller pixels common in the cheaper DSLRs with 22mm x 15mm detectors. In addition, the camera will have a separate built-in IR cut-off filter in front of the detector that transmits only about 20% of the 657-nanometre wavelength H-alpha light so that very little red from emission nebulae gets recorded.

If your objective is to photograph bright emission nebulae, then you will need a modified DSLR - one where the manufacturer's IR filter has been removed and has been replaced with an IR filter that cuts off above 657 nanometres or with a clear filter. You can purchase a new, modified Canon, Nikon or Fuji camera complete with a 12 months guarantee from the range offered by Hutech in the USA. You can also have your own camera modified by Hutech if it is a Canon, Nikon, Fuji, Olympus or Panasonic DSLR. (Refer to hutech.com for details and costs.) The modification can also be carried out by a technician in Melbourne (check Ice in Space website).

The major drawback of DSLRs is that the detector is not cooled to reduce dark current (thermal noise). At shutter speeds of 1/1000 to 1 second, very little dark current accumulates, but in long exposures the dark current can become the major component of what the pixels record, as apart from bright stars, the objects are very dim. This means that there is a limit to the length of individual exposures, even if dark frames are used. (A dark frame is an image of the dark current, taken at the same temperature and for the same exposure time as the main image. It is subtracted from the main image to leave only what was intended to be recorded. A dark frame is taken with the end of the telescope covered unless the camera is used in a dark frame mode.)

New models of DSLRs are offered every year, usually with improved performance and additional features, including lower dark current and, for time exposures, turning off the amplifier that causes "amp glow" at one or more edges of the detector. Because of this rapidly changing market, specific camera models are not discussed in this summary.

DSLRs have the ability to take and subtract a dark frame each time an image is taken. This is convenient if the ambient temperature is likely to vary significantly over the course of an imaging run but it does double the time for each image. If you use the B setting for longer exposures, then it will save a lot of time if you take one set of darks and apply them to all the images during image processing at a later time. However, if there is a significant temperature variation during an evening's imaging, one set of darks may not yield optimal results. (As a general rule, the amount of dark current is halved for every 6-degree drop in temperature.)

Focusing can be a challenge. The camera autofocus cannot be used at night, so focusing must be done visually unless you have a motorised focuser on the telescope and the camera is connected to a computer with an autofocusing program. Such programs make focusing both accurate and quick. In the absence of such equipment, the best option is a knife-edge focusing tool (available from Hutech for about US$140, or Stellar Technologies International for about US$230). Other alternatives include a Hartmann mask and crossed bars in front of the telescope tube to create diffraction spikes. Both of these require close observation of the image to determine best focus. (I have found the diffraction spikes method more accurate than the Hartmann mask.) Some DSLRs now on the market offer a "live view" mode for the camera's LCD monitor, where the image can be highly magnified, creating an effective method of achieving accurate focus.

If you purchase or make a cable to connect the camera to a computer, the whole image acquisition and file storage procedure can be pre-programmed and automated via one of several software packages available. Typically, about six 5-minute images combined in an image processing program would be necessary for bright emission nebulae such as M8, M42 or the Eta Carinae nebula. Much longer total exposure time would be required for fainter targets

One of the most significant advantages that DSLRs have over CCD astronomical cameras is that there is no need to use a computer to acquire the images - all data can be stored on the memory card in the camera, making field trips easier to organise. A laptop computer will need 240V power or reliable, long-lasting battery power (which can be supplied via a 12V car battery with an appropriate DC to DC converter). However, leaving the computer at home makes things much simpler all round.

Cooled CCD cameras

Cooled CCD cameras can be monochrome or one-shot-colour cameras. They are more sensitive than the average DSLR camera and more expensive because they are not produced in manufacturing runs of many thousands. Thermo-electric cooling is used to cool the CCD to 30o to 40o below the ambient temperature. Some cameras have regulated cooling, meaning that the CCD can be set at some predetermined level, eg. -15o C, resulting in a constant dark current and therefore the need for just one set of dark frames, whatever the ambient temperature. Some CCDs, most notably the Sony HAD chips, have very low dark current so that the use of dark frames is unnecessary and regulated cooling not so advantageous.

One-shot-colour CCD cameras reduce the imaging time in the field by about 30% and can reduce the amount of image processing required - only dark frames and flat fields are necessary - but possibly at the expense of resolution, as explained for DSLRs. Hence you should choose a one-shot-colour camera with small pixels. If you want to improve the image, it is possible to create a pseudo luminance image and separate the colour channels using Photoshop or a similar program, then use the processing tools to sharpen, shrink, remove gradients, etc, as you would for monochrome images. However, if you are going to always do this, you may as well use a monochrome camera in the first place and get slightly better results.

In order to take a colour image with a monochrome camera, you will need a set of red, green and blue filters mounted, usually, in a filter wheel. Separate images are taken through each filter in turn and these are combined in an image processing computer program to create the final colour image. Some cameras have internal filter wheels and others need a separate filter wheel. Image acquisition programs will control the filter wheel in a predetermined sequence of exposures, or else the filter wheel can be manually controlled.

Besides considering cost and monochrome or colour, you need to consider whether the camera will be used for scientific measurement such as variable star monitoring. This will dictate whether you get an anti-blooming gate (ABG) or a non-anti-blooming gate (NABG) camera. (Blooming is the name given to the spillage from a pixel into adjoining pixels when a bright star is imaged. In an ABG camera there is, in effect, an overflow outlet in the pixel to bleed off electrons if too many are produced by the incoming light photons.) An ABG camera will not give a linear response to brightness except at relatively low brightness levels, say up to 50% of the full pixel capacity.

Other considerations will include:

• Size of pixels - the smallest is about 6.4 microns square. As a general rule, small pixels should be teamed up with short focal length telescopes and vice versa. Many cameras have pixel sizes of between 6.45 and 9 microns and these sizes suit telescopes with focal lengths from around 750mm to 3000mm or more. For my 1040mm focal length refractor, I have found the 6.45-micron pixels in my SXV-H9 camera to be ideal and the 9-micron pixels in my STL 11000 camera adequate for wide-field images.
• Size of detector - the larger the better as images can always be cropped. However, there are a couple of downsides to large detectors. First, the field of view can really show up the limitations of the telescope optics with oval stars away from the centre of the field of view; and second, a large detector creates large data files that are slow to download from the camera, and large files make image processing much slower. With large CCDs, field flatteners are a necessity and they are not cheap.
• Type and quality of the detector. Most manufacturers use CCDs rather than CMOS chips and they come in a range of qualities - the more defective pixels there are, the cheaper the chip. In general, you get what is on offer and you pay for what you get.
• Self guiding or externally guided. SBIG has a patented self guiding system that uses a second small CCD mounted to the side of the main CCD to take an image of stars that can be used for guiding. QSI Imaging also offers a self-guiding series of cameras but with the guider separate from the camera body and a small pick-up mirror in front of the internal filter wheel to direct light to the guider. Using the same telescope to both image and guide avoids any problems from flexure, mirror shift, etc. Starlight Xpress also offer a self guiding system, but it uses the main imaging chip by, in effect, stealing some of the exposure time to image a guide star, so that exposures are necessarily increased. Other cameras need either a separate guide scope or an off-axis guider. I have used all three methods and would not recommend an off-axis guider. A separate, rigidly mounted guide scope is the easiest to use, but with some preparatory work in a planetarium program to compose an image and locate a guide star, the SBIG and QSI self-guiding cameras are very effective.

In comparing CCD cameras, you will sometimes find that two cameras use the same CCD chip yet one costs 50% more than the other. Why, you might ask. The answer lies in the electronics circuitry that controls the dark current, the number of bits used to convert the signal to digital format, the type of cooling, the ability to upgrade components, etc. As previously stated, you get what you pay for. Nevertheless, the cheaper entry-level cameras are an ideal way to allow someone on a tight budget to get started in astrophotography. You can always later sell the cheap camera and step up to a more expensive model so long as time has not made the original camera obsolete.

Appendix 1 - Focal length to pixel size relationship

The above table lists the pixel size required for a given focal length of telescope based on 1.5 arc seconds per pixel for deep sky imaging (or 3 arc seconds seeing) and 0.5 arc seconds for solar system imaging.  Of course, these relationships are not hard and fast - they are at best a rough guide. With excellent seeing (say 2 arc seconds or less), excellent focus and guiding, smaller pixels will yield superior results for any given focal length, but such conditions are seldom attainable. The same arrangement with poorer conditions will not necessarily cause a poor result, rather the result will be no better than that for larger pixels.

Appendix 2 - Pixel size calculations

Full width half maximum (for perfect seeing conditions) = 1.02 x wavelength x focal ratio

The visible spectrum extends from about 350 nm to about 750 nm.  Because we want to cover the full range, our pixels need to be sized for the shorter wavelengths - say 400 nm.

Hence   FWHM = 1.02 x 400 x f  nm      = 408f nm       = 0.4f microns

For an f8 telescope, this gives a FWHM of 3.2 microns and a pixel size for optimum imaging of 1.6 microns. However, this is for perfect seeing, focus and guiding, which never exist.

Now image size = object angular size x telescope focal length / 206265   (mm)

= "seeing" x telescope focal length / 206.265   (microns)

For reasonable seeing (say FWHM = 3 arc sec), image size = 0.0146  x FL   (microns)

Hence optimum pixel size = 0.5 x 0.0146 = 0.0073 x FL

e.g. for a 1000 mm focal length, pixel size = 7.3 microns, as shown in the table above.

From this we can see that for rare, excellent seeing conditions, say 1.5 arc seconds, a 1000 mm telescope would need 3.6 micron pixels. Such a pixel size is not offered by camera manufacturers, so a longer focal length is needed to take full advantage of such conditions.