As anyone with an active interest in astronomy knows, CCD cameras seem to be all the rage these days, and many of these devices are advertised as being easy to use and virtually "plug in and play" systems. Having used one such camera for over a year now, I want to present a realistic and practical picture of some of problems associated with CCD imaging.
Most readers will have seen CCD images published in magazines, in particular the planetary images taken by Don Parker, and been amazed at the quality. However, I would suggest that the quality of Parker's images say more about the quality (and aperture) of his telescope optics and tracking accuracy than the camera and the technique of using it. Producing high quality CCD images provides a severe test of the telescope mount, drive system, polar alignment and optics which should not be underestimated.
In particular, I believe that the tracking accuracy of the telescope (which includes alignment of course) is the most important factor, and since most of us probably have portable set-ups, this is the area which is least likely to be up to scratch. When using the Pictor 216 in focal plane imaging in an f 10 telescope, one pixel in the CCD array corresponds to about one arc second of sky. Thus, to get pin- point (faint) star images covering no more than say four pixels, the telescope must track to about two arc second accuracy over the duration of the exposure, which may be anything up to five minutes. Any deviation from this degree of accuracy degrades the image, and at the least it looks out of focus. Such accuracy really requires very careful polar alignment coupled with an accurate drive system fitted with a drive corrector.
The camera image itself is of course acquired by computer. To organise a night's imaging, I need to set up the telescope on the mount, provide power to it and the camera, set up the computer outside (with power and shelter) and plug the camera into the computer. In addition, it is necessary to take flat-field calibration images each night, so a portable light-box which attaches to the telescope (which you need to fabricate yourself) is also needed, with power. Setting all of this up in the dark can be a serious deterrent unless a whole night's viewing is in prospect. As you can see, "plug in and play" is something of a misnomer.
The computer itself is worthy of comment. Having acquired the image, it must first be calibrated and then processed to get good quality. This requires some serious computing power, both in hardware and software. The bottom line is to avoid endless waiting time and frustration, a fast Pentium with plenty RAM and a powerful image processing program like Photoshop is virtually essential. Certainly, the processing package that is supplied with the Pictor is quite inadequate.
You may be thinking by now that this all sounds pretty daunting, but all the comments made regarding the telescope are just as relevant to conventional photography, and indeed the constraints on your set up apply equally to photography and imaging. If you can't take clean and sharp photographs through your telescope, then you won't be able to take sharp CCD images either. The only possible exceptions to this are the moon and brighter planets, where CCD exposure times are very short. I can image Jupiter in an 8" SCT at f25 in about 40 milliseconds, and the exposure times for the moon are so short that a neutral density filter is really required to avoid over-exposure, even at the shortest exposure the camera can take of 4 milliseconds. Nevertheless, a high degree of tracking accuracy is still very desirable so that successive images still have the object in the field of view. At f25, the field in the Pictor 216 is only about 3 arc minutes wide, and any significant drift means constantly hand correcting the object back onto the centre of the CCD chip.
Let's suppose that you have ironed these problems out - what's actually involved in obtaining a good quality CCD image on your computer monitor?
You begin by instructing the camera to take an image via the camera software, and when the exposure time is completed, a raw image appears on the screen. This image is actually composed of three elements. The target object itself generates an image because light from the object striking the pixels in the CCD array generates an electric charge in each pixel which is proportional to the amount of light striking it. Charge also builds in the pixels from two other sources, and can be considered contamination of the image. These sources are the thermal image and the bias image. The thermal image is produced by the spontaneous generation of electrons in each pixel by heat, and the intensity of the thermal image is directly proportional to the exposure time and the temperature of the array. CCD cameras are generally cooled to at least 20 C below ambient to keep the thermal image to a minimum. The bias component is caused by electronic noise from the camera electronics, and can be considered to be an exposure of zero time duration. These two components are often dealt with together and are then called dark current. The dark current image is therefore an image taken with the same exposure time as the raw image, but with the telescope optics covered so that no light reaches the CCD chip. There is one further source of error in the raw image. Each pixel has a different sensitivity and this needs to be evened out, particularly with high resolution planetary work. This is known as flat- fielding, and requires you to take an image of a uniformly lit field. The best way to do this is to construct a light-box which illuminates a piece of milk plastic and fits over the end of the telescope. The flat-field image also cancels out the effects of vignetting and dust spots which may be registered during the raw image exposure, but in order to do this, a new sequence of flat-field frames must be taken each time the camera orientation is changed with respect to the telescope optics.
I usually begin a night's imaging by choosing an object to image, and concentrating on obtaining a series of images of that object. Using a high power eyepiece, I visually centre the object in the field of view of the telescope, and begin tracking it. The eyepiece is then replaced with the camera, and a number of trial exposures and adjustments made until the object is centered on the CCD chip, and accurately focussed. Once this is achieved, I attach the light-box to the telescope and take 3 flatfield frames followed by a dark frame and store these images on the hard disc. Having checked that the object is still centered, I then take alternate images of the object and dark current (generally 9 of each) and store them. Given that the raw images of the object look acceptable, I will then try to locate the next object without moving the camera, so that the flat field frames can be used to process the new image as well. It is advisable to take a further 3 flat-fields in the middle of the session, and 3 more at the end, making 9 in all.
Having completed a night's imaging, and having dismantled all the gear, I have saved 9 frames each of the object, the dark current and the flat-field on hard disc. The next step is to calibrate the image. Image calibration is done by the camera software, and basically involves subtracting the dark current frame from the raw image, and dividing the result by the flat-field. However, if only one support frame is used, it tends to introduce random noise into the image, so it is much better to average a number of support frames to create a master dark and a master flat field and then calibrate the raw image with them. Random noise is reduced as a function of the square root of the number of frames averaged, so by making a master dark frame from 9 individual frames, noise is reduced 3-fold. So, I make a master dark and subtract it from the raw image, and divide the result by the master flat field. The resulting calibrated image is saved and is now ready for processing.
Image processing includes sharpening, unsharp masking, contrast enhancement and high/low pass filtering and is best performed by a program like Photoshop. The details of processing are beyond the scope of this article and are obviously dependent on the software used. However, processing is a very powerful tool, and enables images to be greatly enhanced. It is possible to obtain images fainter than sky background, and to image stars in an 8" telescope down to magnitude 20 within 5 minutes.
I began this article by stating that CCD imaging is not a plug-in-and-play system, and no doubt you will have understood what I mean by now. However, it is also true to say that finding and observing visually a 3 arc-minute magnitude 13 galaxy, or obtaining high resolution photographs of Saturn do not fall into that category either. Of all the pros and cons of CCD imaging, there is one big plus for me: on the wind-swept, wet nights following an imaging session, I'm still doing astronomy with lots of image calibration and processing to do.