The pixel size of your CCD should be matched properly with your telescope's effective focal length. There are many considerations that should be accounted for in this coupling of scope and camera. One significant trade off is speed versus resolution.
When imaging a bright object like a planet or the moon you are less concerned about the photographic speed (focal ratio) of the camera than the resolution needed to capture fine planetary detail. In this case resolution should be maximized since the planets/Moon are bright enough that speed is not a significant factor. Nyquist sampling theory states that when transferring spatial information from one medium to another, two samples must be acquired of each datum in order to capture all the information present. In telescope resolution terms, this means that the resolution of your CCD's pixels should be twice as precise as the resolution of your scope. For example a 5-inch scope has a diffraction-limited resolution of about one arcsecond. To capture all the information present in the planetary image, the CCD's pixel's FOV should span no more than 0.5 arcseconds. This assumes that seeing conditions are not a limiting factor. At most imaging locations used by amateurs, seeing conditions limit sky resolution to considerably more than 1 arcsecond for CCD integrations longer than a second or so; but the good news is that the very short integrations used for bright planets and the Moon, usually less than 0.5 second, often "freeze" the seeing so that it approaches the theoretical resolution of the optics.
In deep sky imaging you are concerned more with the speed of the system. Additionally the tracking accuracy of your mount may be a consideration in determining appropriate resolution. While there is no absolute answer to how many arcseconds a pixel should span for deep sky imaging, a good starting place is about 2 arcseconds per pixel for moderate size amateur scopes. This is a compromise of many factors and can be deviated from significantly with good results.
Approaching this from another perspective, you could first determine what field of view is desired and then determine the focal length required. In an SCT telescope one can easily alter the focal ratio to obtain a desired focal length using various telecompressors that are available. In this way the FOV of a single telescope can be adjusted to accommodate deep sky objects of various sizes. Specific issues are as follows:
a. Focal length and f-ratio. The effective focal length of the optical system should be matched to the width of the CCD pixels so that the sky resolution of each pixel (as measured in arcseconds) is at least twice as precise as the FWHM size of the PSF delivered by the optical system under expected atmospheric seeing conditions and exposure times. Most amateur instruments and observing sites would be doing quite well to achieve an FWHM of 3 arcseconds in long-exposure images, so a pixel resolution of 1.5 arcseconds per pixel is usually more than satisfactory. As stated earlier in this section, 2 arcseconds per pixel is often considered a decent compromise for most amateur setups. If the imager can manage it, a system designed with a modest degree of oversampling (e.g., one arcsecond per pixel) is ready for those nights of exceptional seeing and provides more useful data for some image processing algorithms, such as deconvolutions. At the same time, many beautiful and scientifically valuable images have been made by considerably undersampled systems, so imagers should not feel unduly constrained by the "rules" of system design and sampling.
The focal ratio should be kept as fast as possible, given the constraints of required focal length, so that integration times can be minimized. Just as in film work, the speed of the optical system is important to achieving high-SNR images in reasonable exposure times and in minimizing tracking and guiding requirements. Coupling this issue with CCD resolution and sampling requirements may make it seem as though large, fast optical systems (e.g., Richey-Chretiens and Newtonians) with heavy-duty (and expensive) tracking mounts are needed for CCD work, but the many beautiful CCD images made with modern, small, fast refractors and telecompressed catadioptric instruments show that this is not at all the case. Imagers planning their systems will find that CCD cameras with small pixels used in single-binned mode will deliver adequate sampling with these short focal lengths and will often provide excellent SNR in images only a few minutes long.
b. Vignetting (see section 1.c. on vignetting)
c. Stability and balance. Adding a CCD camera to a telescope often creates a weight imbalance in the tube assembly configuration with the mount. Different types of mounts (e.g., German equatorial, fork, altitude/azimuth, etc.) have their own special weight distribution characteristics, but the common requirement is to assure that weight is distributed evenly around all axes of rotation so that slippage or binding does not ruin the excellent tracking required for good images.
d. Tracking requirements and methods. In order to yield no noticeable tracking error in a CCD image, the centers of all PSFs (such as stars) should stay within the same pixels during the entire integration time (total exposure duration). Large image scales from short focal length optical systems reduce the demands of tracking by assuring that pixels are large in relation to the optical PSF, but an inaccurate tracking system will still yield poor results. Poor polar alignment, periodic error in gears, incorrect tracking rate, field rotation, optical shifting, wind, and mount instability are the enemies of good images. Careful, accurate work in polar alignment, drive system design, construction and maintenance, and optical support construction can minimize many problems and allow accurate short integrations (perhaps up to a minute or so) to be made without guiding; but a manual or automatic guiding system is the real answer for long, accurate integrations. More on this in the next section.
Relatively lightweight structures can make fine imaging systems, but to keep wind and potential instability from ruining images, the careful imager will work in still and/or protected environments, refrain from creating ground vibrations, and will assure good weight balance around all rotational axes.
e. Field Rotation. Field rotation is the slow rotation of the FOV around a point within or adjacent to the FOV. It can cause some stars in the FOV to seem perfectly tracked while others may appear elongated in a slight arc. In equatorially driven systems it can result from inaccurate polar alignment, but it is inherent to altitude-azimuth (alt-az or az-el) drive systems. Alt-az drives must incorporate an image plane derotation mechanism to prevent field rotation if they are to be successfully used for accurate integrations longer than a minute or so (depending on where the optical system is pointed on the celestial sphere).
f. Unguided vs autoguiding.
One of the most demanding applications for a telescope mount is imaging with a CCD camera. With the small pixels and high resolution of CCD cameras, few commercial mounts are made with enough precision to accomplish accurate unguided tracking. Even the smallest tracking errors can ruin an otherwise great image. Many imagers choose to augment their mount with a guiding mechanism. This usually takes the form of an additional guide scope or an off-axis guider. Guiding off axis is the preferred method for accuracy since the same optical path is used by both the guider and the CCD camera.
Although off-axis guiding is more accurate, it has the limitation of requiring a guide star close to the center of the same FOV as that being imaged by the CCD. Sometimes, this can be a negative factor because a suitably placed guide star of sufficient brightness may not be available. However, with the advent of more sensitive CCDs this limitation is not always significant. Many of the newer top-of-the-line CCDs have a separate guide chip in the CCD's camera head which is dedicated to guiding the image (SBIG), while other methods include an off-axis camera separable from the imaging camera (Apogee) and the use of an interline transfer CCD, which allows simultaneous imaging and guiding by the same chip (Starlight Express). See http://sbig.com , http://www.apogee-ccd.com , and http://www.starlight-xpress.co.uk for online catalogs of self-guiding cameras and other products from SBIG (Santa Barbara Instruments Group), Apogee Instruments, and Starlight Express, leaders in CCD instrumentation for amateurs.
Another method that is commonly employed is a separate guide scope that is used to track the image. The SBIG ST4 CCD, the home-built Quickcam autoguider, the Cookbook 211/245, and other cameras are commonly used as autoguiding cameras in piggyback guide scopes. The separate guide scope has a limitation in that differential flexure may occur between the optical path of the guide scope and that of the imaging scope. This can result from mechanical displacement due to slippage or bending, but can also be due to optical shifting. In SCT telescopes the primary mirrors are moveable for focusing and the flexible mirror support mechanism can allow the main mirror to shift slightly during a long exposure. This will have the effect of introducing significant guiding errors and can ruin an image.
The focal length of the guide scope should be matched with the resolution of the imaging scope. Too much focal length in the guide scope can result in the guide scope tracking on atmospheric scintillation rather than correcting true drive errors and can reduce overall tracking accuracy, while too little focal length will not allow tracking to the precision required by the resolution of the imaging pixels. Since most autoguiding cameras will guide to subpixel precision on the autoguiding chip, the guide scope's focal length does not need to be as long as the imaging scope's. For example, the ST4 is capable of tracking to an accuracy of 0.2 pixels on the ST4 CCD chip. The focal length of the guide scope should be long enough so that the tracking resolution in the guide scope is about twice as precise as the resolution of the imaging CCD. This will allow for sufficient tracking accuracy without being affected too adversely by atmospheric scintillation.