The sampling represents the area of the sky seen by a pixel of the CCD.
The sampling S (in arc second per pixel) depends on the size P of the pixel (in microns) and the focal lenght F (in mm):
S = 206 P/F
Example : at a focal lenght of 2000 mm, the sampling on a KAF-0400 chip (9 microns pixels ) is S = 206x9/2000 = 0.93"/pixel.
The same formula can be used to determine the focal lenght necessary to reach a given sampling:
F = 206 P/S
In high resolution, theory and practice tell that a good basic value of sampling, in favourable conditions, is about twice the maximum spatial frequency (see What is a MTF curve ?). It is the Nyquist sampling. It depends on the diameter D of the telescope and the wavelenght l, its value is (in radian per pixel):
SN = l/2D
If P is the pixel size (in the same unity as l), the focal ratio corresponding to this sampling is:
RN = 2P/l
Example : for a 250 mm telescope, whose maximum spatial frequency is 2 lines pairs per arc second at 0.6 Ám, SN is 0.0000012 radians per pixel, ie 0.25 arc second per pixel. For 9 microns pixels (KAF-0400), the corresponding focal ratio is about 30.
A higher sampling could bring a little gain, at the expense of a reduction of the field and an increase of the exposure time which can be harmful if seeing is not very good. Only the experience of the amateur can tell him which value of sampling is better, depending on its hardware and the circumstances.
In CCD imaging, as in photography, adjusting the focal lenght is therefore of the highest importance to work at an adequate sampling. For more details, see How to adjust the focal lenght ?