Hyperspectral imaging with a spatial resolution of a few mm2 has proved to have a great potential within crop and weed classification and also within nutrient diagnostics. A commonly used hyperspectral imaging system is based on the Prism-Grating-Prism(PGP) principles produced by Specim Ltd. Finland. One of the novel systems based on the PGP spectrograph (VTTVIS) was build by The Department of Agricultural Sciences, AgroTechnology, KVL, Denmark, in 1995. Several other agricultural institutions havenow implemented the technology in their research. None of these has published any thoroughly work describing the basic principles, potential error sources, and/or adjustment and calibration procedures. This report fulfils the need for such documentationwith special focus on the system at KVL. The PGP based system has several severe error sources, which should be removed prior any analysis. Most of the random noise sources can be minimised by carefully selecting high-grade components especially withconcern to the camera. Systematic error sources like CCD fixed pattern noise (FPN), CCD photoresponse nonuniformity (PRNU), CCD charge transfer efficiency (CTE), slit width variations, changes in off-axis transmission efficiencies, diffractionefficiencies, and image distortion have a significant impact on the instrument performance. Procedures removing or minimising these systematic error sources are developed and described for the system build at KVL but can be generalised to other PGP basedsystems.