- Abstract The relative calibration of pyranometers in the field (as opposed to laboratory calibration) is discussed. Measurements performed both for horizontal instruments and for those at normal incidence are analysed. It is found that the normal incidence geometry results in relative calibration constants with a precision of a few parts per mille (for instruments with tilt-independent outputs). This is considerably more accurate than results obtained using the horizontal geometry. The reason for this difference is discussed. It is pointed out that a combination of calibration measurements performed successively at both geometries can serve to identify common instrument deficiencies such nonlinearity, poor cosine response, and built-in leveling errors.