Estimation of the electrical yield of a PV module is expected to be a more useful predictor of performance for installers than W, alone. A method for the energy rating of PV modules based on performance surfaces under development at the ESTI laboratory uses the module temperature and incident irradiance as independent variables and has been successful in prediction of real energy production for crystalline Si modules. However, it was found to be more difficult to accurately predict the performance of thin film modules and it was therefore necessary to explore the reasons. One potentially significant parameter not included in the standard performance surface is the effect of spectral variations, and this has been studied during indoor and outdoor testing on CIS and a-Si modules. The outdoor measurements were performed on a tracker so as to preclude angle of incidence effects. Module I-V curves and the solar spectrum were measured at frequent intervals over a range of air mass values during the course of a number of days. A crystalline Si reference device and a pyranometer were used as irradiance sensors in order to explore the effect of the choice of reference device used. The spectral mismatch factor is calculated from measurements of the solar spectrum and device spectral responses and is applied to correct the individual module measurement points. The dependence on air mass, i.e., the details of the solar spectrum of these devices, has also been shown, so employing only total irradiance and device temperature may not be sufficient when an energy rating is being made. This effect is most pronounced for the a-Si module tested, for which a significant part of this dependence was corrected by the application of the relevant mismatch factors. (c) 2005 Elsevier B.V. All rights reserved.