The SMARTS spectral model can advantageously be used to predict clear-sky irradiance spectra on surfaces of any tilt and orientation, e.g., for the simulation of spectrally selective technologies like photovoltaic devices or coated fenestration systems. To evaluate the intrinsic accuracy of the model, its current version undergoes here a three-step validation exercise, involving reference radiative transfer codes, and two series of sophisticated spectral and ancillary measurements performed at different locations. Provided that the most important inputs are known with sufficient accuracy, it is concluded that the model performance is very high, with typical differences of 1-2% when compared to reference models, and uncertainties largely within the overall experimental error (approximate to 5%) when compared to spectroradiometric measurements. The effect of imprecise spectral reflectance data of the foreground on the diffuse irradiance incident on shaded vertical surfaces is discussed. (C) 2007 Elsevier Ltd. All rights reserved.