The daylighting softwares RADIANCE, SUPERLITE, GENELUX, ADELINE and LESO-DIAL have been used in an international task of comparison of results of simulations as well as comparisons with data obtained from measurements in scale models located in artificial skies. Three major building configurations have been tested, and surface reflectances have been varied to test the sensitivity of the results to surface reflectance. The simulation of the specularity of the surfaces has also been included. Results show that the calculation of illuminances inside buildings due to daylight require a careful description of the luminances of the sky section seen through the window. Direct illuminances inside buildings can be computed with an accuracy of about 5%. Then the accuracy of the calculations of the multiple reflexions depends on the assumptions on the optical caracteristics of the surfaces. When diffusing finishes were used in the scale models, we have found that the assumption that they are perfectly diffusing is satisfactory (error below 10%) as long as the point of comparison is not too far from the window. For locations far away from window, the assumption of perfectly diffusing surfaces tend to lead to an underestimation of the illuminances compared to reality. This is due to the fact that most diffusing surface are not "perfectly diffusing" and tend to reflect more light in the direction opposite to the light source. Finally, it appears that the programmes should be used with a lot of care, since lighting calculations are very sensitive to the quality of the input file : light source description, material photometry, building geometry and simulation parameters.
展开▼