Approximations to total and diffuse horizontal and direct normal, broadband solar irradiance (280-4000 nm) can be obtained from the multi-filter rotating shadowband radiometer (MFRSR) using the unfiltered silicon channel of this seven-channel instrument. However, the unfiltered silicon channel only responds to wavelengths between 300 and I 100 nm and does not have a uniform spectral response. In contrast, the best, more expensive, first-class, thermopile-based radiometers respond fairly uniformly to all solar wavelengths. While the total horizontal and direct normal solar irradiance measurements made with the MFRSR unfiltered silicon channel are reasonable if carefully calibrated with a thermopile radiometer, the diffuse horizontal irradiance calibrated in this way has a large bias. These issues are common to all inexpensive, silicon-cell, solar pyranometers. In this paper we use a multivariate, linear regression technique for approximating the thermopile-measured total, diffuse, and direct broadband solar irradiances using the six, narrowband filters and the open-channel of an MFRSR. The calibration of the MFRSR for broadband solar by comparing various combinations of MFRSR channels to first-class thermopile instruments is illustrated, and methods to track the instrument response during field deployments are investigated. We also suggest an approach to calibrate the open-channel for all three components that could improve measurements that are made using typical, commercial, silicon-cell pyranometers. Published by Elsevier Ltd.