When working with moderate bandwidth radiometers in regions of the solar spectrum with large variations in intensity as a function of wavelength, significant errors can be introduced from the difference between the lamp spectrum and the solar spectrum. Under these conditions, a `perfect' radiometer can grossly over- or under-estimate the solar irradiance, while several otherwise identical radiometers with center bandwidths ranging over +/- 2 nm can report significantly different irradiances. A direct instrument intercomparison experiment was conducted in San Diego, Calif. using 15 five-channel filter radiometers. The instruments were operated for approximately one week alongside a 0.7 nm bandwidth scanning spectroradiometer under a variety of solar conditions. All of these instruments had been calibrated with lamp standards traceable to NIST. In an effort to reduce the uncertainties of transferring calibration made with standard lamps to filter-based radiometers, we suggested that `source-specific' calibration constants, referenced to both lamps and reference solar spectra, be used. This investigation explores how well this procedure can work, and contrasts this result with intercomparison- and regression-based calibrations.© (1994) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.