Abstract. The properties and performance of charge-coupled device (CCD) array spectroradiometers for the measurement of atmospheric spectral actinic flux densities (280-650 nm) and photolysis frequencies were investigated. These instruments are widely used in atmospheric research and are suitable for aircraft applications because of high time resolutions and high sensitivities in the UV range. The laboratory characterization included instrument-specific properties like the wavelength accuracy, dark signal, dark noise and signal-tonoise ratio (SNR). Spectral sensitivities were derived from measurements with spectral irradiance standards. The calibration procedure is described in detail, and a straightforward method to minimize the influence of stray light on spectral sensitivities is introduced. From instrument dark noise, minimum detection limits ≈ 1 × 10 10 cm −2 s −1 nm −1 were derived for spectral actinic flux densities at wavelengths around 300 nm (1 s integration time). As a prerequisite for the determination of stray light under field conditions, atmospheric cutoff wavelengths were defined using radiative transfer calculations as a function of the solar zenith angle (SZA) and total ozone column (TOC). The recommended analysis of field data relies on these cutoff wavelengths and is also described in detail taking data from a research flight on HALO (High Altitude and Long Range Research Aircraft) as an example. An evaluation of field data was performed by ground-based comparisons with a double-monochromatorbased, highly sensitive reference spectroradiometer. Spectral actinic flux densities were compared as well as photolysis frequencies j (NO 2 ) and j (O 1 D), representing UV-A and UV-B ranges, respectively. The spectra expectedly revealed increased daytime levels of stray-light-induced signals and noise below atmospheric cutoff wavelengths. The influence of instrument noise and stray-light-induced noise was found to be insignificant for j (NO 2 ) and rather limited for j (O 1 D), resulting in estimated detection limits of 5 × 10 −7 and 1 × 10 −7 s −1 , respectively, derived from nighttime measurements on the ground (0.3 s integration time, 10 s averages). For j (O 1 D) the detection limit could be further reduced by setting spectral actinic flux densities to zero below atmospheric cutoff wavelengths. The accuracies of photolysis frequencies were determined from linear regressions with data from the double-monochromator reference instrument. The agreement was typically within ±5 %. Because optical-receiver aspects are not specific for the CCD spectroradiometers, they were widely excluded in this work and will be treated in a separate paper, in particular with regard to airborne applications.