Purpose: Microbeam radiation therapy is a preclinical concept in radiation oncology. It spares normal tissue more effectively than conventional radiation therapy at equal tumor control. The radiation field consists of peak regions with doses of several hundred gray, whereas doses between the peaks (valleys) are below the tissue tolerance level. Widths and distances of the beams are in the submillimeter range for microbeam radiation therapy. A similar alternative concept with beam widths and distances in the millimeter range is presented by minibeam radiation therapy. Although both methods were developed at large synchrotron facilities, compact alternative sources have been proposed recently. Methods and Materials: A small-animal irradiator was fitted with a special 3-layered collimator that is used for preclinical research and produces microbeams of flexible width of up to 100 μm. Film dosimetry provided measurements of the dose distributions and was compared with Monte Carlo dose predictions. Moreover, the micronucleus assay in Chinese hamster CHO-K1 cells was used as a biological dosimeter. The focal spot size and beam emission angle of the x-ray tube were modified to optimize peak dose rate, peak-to-valley dose ratio (PVDR), beam shape, and field homogeneity. An equivalent collimator with slit widths of up to 500 μm produced minibeams and allowed for comparison of microbeam and minibeam field characteristics. Results: The setup achieved peak entrance dose rates of 8 Gy/min and PVDRs >30 for microbeams. Agreement between Monte Carlo simulations and film dosimetry is generally better for larger beam widths; qualitative measurements validated Monte Carlo predicted results. A smaller focal spot enhances PVDRs and reduces beam penumbras but substantially reduces the dose rate. A reduction of the beam emission angle improves the PVDR, beam penumbras, and dose rate without impairing field homogeneity. Minibeams showed similar field characteristics compared with microbeams at the same ratio of beam width and distance but had better agreement with simulations. Conclusion: The developed setup is already in use for in vitro experiments and soon for in vivo irradiations. Deviations between Monte Carlo simulations and film dosimetry are attributed to scattering at the collimator surface and manufacturing inaccuracies and are a matter of ongoing research.