This report documents tests of a method for object-free calibration of radar cross section (RCS). The advantages of this method include that errors in the alignment and in the theoretical RCS estimates of calibration targets are eliminated, the influence from multipath propagation, cross-talk and background can be reduced, the signal levels can more easily be optimised and the measurement procedure can be simplified. The method is particularly suited for the calibration of bistatic measurements and has been developed for use in ongoing studies of the bistatic scattering properties of ground surfaces. Methods for the reduction and correction of drift errors have also been tested. The results confirm the applicability of the methods and provide information on what accuracy level that can be expected. During the tests the measured and theoretically expected RCS values agreed with differences up to slightly above 1 dB. Conditions affecting the accuracy are analysed. Among the significant sources of measurement errors are probably alignment errors for antennas and reference targets. The optical methods used for alignment should certainly be sufficient but could need improvements. It is recommended that the stability of the system is always monitored by the collection of a reference signal provided through a shunt path. This is particularly important when equipment such as the high-frequency fibre-optic links we use e.g. for bistatic measurements since they exposed and sensitive to thermal drift. The frequency response of the reference targets also differ from that expected. This could be due to alignment error, but the approximations of the analytical expressions may play a role here too. Some of the differences between measured and expected values could also be due to approximate analytical RCS expressions for the reference objects. The methods could improve ordinary monostatic RCS measurements too, but some parts may require further developments before application. Further improvements in the measurement procedure and the RCS calculations for reference objects would be expected to reduce the difference between measured and calculated values