Abstract: A calibration step is carried out on a calibration tube in which magnetizations which cause beam displacements in x and y directions are determined for all the magnet poles and beams, and the adjustments of sensitivity of each beam from the x/y values is determined. A beam deviation of a production tube is then adjusted by determining its initial error in x and y directions and then starting a calculation sequence in which the magnetization currents are calculated by linear superposition of the individual currents on the basis of the known adjustment sensitivities. An additional rotating and decreasing auxiliary field is active during the magnetization of the poles in the rings of the calibration tube and of the production tube. It is also possible to correct twist errors using the invention.