Stacks of superconducting tapes can trap much higher magnetic fields than conventional magnets. This makes them very promising for motors and generators.However, ripple magnetic fields in these machines present a cross-field component that demagnetizes the stacks. At present, there is no quantitative agreement between measurements and modeling of cross-field demagnetization, mainly due to the need of a 3D model that takes the end effects and real micron-thick superconducting layer into account. This article presents 3D modeling and measurements of cross-field demagnetization in stacks of up to 5 tapes and initial magnetization modeling of stacks of up to 15 tapes. 3D modeling of the cross-field demagnetization explicitly shows that the critical current density, J c , in the direction perpendicular to the tape surface does not play a role in cross-field demagnetization. When taking the measured anisotropic magnetic field dependence of J c into account, 3D calculations agree with measurements with less than 4 % deviation, while the error of 2D modeling is much higher. Then, our 3D numerical methods can realistically predict cross-field demaga This article has been published in Supercond. Sci. Technol. with netization. Due to the force-free configuration of part of the current density, J, in the stack, better agreement with experiments will probably require measuring the J c anisotropy for the whole solid angle range, including J parallel to the magnetic field.the probe is at least 10 mV/T. Cross-field demagnetization consists on the following three main steps: magnetization by field cooling (FC) method, relaxation time and cross-field demagnetization. The detailed process is the following:• The sample is placed into the electromagnet at room temperature.• The electromagnet is ramped up to 1 T.• The sample is cooled down in liquid nitrogen bath at 77 K.• The electromagnet is ramped down with ramp rate 10 mT/s. 400 405 410 415 420 425 B t /B 0 [-] t [s] cal 50 mT cal 100 mT cal 150 mT cal 50 mT n(B,θ) cal 100 mT n(B,θ) cal 150 mT n(B,θ) (b) FIG. 19: (a) The n(B, θ) measured data on a 4 mm wide SuperOx tape, measured in Bratislava by the set-up in [46]. (b) The comparison of calculation with constant n=30 and n(B, θ) dependence, both cases use J c (B, θ) dependence. Using n(B, θ) slightly reduces the demagnetization rate for a few number of cycles, but later on it is increased slightly.