Observations: around 5% of main-sequence A stars have a strong, steady, large-scale magnetic field. Possible explanations: Core-dynamo or fossil-field? Problem: fossil-field model requires stable field, but there is no known stable field. Solution: use numerical magnetohydrodynamics to follow evolution of arbitrary initial field into a stable configuration. How magnetic fields are measured Circular polarisation in wings of spectral lines (via Zeeman effect) yields average line-of-sight component of magnetic field, B s . If field > 2 kG, and line splitting exceeds thermal and rotational broadening, modulus of field can be obtained, B mod . These (and two or three related quantities) are measured as the star rotates. A model is made of the magnetic field, (e.g. dipole + quadrupole) and the free parameters are found by fitting to the observations.