Taken from Andy's e-mail explaining the simulations:
Hi Christoph, I've just taken a look (as we agreed at Monday's meeting) at the field integrals through the EPU in RADIA. I attach a plot of the results. This shows the vertical kick in urad experienced by a particle entering on the "design orbit" as a function of the shift parameter in units of the undulator period. The points show the values obtained by integrating the horizontal field along the magnetic axis of the undulator. Of course, one should integrate along the trajectory, but the amplitude is small enough that the difference is negligible. Since RADIA takes ages to integrate along the trajectory, but gives the integral along the straight line almost instantly, I cheated. The line shows the values obtained by actually calculating the trajectory, and looking for the value of z' (in RADIA co-ordinate system!) at the exit (i.e. far enough out of the undulator that z' isn't changing anymore). The agreement is pretty good, so my assumption that I can replace the real trajectory by a straight line seems to be reasonable. The non-zero values that I find are a consequence of the permeability varying from 1. The vertical field integral seems always to be in the noise. I also checked the amplitude of the orbit with zero shift parameter, and it agrees well with the analytical result, so I must be doing something right with RADIA, even if I don't have it fully worked out yet. It will be interesting to see how these results compare to the feed-forward tables that are actually used... Andy.
P.S.: Oleg Choubar later found that the Radia model was not sliced fine enough. Therefore the field integrals came out small in these simulations. Oleg's results will be posted soon.