The simulations were performed using LOCO for Matlab (James Safranek, Greg Portmann, Christoph Steier), which is programmed on top of the accelerator toolbox AT (Andrei Terebilo).
The orbit response matrizes were measured for five different conditions at 1.9 GeV:
The plot shows cases 2-5 with the results from case 1 (the baseline) subtracted. The vertical scale is the k-value in m^(-2), i.e. the gradient divided by B*rho. The blue bars indicate the results for EPU4, the red ones the results for EPU11.
Conclusion: The fit results for the skew quadrupole strength are very consistent with the measured variation of the vertical beamsize. The resolution of the fit is good enough, that it should be possible to localize which part of the EPU (ends, bulk, ...) are causing the effect. The measurements for both devices indicate that there is also a gap dependent skew term, which is stronger in sector 4, whereas the phase dependent skew term is stronger in sector 11.
We later measured orbit response matrizes for the sector 4 EPU for many additional shift parameters (every 5 mm). In addition, we modified the LOCO fit routine, as to split the EPU in 3 sections (first period, core and last period). The following plots show the results.