Experience with CST Eigenmode Solver for the LHCb Velo Upgrade Project N.Biancacci, B.Salvant Impedance Meeting 12 May 2017

Context LHCb detector Velo

Context LHCb – Ve(rtex)lo(cator) LHCb detector RF box Velo Picture from Paula Collins, 2nd RF foil checkpoint Meeting, March 2017

New pixel sensor modules Context The LHCb Velo detector will be upgraded for HL-LHC in order to increase its resolution. New pixel sensor modules Pictures from Paula Collins, 2nd RF foil checkpoint Meeting, March 2017

New pixel sensor modules Context The LHCb Velo detector will be upgraded for HL-LHC in order to increase its resolution. Change in RF box shape is needed to host the new Velo pixel detectors. New pixel sensor modules Pictures from Paula Collins, 2nd RF foil checkpoint Meeting, March 2017

Context The LHCb Velo detector will be upgraded for HL-LHC in order to increase its resolution. Change in RF box shape is needed to host the new Velo pixel detectors. RF box embedded in a large empty volume: presence of resonant modes to be studied. Problem addressed with the help of CST Eigenmode Solver

First simulations Model simplified for EM simulations by Benoit. Resolution of the RF box edges and convolutions looks accurate with 97k tetrahedrons

First simulations Eigenmode Solver Wakefield Solver Eigenmode field post-processing allows for automatic calculation of shunt impedance. Shunt impedances in the order of 10-100kOhm: worrying for longitudinal stability. Not seen as high in the Wakefield Solver -> hint of potential issue!

Field cross-check The field is auto-sampled automatically during the integration. The appearance looks spiky and abruptly cut

Field cross-check Auto -sampled 1mm-sampled Increasing the sampling to 1mm step size ameliorates the situation. The rapid change of field sign is correlated with the RF box convolution periodicity.

Field cross-check 1mm-sampled Increasing the sampling to 1mm step size ameliorates the situation. The rapid change of field sign is correlated with the RF box convolution periodicity. Shunt impedance decrease already of 3 orders of magnitude.

Mesh improvement Higher sampling on the convolution areas can improve the field quality. The CST-Eigenmode Solver convergence tests are done in frequency which is a global parameter (i.e. depends on the full geometry). A change in a small detail does not affect too much the result. The field (i.e. the eigenvector) is a function of the 3D space and may have strong values in narrow regions -> dense mesh needed to describe these details. Achieved with local higher meshes in sub-volume -> 200k mesh

Post processing Auto sampled 1mm sampled Sampling with 1mm, the filed variation looks much smoother

Shunt impedance Shunt impedance Impedance decreased of other 2 orders of magnitude. Now closer to Wakefield Solver and negligible impact on longitudinal impedance budget. Still the question remains on how to converge automatically for this kind of structures.

Conclusions and outlook CST Eigenmode solver ensures convergence test on the resonant frequency. Convergence in the field calculation is not automatically ensured. The shunt impedance post processing module may lead to wrong results if the automatic sampling is too coarse with respect to the field variation. In the case of LHCb Velo, increase in mesh cells close to the RF box and finer field sampling improved the precision of field calculation. Is there a way to loop a CST simulation over the accuracy in field calculation? Do we have similar structures in which this kind of problem may arise?