PyHEADTAIL Meeting #22
Michael presented first results from the new multi bunch PyHEADTAIL modules:
- Michael presents objectives and outlines
- Several benchmarks were carried out comparing with semi-analytical models as well as to the original HEADTAIL code.
- Hyper-threading had to be disabled for the HPC cluster for PyHEADTAIL to run efficiently. Giovanni suggested to iterate this issue with IT to understand why hyper-threading is actually penalizing to the simulation time (ACTION). The sightly discrepancies between theory and codes could come from the fact that the different methods are not yet fully converged.
- Michael reminded that in general, convergence was a major issue rather often and needs to be checked carefully.
- Michael showed the different methods that can be used for multi-bunch PyHEADTAIL simulations. A very good agreement with HEADTAIL is obtained in general. A problem was found when using quadrupolar wakes in combination with the FFT optimization algorithms.
- Michael presented an important slide on the lessons learned: among others, fixed longitudinal cuts are important to be used. It was suggested whether this should not always be the case in PyHEADTAIL and to omit variable slicing (ACTION).
- There is a problem remaining in the computation of quadrupolar wakes with circular FFT. Nicolas suggested this to be due to the case correction implied for the computation of the wake kicks. It may be correct for dipolar wakes which needs to take into account the difference between source and witness particles. This could be different for quadrupolar wakes. This will be followed up offline. (ACTION)
- A very good agreement was achieved between bbb tune shifts. Some discrepancies are still present and could be due to the fact that the impedance model used is highly simplified.
- An interesting effect was observed for measurements in the horizontal plane which was not reproduced in simulations. One possible explanation for this observation could be e-cloud effects.
- The instability observed in the horizontal plane was tackled using an impedance model consisting resistive wall wakes along with a set of resonators that should resemble the different kickers. The first attempt of simulating these case gives rise to a few questions. A string coupled bunch mode with equal rise times for all bunches along the batch seems to evolve.
- One idea was to add an aperture to avoid both resonances as well as time resolve the occurring losses.
- Hannes suggested to put come quantitative comparison in place, checking simulations with actual measurements.
- Convergence study of slices is still an open issue.
There are minutes attached to this event.
Show them.