Participants: S. Albright (SA), F. Antoniou (FA), H. Bartosik (HB), J.P. Burnet (JP), L. De Mallac (LD), G.P. Di Giovanni (GP), J.L. Gomez Costa (JL), V. Forte (VF), Q. King (QK), K. Motala (KM), R. Murillo (RM)
Discussion:
RM: is the LN4 pulse length fixed? No, the length could change from 1-150 us (i.e. 1-150 turns) per ring.
JP: Where is the painting done? The transverse is done at injection (BSW-KSW). The longitudinal is done before with the Linac4 (last 2 PIMS modules + Debuncher)
QK: Is the chopper the same equipment that is responsible for the 1.5 us chopping between rings? Yes, the chopper must be able to provide a finer chopping on top of the larger ring separation of 1.5 us
RM: After 'End of Injection' how fast is the KSW ramp down to ~-9 mm? It will take ~10 us. And it will bring the LHC type beam completely away from SF, but not completely for ISOLDE type beam.
QK: Concerning the Bdl could the Linac4 ramp their energy for the different rings? Unfortunately not that fast and not that far (SA).
QK: How much time is needed until one could get the interlock notification? It should be possible at the beginning the cycle (BM). (CB) For the KSW we need to know the function even the previous cycle.
It is clear that BE-OP has two major requests:
QK reminded that what we incorrectly label 100 us jitter between the different FGCs; it is in reality a sampling of the FGC at 10 kHz, i.e. 100 us sampling. The situation will remain unchanged until 2022 (sampling at 10 kHz). So this issue cannot be solved now, but only in the future. TE-EPC is working on a new version of the FGC3.2 which will support sampling and regulation at more than 10 kHz. The target is 100 kHz, but this may not be achieved with the processor currently under evaluation. The higher sampling rate will reduce the regulation error during fast reference changes.
Concerning the issue with the ring interlock and ring pulse length, RM proposed to finally add an LTIM which will tell when the function has to start.
TE-EPC will need 4 Front-End computers, 1 per ring. Is there space in the rack? A meeting with the timing support is needed. Another modification would be to change the protocol of the gateway, which is currently at 1 ms, to be at 1 us. Work to be done by TE-EPC (RM+QK). QK said that the information about the start of the function will have to be received by at least 40 ms (so a warning of at least 40 ms). QK reminded that the sampling will be still fixed and the start of the sampling cannot be changed with the start of the function. So any timing shift may produce a slightly different function shape.
As any shift in timing for the start (given that the sampling is fixed) will alter the shape of the function TE-EPC is investigating a learning algorithm as currently the measured PPM error is outside the specification. The reference reproducibility would be good only for shifts of multiples of 100 us, but any learning algorithm will anyway take longer for any other shift in time. This is unfortunately the case for the PSB where the timing start will not only shift by multiples of 100 us.
Action: BE-OP to organise a meeting with the timing support and TE-EPC experts to discuss the possibility to implement LTIMs to to provide fast timings to start the functions.
JP said that it he was afraid that the PSB needs to consider 1 us for the magnetic field. It will be difficult to reach the required stability taking into account external influences e.g. from the mains.
QK: How can one isolate the Bdl from the influence of the main field? It is not tested as of today. This is a shared worry.
JP pointed out that there will certainly be different effects with much more impact than the 100 us FGC sampling.
HB: How large is the energy error? Not clear. How large would the dp/p change as the quadrupoles will change? What is the effect concerning the timing shift? Action: SA to check.
HB: Why the ring blanking is not possible using the chopper? Because of Linac4 RF limitations (feed-forward and space-charge compensation).