Tuesday October 9th,
14:00
CCC conference room
874/1-011
Persons present
Roger announced that the agenda of this meeting had slightly been modified, since more data still needed to be taken as input for Magali’s talk, which would therefore be postponed to a later date.
Concerning the minutes from the 31st meeting, Luca Bottura commented on the discussion about the availability of tools for the analysis of magnetic field measurements, which had taken place at the last meeting. He regretted that neither Laurent nor he himself had been present in that discussion, as they were the persons actually responsible. A FIDEL meeting prior to the 31st LHCCWG meeting had concluded that from the end of 2007 one would not be able to use the automatic data reduction and data analysis which had been put in place for the magnet series measurements. However, Luca pointed out that the measurement system proper would always be available. Also the database storage would stay as it is now, except for a change in the ORACLE version to be performed by AB/CO. It is only the data reduction, a third tool, which will cease to be available from the end of 2007, as its maintenance would imply porting to linux SLC4, which is not foreseen. As a substitute, Luca and his team will provide a minimum set of necessary analysis tools. Luca reemphasized that the measurement possibility itself would always exist, and that only the automatic data reduction was concerned. Responding to a question by Frank he confirmed that more man power would always overcome this potential bottleneck if ever the need arose. He also stressed that the amount of data expected for 2008 was not enormous. There were no other comments on the minutes of the 31st meeting.
Roger presented a new LHC schedule which was shown during the cost & schedule review the same morning. Looking at this schedule he had initially concluded that a sector test would be possible in May 2008. Lyn’s reply had, however, been that he would insist on taking the beam all the way around the LHC. Massimiliano asked for the minimum number of sectors needed for a sector test. Roger answered that in the original version only sectors 7 and (parts of) 8 were required. The schedule represents a planning for success, in the sense that it takes into account all available information, but that it does not contain any contingency in case some problem should be encountered. Ralph remarked that any vacuum leak might cause a delay of 3 month (more precisely, 3 months are required per thermocycle). Stefano and Roger commented that the planning must provide for two months to commission with beam. Also at the cost & schedule review, Mike Harrison recalled that interleaved hardware and beam commissioning had successfully been done at RHIC. Massimiliano posed the question whether beam operation would be possible in part of the machine, while hardware commissioning hardware in the rest. Roger replied yes, in principle, but that details would need to be checked with regard to safety.
Massimiliano asked about the need for a radioactive cooldown in case of interleaved commissioning. Roger answered that during this early phase of commissioning the beam was only of low intensity, e.g. a single bunch, and that activation levels were consequently expected to remain insignificant. Ralph added that nevertheless the LHC would belong to a different control zone afterwards, with more restricted access rights, which could be a problem e.g. for workers from outside companies. Gianluigi asked whether the helium problem was taken into account in the schedule. Roger reported that there would not be any helium problem according to Lyn.
Verena discussed the interleaved filling possibilities for the LHC. Her talk was a follow up of a topic which was raised in the 25th meeting of the LHCCWG. Verena reminded the WG that there are two types of extraction needed from the SPS to fill the LHC. One refers to injection of beam1 through TI2 into LSS6, and the other to injecting beam2 through TI8 into LSS4. The nominal filling comprises 12 injections per ring. The question to answer was whether the two LHC rings could be filled in an interleaved manner. The short answer is yes. A more difficult question is whether this can be done easily and efficiently. One obstacle is the large horizontal bumps of more than 35 mm, required at the SPS extraction septa.
The central timing system (CTS) controls the production of the LHC beam by means of the central beam and cycle manager (CBCM). Beams are associated with “users”. Static and dynamic destinations exist. For example, for the LHC beam in the SPS the static destination is always the SPS dump, while the dynamic destination is set on the fly (TI8 or TI2). The dynamic destinations TI2/LHC or TI8/LHC can only be set by the LHC sequencer. Three basic SPS supercycles are foreseen, namely nominal LHC (~20 s), FT+CNGS+pilot (~43 s) and pilot (~8 s).
The SPS power converter control (“mugef”) is run by the ROCS software, which cannot understand “destination”. The extraction kickers on the other hand can understand destination. Therefore, one possible approach would be to always pulse both extraction lines plus both extraction bumps and work in the same scenario as outlined above with interleaved injections into the two rings. This is expensive in terms of electric power and, in particular, it aggravates safety issues due to the bumps in the SPS, with the risk to damage both extraction channels simultaneously. The ROCS system will eventually be superseded by another system that understands “destinations”, e.g. FESA, but this will not yet be implemented for next year.
Verena then proposed an alternative way of realizing interleaved injection, based on two LHC users and two LHC cycles (one for each LHC user) per LHC supercycle. This scheme was tested at the SPS this year, by introducing 2 LHC cycles in the SPS supercycle. The total length of the supercycle was 16.8 s. If one uses the same supercycle to fill a single ring, it would take twice more time than with a dedicated cycle. This problem could be overcome by changing the supercycle in such a case. In other words, more types of supercycle would be required. However, all supercycles would have to be ready, and to be really efficient, they should be resident in the hardware. But there still is a problem, namely the total number of users is limited – only 24 are possible in total (of these three are already used for CNGS, three for coast, others for pstart/pstop, SFTPRO, etc.), and the ROCS do not have enough memory. Changing the supercycle takes 1-2 minutes if the underlying cycle is resident, and up to half an hour if the cycle is not resident.
Verena explained that for next year we can live with the present situation Her proposal for next year is to have two permanently resident LHC supercycles, one consisting of FT, 3 CNGS and pilot (43 s) and one consisting only of the LHC pilot (8 s) for single injections from the PS with low intensity of up to 72 bunches or up to 12 bunches with both extraction lines and bumps pulsing. Nominal LHC supercycles (single and double cycle) would also exist but not be resident.
Roger remarked that this scheme was OK for commissioning, but not for routine operation. Gianluigi noticed a possible problem with the 43 bunch operation mode which may require the nominal cycle. Roger, Ralph, and Jorg commented that during LHC physics filling there would be no CNGS beam. Paul observed that he would rather select FT+3 CNGS+LHC cycle and the nominal LHC cycle as the two resident cycles. Jorg confirmed that presently we are already at the very limit of memory/capacity, and that in order to include additional supercycles we must optimize the number of points on the ramp. The primary limit is a single crate for the main power supply memory of the SPS, which must be taken care of. The corrector limits could presumably be dealt with.
A general consensus emerged that if it takes not much more than 10 minutes to switch from one operational mode to another, this would be fine. Jorg commented that we were only discussing the SPS, but he pointed out that there are also all the other injectors which might as well need time for switching between modes. Helmut commented that experimental tests on beam scraping indicate that the cycle duration of 8 s would be too short, and that 9 or 10 s SPS cycle times are needed for scraping if cleaning is required. Ralph recalled the need for displaced bunches in view of LHCB and he asked whether displaced bunches were compatible with the proposed cycles. Verena replied in the affirmative.
Verena continued her presentation, observing that the ROCS will migrate in the shutdown 2008/09. The next generation software will understand destination. Verena stressed that we would like to have more “users”, at least 8 more (giving a total of 32), despite the memory limitations in the ROCS. The functions for the corrector magnets need to be cleaned up. Jorg remarked that the present number of 24 users comes from the PS. This number seems to be hardcoded somewhere and not easy to change. Also RF and BI would profit from a change.
Paul suggested that this topic should be brought up at the LTC, which should issue a request based on the real LHC needs.
The solution for interleaved injection in 2008 will be to use two LHC cycles in the same supecycle. Verena illustrated this scheme once more by a snapshot from the SPS page 1 screen. In her experimental test, the beam consisted of about 1.3e13 protons per extraction (5e10 protons per bunch, 72 x 4 bunches).
Paul asked for the duration of the SPS test. Verena replied that they executed 12x2 extractions. Jorg mentioned that this had taken at least 8 hours. Paul expressed concern about the stress on the PS pole-face windings. Verena added that another 3 extractions were missed due to interlocks
Gianluigi inquired how the extraction to the SPS could be stopped. Jorg replied that one could change the destination in the PS to “dump”, and also activate the tail clipper, which is to be tested.
Roger asked whether a demonstration of the switch from the pilot cycle to two LHC cycles had been tried. Verena answered that, yes, this switch had been explored during the MD, and that the second time it had been quite fast. With the SPS sequencer running, this switch should be OK.
Stefano asked why we need to go back to the pilot cycle. Jorg responded that RF and BI settings need to be changed depending on beam intensity. Stefano asked whether this switch cannot be avoided. Jorg answered that, yes, indeed another possibility would be to redirect the settings for BI and RF and keep the identical magnet settings.
Jorg also pointed out that, in case of different supercycles, when using the pilot for steering the transfer line, one would need to copy the magnet settings to the other cycle.
Summarizing her presentation, Verena stressed the we can fill the LHC in an interleaved way, that the current limitations for an efficient and flexible filling of the LHC are the SPS mugefs (mainly the extraction bumpers), and that the proposed solution consists of migrating the ROCS to FESA or similar (which will come), and of adding more users, at least 8 more. Verena finally presented some details on the additional complexity of joint LHC and CNGS extraction in TI8/LSS4, where an electronic switch will be employed.
Paul emphasized that we will need to train the various schemes before LHC commissioning. An interesting question is the remanent fields, which may differ between cycles. As the resynchronization between rf systems is done automatically, Massimiliano wondered what would happen if one beam is not sent. The answer is that this will be taken care of by the LHC beam quality check. Jorg outlined that there would be a fast detection of the problem and a quick decision to refill the missed injection.
Roger concluded that we know what to do for next year, namely inject the pilot beam first into one ring, then into the other ring, followed by interleaved injection.
Stephane presented “pros” and “cons” for and against interleaving the filling of the LHC. He first recalled the general prerequisites that any filling scheme must provide, namely a measurable circulating beam before the injection of nominal bunches and a check plus readjustment of the main beam observables before injecting new nominal bunches.
There are no arguments in favor of an interleaved filling scheme from either RF, BT, or BI. From ABP the only argument put forward in favor is the LEP-II experience, which, however, is not completely understood and where parasitic beam-beam effects might have been much stronger than they will be in the LHC.
RF and BI also do not provide any arguments against an interleaved filling scheme. BT has a weak preference for sequential filling e.g. in case one beam aperture is limited by the TDI of the other beam. ABP brought up two weak arguments in favor of sequential filling. One is the emittance blow up for ions in proton-ion operation, i.e. ions must be injected after protons provided that no beam-beam related instabilities are preventing this. The other argument is the aperture bottleneck in D2 for off-bucket particles (see Stephane’s comment at the LTC 2006-15). In the case of interleaved filling two times more particles could hit D2 than for sequential filling.
Stephane concluded that the final answer is likely to come from an optimization of the turnaround time and the average luminosity. In particular, he noticed that the interleaved scheme would be better in case of frequent injector failures.
Roger suggested that we should choose the scheme which is more reliable from the viewpoint of the injectors. Ralph commented that in case of sequential filling, decay on the injection plateau will lead to an asymmetry in the currents of beam 1 and beam 2, recalling that HERA for example had had a low beam lifetime at injection. Oliver commented that, whatever the scheme, always some weak bunches will collide with some strong bunches. John remarked that a distribution of intensity in both beams may be better than a large asymmetry between the two beams. JPK commented that the ISR had preferred interleaved injection, due to the limited stability of the injector system. But he also added that whenever it had been uncritical, the much easier sequential injection had been chosen.
AOB
Oliver recalled that there are several open actions for the LTC which should be addressed by the LHCCWG.
For the next LHCCWG meeting originally a presentation of the aperture model by Stefano had been foreseen. However, since the date coincided with a “school vacation week” where Massimo was not likely to be around, it was considered to postpone Stefano’s talk.
Tuesday October 23rd, 14:00
CCC conference room 874/1-011
Provisional agenda
Minutes of previous meeting
Matters arising
LTC open actions
AOB
Reported by Frank