Wednesday February 28th,
14:30
CCC conference room
874/1-011
No comments had been received on the minutes of the 20th LHCCWG meeting. Referring to the statement that a 4% increase in bunch intensity can recover the nominal luminosity for the alternative filling scheme presented by Werner and Elias, Massimiliano pointed out that this were not a fair comparison if the 4% could also be gained in the nominal scheme. Gianluigi replied that the parameters of the nominal scheme are “at the edge”, and that by contrast the alternative scheme provides some margin. Massimiliano asked for an explicit confirmation that the limitation in the present scheme is related to the bunch intensity. Gianluigi clarified that the present limit is related to the product of the number of bunches and the bunch intensity.
Luca started with a general overview of the Field Description of the LHC (FiDeL). FiDeL aims at predicting the magnetic state of any magnet following arbitrary operating cycles to a practical accuracy. FiDel clients are LSA, which requires a parameterization of transfer function and harmonics, and the off-line LHC model (MAD) through WISE, for which FiDeL provides a snapshot of the machine. The overall view and scope of FiDeL can be understood by its interrelations with several databases, LHC configurations, and applications. LSA, WISE and MAD belong to the set of relevant applications. Databases include the layout, magnetic field data, geometry, and survey data. So far the cycle and recovery procedures for the various types of magnets are not defined. Different departments and groups are involved in the numerous aspects of the magnet modeling. Luca stressed that FiDeL forms the core of all field modeling activities.
=> ACTION: Define cycle and recovery procedures for all types of LHC magnets (FiDeL team)
The FiDeL modeling concept consists of two main objects: a parametric model with 7 physical components, and about 20 field parameters (referring to “shapes” for magnet families and “amplitudes” for individual magnets). Examples for the 7 components are saturation, magnetization, decay, or ramp. The interface to LSA will be a JAVA implementation of the FiDeL equations and the ORACLE database of circuit parameters, magnet trims being taken into account. The interface to MAD will allow generating a snapshot of the machine at any time during the cycle.
Luca now turned to the status of FiDeL. The Field Quality Working Group was used to follow up the FiDeL progress. Three groups from AT and AB-OP are participating in this endeavour. The FiDeL web site www.cern.ch/fidel contains pertinent project information, including minutes of meetings and documentation. As part of the FiDeL project, new reference magnet tables are being generated which will supersede the present tables. Luca discussed the detailed implementation plan of FiDeL. FiDeL data structures and a test version of the FiDeL engine will be available at the end of March. In the following months, through September, normalization cycles, magnet data consolidation, powering and tracking tests, and sector powering tests will be addressed, and the WISE interface will be adapted to FiDeL. Luca then presented some further details of the magnet data consolidation. Issues to be resolved in the next three months include the reference frames and conventions, magnetic data and modeling. So far only manual tests of FiDeL were performed on a small selection of magnets. Now a semi-automated fitting procedure for massive data treatment is needed.
A first FiDeL validation milestone is the tracking test of 2 MBs and 1 or 2 MQs in SM-18, where nominal ramps in the main magnets and correctors are generated with the goals (1) to test the model and verify its accuracy, (2) to determine the effect of deviation from the nominal operating conditions, and (3) to assess the effectiveness of recovery procedures. A second validation milestone consists in testing the LHC ramps during the hardware commissioning, where the plan foresees generating the nominal ramp, run ramps from CCC, and diagnose the results, with the goals to prove the readiness, to quantify the ramp-down load and to explore real time trim mechanisms, as well as to check hardware limits of correctors and quench protection.
Responding to a question by Ralph on the completion schedule, Luca announced that a FiDeL test version will become operational in April. Paul then asked for the current limits during the 450-GeV run. Luca replied that the magnets can be ramped up to 1 TeV.
Luca stressed that significant resources are required for this work. He suggested that a formal assignment of work packages be considered now. His talk concluded with some acknowledgements.
Oliver
recalled that in SM-18 only about 20% of the
Ezio presented the WISE input to the MAD code. He acknowledged
contributions from many people. The WISE code has been written by Per Hagen.
The idea itself was launched by
Ezio emphasized that the addition of uncertainties allows generating several instances of the LHC and thereby estimating the error of the predicted quantities. With WISE one can for example check whether the magnet sorting was effective or not. Magnetic measurements, geometric measurements, and the installation database all provide information to WISE, from which the MAD input file is then constructed. WISE provides MAD inputs for the machine as built as well as for the study of magnet replacement scenarios. Another important outcome of WISE is the validation of field-model information. A large amount of data is stored at different places. WISE combines magnet data from different sources and allows a check of data consistency. A useful spin off of WISE is its use for the validation of FiDeL and LSA.
WISE has been implemented in EXCEL format. A script downloads the data from the databases. Macros carry out reconstruction of missing data and in particular the extrapolation from warm measurements. Another macro writes text files of MAD input and MAD scripts. The WISE EXCEL worksheets contain all the information (magnetic measurements, geometric measurements, slot allocation). The code has been built as a “transparent box”.
Replying to a question, Ezio explained that the layout and slot-allocation data are obtained from the manufacturing and test folder (MTF) and do not need to be downloaded repeatedly.
Ezio next sketched the WISE origin and evolution. He then took the working group on a walk through WISE. The main input menu allows selecting the type of magnet errors to be included in the MAD input file. A number of instances can be chosen (called ‘iterations’). At the moment, MAD files for injection and collision are available, based on measurements. Once FiDeL is fully operational, any mode of the machine can be considered.
Frank asked about the two options of using either FiDeL predictions or the magnetic measurements as basis for generating the MAD files. Ezio responded that the default will be FiDeL in the future.
WISE’s intermediate steps make all magnetic measurements available for inspection. The same transparent approach is adopted for the geometric measurements. Misalignments were listed as an example. The most complicated part is the slot allocation. The pertinent information is again downloaded from different sources to allow for a crosscheck. As stated earlier the final WISE output is a MAD input file including appropriate field errors.
WISE applications considered so far include beta beating (see the talk on “transfer function of the quadrupoles and beta beating” by S. Sanfilippo at Chamonix 2006) and dynamic aperture estimates (LHC Project Report 926). Studies in progress comprise the gain from sorting, and the impact of magnet geometry on the beta beating.
More information can be found on the WISE web site http://www.cern.ch/wise . An official documentation will be released in a few weeks. Further validation of the geometry will be done by the users. In the future, when WISE will use the field model values provided by FiDeL, downloading magnetic measurements will cease to be necessary.
Ralph expressed interest in using WISE for collimation. He asked which version of EXCEL is used. Per replied that WISE should work also on Windows XP. Ralph then asked whether there would be any concern about operating-system upgrades. Ezio answered reassuringly that no problem had been encountered so far. Oliver commented that the main achievement of WISE is the slot assignment of the errors. Ezio pointed out that the option to see the measurements is also important. Small MTF-ORACLE mismatches were already found thanks to WISE.
Stephane summarized his view of the WISE status as follows: The geometry, misalignment, nonlinear errors etc are all OK, but an aperture model of the machine is still missing. Arguing that it should not be too complicated to assign to each magnet a number of misaligned slices with aperture information, he then proposed adding for each magnet 20 slices of aperture data. Per conceded that indeed an aperture model has not yet been implemented. John remarked that a summer student had developed tools for generating MAD apertures from the database in 2003 and that this tool would still be available. Stephane commented that a magnet is not a straight line in general, but has internal deformation, which must be accounted for. Luca considered this to be an interesting demand, before underlining that also survey and movement data would need to be added, since the data from the construction measurement are not sufficient. Per added that the survey information is not included at present. Elaborating on his proposal, Stephane recommended developing a tool to set up the alignment and aperture data with 10 cm steps. He reiterated that it would be extremely helpful if WISE could provide such data. Oliver highlighted that this was a very good request from Stephane, which could also help in the understanding of observed beam losses during commissioning. Ezio replied that this request can be inserted in the WISE development pipeline. Massimo agreed with this plan and he proposed that the measured shapes of the magnets be included as well, as they have an important effect on the geometry.
Stefano stated that a program to assign aperture alignment errors with 10 cm longitudinal resolution does exist. It was developed for the collimator efficiency studies and it is being used to simulate the effect of static misalignments including the measured alignment profiles (x,y).
After the meeting, Massimo clarified that the misalignments between magnets in the same cryostat as well as the misalignment between the cryostat and the beam line can be already obtained via WISE, though he presumed that some cross-check/debugging of this part may still be necessary. Massimo had also started discussions on the addition in WISE of the cold-bore profiles, and the implementation did not appear to be straightforward.
=> ACTION: Add the measured cold bore profile and cross-check the misalignment part in the WISE LHC model (WISE team)
Marek discussed the LSA implementation of FiDeL. After a short introduction, his talk covered an overview of FiDeL models, the implementation within LSA, deliverables, status and milestones. FiDeL is a set of equations describing the functional dependence of field and field errors, using a set of parameters. Examples of dynamic effects represented by FiDeL include decay and snapback. LSA stands for LHC Software Architecture, which is a set of supervisory control applications. A chart of modules and data flow illustrated that the main LSA FiDeL model can be separated into 4 submodules related to calibration, harmonics, snapback and decay. All of these submodules provide an input to the setting generation.
Marek next addressed the implementation in greater detail, showing the components in an entity relationship diagram (ERD). Subsequent flow diagrams described the handling of static and dynamic errors. An example for a static error showed a non-constant transfer function of B/I vs I. The handling of dynamic errors takes into account the cycle phase and the powering history. LSA computes the corrections needed and applies trims based on the FiDeL model.
On the injection plateau, the decay corrections are applied as a function of time. The predicted snapback is computed, and its correction is applied every 10 s. LSA-FiDeL deliverables are the calibration curves, the harmonics, the prediction as well as the correction of decay and snapback, and a result exported to MAD.
Marek displayed the status and milestones. The main parts of the application are ready. Work is still ongoing on the powering history and on the transfer of FiDel data to LSA (the completion of which is foreseen for April 2007).
Paul asked how the hysteresis in the calibration curves is taken into account. Mike replied that one needs to know on which hysteresis branch the magnet is presently located. Luca commented that the branches are already included in FiDeL, but that the details still need to be worked out. Gianluigi whether one needs to know how much later the ramp will be started at the time of calculating and loading the ramp functions. This would be necessary for a proper matching between the end of the injection plateau and the start of the ramp for those magnetic elements whose settings are varying during the injection plateau in order to compensate for the persistent current decay. Marek responded that an underlying assumption is that the injection plateau is long enough and that the injection effects have settled by the time the ramp starts. Luca remarked that some of the dynamics is dealt with as a trim correction, and that one still needs to determine the start of the ramp in advance. Mike commented that some lee time may be needed. In particular, after loading the ramp functions, the latter cannot be changed. The sequencing must therefore be done carefully. Verena asked how accurate the predictions would need to be. Mike explained that after 20 minutes most of the decay is over, but that a ramp after 10 minutes would be more difficult. Stephane stressed that it is important to identify the source of measured chromaticity changes. He recommended forming the sum and difference of the horizontal and vertical chromaticity changes in order to infer the contributions from the b3 decay.
=> ACTION: Include hysteresis model in FiDeL field predictions (Luca, Marek et al?)
Walter reviewed the behavior of the magnets during the low-beta squeeze, which amounts to a sequence of current ramps, eventually with changes of signs and intermediate stops.
Replying to a question by Ralph, Walter confirmed that persistent-current decay and snapback may also occur during the squeeze.
After listing all the magnet classes concerned by the squeeze (MQM, MQY, MQT, and MQTL), he described the assumptions and approach taken to address this problem: Massimo had provided him with the magnet strengths for the squeeze. A linear transfer function was then assumed to generate the magnet currents, and then the field deviations due to hysteresis as well as the field decay on the last steps of the squeeze were measured. For the squeezing sequence chosen, the beginning of the squeeze corresponds to the injection optics scaled at 7 TeV. For some initial tests it had been assumed that the nominal current reached by each magnet was 5390 A. This turned out to give representative results for Q5L5.
First examples presented were the MQMs Q7L5 and Q5L5, for which the squeeze was applied after a cycle to nominal current. The amplitude of the field error was 2-4 units. The squeeze here implied one hysteresis crossing. Another example was Q6R5, for which the field decay was found to be below the measurement noise. Further cases included Q9R8, Q6B5R2F, Q11R8B2, Q11R8B2, Q11L5B1, Q11L5B2, as well as Q11 and Q12 in IP5. For MQM and MQY magnets, the crossing of zero field is avoided. However, some of the MQTLs and MQT cross zero during the squeeze, resulting in a large measurement error.
Stephane commented that if the hysteresis branch were not taken into account the error would be much higher. He asked whether the hysteresis would be easy to model. Walter replied that the modeling would indeed be easy for the main hysteresis loop in case of single crossings at relatively high field. Massimo commented that the IR8 squeeze and trims are not yet fully optimized. The 3 examples from IP8 shown might be a worst case. IP1 and 5 are in better shape. This is why most measurements used cycles from IP5.
Walter stressed again that the behavior near zero is problematic, and that transfer function values are greatly variable at low current. From the shown measurements, Stephane inferred a 50 unit variation at 10 A current, which he did not remember having seen at Chamonix 2006. Oliver recalled that the trim quadrupoles are meant to be powered only up to 10% so as to avoid large beta beating, and that therefore they are likely to be operated at low current in the LHC. Stephane commented that at injection a single-quadrupole strength change of 0.001 Tm causes a typical tune shift of order 1e-3. The LHC beam should be less sensitive at top energy. Oliver cautioned that this tune shift was due to the effect of a single magnet only, and that contributions from many magnets need to be faced.
The maximum setting errors can be estimated from the hysteresis loops (without modeling hysteresis crossing). The decay of MQY and MQM for the reference cycle is -6 or -4 units. On the positive side, the MQT and MQTL fields were found not to decay at all. Walter pointed out that the error bars can be extracted from the hysteresis loops. He also reminded the working group that the present FiDeL prediction does not yet contain the information on the hysteresis crossing branches. A FiDeL model of the hysteresis crossing should reduce the errors of the prediction.
Roger asked whether such an extension of FiDeL is foreseen. Luca replied yes, but that it had recently been put into question. A related action was initiated further above.
Walter drew the following conclusions: Very low settings for MQTL and MQT are a problem. However, there is no decay in MQT and MQTL. Extracting the full decay characteristics for MQM and MQY would require more data. A squeeze on the ramp would reduce the number of hysteresis crossings, and would probably also reduce the decay. Both these consequences would simplify the magnetic model.
Luca posed the question how critical the effect of hysteresis crossing is. A mechanism for storing the cycles could be put in place. Stephane replied that the accuracy target is known. This target could be compared with the size of the hysteresis-crossing effect. Stephane asked for an estimate of the error that could be obtained by a cycle from 100 A over -100 A back to 100 A. Walter answered that this error would be of the order of the size of the hysteresis loop at 100 A.
Ralph S. recalled that a related phenomenon, namely the hysteresis of steering correctors and their effect on the orbit feedback, had been discussed in his presentation at Chamonix 2006 (“What is the Impact of Hysteresis on Orbit Correction and Feedback?”).
Paul remarked that the tune feedback would take care of tune changes. Ralph S. cautioned however that the tune feedback corrects the tune but not the beta beat. Paul asked for the behavior of the lattice sextupoles during the squeeze. Sextupole magnets had not been measured by Walter. Stephane remarked that certain phase advances between IRs could alter the sign of one sextupole family, and that this situation can, however, be avoided by changing the phase advance. Oliver commented that the octupoles for Landau damping should be zeroed to the per mil level at injection. Walter answered that indeed this is the reason why the octupoles are degaussed at injection.
=> ACTION: Follow up sextupole behavior during the squeeze, and the problem of low settings for MQT and MQTL. Take additional data as needed to fully characterize the MQM and MQY decay characteristics (Walter et al)
Stefano went through the procedures and documentation for the squeeze, phase A.8 at 7 TeV. Already many discussions were held off line. Here he could not go too much into the details. The procedures were developed by the EICs plus Verena.
Entry conditions for the squeeze include the following considerations. For the moment, it is assumed that the squeeze will take place at 7 TeV. The number of steps in the squeeze is still to be agreed upon. Other requirements refer to orbit and optics. In particular, a golden orbit and sufficient beam stability are required at the collimators and in the IRs. Detailed measurement procedures at 7 TeV are to be defined, as well as the accuracy of these measurements and the applied corrections. In addition, well defined procedures for switching on or off the feedback must be in place. Collimators and dump must have been tuned. Landau damping by octupoles should be active. BI requirements are similar to those in previous phases, except that, possibly, additional set up time may be needed for the synchrotron light monitor.
Oliver asked whether the squeeze should stop at 1 m. He posed a similar question for the crossing angle. Massimiliano remarked that according to his understanding the solenoids do not matter at 7 TeV. Stephane answered that the coupling introduced by the solenoids is smaller than other effects but nevertheless needs to be corrected. This can be done by trimming the skew quadrupole correctors. Mike asked whether a special treatment would be needed for LHCb.
Stefano then discussed the procedure itself. Questions concern the compensation of the field decay at each step of the squeeze, and its implementation in LSA. Checks are needed of the procedures for optics and orbit.
Ralph asked whether we want to squeeze one IP after the other, or e.g. the two high-luminosity IPs simultaneously. Stefano replied that it might perhaps be best to start with one beam in one IP, and then go to the other IP or to the other beam. Nothing had yet been decided. Roger and Stephane suggested the squeeze to be commissioned down to 2 m or below. Mike warned that we should be prepared to back off if needed.
Stefano commented on the final values of beta*, beam quality checks, and the detailed measurement programme required for each beta* step.
Oliver recommended exciting the trim power supply at Q1 in bench measurements, so at to check the response of the quadrupole. Walter responded that triplet magnets cannot be measured, but perhaps this study could be done with an MQY magnet which is similar.
In the machine, the beta beat can be inferred from consecutive BPMs.
Ralph S. asked for the speed of the squeeze. Its ideal duration is 6 min. The squeeze will occur more slowly in the commissioning. Another question was related to the constancy of the golden orbit. Stephane commented that the orbit cannot be stable during the squeeze, as e.g., the crossing angle will change. Ralph S. reminded the working group that the reference orbit needs to be updated in the LSA database whenever it changes. Frank asked whether the beams are separated during the squeeze. Stephane replied yes, they are kept separated with a constant transverse distance of 1 mm. Oliver clarified that the separation is only constant if the steering magnets are changed during the squeeze. Ralph A. asked whether there will be a crossing angle. Stefano answered that for the time being no crossing angle is foreseen in this phase. Roger recalled that the commissioning step of colliding physics is followed by a squeeze with 43 or 156 bunches. The latter corresponds to the phase discussed by Stefano. Later phases will include squeeze with crossing angle. Ralph concluded that we later may need another procedure for commissioning the squeeze in the presence of a crossing angle. Stephane remarked that in any case knobs for separation and angle bumps are needed for adjustments, even if the nominal crossing angle is zero. In particular, all steering correctors in the IR with one possible exception will need to be commissioned. Paul commented that a correction around zero is not the same as generating a large angle.
Oliver warned that the steering magnets inside the triplet are problematic quoting K.-H. Mess, in particular in view of a possible x-y coupling. Stephane remarked that, for the Q2 quadrupole, a significant discrepancy exists between magnetic and mechanical center, which must be compensated by orbit correctors.
Paul highlighted that in his opinion no part of the squeeze should be commissioned with single beams. Instead we should commission the squeeze for both beams at the same time, just separating the two beams transversely. A side benefit would be that in this way we acquire the beta beat information simultaneously for both beams. Mike asked what would be wrong with 1 beam, arguing that a single beam could simplify the early commissioning. Ralph agreed, stressing potential advantages of single-beam commissioning for machine protection, e.g., the absence of any cross talk in the beam loss maps. Oliver cautioned that single-beam commissioning would require a good reproducibility of optics and orbits. Paul discarded the machine protection argument, pointing out that the commissioning of the ramp is anyhow done with a single bunch per beam. Ralph maintained that at higher beam current it may prove advantageous to operate with only a single beam. Stephane suggested that the commissioning be done with the smallest possible intensity.
Stefano asked whether the pilot bunch at 7 TeV would be above the damage limit. He explained that according to the present working assumption all machine protection elements are thought to be in place and active at the start of the ramp. He would appreciate a confirmation of this assumption.
=> ACTION: Decide between the options of single-beam and two-beam squeeze commissioning, and choose an IP squeeze sequence. Clarify whether in commissioning phase A.8 the machine protection is fully operational at the start and at the end of the ramp (Stefano, EICs and Verena?).
Concerning the choice between single-beam and two-beam squeeze commissioning, after the meeting Ralph S. remarked that a correction of the orbit in the experimental IRs requires the presence of both beams if common correctors are to be used. Otherwise, a potentially beneficial beam 1 orbit correction might turn out to be worse for the beam 2 orbit and vice versa. However, if the primary beam consists of pilot (nominal) bunches, already one pilot (nominal) bunch in the secondary beam is sufficient to take the cross-dependence of the common correctors into account.
Roger reported that in 2007 similar discussions on the commissioning phases are foreseen to take place at the LTC. These LTC discussions would then amount to a second round through all the procedures. Following the meeting it was decided that, to avoid any possible time conflict with the 2007 LTC schedule, future LHCCWG meetings will be held on Tuesday afternoons, starting on March 27 in a bi-weekly pattern.
Tuesday March 27th, 14:30
CCC conference room 874/1-011
Provisional agenda
Minutes of previous meeting
Matters arising
Optics measurements needed at top energy
(Frank)
Collisions at top energy, luminosity
determination and optimization (Helmut)
Beam losses and radiation monitoring for
steering the beams in collision during the engineering run at 450 GeV (and at high energy?) (Thijs)
Commissioning of accelerator system -
BLM (Laurette)
AOB
Reported by Frank