Tuesday September 11th,
14:00
CCC conference room
874/1-011
Mike chaired this meeting. There were no comments on the 29th meeting held two weeks earlier. The latter had been a kick-off meeting discussing the status and the programme for the second half of 2007. Frank reviewed the planned talks through the start of November. Oliver then drew attention to the LTC open actions of relevance for the commissioning working group. The main open actions related to the LHCCWG concern the partial squeeze during the commissioning, the limits on momentum errors for the beam-dumping system, the safe beam for smaller than nominal emittance, and the definition of a bad BPM. A complete list of LTC open actions is available on the LTC web site.
=> ACTION: Follow up of LTC open actions (All)
Rob announced that a comprehensive engineering specification for all magnet cycles presently is under development. Rob was speaking for the FQWG as a whole. His presentation was structured as follows: (1) overview, (2) details of individual cycles, and (3) synchronization.
The aims of the magnet setup cycling are (1) to put all LHC magnets in a known magnetic state, well characterized by the associated field descriptions - only with such proper magnet characterization will we be able to accurately predict the corrections needed for decay and snap back; (2) to assure magnetic reproducibility; (3) to limit field decay; and (4) to limit field errors due to hysteresis, coupling between magnets or apertures, etc.
Ensuring magnet reproducibility is of crucial importance for rapid initial commissioning. In later years, more emphasis will be placed on computing corrections based on magnet history and on arriving at the shortest possible setup time.
Rob pointed out that the interaction between magnets requires the cycling of individual magnet groups to be synchronized. In particular, the two-in-one design of the main LHC magnets implies that the two magnet apertures have to be cycled simultaneously. Other examples of interacting magnets are the spool pieces at the end of the main dipoles (MCS - sextupoles, MCDO – combined decapole-octupole correctors), MS and MCB in MSCB (combined sextupole-dipole correctors), as well as nested magnets: MCDO, MCBX(A) – combined horizontal-vertical-dipole (+sextupole+dodecapole) correctors, MCSOX – combined sextupole-octupole correctors, and accelerator magnets coupling with detector magnets. In addition, the LHC superconducting bus bars could slightly magnetize the magnet yokes. The resulting small pertinent fields could also depend on the sequence of magnet cycling.
Next reviewing powering and protection, Rob emphasized that the LHC rings comprise 1612 circuits in total. The maximum attainable ramp rate will limit the minimum length of the setup. Any quenches will cause important delays. To limit the risk of a quench the maximum cycling current should not much exceed the maximum operating current (MQM, MQT, MQTL). The quench protection system may impose limits on the ramp rate, in particular on the gradual stop and start of a cycle. A parabolic slow stop and start of 10 s is in the right ballpark.
Many power converters (PCs) are monopolar. The maximum ramp rate down is determined by the circuit time constant. The longest time constants of 240-400 s are encountered for the MQs (it appears that the MQs precycles could be longer than those of the main dipoles).
A very extensive measurement programme is still underway to understand the influence of the cycle. The FIDEL modeling is also applicable for the setup cycle. It can predict the size of the pre-injection snapback. Choices to be made for numerous cycle details are guided by another program, CUDI. Rob stressed that for the LHC we have the best knowledge of the precycle impact ever obtained for any hadron machine to date, but that even more knowledge is needed for optimum machine operation.
Concerning
the cycle types, Rob distinguished magnets with and without field decay. The
latter can be treated similarly to LEP magnets. For the former, time is an
important parameter. Nested
Oliver asked whether constant ramp rates should be used. Rob replied that we have no choice for the ramp rates. The single quadrant PCs should follow the natural exponential decay of the magnet excitation. One should not attempt to implement a linear ramp, which would take longer.
Demagnetization cycles for the main magnets which minimize persistent-current decay but lead to large snapback are also available optionally.
Rob next addressed the question which cycles to use in case of an abnormal termination. Recovering magnets with field decay may require a pre-cycle, which basically imitates a physics run. Such standard setup cycle can take a long time.
Now
some reference cycles for magnets with field decay were presented. These
magnets are MB, MBRB, MBRS, MBX, MQ, MQM, MQMC, MQML, MQXA, MQXB, and MQY (all
magnets with
The nominal cycle comprises a ramp down of about 20 minutes followed by about 20 minutes pre-injection plateau, and also about 20 minutes injection proper. Afterwards the magnets are ramped to physics energy. The total turnaround time could be about 2 hours. For the main dipoles the pre-injection current is chosen as about 600 A (25% below the injection value). One can only go to the injection plateau when all magnets have reached their pre-injection values. All magnets together ramp to the injection energy, were the beam is injected after a short waiting time.
Ralph (A.) noticed that a waiting time of 20 minutes for injection was shown. He asked how much this value could vary and which impact such variation would have on the cycle. Oliver recalled from HERA that if the injection time exceeded 1 hour, HERA could not ramp anymore. In such a situation, it proved necessary to perform a new magnet cycle. Karl-Hubert added that another prerequisite, before going to the injection plateau, would be the readiness of the injectors. Concurring with these comments, Rob responded that indeed there are several reasons for staying on the pre-injection plateau.
Stephane asked for the general usefulness of the pre-injection plateau, and, for example, why one could not stay for 20 minutes on the injection plateau prior to injecting. Rob replied that the pre-injection plateau is meant to reduce the snapback during the start of the ramp, and to improve the reproducibility at injection.
Mike inquired the importance of a short minimum of 350 A, just prior to the pre-injection plateau. Rob replied that all PCs and magnets can reach this minimum, but did not give a clear indication of its importance.
Ralph refined his earlier question, now asking for the target length of the injection plateau. Rob answered that only field considerations are taken into account, and no operational aspects. Mike commented that 20-30 min. is a realistic target for adjustments, collimation, etc. Ralph suggested a further optimization of the injection-plateau time. Luca remarked that the typical minimum injection time was taken to be 1000 s, based on information obtained many years ago. All measurements were performed with this duration. Longer plateau times should not be a problem, however.
John Miles posed the question whether the pre-injection plateau length can be extended if injectors are not ready. Rob replied that, yes, exactly this was its purpose. Nicholas Sammut added further details.
Observing that the QRL needs time to remove energy, Ralph S, asked whether this was another constraint. Oliver thought likely not, since the same problem would occur on the regular ramp. Ralph S. recalled that in some cases, like a quench or fast-ramp down of the main dipole circuits, the cryogenic-system would require a given minimum grace time prior to re-filling the LHC in order to dissipate the generated heat inside the system. This recovery time could range from 2 to 6 or more hours.
Luca stressed that the pre-injection plateau allows for the definition of a clear time 0. John wanted to hear more details on the down-ramp time of the quadrupoles. Rob specified that their ramp down is a few minutes longer than for the other magnets. Oliver remarked that the pre-injection stretch will then simply be shorter for the quadrupoles. Mike mentioned that all the proposed cycles contain a parabolic rolloff.
Nicholas explained that, according to the data, the dependence of the decay amplitude on the pre-injection plateau is asymptotic and varies less the longer the plateau. The average time constant is 375 s and the variation is negligible after 1000 s. Standard measurements have been conducted with a pre-injection plateau of 0 s length. So the prediction is best without a pre-injection plateau. However, if we choose a pre-injection plateau of 1000 s and we keep this constant, we can maintain a reproducible decay amplitude. On the other hand, if the pre-injection time is varied, we would rely on the FIDEL pre-injection plateau model, which, though working well, has its own inherent error, particularly due to the small sample of measurements taken. Oliver remarked that the gain in the total cycle length from a shortened pre-injection plateau would be insignificant. Mike agreed and he recommended the approach “first operate, then optimize”.
Coming back to a previous remark of Oliver, Verena asked whether more than one hour waiting time would be possible at the LHC. Karl-Hubert remarked that the dispersion between magnets slowly diverges over the time spans of hours. Oliver repeated the experience at HERA that it was better to recycle than to ramp after waiting for more than hour. Replying to a question by Ralph, Rob explained that 90 minutes recycling time are needed to reset the machine completely.
Stephane commented that the pre-injection reduces by only 0.5 units the decrease in decay out of a total of 2 units, He repeated his question about the usefulness. Rob replied that the pre-injection plateau provides better conditions with basically no loss of time. Stephane asked whether alternatively one could also reduce the ramp rate and give up the pre-injection plateau. Nicholas Sammut responded that the measurements at 50 A/s show a b3 decay of 2 units, while at 10 A/s the decay would be about 1 unit. Results of the planned extended measurement campaign may show even smaller gains and the modeling error might also be smaller. Once these results come in, we will know precisely.
=> ACTION: Review and finalize precycles when extended measurement results are available (Rob and others)
The ramp up during the precycle of the MQs follows the main bends, however, during the ramp down the MQs follows the natural decay of the quadrupole circuit. This ramp down could take up to 35 minutes (five time constants) perhaps exceeding the duration of the main bends ramp down.
Rob presented a summary table of parameters of the magnets with field decay.
Luca asked whether the slowest magnet indeed is the MQ. Rob answered that this appears to be the case, but still needs to be confirmed. Certainly either the MBs or the MQs are the slowest magnets.
In some case the pre-injection level consistent with MB is lower than the proposed pre-injection level (power converter limitations etc.). For example, the official MQM mininum current of 120 A is not much lower than the injection range. The presently considered step is too low to reach the other branch of the hysteresis curve. Stephane asked whether the MQM limit is due to the power supply. Rob replied that, yes, it is imposed by the power supply stability. Massimo remarked that we can go to lower current levels if we do not require a guarantee of the PC performance. It was commented by both Stephane and Massimo that the latter may indeed not be important for the precycle. Luca observed that power supply oscillations could still be a problem. Rob explained that the present official values were derived from the layout database. The measurements were performed down to 50 A excitation.
Stephane pointed out that there are several magnets with peculiar excitation pattern, that might affect the precycle strategy. For instance, Q6 and Q5 carry very low current for the squeezed optics, but earlier during the cycle their actual current reaches a maximum. He asked whether we should ramp these magnets up and down in a special way. As another example he quoted the B5 spool pieces, which have zero field at flat top and are excited at 40% of nominal strength at injection. Rob agreed that the Q5/Q6 case will need to be considered more carefully.
=> ACTION: Precycles for Q5/Q6 and for b5 spool pieces (Rob and his team)
The insertion magnets (triplets) have three independent nested power supplies. Rob next commented on the need of an extended measurement programme, in particular on measurements for practical cycles and the influence of the 10 A/s ramp rate. FQWG will define the necessary programme.
Magnets without field decay can use standard cycles as for LEP. Typical examples for monopolar magnets, dipolar magnets and for demagnetization were shown.
Interaction between magnets will require synchronization of precycles between groups of magnets. Best reproducibility would be obtained if all magnet precycles were synchronized, but this might not always be practical or necessary.
Luca commented that the nominal ramps for MQ need to be synchronized, i.e. their down ramps to be started at the same time after the end of a run. Rob added some explanations.
Nested correctors require a special treatment. For the nested corrector magnets MCDO and MCBX first the inner corrector is set. Then with the outer correctors a specified cycle is performed before these are also set to their injection level. The case of MCSOX and MCBXA with three, respectively four, nested circuits is handled in a similar manner.
In conclusion, the EDMS specifications of pre-cycles are in preparation with details for each circuit. Further magnetic measurements are to be done. FIDEL and CUDI modeling can assist in defining the setup. LHC requires considerable synchronization during set up.
Karl-Hubert remarked that if we have a quench the quenched magnets/circuits will be in a completely different state. Luca replied that a pre-cycle with more than 30 min flat top will bring the magnets back to an identical state. Oliver recalled that in HERA a precycle was always used when anything went wrong. In certain cases it was sufficient to precycle some corrector circuits.
This talk was a sequel to an earlier presentation at the LHCCWG (4 October 2006), serving as a reminder and update. It was prepared with input from Nicholas Sammut and Stephane Sanfilippo. Reproducibility is affected by geometric changes, persistent currents, decay and snapback, and residual magnetization. The geometric uncertainty is 3 units of b3. The persistent current uncertainty ranges between 1 and 10 units for b1, b2 and b3, but this type of uncertainty is only relevant if the pre-cycle is changed. The uncertainty in the predicton of the decay is up to 20%, due to the limited sample of magnets measured (only 10). The resulting error changing from cycle to cycle amounts to 1.2 units of b1, 0.4 units of b3, and 0.14 units of b5. These numbers do not include deviations of the precycle from the nominal one, which would lead to much larger uncertainties.
Luca now presented new elements and a reminder, covering the tracking tests of July 2007, the effect of hysteresis on the correctors, as well as a few other mysteries and miseries.
The tracking test of July 2007 was originally scheduled to confirm B1/B1=const. B2/B1=const. B3 integral=0, B5 integral =0 over the MBs. However, the conditions at the SM-18 benches did not allow operating two benches simultaneously. A first test campaign was performed nevertheless, in which the b3 and b5 corrections were probed on a single MB using a customized LSA application from the SM-18 control room. After tweaking and numerous measurements, b3 was corrected to about 0.5 units. The measurement data accumulated during this test correspond to 10% of the entire raw measurement database. Luca addressed the questions why no better correction could be achieved. The instrument calibration indeed should translate to a tracking accuracy as good as 0.1 units. The completeness of harmonics integral could add a further small contribution. Potential issues are the persistent current model (50 A/s vs 10 A/s ramp rate) and the residual magnetization model. Another factor could be control issues and timing. The ramp speed of the precycle can produce substantial variation in the persistent current. The accuracy of the extrapolation needs to be checked.
Concerning other difficulties, a low-current pre-cycle may have a massive effect for the MQTs. Luca also reminded the LHCCWG of a strong anomaly encountered in one aperture of the Ansaldo-2 magnet, which was observed in a few other magnets as well, and which could hint at magnetic pieces inside the collared coils. The effect depended linearly on the maximum current reached.
Luca’s main conclusions were that the extrapolation from 50 A/s ramp rate (where the magnet measurements were performed) to 10 A/s (where LHC will operate) may have troubles, and that the tracking of the b3 evolution for a single magnet is five times worse than expected.
Oliver asked how this last measurement is to be interpreted. Luca explained the measurement and compensation were done for a single magnet, using a model based on data from the very same magnet. The different precycles (10 A/s vs 50 A/s) could be part of the explanation for the poor tracking. This is one of the reasons why the accuracy of the extrapolation from 50 A/s to 10 A/s requires renewed checking. The new measurement campaign will start in October. Volunteers are being looked for.
=> ACTION: Volunteers for extended measurement campaign (All)
MQT and MQTL are most strongly affected by the effect of low-current cycling. Luca recommended powering all magnets up to their quench current minus “x” Amps to avoid MQT-like problems.
Oliver asked about the status of the extended measurement programme. Luca responded that the latter is not tied to the sector commissioning. A problem is the cryogenic plant. Although the nominal power was provided, only one bench could be maintained instead of the three benches maximum planned. A few dipoles for tracking measurements will be available in October. There is no deadline. Karl-Hubert confirms that people are missing for these measurements. Oliver invited an LTC presentation on the extended program. Karl-Hubert added that several measurement programs are currently underway or in planning. John Miles remarked that the spool pieces cannot be powered independently.
Massimo based his presentation on material by Stephane from an FQWG meting of April 2004 in the framework of RMS system discussions. He now included magnets other than MBs and MQs. The analysis emphasizes the relaxed requirements in certain phases of operation, considering in particular the commissioning needs, pilot bunch running, and nominal intensity.
Commissioning needs imply a systematic b1 of 2-5 units with rf on or off, a systematic b2 in MB of 0.2 units and b2 in MQ of 4 units, an a2 of 2 units, and a b3 of +/- 1 unit over the ring. Setting up for nominal operation calls for a systematic b1 and a1 of +/- 1 unit, random b1 and a1 of 0.6 units, and a systematic b3 of 0.1-0.2 units (correct Q’ to ~5-10 units), etc. Massimo reviewed the different criteria and demands for final nominal operation. He presented a summary table for the MB and MQs requirements in the three phases, as well as a summary for other magnets.
Luca commented that alignment changes during the course of a year will also affect the reproducibility.
Massimo
pointed out that the Landau octupoles need to be zero
at injection. Karl-Hubert asked for how close to zero is zero, stressing that a
Ralph inquired for the size of beta beat with the assumed errors. He emphasized that the sum of static and dynamic beta beating should not exceed 21%. The tolerance for the dynamic beta beat is 8%. Especially, he was interested in the fill-to-fill reproducibility including all contributions. Stephane responded that not more than 3% beta beat was expected. Ralph added that at the Tevatron, HERA and other colliders the collimator settings are not fully reproducible at top energy. Frank noticed that these other machines use collimators for background control, which is not the case for LHC. Ralph called for contributions from other sources to be included in the accounting. Ralph S. pointed out that at top energy 0.3 sigma is equal to the stability of the BPM system (intensity and temperature dependent drifts). Stephane commented that 0.3 sigma orbit change would correspond to 0.15 random b1. Ralph S. replied that the problem of the b1 uncertainty arises only at injection, and that a 30 micron BPM resolution is possible.
=> ACTION: Refined estimates of static and dynamic beta beating with all contributions (LCU, collimation team)
Responding to a question of Oliver, Walter confirmed that the degaussing cycle for the Landau octupoles would be faster than the precycles of the main magnets.
Tuesday September 25th, 14:00
CCC conference room 874/1-011
Provisional agenda
Minutes of previous meeting
Matters arising
Report from MPSC subgroup (Jan)
AOB
Reported by Frank