RF PERFORMANCE AND OPERATIONAL ISSUES A. Butterworth, L. Arnaudon, P. Baudrenghien, O. Brunner, E. Ciapala, W. Hofle, J. Molendijk, CERN, Geneva, Switzerland Abstract During the 2009 LHC run, a number of difficulties were encountered in the operation of the RF system and transverse damper. The main operational issues are presented, along with proposed or implemented solutions. The readiness of both systems for operation in 2010 is presented, including the preparations necessary for running with unsafe beam. switched off, reducing the klystron cathode current and the collector power to around 100kW. KLYSTRON COLLECTOR POWER The nominal scheme for cavity operation requires that the cavity external quality factor is changed before acceleration using the moveable main power coupler. The parameters at injection and top energy are shown in Table 1. Cavity voltage/mv Cavity Q ext RF power/kw Injection 1.0 20000 135 Physics 2.0 60000 185 Table 1. Nominal RF power parameters at injection and physics [1]. The nominal klystron working point is 58 kv cathode voltage and 9 A beam current, giving a DC power of 520 kw, and a peak RF power of approximately 300 kw. However in 2008/9, for operational simplicity, it was decided to use a fixed Q ext of 60000 at injection and top energy. At injection, with 1 MV per cavity, this requires only 45 kw of RF power. In a klystron, the residual DC power not consumed as RF output power is dissipated in the collector, and with the low RF power required in this operational configuration, the collector power was close to the rated maximum, which raised worries of potential overheating of the collector. Traces of overheating were indeed observed when several klystrons were checked in January 2010 (Fig.1). In order to reduce the collector power, it was decided in 2009 to run with fewer cavities, with higher voltage per cavity. The higher RF power demanded per klystron results in reduced collector power. Initially an attempt was made to use 4 cavities at 2 MV per cavity. However this resulted in many RF trips from helium pressure as the cavities had de-conditioned during 2 weeks of running at injection level. Eventually a configuration was found which was more or less reliable using 5 cavities at 1.6 MV per cavity. Another measure to reduce collector heating was to modify the front-end software to automatically switch the power system to the READY state when the RF was Fig. 1. Klystron collector opened January 2010 showing traces of overheating Choices for 2010 running Three possible scenarios are identified for operation in 2010: 1. Reduce the number of cavities used. With the beam intensities expected in 2010, there is no limit from the impedance of the unused cavities, assuming they are detuned by 100 khz. However, we cannot reach the nominal 16 MV per beam in physics with the standard Q ext of 60000. A possibility might be to go to higher Q ext. 2. Use all cavities, but work at lower DC power. There is still very little headroom for feedback at 2MV during physics. We would have to accept a reduction in cavity voltage. 3. Commission the nominal scheme, using the movable coupler to change the Q ext after injection. Approximately 2 additional weeks would be needed before startup to fine-tune the cavity low level loop parameters at different Q ext values. However, this is clearly the only long-term solution for higher intensities, and is the choice strongly preferred by the RF group. HARDWARE ISSUES: SPURIOUS INTERLOCKS Slow interlocks Some trips due to slow control interlocks were encountered, related to waveguide and coupler blowers, water flows, etc., and associated interlock thresholds and detection times were adjusted in the PLCs.
A number of trips due to RF reflected power interlocks (Wattcher) were observed, and the integration time was adjusted on some of the Wattchers. The RF voltage ramp-up rate at switch-on was reduced, which in addition helps the cryogenic regulation stability. Helium pressure Some He pressure interlocks were seen when increasing the cavity voltages above 1MV, assumed to be due to de-conditioning of the cavities when running for a prolonged period at the injection voltage. Crowbar triggers A significant number of crowbar thyratron triggers were observed when starting and raising the high voltage on the power converters. These were diagnosed as spurious triggers, since no signals were seen on the diagnostic channels associated with the individual klystrons. An initial solution was to reduce the HV power converter ramp rate, which improved but did not completely eliminate the problem. A possible source of spurious triggers is the thyratron heater adjustment: the 4 heater voltages must be regulated to within 0.05V out of 4.5V, and any drifts will cause auto-triggers. This will be checked during the technical stop. Another potential source of triggers is earth noise. The earth for the HV bunkers is connected in the RB, far from the bunkers themselves. This makes the system more sensitive to earth perturbations when the machine is running: magnet power converter trips have sometimes been seen to trigger the crowbars. This will be investigated and possible changes to the earth layout will be considered. POWER SYSTEM SWITCH-ON PROBLEMS HV trips Occasionally it was difficult to restart an RF line due to repeated HV trips from crowbars. In this case the power converter post-mortem procedure blocks the communications for around a minute, and the RF application did not inform of this. This will be rectified before the 2010 startup. Several times the 18kV was switched on with the converters on, which in one case damaged some 5V power supplies in HV bunker. This is a dangerous situation and a safety interlock on the 18kV has been requested from EN/EL. Alarms Sometimes the alarm diagnostics failed to display a correct fault. This is due to a bug in the FESA class which will be rectified before startup. In addition, the alarms are level 2 (yellow), so are generally ignored by the operators. They should be level 3 (red); thus will be rectified with the LASER team. Software issues The RF power application was found to block from time to time, due to crashes of FESA classes running on cfc-ccr-cgplrf. This was traced to a hitherto undetected bug in the FESA framework, where an internal message queue used for notification of CMW subscriptions overflowed and resulted in a crash of the server process. The problem has been fully solved in the January maintenance release of FESA [2]. SYNCHRONISATION PROBLEMS If for any reason the synchro loop unlocks, a desynchronisation of the master revolution frequency with the revolution frequency program will occur. This causes the beam to be subsequently injected into the wrong bucket. This occurred repeatedly in the first weeks of operation and was traced to several causes: Coarse frequency program mismatch: The coarse frequency program is used as a reference frequency for the master oscillator (VCXO). With the frequency programs of the two rings locked together, which is the normal mode of operation, the ring 1 coarse f prog should be used for both rings. This had been overlooked, and the ring 2 reference used. In the case of a large trim, the incorrect reference would make the synchro loop unlock (Fig. 2). New FGC channels were added and modifications made to the sequencer to allow use of ring1 coarse f prog for both rings. Phase loop problem with beam but RF off: Injecting beam with RF off gives a random phase in the phase loop, and the resulting phase loop transient kicks the synchro loop out of lock (Fig. 3). A similar effect occurs when when switching the RF off with beam (e.g. for debunching). A threshold on cavity voltage was added in the firmware to disable the phase loop when RF is off. Beam control clock generators: A resynchronisation of the 40MHz sampling clock had been left out of the synchro initialisation sequence. This clock is used for sampling of the beam phase measurement. It comes up with random phase if not properly resynchronised. A sequencer task was added to resynchronise the 40 MHz to f rev. Additional diagnostics were installed (TDC measurements) in the beam control VME crates (f rev master vs f rev prog) and in the synchro system ( cogging between f rev prog 1 and f rev prog 2).
cavities are connected to the master reference before injection. Revolution frequency synchronisation interlock The TDC measurements of the f rev synchronisation (see the previous section) should be interlocked to prevent injection if a desynchronisation is detected. Fig. 2. Synchro loop error (blue) during an 855Hz trim. The excursion of 150 degrees causes the loop to unlock. Fig. 3. Synchro loop unlocking at injection READINESS FOR UNSAFE BEAM RF voltage interlock No beam interlocks on RF voltage or RF trips are currently connected. At half nominal intensity, the beam can survive 1 cavity trip [3], but ultimately we will need an interlock on the RF switch for each cavity controller. This is currently not implemented, but for now, we will add an interlock on the cavity voltage sum to generate a beam dump if the total RF voltage is below threshold. RF frequency interlocks A coarse interlock on the cavity controllers 40MHz clock (which corresponds to +/- 10 khz at the RF frequency) is currently active. An additional fine interlock (+/- 200Hz at the RF frequency) will be installed. The frequency will be measured behind the distribution point in UX45 every 50ms and compared with a central frequency derived from the beam energy distributed via the Safe Machine Parameters system. This system will also allow a ring1/ring2 inversion check before each fill, by trimming the frequency on each ring and measuring the frequencies of the RF in point 4 and the revolution frequencies at the beam dump in point 6. A software interlock should ideally be implemented on the cavities remote/local clock status to ensure the LOW LEVEL RF DEVELOPMENTS FOR 2010 Cavity controller Variable coupler: The use of the variable power coupler for changing the cavity Q will be commissioned before the startup. 1-turn feedback: The 1-turn feedback will be put into operation during 2010, giving a further impedance reduction for increased intensity [4]. This should be transparent for operation. Beam control Injection transients in the phase and synchro loop error signals: A fixed display of these signals is already available in the CCC. An automatic adjustment procedure for the injection phase could be developed based on these signals. A reliable adjustment from injection to injection would reduce the need for longitudinal feedback. Longitudinal feedback: The longitudinal feedback gives bunch by bunch damping of injection errors [4]. Firmware development is still needed and the system is foreseen to be installed in the second half of 2010. Optical fibre cabling is in progress, and some short accesses will be needed after the startup for equipment installation. Emittance blowup by controlled noise: The hardware and firmware is in place, and the software development is in progress. Some MDs will be needed to commission the system. TRANSVERSE DAMPER: READINESS FOR 2010 RUNNIN NG accumulated more The ADT power system has now than 2000 hours of operating time. The damper Low Level hardware was already complete and available in 2009. The feedback has not yet been closed, since no time was yet allocated for commissioning. The application software for the power system has been tested and used operationally. The Low Level software (function generators, expert software, abort gap cleaning) has been tested in dry-runs and during beam operation. The abort gap cleaning functionality has been tested with beam. The software and firmware to generate the cleaning pulses worked very well, and data analysis is ongoing. Further optimization of the cleaning pulse will
be necessary in order not to touch the beam outside the abort gap. Noise issues A reduction of beam lifetime by unwanted excitation through the damper has been observed. The influence of the damper level 3 (signal drive) on/off status on the beam lifetime and beam spectra has been observed by R. Steinhagen [5]. Signal level (dbm) -60-70 -80-90 -100-110 -120 0 200 400 600 800 1000 Frequency (khz) BI input enabled BI input disabled Fig. 4. Measured noise injected from the BI excitation input confirmed in 2010 when the SR4 and UX45 UPSs were switched off. The mechanism by which the 8 khz perturbation enters the damper equipment is thought to be via ground loops between the surface and underground areas. The grounding scheme needs to be looked at, and possibly improved. For 2010 a scheme to suppress the noise is to install common mode chokes at the input of the damper equipment in UX45. Using this scheme a 10 db suppression has already been shown (Fig. 5). Chokes with more windings will be constructed for better suppression. If the problem persists, another solution would be to move some equipment from the surface to UX45, which would be a considerable investment in time, and only justifiable if it is really proven necessary. Pickup cable problems There is an issue with the long cables running from the pick-ups to SR4. 7/8 Flexwell cable is known to have some ripple in the impulse response due to the corrugation. However, the 16 ADT pick up cables have a larger than expected variation of the ripple from cable to cable when excited by the bunch passing the pick up (Fig. 6). A periodic perturbation is visible in reflectometry measurements, with a peak at around 200Mhz. This results in crosstalk between the measurements of adjacent bunches. It is thought that the degradation is due to cable damage during installation. The worst two cables will be replaced. The performance for multi-bunch operation will need to be checked after the 2010 startup. Fig. 5. Reduction of damper noise at 8 khz by common mode choke in UX45 Investigations of the noise spectra at the damper equipment by the RF group led to the identification of different contributions: 1. A contribution from the Beam Instrumentation input signal used for tune measurement, which is found to have a triangular shaped noise spectrum from 50 khz to 250 khz. A solution to suppress this using filtering is underway for 2010 from the BI group. The fix in 2009 was to attenuate the excitation signal (Fig. 4). 2. Lines at 8 khz and multiples were identified as coming from the UPS power supplies. This has been Fig. 6. Signal from single bunch as measured on the surface; the large ripple is due to a periodic perturbation on the cable RF CONCLUSIONS Minor hardware problems with the power system have been understood and addressed. However there is a possible concern over the cooling of the klystron collectors, and the klystron operating mode for 2010
needs to be defined. Time will be needed to commission the variable cavity Q. Some serious operational problems were encountered with front-end software for the power system, and these have now been solved. Various causes for synchronisation problems have been understood and resolved. In order to be ready for unsafe beam, a number of interlocks will be added for the total RF voltage, RF frequency and the revolution frequency synchronisation. A number of developments are still outstanding in the Low Level RF, including the 1-turn feedback, longitudinal feedback and longitudinal emittance blowup. ADT Commissioning of the ADT system with beam has started, and will need dedicated time in 2010. The noise spectrum needs attention. Measures have been taken to remove ground loops by common mode chokes, but improvement of the grounding between the surface and underground areas may be necessary. Two pick-up cables (7/8 Flexwell) will be replaced during the technical stop, and performance for multibunch operation will need to be checked due to residual ripple from the cables. Abort gap cleaning has been shown to be promising, but further optimization of the pulse shape will be required. REFERENCES [1] Hardware and Initial Beam Commissioning of the LHC RF Systems, T. Linnecar et al., LHC Project Report 1172, CERN, Geneva, 2008 [2] Controls issues, Wojtek Sliwinski, these proceedings. [3] Consequences of RF system failures during LHC beam commissioning, T. Linnecar, LHC Project Workshop, CERN, Geneva, 2005 [4] The LHC Low Level RF, P. Baudrenghien et al., European Particle Accelerator Conference, EPAC 2006, Edinburgh, UK. [5] Tune, chromaticity, feed forward and feed back, R. Steinhagen, these proceedings.