Presentation on theme: "Efficient RF sources for Linear Accelerators"— Presentation transcript:
1Efficient RF sources for Linear Accelerators Dr Chris Lingwood
2MotivationFrom CLIC CDR (2012)LARGE numbers of RF sources are required for future linear colliders.According to CDR CLIC requires 15 MWSupply large amount of power at affordable cost (high efficiency)Current state of the art15 MW klystrons can achieve 65% efficiency
3CLIC MBK StudyCollaboration with CERN and Thales (Erk Jensen, Igor Syratchev, Phillipe Thouvenin, Rodohple Marchesin).Efficiency as main targetEvaluated configuration options, multiple beam klystronTargeted a conservative (plausible) designTargeted TESLA/ILC specificationTheoretical efficiency: 80% (beyond state of the art)
4Why many beams? Low perveance leads to higher efficiency. Low current -> lower space charge forces -> better bunching -> higher efficiency20 beams – trade off between beam voltage and complexity due to beams𝑝𝑒𝑟𝑣𝑒𝑎𝑛𝑐𝑒 𝐾= 𝐼 𝑉 3/2Ib = 8.2AVb = 115V
5Cavity choices Comparison of multiple cavity types. 1. Re-entrant2. Recessed Re-entrantTM 0 1Comparison of multiple cavity types.Re-entrant & HOM cavities -> Low R/QRecessed re-entrant and coax cavity ->high R/QTM 10 1TM 0 13. & 4. Coaxial Cavity5. Whispering Gallery
6Interaction structure Optimised 6 cavity(single 2nd harmonic)Low R/Q structure 70%High R/Q structure 20 beam structure up to 80%
7OptimisationDeveloped and published a new way to design klystron amplifiers:~14 Decisions (frequencies, drifts, Qe’s)3-4 objectives (efficiency, length,bandwidth, slowest electron,000 evaluationsNovel publishable optimisation concepts(recombination operator).Impractical without high throughput computing (CI HTCondor Pool)Use spare clock cycles of desktop pcs
8So it’s all done then? Conservative approach lead to complex tube Many, many, many beamsPush the voltage (always the plan)Don’t rule out newer techniquesBe braver on layoutFundamentally: do you want 50MW in an MBK?
9Cavity HOMs Can model coaxial cavity as a piece of ridged waveguide CPI(estimated)OursQuadrupoleCavity radiusNormalised frequencyCan model coaxial cavity as a piece of ridged waveguideVery good agreement even for HOMsRLarge diameter (35cm) at 15MWMore power -> more beams -> larger stillDipole mode gets closer for larger cavities
10Current Collaboration with CERN Working with I. Syratchev, C. MarelliAttempting to formalise empirical relationship between efficiency and perveance
11Proposed task Many questions still surround the RF sources Requirements still push state of the artPower, efficiency, cost, lifetimeEvaluate options (SBK, MBK, MBIOT…)Solution is probably klystronsConfiguration (MBK/SBK)Work towards helping CERN become the intelligent customer
12Maximising Efficiency Tight bunching isn’t enoughThe key to higher efficiency is:slow all your electrons down as much as possible in the output gap without stopping or reflecting themThis isn’t trivialPotential improvementsLow frequency penultimate cavitiesTravelling wave output structures
13Higher harmonic cavities To get the tightest bunch from a single cavitysawtooth waveform (includes high harmonics)2nd harmonic cavities well understoodWill 3rd or higher harmonic cavities help more?Effect on bandwidth?Effect on velocity spread?
14Reduce velocity spread IVEC 2013Detune penultimate cavity to achieve π phase changeWhen phase change is good, coupling to the beam is badTwo gap cavity can help with controlTested in 63 W C-band tube, increase efficiency by 8%, 25% reduced voltage.Does this approach scale to MW?
15Klystron Configuration Some interesting configuration options under consideration.Some beyond state of the artSome just beyond ….
16New softwareA number of klystron specific codes exist. Only one generally availableAJDiskFor instance cavity voltages can be unreliable disagree as much as 50% between GdfidL and AJDisk and 1500% (!) with klys2D (Thales).Closed source so difficult to identify the issueAlso difficult to integrate with other codesPropose to develop new “open” disk model code for klystron research from existing code at Lancaster.
17PIC Simulations Simplified models can only get us so far Detailed verification of designs to demonstrate improvements1 week simulation time for klystron 1GHz up to 10 μs8 coresScope for improvement with HPC (available at Hartree Centre, Sci-Tech Daresbury)Careful benchmarking of code (V-SIM) against MAGIC
18MB-IOTs for Linear Accelerators Reduce power lost to collector by switch beamShort “excursions” above rated power allowable.Very useful when high efficiency and variable output power are needed.Even better if “headroom” is neededIOTs exist up to about 100kWNot a great deal in the context of proposed large LINACsJust like klystron MBK -> MB-IOT10 beams -> 1MW, more plausible.ESS seriously considering.Road block - gunsWorth doing the sums.KlystronIOTFigures from IOT based High Power Amplifiers, Morten Jensen, TIARA Workshop on RF Power Generation for Accelerators – Uppsala June 2013
19Magnetrons Injection locked magnetrons to achieve phase control using Small, cheap, efficient, low voltagerun at saturationPerhaps lock with IOTProbably not costeffective at CLIC scale
20Milestones2014 Investigate suitability of single and multi-beam klystrons, MBIOTs, magnetrons.2014 Produce a bunched beam vacuum tube model. Transcode and improve existing code to interface with PIC codes for output cavity. Benchmark against existing codes.2015 Evaluate new and existing techniques to improve efficiency.2015 Benchmark V-SIM and CST against MAGIC for bunched vacuum tubes.2016 Design most appropriate tube and validate using PIC.
21Deliverables 2014: Tube model complete and open sourced 2015: Design of tube interaction structure for drive beam - report2016: PIC simulations and verification of proposed interaction structure - report
22Financial Staff (Lingwood) 1 3 RA3 6+6 18+18 Student (TBD) 12 36 Materials5+74+50+09+12Travel1+23+61 RA and 1 Phd student for three years
23Summary Re-evaluate options for RF sources Push efficiency and plausibilityScope for improvement in simulation timesDevelop more flexible and open klystron simulation tools.Produce candidate structure using lessons learned.