1. RF SYSTEM FOR THE DRIVE BEAM LINAC Erk Jensen BE-RF TDR Task force meeting 21-April-2009

2. Reminder: What I was asked to do: 2 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

3. What had been done before: Thales study CA1174974 In 2001, Thales had made a study on an MBK for this same purpose, 937 MHz, 50 MW · 100 μs, 50 Hz. MBK’s with 6 and 7 beamlets were studied; 7 beamlets (6 on the circumference and 1 central) were excluded (e- reflection!). Result: With 215 kV, 284 A: Pout ≈ 40 MW, η ≈ 65%. Size: 5050 mm. Estimated development time: 30 months from order (would require 3 prototypes; PT1 after 16 months, PT2 after 28 months). This is reported in CERN CA 1174974 reports RT5213T and RT5141T. Cost estimate: I don’t know. 3 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

4. Reminder: what does CLIC need? Over the last few years, parameters have changed quite a bit: the drive beam accelerator frequency changed from 937 MHz to 1333 MHz in Nov 2006 to 999.52 MHz in Sep 2007. The total peak RF power required (per linac) is about 12 GW (from 4.21 A · 2.38 GV / 93.5% / 91.2%). With a rep. rate of 50 Hz and an RF pulse length of – say – 150 μs (total CLIC length/c), we get: Duty factor 150 μs · 50 Hz = 0.75 %, average power 90 MW! *) Of major importance for the RF power source in the specifications are the phase stability, the power conversion efficiency. *) This is consistent with the average beam power of 70 MW in Fig. 30 of the CLIC 2008 Parameters.CLIC 2008 Parameters 4 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

5. Average powers from CLIC 2008 Parameters 5 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

6. Klystrons or something else? 6 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

7. Existing ILC 1.3 GHz MBK’s (10MW, 1.5 ms, 10 Hz) 1.CPI: VKL-8301B (6 beam): 10.2 MW, 66.3 %, 49.3 dB gain (to be published)VKL-8301B 2.Thales: TH 1801 (7 beam): 10.1 MW, 63%, 48 dB gainTH 1801 3.Toshiba: E3736 (6 beam): 10.4 MW, 66 %, 49 dB gain 1. 2. 3. 7 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

8. Existing idea: Cf. Jensen, Syratchev: “CLIC 50 MW L-Band Multi-Beam Klystron”, CLIC-Note-640Jensen, Syratchev: “CLIC 50 MW L-Band Multi-Beam Klystron”, CLIC-Note-640 The main idea: use a mode like the one depicted as #4 (whispering gallery mode) for many beams; the advantage of this mode: It can be made very pure! The problem: This device became really big: how do you braze this? Imagine a little problem in one of the 27 beams! There is a study ongoing in collaboration with Thales and Lancaster University; Chris Lingwood is finishing his PhD on this. His present conclusions are not supporting our claims though! 8 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

9. What’s in the Parameter list Given the total peak power, it had been assumed that 33 MW peak could be made available at the input of each accelerating structure. This resulted in 326 klystrons and 326 accelerating structures per linac. The accelerating structures were scaled from the existing 3 GHz structures (  results in a beam pipe diameter of 102 mm, which seems excessive…). The number of cells is then adjusted to be fully beam-loaded for the nominal current and power. Keeping the beam current at its nominal value of 4.21 A, here the input power requirements for structures with different cell numbers: 9 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

10. Input power for full beam loading for different cell numbers # cellsP in [MW]  length [m]τ [ns]acc [MV]# structtotal length [km]P tot [GW] 156.3398.81%1.500142.01.4916022.40210.139 167.2198.73%1.600151.41.6914072.25110.146 178.1498.66%1.700160.81.9112472.11910.154 189.1298.58%1.800170.22.1411142.00510.162 1910.298.51%1.900179.52.3810001.90010.170 2011.398.43%2.000188.92.649011.80210.150 2112.498.36%2.100198.32.908211.72410.185 2213.698.28%2.200207.73.187491.64710.193 2314.998.21%2.300217.13.486851.57510.201 2416.298.13%2.399226.53.786301.51210.208 2517.698.06%2.499235.94.105801.45010.216 2619.197.98%2.599245.34.455351.39110.224 2720.697.91%2.699254.74.794971.34210.232 2822.197.84%2.799264.05.144631.29610.239 2923.797.76%2.899273.45.504321.25310.247 3025.497.69%2.999282.85.894041.21210.255 3127.197.61%3.099292.26.283791.17510.263 3228.997.54%3.199301.66.703551.13610.270 3330.897.47%3.299311.07.133341.10210.278 3432.797.39%3.399320.47.573151.07110.286 3534.697.32%3.499329.88.002981.04310.294 3636.697.25%3.599339.28.462811.01110.302 3738.797.17%3.699348.68.932660.98410.309 3840.897.10%3.799358.09.412530.96110.317 3943.097.03%3.899367.39.912400.93610.325 4045.396.95%3.999376.710.432280.91210.333 ≈ 28.3 kW · n cell 2 present “nominal” 10 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

11. Conclusion (from this table) The number of cells can be adapted to the available RF power. Shorter accelerating structures are even more efficient (less ohmic losses – small effect) With shorter structures, the linac overall length gets large. These are preliminary structures, just scaled from CTF3. As a consequence, the beam aperture is a factor 3 larger than at 3 GHz (34 mm -> 102 mm, while the beam current is only 20 % larger (3.5 -> 4.2 A). My guess would be that an aperture of 60 mm should be fine. This would increase the impedance and make the linac shorter again.  Action: get data from ABP and redesign! 11 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

12. Efficiency The CLIC 2008 Parameters assume a tube efficiency of 70%, existing tubes reach 66%.CLIC 2008 Parameters It is generally accepted that maximum obtainable efficiency is a function of the perveance I/V 3/2. Using an empirical model, here is what one could expect (numbers for 13 MW DC): For practical reasons, the voltage should be kept low (say below 140 kV). To limit the complexity, the number of beamlets should remain reasonable. I marked a point which I find interesting: 12 beams, 140 kV; it could reach above 70%. klystron DC voltage [V] MBK # of beamlets 84 % 82 % 80 % 78 % 70 % ILC klystrons 140 kV, 12 beams 12 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

13. How does the cost of a klystrons scale with peak power? Probably: cost per klystron proportional to (peak power) 1/2 (*) At a level of around 15 MW peak, the slope will become steeper due to increased system complexity. This leads to the following model: Blue: present state of the art Red: assuming a major investment into the development of a dedicated 30 MW tube (*) rule of thumb given by T. Habermann/CPI. Rees/LANL estimates P 0.2 for 0.5 to 5 MW tubes. 13 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

14. Cost per MW Using the above model, here’s the klystron cost per MW (peak) Blue: present state of the art Red: assuming a major investment into the development of a dedicated 30 MW tube 14 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

15. Tube lifetime In spite of its price, a klystron is a consumable! A klystron has a finite lifetime; this will also depend on its internal complexity (and on the peak power!). The lifetime will depend on many parameters, primarily the current density, but here’s one estimate... 15 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac What about an MBK?: is the tube dead if one of n beams fails? If the design is good, the n beams would fail at around the same time...

16. Cost for 100,000 operating hours and MW Even if this model may be wrong, there will be a cost per MW and per operating hour: With the above model, this becomes: Blue: present state of the art Red: assuming a major investment into the development of a dedicated 30 MW tube 16 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

17. Conclusion (from the cost per tube per operating hour) The lifetime model presented here may be wrong; the scaling for the unit cost may be wrong, but for a correct cost estimate, both these influences must be included. Assuming that the models used above are somehow reasonable, the optimum size of an individual tube would be not significantly above 10 MW. This conclusion may change depending on a better model. It may also change after dedicated R&D, but in my opinion this R&D should rather address the reliability, cost and lifetime than the peak power. I cannot comprehend the extrapolation from 33 MW to 78 MW klystrons presented in the parameter meeting on June 5 th, 2007 by Hans. Page 6 of http://indico.cern.ch/conferenceDisplay.py?confId=16292. With all due respect a personal remark: Would CLIC management please reconsider the presently applied secrecy policy of the CLIC Cost WG? I think it would help progress of the project and the people concerned to have some transparency! 17 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

18. Bigger is better? I’m not the only one who believes that bigger is not always better....even though I believe that this is too small 10 beam MBK 18 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

19. Concerning modulators R&D is going on for ILC, SPL,... we should take full advantage! Our CERN modulator experts explain: A classical “bouncer” type modulator for a size of 12 kV, 2 kA can be considered feasible. It would look like this (just the topology, picture taken from ILC): A larger modulator would combine a number of these; it’s cost would scale at best linearly with peak power – the “modular modulator” – no saving from making it bigger. This (20 MW peak or so) seems to be the natural module size. A modulator with 3 modules would cost around 1 MCHF. The numbers given here would be consistent with a 15 MW MBK. 12 kV, 2 kA 140 kV, 160 A 19 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

20. Modulator Commercial modulator 20 MW, average power around a factor 10 too small. This is some really big object! HVPS and pulse forming unit: IGCT stack: Pulse transformer: Some ILC examples: You would need 1 of those every 2 to 3 m for the total length of the DBL! ILC estimate: 300... 400 k$/unit 20 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

21. Conclusion (modulators) Base line: bouncer type modulator is quasi “commercial” 12 kV/2kA is a natural module size (24 MW DC); pulse transformer 12:1,  140 kV/160 A Larger modulators of this type would not be cheaper per MW. Modulators of other types require R&D! With a 70% efficient klystron, this would correspond to 15 MW RF. TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac21 21/Apr/2009

22. My conclusions 1.For CDR, stay with 10 MW MBK’s – we know (from ILC) that they can be done. Reducing pulse width is trivial – but extrapolating to higher peak power at the same time requires R&D. 2.  Concentrate R&D on a modular RF system with peak powers of 10...15 MW peak, addressing – in addition to the RF parameters – cost, reliability, tube lifetime, serviceability, graceful degradation. 3.  Include the modulator in this design. 4.So – some of this R&D is required for CDR, but most for TDR. 5.For reference,  re-evaluate the potential of SBK’s and PBK’s! 6.The numbers presented above for cost scaling and MTBF are the result of some emails, telephone calls and google searches; I believe however that they indicate which way to go...  One should dig deeper and improve the simplified models I’ve used – maybe this will even change the conclusions I’ve made! 7.  Re-adapt the beam pipe diameter of the accelerating structure for higher impedance to stay below say 1 km. (Considering the probable size of the modulators, this may not help too much) and finally: 22 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac

23. What can happen if you set your target too high: An experience communicated by Dan Rees, LANL & SNS, with their 5 MW peak, 400 kW average(!) 805 MHz tube. “…When we bid the 5 MW, 805 MHz tube, Thales was the low bidder by more than $200 k per tube but again they were non compliant because they couldn't factory test the tube to the full peak and average power. The cost savings per tube was so profound that we changed the specification to allow for partial factory testing and then rebid. Thales then won. The tube development for the 5 MW klystron went on to be a disaster. …” Thank you very much! 23 21/Apr/2009 TDR task force meeting: Erk Jensen - RF system for the Drive Beam Linac