1. Rev 0.3: Aug 7, 20061 Notes from TRISTAN Experiences N.Toge (KEK) for GDE RDB S2 Rev.0.3 Aug. 7, 2006

2. Rev 0.3: Aug 7, 20062 References Experience on the Superconducting RF System in TRISTAN, S.Noguchi, Particle Accelerators, 1996, vol.54 pp [325-336]/25-36. Couplers - Experience at KEK, S.Noguchi et al, Proc. 4th Workshop on RF Superconductivity, ed. Y.Kojima, KEK Report 89-21 (Jan. 1990). –Both posted at http://www.linearcollider.org/wiki/doku.php?id=rdb:rdb _external:rdb_s2_home http://www.linearcollider.org/wiki/doku.php?id=rdb:rdb _external:rdb_s2_home –This report mostly concerns the first write-up only.

3. Rev 0.3: Aug 7, 20063 TRISTAN SCRF System Basic Unit –Each RF Station (508MHz) drives 2 Cryostats, Each CM contains 2 units of 5-cell Cavities with Eacc ~ 3.6MV/m. –So, 1RF = 2 CMs = 4 Cavs TRISTAN System –8 RF Stations were operated in its final installation, –So, in total, TRISTAN: 8RFs = 16 CMs = 32 Cavs = 160 Cells –Compare this with an ILC unit, where 1RF = 3 CMs = 24 Cavs = 216 cells. –Beware, TRISTAN is CW (so is ILC DR); ILC ML is pulsed. Operation Period –Cold (4.4K) Time = 33,000 hrs (~ 1375 d) over 7 yrs –13 heat cycles –Beam Time ~ 870 d ~ 20,900 hrs –Total Cavity Beam Time ~ 20,900 d ~ 556,100 hrs

4. Rev 0.3: Aug 7, 20064 Failure Statistics 19881989199019911992199319941995 HOM4 IC Leak Ceramic Arc 2 1 Polyethyl ene 11 Water Leak 21 Cavity Leak 1 32 1 Piezo Tuner 1365121 # of CMs continuously opearted for 6-7yrs = 7 # of CMs that had to be serviced in one way or others = 9

5. Rev 0.3: Aug 7, 20065 Failure Issue 1 : HOM Design revision was found necessary for the connectors. –Case A: Abnormal heating due to poor pin contact –Case B: Quenched HOM coupler allowed the fundamental mode to leak. All units were replaced soon afterwards. Lesson: –We need testing of all components in realistic operating (power) conditions, prior to design freeze. Comment: –Might be less acute for ILC ML whose RF is pulsed. –Same precaution is mandatory for ILC DR whose RF is CW.

6. Rev 0.3: Aug 7, 20066 Failure Issue 2: Input Coupler Type 1: Leaks at ceramic windows –Causes included cracking; arching from a coupler port flange; and unidentified reasons. Type 2: Burning of polyethylene backup disk leading to a window leak –Polyethylene in use, in stead of teflon, for fear of radiation damage was slowly degrading, although catalog implied OK. Went back to teflon. Type 3: Water leak from an outer cooling jacket of the window –Lucky to have them happen when warm (potentially a disaster scenario) Lesson: –Pay attention to long-term use of seemingly innocuous components in realistic operation conditions. Do not simply trust the catalog specifications. –Avoid use of water (ILC ML SCRF will not use LCW).

7. Rev 0.3: Aug 7, 20067 Failure Issue 3: Cavity Leak at Indium Joints This failure mode occurred only after long-term use –Indium joints got loosened after going through thermal cycles. –Design improvement applied. Lesson: –Look hard for the optimum flange joint design. –Pay attention to effects of thermal cycles. –If possible, try thermal cycle testing before design freeze.

8. Rev 0.3: Aug 7, 20068 Failure Issue 4: Piezo Tuners Short MTTR (since the drivers placed outside CM) but short MTBF. Radiation damage of plastic bolts pushing the tuner disks. No major issues of heat load due to the tuner components being outside CM (for TRISTAN) Lesson: –Radiation damage stability of components. Remark: –Need dedicated reliability study for ILC, since HW implementation could to be substantially different. Illustration from KEK Preprint 89-184

9. Rev 0.3: Aug 7, 20069 Discussion (1) Quite educational pieces of data. BUT, notice that if the SCRF system gives failure rates similar to what was observed at TRISTAN, operation of ILC will be in a serious trouble. As indicated by the KEKB SCRF performance, situation might be actually better if “TRISTAN lessons are faithfully respected”, or similar learning period is specifically allocated for ILC SCRF systems. We see, overall, components that can be tried out before design freeze and production launch should be tried out in realistic stress testing conditions thoroughly. Parameters to pay attention to in test environments : –Power –Mechanical –Thermal cycles –Radiation

10. Rev 0.3: Aug 7, 200610 Discussion (2) In terms of the # of HW components x time, 7 yrs operation at TRISTAN, in a very crude approximation, corresponds to ~1 ILC unit (1RF = 3CMs, etc) for a similar period. So, to time-compress the testing period, one MIGHT imagine –2 yrs operation of 3 ILC units (9CMs) –1 yr operation of 6 ILC units (18CMs) could give us operation experiences of a similar magnitude. However, we must be aware of the danger of simple scaling. –Some failure modes, before showing up, require a certain operation time on each individual component, rather than just a big total accumulated operation time. –Seeing a failure mode is one thing. Understanding it and developing a cure is another (i.e. takes time). We should repeat similar analysis of colleagues’ experiences from KEKB, CEBAF, LEP, SNS, TTF. E N D