Breakdown position analysis Alberto DEGIOVANNI W. Farabolini, J. Giner-Navarro, L. Navarro, R. Rajamaki, B. Woolley High Gradient Workshop 2015 Tsinghua University Beijing, 16-19 June 2015
A. Degiovanni - HG2015 workshop - Tsinghua University, China Outline Motivation BD cell location methods: Transmitted and Reflected signals delay Incident and Reflected signals delay Shock wave propagation (accelerometers) Data analysis T24 dogleg Results from Crab cavity Conclusion 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
A. Degiovanni - HG2015 workshop - Tsinghua University, China Motivation and goals BD position analysis is important for: better understanding structures performance monitoring progress in conditioning process extract information to study more in details BD phenomena (temporal-spatial correlation of subsequent BD) GOAL: Reconstruct position of BD events inside the structure Study evolution of BD Long term (accumulated BD in structure) During pulse (BD onset time and drifting) 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
Methods for BD cell location based on RF signals 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
BD Location based on RF signals Reflected rising edge Transmitted falling edge NOTE: the 1st method detects the BD location at its onset, the 2nd method at the end of the RF pulse (meanwhile the BD can moved). Pros: accurate determination by signals correlation (no thresholds used for edges determination). Cons: requires a time structure at the end of the incident pulse (not always present). 1st method: EDGE detection By Dt between Reflected rising edge and Transmitted falling edge (BD start) Incident power falling edge Reflected falling edge 2nd method: CORRELATION method By Dt between Input falling edge and Reflected falling edge (echo) (BD end) W. Farabolini 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
BD Location based on RF signals 1. Delay between rising edge of REF signal and falling edge of TRA signal 2. Delay that maximizes correlation between INC and REF signal in the tail 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
BD Location based on RF signals 1. Delay between rising edge of REF signal (t1REF) and falling edge of TRA signal (t2TRA) 2. Delay that maximizes correlation between INC and REF signal in the tail (t2REF - t2INC) t1REF t2TRA tfill ΔtREF = t2REF -t2INC t2INC t2REF 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
BD Location based on RF signals 1. Delay between rising edge of REF signal (t1REF) and falling edge of TRA signal (t2TRA) 2. Delay that maximizes correlation between INC and REF signal in the tail (t2REF - t2INC) Time delays produced by both estimators can be made comparable: t1REF t2TRA tfill ΔtREF = t2REF -t2INC t2INC t2REF t2TRA – tfill = t1REF – ΔtREF 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
A. Degiovanni - HG2015 workshop - Tsinghua University, China Comment Time delays produced by correlation and edge detection methods can be directly compared tfill x ΔtREF o t2TRA – t1REF o t1REF – t2TRA t2TRA – tfill = t1REF – ΔtREF t2TRA – t1REF = tfill – ΔtREF t1REF – t2TRA = ΔtREF – tfill 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
A. Degiovanni - HG2015 workshop - Tsinghua University, China Edge detection Edge detection in two stages: 1. threshold detection t2TRA: 90%of flat top avg. t1REF: 10% of flat top avg. 2. detection of peaks in the derivative of the signals (with a 5 samples moving average to avoid spikes detection) Comparison with previous pulse also considered: TRA(n)/TRA(n-1) REF(n-1)/REF(n) Power INC TRA REF Log detectors signals 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
Correlation method details A time slot of the signals is determined after the end of the klystron pulse. Replicas of the REF signal with various delays and amplitude are correlated with the INC signal. Correlation peak (delay and amplitude) is determined, with confidence estimators (the sharpness of the peak) and integrated error between the superposed signals. 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
Data from T24 dogleg experiment BD index t1REF – t2TRA [ns] Filter for good BD in range -100 ns to 400 ns Good BDs: 14441 / 15835 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
Data from T24 dogleg experiment 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
Full history - time delays Edge detection method Correlation method (evaluated in the tail of the pulse) 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
From measured delays to cell location 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
From measured delays to cell location 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
Full history – BD cell location evolution 8 0 ns 120 136 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
A. Degiovanni - HG2015 workshop - Tsinghua University, China Aug 2014 (unloaded) 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
A. Degiovanni - HG2015 workshop - Tsinghua University, China Apr 2015 (unloaded) 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
A. Degiovanni - HG2015 workshop - Tsinghua University, China BD drifting Phase signal jumps Pattern in the tail of the pulse moving 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
Early BD in structure + BD in waveguide TRA signal drops REF rises INC signal drops 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
Possible explanation of the BD drift TRA BD ignition. Plasma development. Wave absorption . TRA falls 1. REF Plasma density sufficient to reflect the INC wave. REF raises 2. REF Another BD appears upstream. The second reflected wave combines with the first one in power and phase and can cause destructive interference. 3. REF The upstream BD plasma density is sufficient to reflect the INC wave. REF raises again but with a different phase and higher amplitude. 4. Power is likely to be absorbed before it is reflected. This is impairing the accuracy of BD location methods based on REFup – TRAdown thresholds W. Farabolini 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
Shock wave propagation 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
Measurement setup on crab cavity in Xbox2 RF in RF out N. Catalan Accelerometer 6 accelerometers with dynamic range ± 500 g #1 cell 3 0 deg. (top of the structure) #2 cell 3 45 deg. #3 cell 3 90 deg. (cooling block) #4 cell 3 180 deg. #5 cell 7 0 deg. #6 cell 10 0 deg. 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
A. Degiovanni - HG2015 workshop - Tsinghua University, China 1. BD at RF input Strong shock on the cell #3 accelerometers (300 g) Same phase for all polar angle position (circle inflates) Slight delay on accelerometer #3 due to cooling block thickness Shock propagates to accelerometers #5 and then #6 Lower signal than regular RF pulse on downstream accelerometers since the BD has interrupted a part of the RF travelling wave 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
A. Degiovanni - HG2015 workshop - Tsinghua University, China 2. BD at RF output Strong shock on the cell #10 accelerometer (200 g) Shock propagates towards upstream accelerometers Higher signal than regular RF pulse on upstream accelerometers since the BD has reflected a part of the RF travelling wave 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
3. BD in the middle of the structure Shock first and larger on the cell #7 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
4. BD before the structure All accelerometers amplitudes lower than regular RF pulse Same waveform shape as for regular RF pulse 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
A. Degiovanni - HG2015 workshop - Tsinghua University, China Spectrum analysis There is a common frequency line for accelerometers at 45.3 kHz Otherwise the spectrums of both accelerometers are slightly different (Eigen-modes ?) Mechanical characterisation of the structure should be done (mode analysis, measure with mechanical excitation) 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
A. Degiovanni - HG2015 workshop - Tsinghua University, China Summary BD position analysis gives a better picture of BD phenomena in high gradient structures and help in understanding the evolution Different methods have been presented based both on RF signals and on accelerometers signals Accelerometers can deliver a clear signature of BD, and delays between array of accelerometers can be used to locate BD position (eventually 1 sensor per cell ?) There are hints of BD drifting through the structure during BD pulse Further analysis of the data is needed 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
THANK YOU FOR YOUR ATTENTION ! 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
A. Degiovanni - HG2015 workshop - Tsinghua University, China EXTRA SLIDES 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
A. Degiovanni - HG2015 workshop - Tsinghua University, China BD INTERLOCKS REF BD KREF BD TRA en REF en 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
A. Degiovanni - HG2015 workshop - Tsinghua University, China 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
A. Degiovanni - HG2015 workshop - Tsinghua University, China 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China
A. Degiovanni - HG2015 workshop - Tsinghua University, China 3 in a row 19/06/15 A. Degiovanni - HG2015 workshop - Tsinghua University, China