Transcript TBTS data analysis

Comparison of breakdown
behavior between klystron and
beam driven structure
W. Farabolini
With the support of J. Kovermann,
B. Woolley, J. Tagg
HG2013 3-6 may 2013 Trieste
W. Farabolini
1
Contents
•
•
•
•
•
•
•
Main test characteristics of TBTS vs. X-Box 1
BD locations
BD precursor research
BDR as function of RF power
BD distribution within time
BD ignition and transmitted RF falling time
Structure RF analysis after removal
HG2013 3-6 may 2013 Trieste
W. Farabolini
2
Typical RF signals
Drive beam generated with PETS
Klystron generated with pulse compressor
• Pre-pulse
• Triangular shape (recirculation)
• Quite stable pulse 24/7
• Often instable pulse (and trips)
• Great flexibility in pulse
• Pulse length and power not really
length and power
flexible
After-pulse in case of BD: reflected power perturbation on RF generator
HG2013 3-6 may 2013 Trieste
W. Farabolini
3
Stability of the RF power
Many
beam
trips
Two Beam Test Stand : beam generated power with RF recirculation
Energy
reduction
after BD
detection
X-Box1 : Klystron generated power with pulse compression
HG2013 3-6 may 2013 Trieste
W. Farabolini
4
Data production
• Total number of RF pulses
– ACS 1 in TBTS
about 3 millions (0.8 Hz repetition rate)
– T24 :
over 98 millions (50 Hz repetition rate)
– TD24R05 :
over 144 millions (4.3 millions per day max )
• Total number of BDs
– ACS 1 in TBTS
about 10000 (?) (10-2 < BDR < 10-3)
– T24 :
3502 (BDR = 3.6 10-5)
– TD24R05 :
7278 (BDR = 5.0 10-5)
• Total number of 8 hours data log (about 40 Mbit each) processed
– ACS1 in TBTS : few 10’
– T24 :
116
– TD24R05 :
228
HG2013 3-6 may 2013 Trieste
W. Farabolini
5
T24 test condition summary
Power ramping
Pulse length to
keep BDR around
10-5
Conditioning not
achieved
HG2013 3-6 may 2013 Trieste
W. Farabolini
6
TD24R05 test condition summary
Power and pulse length
ramping strategy. (limit
the available energy in
case of BD)
Full gradient 100 MV/m
and pulse length 220 ns
achieved with BDR = 10-5
HG2013 3-6 may 2013 Trieste
W. Farabolini
7
BD location determination
Reflected
rising edge
time
Transmitted
falling edge
time
1st method
Input
falling edge
Dt between Reflected rising edge and
Transmitted falling edge
(BD start)
time
Reflected
falling edge
time
2nd method
(echo)
Dt (correlation) between Input falling
edge and Reflected falling edge
(BD end)
•
•
Edge detection is always tricky especially for the transmitted signal (BD ignition time)
Cross-correlation method is much more robust but possibly biased (needs strong
and structured Reflected signal)
HG2013 3-6 may 2013 Trieste
W. Farabolini
8
Delays as function of cell #
Effect of tapered cells
Accuracy : 3.5 ns per cell (RF input side) / 7.5 ns per cell (RF output side)
Sampling rate: 1 ns on TBTS,
4 ns on X-Box (log detector), but 1 ns available
HG2013 3-6 may 2013 Trieste
W. Farabolini
9
Hot spot at cell #6 in the 1st TBTS structure
Ref -Trans method
Ref –In method
Evenness = 1 for
equally distributed BDs
Evenness = 0.66
HG2013 3-6 may 2013 Trieste
Evenness = 0.33
W. Farabolini
10
No hot spot in the 2 present TBTS structures
Present 2
ACSs in TBTS
compilation
Evenness = 0.96
HG2013 3-6 may 2013 Trieste
Evenness = 0.95
W. Farabolini
11
T24 BD locations evolution in X-Box1
Hot cell(s) from the beginning
Nota: possible positions absolute shift due to line delays uncalibrated
HG2013 3-6 may 2013 Trieste
W. Farabolini
12
TD24R05 BD locations evolution in X-Box1
Hot cell has appeared after 2 months
HG2013 3-6 may 2013 Trieste
W. Farabolini
13
Histogram of all BDs location (X-Box1)
No BD in this cell !
T24 during 6 weeks
Evenness_1st = 0.77
Evenness_2nd = 0.78
HG2013 3-6 may 2013 Trieste
TD24R05 Feb. & Mar.
Evenness_1st = 0.97
Evenness_2nd = 0.82
W. Farabolini
TD24R05 May. & Jun.
Evenness_1st = 0.83
Evenness_2nd = 0.45
14
2 examples of 8 hours sequences
• During BD cluster a hot cell (# 4 or # 5) appears
• Blue marks show failures in BD location, often related to no current in FCU (red dots)
HG2013 3-6 may 2013 Trieste
W. Farabolini
15
A proposed diagnostic for BD location
Franck Peauger – IRFU 2009
Segmented PMT
rising time < 1 ns
A. Grudiev Plasma modelling
in RF simulations, this WS
RF
output
RF
input
Additional passive or/and
active diagnostics via
damping waveguides
• Possible to observe
plasma oscillation
Plasma ignited
by the
breakdown
HG2013 3-6 may 2013 Trieste
W. Farabolini
16
Research of precursors in FCU and Reflected RF
peak values
Uncalibrated data
Motivation: Y. Ashkenazy, using stochastic theory for RF breakdown analysis, this WS
•
•
•
•
Faraday cup currents are negative (either dark current or BD burst). -1: saturated .
Reflected RF power are positive.
Background levels (offset) are suppressed.
All these signals are used to detect BDs and the 2 previous pulses are also data logged.
HG2013 3-6 may 2013 Trieste
W. Farabolini
17
Zoomed data from the 4th March
Still no evidence of any precursor
HG2013 3-6 may 2013 Trieste
W. Farabolini
18
More subtle data processing to be applied
RF signals
Real BD
Possible BD outside the structure
Faraday cups signals (zoomed)
Dark current only
Burst of electrons
• Look for power spectral density of the dark current (to be done)
HG2013 3-6 may 2013 Trieste
W. Farabolini
19
BDR as function of RF Power in TBTS
But conditioning is still
under progress
Date
2012_11_16
2012_11_19
2012_11_23
2012_11_29
2012_12_04
2012_12_05
2012_12_06
2012_12_07
HG2013 3-6 may 2013 Trieste
W. Farabolini
Mean sigma
power power Pulse BD ACS BD ACS
[MW] [MW] number up down
29.2
2.2
14807
3
2
30.3
1
36955
5
15
29
2.1
10932
1
1
37.2
2.6
45535 102
60
38.4
2.9
10174
12
14
46.1
1.8
13394
16
20
46.5
2.1
21622
27
8
36.2
3
9311
3
6
20
BDR as function of Power (2)
Upstream new ACS
Previous ACS
RF power density of Probability of all RF pulses (blue), of RF pulse with BD (red)
and power law fit of BD probability (green)
• Fitting the Power distribution when BD by a power law of the power distribution of all
pulses provide an exponent between 12 and 18.
HG2013 3-6 may 2013 Trieste
W. Farabolini
21
Distribution of the number of RF pulses
between BDs (clusters problem)
BD count evolution shows several period
of intense BDs activity: clusters
HG2013 3-6 may 2013 Trieste
• Inside clusters the BD probability
becomes very high.
• Discarding BDs within clusters allows to
focus on the stationary BD statistics, well
fitted by a Poisson law
W. Farabolini
22
Ignition and falling edge duration
Ignition
Falling
Can it be related to
neutrals and ions
growth as shown by
K. Sjobak , this WS ?
HG2013 3-6 may 2013 Trieste
Two categories of
Mean falling edge 50 ns
BDs : fast/slow
(for commuting 50MW)
ignition time
Mean ignition
duration 40 ns
W. Farabolini
23
Structure RF analysis after removal
Jiaru Shi, analysis of T18
• R. Wegner found identical results for
the 1st structure tested in TBTS
• However cutting it with wire is delicate
since activated (pb. of the TBTS)
HFSS result: Iris deform 10um  ~ 2MHz
• Great interest in the “internal geometry
measurement tool” presented by M.
Aicheler, this WS
HG2013 3-6 may 2013 Trieste
W. Farabolini
24
Conclusion
• Stand alone test stand provide a incredible
capability of massive results production.
• Fitting on of them with beam capability will be
ideal.
• BD theory / modeling and experimental
activities can gain a lot in exchanging ideas
and suggestions of tests.
HG2013 3-6 may 2013 Trieste
W. Farabolini
25
HG2013 3-6 may 2013 Trieste
W. Farabolini
26
Typical RF signals during BDs in TBTS
Ignition
Exposure
Falling
• Transmitted falling edge and Reflected rising edge supposed to be produced synchronously
(ignition only absorbs power, not reflects)
• Early BDs reflected power disrupts Input power (recirculation process in PETS)
• Transmitted phase quite stable up to the falling edge (even during ignition)
• Reflected phase can drift or jump a lot (Input phase disruption and/or BD displacement?)
HG2013 3-6 may 2013 Trieste
W. Farabolini
27