test of a bac klystron - slac.stanford.edu
Post on 24-Oct-2021
21 Views
Preview:
TRANSCRIPT
Test of a BAC Klystron
R. Kowalczyk, B. Weatherford, A. Haase, E. Jongewaard, M. Kemp, and J. Neilson
SLAC National Accelerator Laboratory
A. Jensen
Leidos
IVEC 2017
SLAC-PUB-17101
This material is based upon work supported by the U.S. Department of Energy, Office of Science, under Contract No. DE-AC02-76SF00515.
Background – SLAC 5045
• The 5045 klystron is used in the SLAC linac
- ~250 in the accelerator
- Frequency: 2.856 GHz
- RF Power: 65 MW peak, 45 kW avg.
- Nominal efficiency: 45%
- Beam tunnel radius: 1.59 cm
• γa= 0.70
- Over 800 klystrons produced since 1983
- More than 30 million operating hours
• Ongoing effort at SLAC to increase energy
efficiency of RF sources and accelerators
- Reduced cost per kWh
- Sustainability
3
Existing 5045 Klystron (65MW peak power)
Gun Collector
Gain Cavities Penultimate & Output Cavities
Drift Space
4
High Efficiency Klystron Design
Conventional
COM
Preliminary cavities are tuned
conventionally for gain
One or more “penultimate” cavities
are tuned high to squeeze the bunch
Most of the circuit sweeps electrons at the
wrong phase into the right phase
Only once most of the electrons are at
the right phase is the bunch compressed
I. Syratchev
I. Guzilov
A. Baikov
5
High Efficiency Klystron Philosophy
Conventional
COM
One or more “penultimate” cavities
are tuned high to squeeze the bunch
Only once most of the electrons are at
the right phase is the bunch compressed
“core” remains
the same size –
space charge
balances RF field
Peripheral
electrons
brought in
Preliminary cavities are tuned
conventionally for gain
Most of the circuit sweeps electrons at the
wrong phase into the right phase
I. Syratchev
I. Guzilov
A. Baikov
The Bunch Align Collect Method
COM
BAC
In the BAC (bunch-align-collect), the same
effect is achieved by BAC triplets
B A C B A C B A C
7
The Bunch Align Collect Method
COM
BAC
B A C B A C B A C
Bunch cavity
kicks all
electrons
toward center
B A C
Align cavity,
tuned low,
defocuses center
Collect cavity, at second
harmonic, primarily
affects the peripherals
In the BAC (bunch-align-collect), the same
effect is achieved by BAC triplets
8
Overall Goals
• A form, fit, function
replacement for the
5045 with higher
efficiency
• Fit into existing
solenoid
Standard 5045 BAC 5045
5045 BAC Approach
• Targets: 80 MW peak output power; 55% efficiency
• New design utilizes BAC (Bunch Align Collect) method
Input and three gain cavities remain
nearly unchanged, to simplify retrofit
Modified form of the BAC method used
to bunch beam
Core remains fixed size, periphery
electrons drawn in
B A C C
Two Dimensional Simulations
• Drift tube diameter reduced in output section so that drift tube is cut off
to second harmonic
• Reduced beam tunnel radius = 1.31 cm (γa = 0.57)
Standard 5045
BAC 5045
• MAGIC2D used for large signal simulations
11
BAC Simulations
0.0
0.5
1.0
1.5
2.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Har
mo
nic
Cu
rren
t
Position [m]
Fundamental
Second
Third
Circuit Assembly
Input cavities nearly
identical to standard
Output
Cavity
Modifed circuit:
13
Test Data
• Pulse length limited to 100 ns to prevent oscillation
• Peak output power of 72 MW, for efficiency of 54%
• Solenoid settings were optimized for stability, not
efficiency (standard 5045 is 45% efficient)
0
10
20
30
40
50
60
70
80
0 200 400 600 800 1000
Po
ut
[MW
]
Pin [W]
54% efficiency
Oscillations
• Oscillation measured in output waveguide at 6.65 GHz
-70
-60
-50
-40
-30
-20
-10
0
6.40 6.50 6.60 6.70 6.80 6.90
Am
plit
ud
e [
dB
]
Frequency [GHz]
15
Oscillations
• MAGIC simulations run to longer time show oscillations
Design power Onset of oscillation
16
Oscillations
• MAGIC simulations showed that there are two potential oscillations:
• Cavities 6, 7, 8 oscillate at 6.7 GHz
• Cavities 2, 3, 4 oscillate at 4.2 GHz
4.2GHz
6.7 GHz
17
Oscillation Prediction
• Would like a standard method to find oscillation before building klystrons
• Initial plan:
• Use ACE3P, the SLAC parallel electromagnetic code suite
• Solve all cold modes in the circuit structure
• Push particles through each mode, observe energy balance to
determine if mode is stable
18
6.7 GHz Oscillation
• Drift tube is not cut off at this frequency, so
individual cavities couple together
• Mode trapped between cavities, with transverse
field in tunnel, axial field on gaps
Zero axial E field
on axis (for perfect
alignment, no net
energy transfer (to
zeroth order)) Beam receives
transverse kick
If phasing is right,
off axis beam
transfers energy to
the mode, driving
the oscillation
This is similar to a
conventional monotron
oscillation, and can readily
be found by analyzing for
net energy gain or loss
19
4.2 GHz Oscillation: Introduction
• Driven by TM11 modes in cavities
• Unlike previous example, drift tube is cut off,
although barely
• Cutoff frequency is 4.4 GHz
• Cavities still couple through cut-off drift tube,
forming a 3-cavity mode
• BUT, mode is stable to the ACE3P analysis
20
4.2 GHz Oscillation: Explanation
Deep Science Mining for Matter
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
0.00 0.05 0.10 0.15 0.20 0.25 0.30
I1/I
0
Distance [m]
Nominal
With Bagel
• Measure of harmonic current (measured spatially) shows gain due
to the self-excitation of the 4.2 GHz mode
• Drift tube is cut off, but
barely, so provides
feedback path, with loss
due to exponential decay
• If loop gain exceeds 1,
system will be unstable
• This type of oscillation relies
on forward gain, thus the
assumption of a single
“mode” with constant
amplitude, as in a monotron
type oscillation, is not valid
• Large signal klystron
analysis required to predict
gain
loss
21
4.2 GHz Oscillation: Explanation
Deep Science Mining for Matter
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
0.00 0.05 0.10 0.15 0.20 0.25 0.30
I1/I
0
Distance [m]
Nominal
With Bagel
• Measure of harmonic current (measured spatially) shows gain due
to the self-excitation of the 4.2 GHz mode
• Drift tube is cut off, but
barely, so provides
feedback path, with loss
due to exponential decay
• If loop gain exceeds 1,
system will be unstable
• This type of oscillation
relies on forward gain, thus
the assumption of a single
“mode” with constant
amplitude, as in a montron
type oscillation, is invalid
• Large signal klystron
analysis required to predict
gain
loss
Tuning
ring
Beam
axis
Ring added to tune the TM11, not
the TM01, kills gain and oscillation
Oscillation Prediction
• Need to perform 3D PIC
simulation to be sure of stability
• Compared Vsim and MAGIC in
a model of the first four cavities
• Both predict the 4.2 GHz
oscillation
• On a single core, both run at
similar speed
• Vsim designed to be parallel
• A Vsim test with up to 128 cores
will be performed soon, to
assess scaling
23
Next Design
• A next iteration of the BAC 5045 is being designed
• Number of cavities have been reduced to 9 from 10
- Increased spacing between cavities reduces likelihood of multicavity oscillation
- Further cavity adjustments with bagels if oscillations seen in MAGIC
• Circuit topology:
- Cavities 1-4 and 9 (output) same as previous version
- Cavities 5-8: Bunch, Collect, Align, Bunch
• Similar performance as first design seen in 1D: 81 MW, 56%
24
Future Steps
• Finish translation of new 1D design to 2D MAGIC and tweak
design for maximum efficiency
- Address effects of radial beam distribution on efficiency, if
possible
• Screen for oscillations using particle push method and
MAGIC3D
• Finalize mechanical design of new cavities
• Retrofit previously built tube with new BAC cavities
• Begin testing of second iteration
25
Conclusions
• BAC method used to improve efficiency of SLAC 5045
klystron
• Test results confirm that efficiency increases by ten
percentage points
• Oscillation prevented running at design pulse length
• Approach developed to predict oscillation before fabrication
• Second prototype underway to achieve the same efficiency
and eliminate oscillations
top related