See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/328827427

High Power and High Repetition Rate X-band Power Source Using Multiple
Klystrons

Conference Paper · May 2018

DOI: 10.18429/JACoW-IPAC2018-THPMK104

CITATIONS READS

0 6

16 authors, including:

Nuria Catalan Lasheras Thomas Geoffrey Lucas

CERN University of Melbourne
90 PUBLICATIONS   394 CITATIONS    13 PUBLICATIONS   50 CITATIONS   

SEE PROFILE SEE PROFILE

Gerard McMonagle i. Syratchev

CERN CERN
23 PUBLICATIONS   170 CITATIONS    83 PUBLICATIONS   461 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Low temperature plasmas View project

Vacuum arc breakdowns View project

All content following this page was uploaded by Thomas Geoffrey Lucas on 09 November 2018.

The user has requested enhancement of the downloaded file.

 9th International Particle Accelerator Conference IPAC2018, Vancouver, BC, Canada JACoW Publishing
ISBN: 978-3-95450-184-7 doi:10.18429/JACoW-IPAC2018-THPMK104
Content from this work may be used under the terms of the CC BY 3.0 licence (© 2018). Any distribution of this work must maintain attribution to the author(s), title of the work, publisher, and DOI.
HIGH POWER AND HIGH REPETITION RATE
X-BAND POWER SOURCE USING MULTIPLE KLYSTRONS
M. Volpi1∗ , N. Catalan-Lasheras, A. Grudiev, T.G. Lucas1 , G. McMonagle, J. Paszkiewicz2
V. del Pozo Romano, I. Syratchev, A. Vnuchenko3 , B. Woolley,
W. Wuensch, European Organisation for Nuclear Research (CERN)
P.J. Giansiracusa, R.P. Rassool, University of Melbourne, Parkville, Australia
C. Serpico, Elettra-Sincrotrone Trieste S.C.p.A., Basovizza
M. J. Boland, Canadian Light Source and Department of Physics and Engineering Physics,
University of Saskatchewan, Canada
1 also at University of Melbourne, Parkville, Australia
2 also at University of Oxford, Oxford, United Kingdom
3 also at IFIC, Valencia

Abstract Xbox 3 uses twin low peak power, 6 MW klystrons. High

In July 2016, the first X-band test facility operating with peak powers are achieved by combining the output power of
two interwoven RF pulses produced from 6 MW klystron these klystrons and using pulse compression. This process
was commissioned at CERN. Outputting up to 49 MW after allows the production of 47 MW for a 200 ns pulses at much
pulse compression [1], the new test stand allows testing of higher repetition rates than would be possible with the single,
two structures concurrently with a combined repetition rate 50 MW klystrons used in previous Xboxes.
of up to 400 Hz. Commissioning of two of the four lines has CLIC prototype structures require 40–50 MW of RF
been completed and testing of high gradient accelerating power to reach a unloaded gradient of 100 MV/m. High-
structures for the Compact Linear Collider has commenced. gradient accelerating structures take several hundred mil-
Operations have been ongoing for more than a year, where lions pulses to condition to nominal operating conditions [4].
dedicated control algorithms have been developed to con- It is therefore advantageous to make testing facilities that
ditioning the structure and to keep the pulse compressors can pulse at high repetition rates, in order to reduce the time
tuned. Significant progress has been made in understand- needed to condition structures.
ing the conditioning of two structures that are sharing an In this paper, we will describe the combination scheme
interlock and vacuum system. The higher pulse repetition used in Xbox 3 and the experience of commissioning [5–7]
rate has demonstrated a significantly reduction in the time the facility under high power.
required to condition accelerating structures.
XBOX 3 LAYOUT
INTRODUCTION
The Compact Linear Collider (CLIC) [2] project at CERN
Amplifier

has led to the development of numerous high-gradient X-

band accelerator technologies. Such advancements are at-
tracting great interest from the light source and medical
communities where linear accelerators are foreseen as the
Amplifier

next generation of accelerators. Critical to CLIC’s devel-

opment of high-gradient X-band technology has been an
investment in test stands, which allowed investigations of Figure 1: Schematic of the high power RF network of half
the complex, multi-physics effects that affect high-power the Xbox 3 facility.
behavior in operational structures.
Increasing the number of available testing slots is impor-
tant because within the CLIC project there is an extensive Depicted in Figure 1 is the layout of one half of the Xbox 3
program aiming to test 40 structures by 2019. CERN has waveguide network. To achieve the peak power necessary,
built three test facilities called Xbox 1, Xbox 2, and the two klystrons are combined through a hybrid and subse-
newest and subject of this report, Xbox 3 [3]. quently compressed using a SLED-I pulse compressor (PC).
Xbox 1 and Xbox 2 use similar technologies to the original The pulse repetition rate of each klystron can be increased up
klystron-based test facilities in Japan and the US. These use to 400 Hz. The whole system counts two symmetric layouts
costly, high peak power, 50 MW klystrons which require with a total of four test stands, consequently each test stand
modulators capable of producing up to 450 kV. However, can operate up to 200 Hz.
The Low Level RF and control system is based on a Na-
∗ matteo.volpi@cern.ch tional Instruments PXI crate. Each pair of klystrons shares
THPMK104 02 Photon Sources and Electron Accelerators
4552 A08 Linear Accelerators
9th International Particle Accelerator Conference IPAC2018, Vancouver, BC, Canada JACoW Publishing
ISBN: 978-3-95450-184-7 doi:10.18429/JACoW-IPAC2018-THPMK104

Content from this work may be used under the terms of the CC BY 3.0 licence (© 2018). Any distribution of this work must maintain attribution to the author(s), title of the work, publisher, and DOI.
one PXI crate and controller. The RF drive signals are pro- for injected bunches and it also decrease the radiation level
duced using NI 5793 IQ generators. Due to inherent non- on the structure. No correlation between BDR and flat phase
linearities in the klystrons, they are run in saturation and the have been observed.
power controlled by changing the drive signal.
After the hybrid combination each line has a PC, test slot
Transmitted power
for the device under test and an RF stainless steel load to Transmitted phase
Power after the structure
terminate the waveguide network. Directional couplers are
placed at critical locations in the network for monitoring
purposes. Vacuum pumping ports complete the network.
There are several control loops controlling the stability of
the RF power distributed to the waveguide network. Pulse-
by-pulse changes can be made to the drive signal which
allows the two test slots to operate independently. Sharing
an interlock system, for safety reasons, an interlock in one
line arrests the RF pulsing in both lines.
Breakdown rate (BDR [8]) of the structure has been
demonstrated that is not affected by the previous history Time

like power pulse ramping after a BD. That means in the next
future we can still pulsing in one line if the BD/interlock Figure 3: Flat top pulse compressor with flat phase.
happen in the other line [4].

COMMISSIONING Xbox 3 uses Toshiba E37113 klystrons and ScandiNova

K1 modulators which can each produce up to 5 µs pulses
Initial pulsing was performed at a low repetition rate to with 6 MW of peak power. At high repetition rate, greater
check the performance of the subsystems, after that the rep- than 200 Hz, the waveguide temperature increases up to the
etition rate was increased to 200 Hz in each line. interlock limit (55◦ C). A dedicated cooling system based on
An algorithm [9] was developed to flatten the pulse com- clamps was design and installed to allow pulsing below the
pressor’s output pulse. Changes to the slope of the pulses interlock limit. Any change in the average RF power (such as
flat-top can be performed through the cavities’ temperature pulsing interruption due to an RF breakdown) caused the PC
controller or by manipulating the input pulse’s phase profile. to cool and, as a result, detune. A frequency shift algorithm
By changing the input phase profile the response will be was added to the RF control GUI to tune the PCs after an RF
immediate. The algorithm uses a simple feedback loop and interlock. The algorithm uses transmitted phase to choose
fitting the flat-top with a 5th order polynomial (Fig. 2 [9]). the frequency shift as shown in Fig. 4. When the PC is
perfectly tuned, the phase is exactly 180◦ at the minimum
3,5E+07 -180
point of the transmitted power is zero. This occurs when
-190 the RF power leaking from the cavities begins to exceed the
3,0E+07
-200 input power. Over tuned cavities mean that the filling phase
2,5E+07 shift is negative so that the phase flip is smaller than the
-210

2,0E+07
desired value of about 180◦ , while under tuned means that
Phase (deg)

-220
Power(W)

the filling phase shift is positive so that the phase flip is larger
1,5E+07 -230 than the desired value of 180◦ . A shift of drive frequency
-240 up to ±1 MHz can be applied to tune the PC instantaneously
1,0E+07
-250
without waiting for temperature stabilisation. Within this
5,0E+06
range the klystrons are still stable, 1 MHz corresponds to
-260
about 5◦ C. The frequency shift algorithm is activated until
0,0E+00 -270 the temperature of the cavities reach the working temperature
2,E-06 2,E-06 2,E-06 3,E-06 3,E-06 3,E-06
stabilized with high power chillers.
Time(s)
Calibration of the entire RF system from waveguide to
Power - Output Power Linear Ramp Power - Output Power PD
Phase - Output Phase Linear Ramp Phase - Output Phase PD
detector is very important, in order to reduce uncertainty
in the RF measurements. There are several stages needed
Figure 2: Power and phase of the incident and compressed in order to fully calibrate the RF network. RF signal ca-
pulses when tuned using the pulse flattening algorithm. bles are split and sent to the low level RF diagnostic crate
for down-mixing/log detection and simultaneously to a cali-
A frequency shift due the phase ramp (line blue of Fig. 2) bration system. Due to the temperature difference between
have been observed that provoke an increase of the the dark summer and winter two sets of attenuation constants have
current [10]. Thanks to the phase algorithm, Xbox 2 is been measured, where positive shift of 0.15 dBm is observed
running with a flat phase (Fig. 3). Flat phase is mandatory in summer.
02 Photon Sources and Electron Accelerators THPMK104
A08 Linear Accelerators 4553
 9th International Particle Accelerator Conference IPAC2018, Vancouver, BC, Canada JACoW Publishing
ISBN: 978-3-95450-184-7 doi:10.18429/JACoW-IPAC2018-THPMK104
Content from this work may be used under the terms of the CC BY 3.0 licence (© 2018). Any distribution of this work must maintain attribution to the author(s), title of the work, publisher, and DOI.
70
INC
60 TRA
REF
50

Amplitude [MW]
40

30

20

10

0
1.1 1.2 1.3 1.4
time [s] 10 -6
Figure 5: BD reflection interference.

80

60
Input Power [MW]

40

20

Figure 4: Simulation of tune/detuned cavity.

0
0 50 100 150 200
REFLECTION INTERFERENCE BD Position in Pulse [ns]
In Xbox 2, it was observed that an interference from break- Figure 6: Power peak BD position in pulse.
down reflection resulted in up to a 40% increase in the input
power (Fig. 5).
Xbox 2 and Xbox 3 have approximately 10 metres of power RF sources into two separate high power testing slots
waveguide that gives around 150 ns of delay minimum. If a has been successfully demonstrated. New algorithms and
BD occurs at the start of the structure and at the start of the new effects due to the peculiarities of this system have been
pulse, the reflected signal from the structure can be interact observed and studied. Four CLIC prototype structures have
in constructive way with itself. If the high power RF pulse already been tested in line 3 and 4 of Xbox 3.
is longer more than 150 ns the pulse can reenter the klystron In March of 2018, we installed a new klystrons in line
and constructive interfere with the end of the same pulse 1 and 2. These lines are currently used to test RF compo-
(Figure 6). nents [11] such as 3D printed loads, a high power phase
This same effect is also observed as a small amplitude blip shifter, and variable power splitter. After testing the RF
in the incident RF pulse flat-top. A correction in the phase components the lines will be commissioned in the same way
program allows flattening of the pulse though this brings of line 3 and 4 with CLIC prototype structures. The high
further challenges through a discontinuity in the incident repetition rate of Xbox 3 is already showing the significantly
phase. A more permanent solution to this is to have long reduced time needed to condition accelerating structures as
waveguides. At least 15 meters in the CLIC design. This will shown in [4, 12].
of course lead to greater power loss compared to a shorter
waveguide. No impact on the BDR is observed despite the REFERENCES
structure seeing more power. [1] Z. D. Farkas et al., “SLED: A method of doubling SLAC’s
energy,” in Proc. 9th Int. Conf. on High Energy Accelerators,
CONCLUSION AND ONGOING WORK SLAC, 1974.
The commissioning of the most recent X-band test facility [2] “A Multi-TeV linear collider based on CLIC technology: CLIC
at CERN has been completed. The combination of two low Conceptual Design Report,” edited by M. Aicheler et al.,
THPMK104 02 Photon Sources and Electron Accelerators
4554 A08 Linear Accelerators
9th International Particle Accelerator Conference IPAC2018, Vancouver, BC, Canada JACoW Publishing
ISBN: 978-3-95450-184-7 doi:10.18429/JACoW-IPAC2018-THPMK104

Content from this work may be used under the terms of the CC BY 3.0 licence (© 2018). Any distribution of this work must maintain attribution to the author(s), title of the work, publisher, and DOI.
CERN-2012-007. [8] W. Wuensch et al., “Statistics of vacuum breakdown in the high-
[3] B. Woolley et al., “High Power X-Band Generation Using gradient and low-rate regime,” Phys. Rev. ST Accel. Beams vol.
Multiple Klystrons and Pulse Compression,” in Proc. IPAC17, 20, p. 011007, Jan 2017.
Copenhagen, Denmark, 2017.
[9] B. Woolley et al., “Control and performance improvements of
[4] T. G. Lucas, Results and Observations from High Gradient
a pulse compressor in use for testing accelerating structures at
X-band Tests at CERN, peesented at IPAC2018, Vancouver,
high power,” Phys. Rev. Accel. Beams 20, 101001, 2017.
Canada, 2018.
[5] J. Kovermann et al., “Commissioning of the First Klystron [10] T. G. Lucas, “Dark Current frequency dependence: measure-
based X-Band Power Source at CERN,” in Proc. IPAC2012, ment and simulation”, CLIC workshop, CERN, 2018.
New Orleans, LA, USA, May 2012.
[6] N. Catalan-Lasheras et al., “Experience Operating an X-band [11] N. Catalan Lasheras et al., “High Power Conditioning of X-
High Power Test Stand at CERN”, in Proc. IPAC2014, Dresden, band RF Components”, presented at IPAC2018, Vancouver,
Germany, June 2014. Canada, 2018.
[7] N. Catalan Lasheras et al., “Construction and commissioning
Xbox 3: a very high capacity X-band test stand,” in Proc. 2016 [12] Zennaro, R et al., “High Power Tests of a Prototype X-Band
IEEE Power Modulator and High Voltage Conference, San Accelerating Structure for CLIC”, in Proc. IPAC2017, Copen-
Francisco, USA. hagen, Denmark, May 2017.

02 Photon Sources and Electron Accelerators THPMK104

A08 Linear Accelerators 4555