MOC3O07
Proceedings of ICALEPCS2015, Melbourne, Australia
LOW LEVEL RF CONTROL IMPLEMENTATION AND SIMULTANEOUS
OPERATION OF TWO FEL UNDULATOR BEAMLINES AT FLASH
V.Ayvazyan*, S.Ackermann, J.Branlard, B.Faatz, M.Grecki, O.Hensler, S.Pfeiffer, H.Schlarb,
C.Schmidt, M.Scholz, S.Schreiber, DESY, Hamburg, Germany
A.Piotrowski, Fast Logic, Lodz, Poland
Abstract
The Free-Electron Laser in Hamburg (FLASH) is a user
facility delivering femtosecond short radiation pulses in
the wavelength range between 4.2 and 52 nm using the
SASE principle. The tests performed in the last few years
have shown that two FLASH undulator beamlines can
deliver FEL radiation simultaneously to users with a large
variety of parameters such as radiation wavelength, pulse
duration, intra-bunch spacing etc. FLASH has two
injector lasers on the cathode of the gun to deliver
different bunch trains with different charges, needed for
different bunch lengths. Because the compression settings
depend on the charge of bunches, the low level RF
(LLRF) system needs to be able to supply different
compression for both beamlines. The functionality of the
controller has been extended to provide intra-pulse
amplitude and phase changes while maintaining the RF
field amplitude and the phase stability requirements. The
RF parameter adjustment and tuning for RF gun and
accelerating modules can be done independently for both
laser systems. Having different amplitudes and phases
within the RF pulse in several RF stations simultaneous
lasing of both systems has been demonstrated.
Copyright © 2015 CC-BY-3.0 and by the respective authors
INTRODUCTION
FLASH has been in operation as a user facility since
summer 2005 [1]. In the mean time as a test facility,
FLASH is in use for testing the superconducting
accelerator technology for European XFEL [2] and ILC
[3] projects. The first effort to operate accelerating
modules at different gradients with alternating RF pulses
was motivated by the ILC and European XFEL and
initially has been demonstrated in year 2008 [4].
FLASH2 [5] is the second undulator beam line with
variable gap built in a separate tunnel. It will make full
use of the existing FLASH accelerator. Part of the bunch
trains are kicked from the main beamline (FLASH1) into
the new undulator beamline. In order to double the
beamtime, users of both beamlines would need the 10 Hz
repetition rate. A fast kicker in combination with a DC
septum is used to deflect the beam into the second
undulator line. In addition, the large variety in beam
parameters should be available at both beamlines
independently in order to ensure a maximum flexibility in
planning of the beamtime. For this reason two cathode
lasers are in use, each with its own bunch train. A variable
delay between the two lasers within the RF pulse gun and
accelerating modules ensures that users from both
beamlines get their own set of parameters.
___________________________________________
* valeri.ayvazyan@desy.de
In addition, the start time of the kicker pulse is shifted
with respect to the start time of the laser pulse and the RF
amplitude and phase of the gun and each of the modules
can be tuned for optimal conditions for both users. During
the initial tests it was shown that the RF system is able to
handle the flexibility needed to compress the beam
independently for FLASH1 and FLASH2 beamlines.
LOW LEVEL RF CONTROL FOR MULTIBEAMLINES
RF Control Specifics for Multi-beamline Case
During the multi-beam line operation, within limits
given by the beam delivery system, the bunch pattern and
beam energy should be adjusted independently for each
beam line, suggesting a time-sliced operation. For the
FLASH2 beamline RF amplitude and phase changes
within the pulse are required within a short time (less than
50µs). Different beam loadings are foreseen for different
beamlines. Particularly the ability of gradient tuning of
the last two RF stations is needed for wavelength scans
for the FLASH1 beamline and the ability of phase tuning
at injector for variation in compression at FLASH1 and
FLASH2. From the operation point of view, the most
important requirement is to have the ability of
independent RF operational parameters adjustments for
both beamlines.
LLRF System Overview
The layout of the FLASH facility including the new
beamline is described in [6] and shown in Fig. 1. Its
accelerator comprises a normal conducting RF gun, a first
8-cavity cryomodule, a third harmonic cryomodule with 4
cavities, a first bunch compressor, a second accelerating
RF station (16 cavities), a second bunch compressor, and
another two RF stations, with 16 cavities each. FLASH is
operated in pulsed mode with repetition rate 10Hz. RF
pulse duration is 1.3ms, 500µs for filling and 800µs
flattop (beam acceleration). The RF power coming from
10MW klystrons is equally distributed to all cavities
through the waveguide distribution system. The current
stage of the LLRF control is implemented based on the
MTCA.4 system [7]. The goal of the system is to control
the accelerating gradient in amplitude and phase for each
RF station based on vector sum control of cavity gradients
[8]. The main components of the LLRF system are
depicted in Fig. 2. For every cavity, the forward (PFWD),
reflected (PREF) and transmitted signals or probes (PRB)
are first down-converted to an intermediate frequency (IF)
by the down-converters (uDWC) and then digitized
ISBN 978-3-95450-148-9
42
Feedback Systems, Tuning
Proceedings of ICALEPCS2015, Melbourne, Australia
MOC3O07
(uADC). The sampled signals are pre-processed by the
uADC and then sent over the MTCA.4 backplane to the
main LLRF controller (uTC) which performs all control
computations. The uTC then generates the drive signal
which is up-converted to RF frequency by the vector
modulator (uVM). The LLRF drive signal is preamplified and then sent to the klystron (KLY). The master
oscillator (MO) provides the 1.3 GHz reference signal
(RF), required by the local oscillator generation module
(LOGM) to generate the LO and clocks (CLK) signals
used by the down-converters, and to distribute the
reference signal to the uVM. The power supply module
(PSM) provides DC voltages to external modules. Finally,
the piezo driver module (PZ16M) digitizes the piezo
sensor data and drives the piezo actuator for detuning and
microphonics compensation. Communication between the
PZ16M and the MTCA.4 system is performed through an
optical link to the main LLRF controller, the uTC. This
setup can be easily extended to control the sum of 16
cavities.
to measure the accelerating field in the individual cavities.
These 1.3 GHz (3.9 GHz) signals are down-converted to
54 MHz and sampled by uADCs at 81.25 MHz. The
digitized signals are going to the digital field detector
which extracts the real and imaginary parts of the cavity
field vectors from the input stream. The resulting field
vector of each cavity is multiplied by a rotation matrix to
calibrate amplitude and phases. Then the sum of
individual field vectors is calculated and rotated to adjust
the loop phase. The vector sum of the cavities fields
represents the total voltage and phase seen by the beam.
This signal is regulated by a feedback control algorithm
which calculates corrections to the driving signal of the
klystron: the measured vector sum is subtracted from the
set-point table and the resulting error signal is amplified
and filtered to provide a feedback signal to the vector
modulator controlling the incident wave. A feed-forward
signal is added to correct the averaged repetitive error
components. Beam current information is used to scale
the feed-forward table to provide fast feed-forward
corrections if the beam current varies. The controller
server software handles: generation of control tables from
basic settings, rotation matrices for the cavity field
vectors, start-up configuration files, feedback and
exception handling control parameters, etc. The interrupt
service routines are used to start the data reading from the
controller boards. The parameters of the feedback
algorithm are modified by the FPGA programs in the time
slot between pulses.
Application Software
Figure 2: LLRF system block diagram for one cryogenic
module.
Brief Description of Control Algorithm
The feedback algorithm is implemented in FPGAs
firmware and DOOCS [9] control system server. The
control algorithm employs tables for feed-forward, setpoint and feedback gain settings to allow time varying of
those parameters. High frequency probe signals are used
A set of generic and especially devoted programs
provides the tools for the operators to control the RF
system. Some of them are created based on the
MATLAB, others as DOOCS middle layer servers. The
adaptive feed-forward is implemented on a front end
server, to allow pulse to pulse adaptation. The application
software includes vector sum calibration, automated
operation of the frequency tuners, phasing of cavities,
adjustment of various control system parameters, etc.
Extensive diagnostics inform the operator about cavity
quenches, cavity detuning, and an excessive increase in
control power.
ISBN 978-3-95450-148-9
Feedback Systems, Tuning
43
Copyright © 2015 CC-BY-3.0 and by the respective authors
Figure 1: Schematic layout of the FLASH facility. The electron gun is on the left, the experimental hall on the right.
Behind the last accelerating module, the beam is switched between FLASH1 and FLASH2 beamlines.
MOC3O07
Proceedings of ICALEPCS2015, Melbourne, Australia
RF control functionality extension for FLASH2
Copyright © 2015 CC-BY-3.0 and by the respective authors
The RF pulse is shared between the electron bunch
trains for FLASH1 and FLASH2. For the full RF pulse
length, the total maximum number of bunches, with
bunch repetition rate of 1 MHz is 800. The bunch pattern
(number of bunches and intra-train repetition rate) and
bunch charge can be different for FLASH1 and FLASH2,
which is realized by using two independent injector
lasers. Timing events from FLASH accelerator are used to
synchronize all accelerator subsystems and are managed
by DOOCS timing server. Programmable timers are
triggered by these events to generate the start pulses for
the klystrons, FPGAs or uADCs. A timer unit provides
several independent output channels. Some machine
parameters that change from macro pulse to macro pulse
are delivered to run all digital feedback loops in parallel.
This information is available for the LLRF controller
server. The control tables are generated through the LLRF
library from the operational setting according to the
provided timing information. Parameters such as beam
start time, number of bunches, bunch repetition rates are
extracted from the timing server for both beamlines.
Other parameters like RF transition time, offsets with
respect to beam pulse, etc. are adjustable. LLRF control
software follows any changes of the timing parameters.
Fig. 3 illustrates the timing setup for two beamlines.
levels of RF pulse (Fig. 4). Background colours show the
beam start time and bunch train duration time for
FLASH1 and FLASH2 beamlines. In order to avoid
operational settings changes in unacceptable ranges,
limits for FLASH2 operational parameters with respect to
FLASH1 are implemented as well. An adaptive learning
algorithm which minimizes repetitive control errors tends
to limit these transition changes. In order to avoid
oscillations, it is possible to deactivate adaptations for
certain regions.
PERFORMANCE OF SIMULTANEOUS
OPERATION OF FLASH1 AND FLASH2
In the initial stage of the project several tests have been
performed to show that simultaneous operation of both
beamlines is possible [10]. Fig. 5 shows lasing of two
bunch trains which were generated with separate injector
lasers, separated by 80µs: a) with the same charge of
about 0.5 nC and b) with a factor of 2 difference in
charge, e.g. of 0.5 nC for the first and 0.25 nC for the
second bunch train. The blue line indicates the actual
SASE pulse energy produced by each individual electron
bunch in the macro-pulse. The green line is the time
average of this signal. The yellow line indicates the
maximum SASE level which occurred since the
measurement was started. Both bunch trains have a
repetition rate of 1 MHz. The number of bunches in this
case was 30 and 20 respectively. During this test only RF
parameters were changed within the RF pulse and orbit
was adjusted behind the FLASH2 extraction point.
Because the FLASH2 beam line was under construction
in that time, this test has been performed at FLASH1
beamline. Careful adjustment was done to make sure that
both injector lasers hit the cathode under the same angle
to make sure that the electron beams have the same
trajectory. This condition was relaxed in the later
situation, since the orbit in the FLASH1 and FLASH2
undulators can be adjusted independently to optimize
lasing.
Figure 3: An example of FLASH timing settings.
Figure 4: Energy and phase variation within RF pulse.
Energy and phase variations are possible to allow for
charge dependent compression and wavelength fine
tuning. A future extension with an additional beamline is
already foreseen, as can be seen by the three different
Figure 5: Lasing of two independent bunch trains with
different charges and a variable delay in time.
ISBN 978-3-95450-148-9
44
Feedback Systems, Tuning
Proceedings of ICALEPCS2015, Melbourne, Australia
Since 2014 tests have been performed with FLASH1
and FLASH2 in operation. First SASE operation in
FLASH2 was achieved at wavelength 40 nm on August
20, 2014 [11] during FLASH1 operation with 250
bunches at 13.5 nm.
After the demonstration of the first lasing at FLASH2
the SASE operation was established at various
wavelengths. So far, the maximum number of bunches
per burst during a parallel SASE operation of both
beamlines has been 400 bunches in FLASH1 and 30
bunches in FLASH2, both with a bunch repetition rate of
1 MHz. Fig. 6 shows the SASE pulse energy along the
bunch trains in FLASH1 and FLASH2. During parallel
operation achieved pulse energy at FLASH1 was about
200µJ and about 100µJ at FLASH2 in the period from
June 2015 to August 2015.
MOC3O07
determined - due to the fixed gap undulator - by the
electron beam energy. Variable gap undulators in
FLASH2 allow different photon wavelength at fixed
beam energies.
SUMMARY AND OUTLOOK
Low level RF control functionality has been extended
which makes simultaneous RF operation of multibeamlines possible. Changes in RF settings within RF
pulse can be achieved which allow different compression
settings for different charges (and therefore different
bunch lengths) while maintaining the RF field amplitude
and the phase stability requirements.
Simultaneous operation of two beamlines and lasing of
FLASH2 at the wavelength 40 nm was achieved, while
FLASH1 was lasing simultaneously with multi-bunch
mode at different wavelength (13.5 nm). SASE operation
at various wavelengths was established.
Gained experience with simultaneous operation of two
beamlines at FLASH is a good basis for successful
commissioning of multi-beamline facility European
XFEL which is foreseen for the second half of 2016. At
XFEL, operation of alternating RF pulses (from pulse to
pulse) is foreseen as well. This operation mode requires
additional changes in the firmware/software structure, e.g.
extending the amount of control tables. Furthermore,
requirements to the timing system are increased to
reliably trigger the correct mode of operation.
Figure 6 : SASE pulse energy per bunch (in a.u.). Top:
400 bunches in FLASH1. Bottom: 30 bunches in
FLASH2. Blue: actual value, green: average, yellow:
maximum.
Figure 7: Photon wavelengths achieved in the FLASH1 and
FLASH2 beamlines during parallel SASE operation.
In Fig. 7 [12] are shown photon wavelengths achieved
in the FLASH1 and FLASH2 beamlines during parallel
SASE operation in the period from August 2014 to
August 2015. The photon wavelength in FLASH1 is
ISBN 978-3-95450-148-9
Feedback Systems, Tuning
45
Copyright © 2015 CC-BY-3.0 and by the respective authors
REFERENCES
[1] V.Ayvazyan et al., Eur. Phys. J. D 37 (2006) 297303.
[2] http://www.xfel.eu
[3] https://www.linearcollider.org/ILC
[4] V.Ayvazyan et al., “Alternating Gradient Operation
of Accelerating Modules at FLASH”, TUPP001,
EPAC’08, Genoa, Italy (2008).
[5] B.Faatz et al., “FLASH II: Perspectives and
chalanges”, Nucl. Instr. Meth. A 635, S2 (2011).
[6] M.Vogt et al., “Status of the Soft X-ray Free Electron
Laser FLASH”, TUPWA033, IPAC’15, Richmond,
USA (2015).
[7] J.Branlard et al., “Equipping FLASH with MTCA.4based LLRF system”, THP085, SRF’13, Paris,
France (2013).
[8] T.Schilcher, “Vector Sum Control of Pulsed
Accelerating Fields in Lorentz Force Detuned
Superconducting Cavities”, PhD Dissertation, DESY
Hamburg, Germany, 1998.
[9] http://doocs.desy.de
[10] S.Ackermann et al., “Simultaneous Operation of
Tow Undulator Beamlines FEL Facility”, TUPD32,
FEL’12, Nara, Japan (2012).
[11] S.Schreiber, B.Faatz, “First Lasing at FLASH2”,
MOA03, FEL’14, Basel, Switzerland (2014).
[12] M.Scholz, B.Faatz, S.Schreiber, J.Zemella, “First
Simultaneous Operation of Tow SASE Beamlines in
FLASH”, TUA04, FEL’15, Daejeon, Korea (2015).