ISBN 978-3-95450-180-9 Proceedings of NAPAC2016, Chicago, IL, USA MOPOB17

RESONANT FREQUENCY CONTROL FOR THE PIP-II INJECTOR TEST

RFQ: CONTROL FRAMEWORK AND INITIAL RESULTS*
A. L. Edelen†, S. G. Biedron, S. V. Milton, Colorado State University, Fort Collins, CO, USA
D. Bowring, B. E. Chase, J. P. Edelen, D. Nicklaus, J. Steimel, Fermilab, Batavia, IL, USA

Abstract In this paper we describe the operational needs for the

For the PIP-II Injector Test (PI-Test) at Fermilab, a RFQ, the design of the control framework, and initial
four-vane radio frequency quadrupole (RFQ) is designed control results. This work represents a first test of reso-
to accelerate a 30-keV, 1-mA to 10-mA, H— beam to 2.1 nant frequency control over the RFQ, a first test of the
MeV under both pulsed and continuous wave (CW) RF framework, and a first test of using a dedicated Python
operation. The available headroom of the RF amplifiers program at Fermilab interfaced with the main control
limits the maximum allowable detuning to 3 kHz, and the system via the protocol compiler.
detuning is controlled entirely via thermal regulation.
Fine control over the detuning, minimal manual interven- SYSTEM DESCRIPTION AND
tion, and fast trip recovery is desired. In addition, having OPERATIONAL GOALS
active control over both the walls and vanes provides a More details on the system and control challenges
wider tuning range. For this, we intend to use model pre- therein are described in [1, 5]. The low-level radio fre-
dictive control (MPC). To facilitate these objectives, we quency (LLRF) system can compensate for detuning of
developed a dedicated control framework that handles the cavity only up to 3 kHz by taking advantage of the
higher-level system decisions as well as executes control available overhead in the RF power amplifiers. This limi-
calculations. It is written in Python in a modular fashion tation translates directly to challenges for the resonance
for easy adjustments, readability, and portability. Here we control system: detuning beyond 3 kHz occurs rapidly
describe the framework and present the first control re- under changes in average RF power, particularly in CW
sults for the PI-Test RFQ under pulsed and CW operation. mode, due to the frequency response of the vanes [8, 9].
In addition, the RFQ operates in both CW and pulsed RF
INTRODUCTION modes, resulting in variable RF heating. Other challenges
The resonant frequency of the RFQ may be maintained are imposed by the architecture of the water-cooling sys-
despite changes in RF heating through thermal control. tem. Transport delays and thermal time constants result in
For the PI-Test RFQ, we use an internal water-cooling open-loop settling times on the order of tens of minutes
system [1]. Both the thermal time constants and transport during normal operation (e.g. see Figure 1). Finally, the
delays present in such systems limit the efficacy of stand- coupling between the wall and vane circuits, the transient
ard PI control. The control problem is further complicated frequency response, the nonlinear valve flow curves, and
by the cavity geometry: different rates of thermal expan- fluctuations in the temperature of the cold water supply
sion and contraction of the main internal components (the make the system more difficult to control.
walls and the vanes/pole tips) result in a large transient The LLRF system [10] is capable of operating in either
frequency response under changes in average RF power. SEL mode (in which the drive frequency follows the
At present, the resonant frequency of other RFQs is regu- cavity resonant frequency) or GDR mode (in which the
lated with a PI loop around the vanes, while the walls are drive frequency is set). In SEL mode, the use of RF over-
held constant [2,3]. In contrast, a joint control loop that head is minimized due to the changing of the drive fre-

Copyright © 2016 CC-BY-3.0 and by the respective authors

governs both the wall and vane temperatures enables quency to match the RFQ, thereby also minimizing the
simultaneous exploitation of their individual impacts on reflected power. As such it is useful to switch into SEL
the resonant frequency. As discussed previously [1,4,5], mode automatically when the detuning increases beyond a
these system characteristics motivate the use of MPC. tolerable threshold.
In support of this, a dedicated control framework was Additionally, accommodation of multiple control algo-
developed to handle high-level decisions and execute rithms is desired. MPC frequency control will be the main
control calculations. Because multiple operational modes method; however, PI frequency control using the vane
are required, the framework is written in Python and in a valve is also desired as an auxiliary mode. In addition, for
modular fashion to facilitate easy modifications to the fast trip recovery it is useful to control the water tempera-
code. The framework interfaces with ACNET (Fermilab’s ture directly. Next, another desired mode is control of the
main control system) and the RFQ/cooling system via a RF forward power magnitude during a cold start or recov-
custom protocol generated with a novel protocol compiler ery from a trip. This would start out as a simple ramp, but
[6,7]. This framework is operational for the RFQ and could eventually be incorporated into an MPC routine.
could be modified for similar control tasks at Fermilab. In the event of an RF trip, the required recovery time
___________________________________________
for the RFQ is no more than 10x the length of the trip,
*Fermilab is operated by Fermi Research Alliance, LLC under Contract with a target requirement of 2x the duration of the trip.
No. De-AC02-07CH11359 with the United States Department of Energy. These constraints and desired system flexibility motivate
† [email protected] the development of a modular control system architecture
7: Accelerator Technology Main Systems
T06 - Room Temperature RF 109
 MOPOB17 Proceedings of NAPAC2016, Chicago, IL, USA ISBN 978-3-95450-180-9

that can be easily modified and would facilitate the use of nance controller sets the flow control valve apertures.
multiple control algorithms, including MPC. Figure 2 shows a simplified view of the system interfaces.

Figure 2: Simplified view of the system interfaces.

Framework Structure
Figure 1: Example of uncontrolled detuning in CW mode The framework itself consists of three main units: a
under a small change in cavity field (55 kV to 58 kV). main program, an action module, and a control module
(see Figure 3). The control module contains the control
FRAMEWORK DESCRIPTION algorithms and their helper functions. The action module
contains several action functions with computations and
Framework Tasks checks that need to be completed for different operational
The control framework handles a variety of high-level modes. The main program includes execution of initiali-
decisions related to RFQ operation, including: switching zation tasks, interpretation of user requests, receiving of
between SEL and GDR mode automatically based on inputs from the front-end, selection of action module
reflected power, measured detuning, and cavity field set- functions based on user requests and readings from the
point vs. read-back; setting the LLRF averaging windows; machine, and collection and sending of commands to the
switching between user-requested operational states; front-end.
detecting RF trips and taking appropriate action (e.g.
water temperature control) until power returns; and calcu-
lating the resonant frequency using a method appropriate
for the current operational state (i.e. SEL vs. GDR mode).
Framework Interfaces
The controller communicates with several other sub-
systems, including LLRF, a Cryo-con 18i temperature Figure 3: Modular control framework structure.
monitor [11], and a programmable logic computer (PLC)
for the water system. All of these communications occur Main Program Flow
through the ACNET control system via an ACNET Erlang The main program flow is as follows. (1) Upon startup,
front-end. This front-end is connected to the resonance the program builds a data buffer for MPC, sets the LLRF
controller via a User Datagram Protocol (UDP). We use a windows, and initializes variables. (2) At each execution
custom protocol (i.e. specific to this controller) generated
Copyright © 2016 CC-BY-3.0 and by the respective authors

cycle of the program, it gathers new readings, checks the

with Fermilab’s protocol compiler [6,7]. The front-end requested state, pre-processes some of the data, makes
then handles the communication with the variety of other relevant calculations (e.g. resonant frequency), and selects
subsystems through standard ACNET messages, thus the appropriate action function. (3) The action is execut-
hiding that complexity from the controller ed, which also involves calling the control module when
An operator can specify the desired operational mode applicable. Mode-specific settings are then sent back to
as an input to the controller. Error read-backs, a heartbeat, the main program. (4) Finally, the information is collected
and the present operational mode are sent back from the and commands are sent to the Erlang front-end. At pre-
framework to the user. From the LLRF system, readings sent, steps 2-4 repeat at a 1 Hz rate, but this rate may be
include the present SEL/GDR state, an indicator for increased in the future if need be. Figure 4 shows the
pulsed or CW mode, the forward and cavity phase read- program flow for resonance control specifically.
ings, RF repetition rate and pulse length, forward and
reflected RF power magnitudes, cavity field magnitude, INITIAL CONTROL RESULTS
and RF timing and averaging windows. The Cryo-con 18i
and PLC return temperature sensor readings, pressure and Recently, the main framework was commissioned and
flow readings, and flow control valve readings. The con- PI control over the frequency in both pulsed and CW
troller sets the LLRF system to SEL or GDR mode and mode was tested. This included frequency recovery after
sets appropriate window settings. For the PLC, the reso- RF trips, user-driven state switching, and automated

7: Accelerator Technology Main Systems

110 T06 - Room Temperature RF
ISBN 978-3-95450-180-9 Proceedings of NAPAC2016, Chicago, IL, USA MOPOB17

switching between SEL and GDR modes. At present, RF

recovery after a trip (that is, returning RF power to its
original level) is conducted manually.

Figure 5: PI frequency control during pulsed RF operation

for a 2-ms increase in pulse duration and a cavity field of
65 kV.

Figure 4: Overview of the program flow.

The PI controller regulates the resonant frequency of
the RFQ by adjusting the vane flow valve aperture. Be-
cause the flow responses of the valves are coupled and
nonlinear [1], a weighted correction is applied to the PI
control output. The correction is inversely proportional to
predicted change in flow for a given change in valve
setting. Because of the transport delays in the system, a
corresponding delay is placed into the frequency and
temperature PI controllers. The weighting of the PI gains,
the flow valve correction, and amount of delay were all
tuned during commissioning. Figure 6: Frequency control under CW operation under a
Figures 5, 6, and 7 show examples of the PI frequency small change in cavity field. This is roughly the same step
controller performance. In Figure 5, the pulse duration size as used for the uncontrolled response shown in
was increased by 2 ms while the cavity field was at 65 kV. Figure 1.
Figure 6 shows frequency control under CW operation for
a similar step in cavity field to that shown in Figure 1.
Figure 7 shows frequency recovery from a 10-second trip
at full specified field (60 kV), in which the detuning is
reduced to 3 kHz in 140 seconds. As such, PI control does
not fully meet the specification for trip recovery time at

Copyright © 2016 CC-BY-3.0 and by the respective authors

full field, although frequency trips at lower field values
can be recovered in under 10x the trip duration. Opera-
tional experience indicates that PI control may be brought
into specification by slightly delaying the activation of the
PI loop after power is restored.

CONCLUSION AND FURTHER WORK

PI resonant frequency control over the RFQ makes Figure 7: Frequency recovery from a 10-second RF trip at
stable CW and pulsed operation in GDR mode possible. 60 kV using PI control over the vane flow valve.
The next step of this work is to commission MPC. This
should provide compensation for water temperature fluc- ACKNOWLEDGMENTS
tuations, will enable finer control over the resonant fre- The authors thank R. Neswold and C. King for their
quency by leveraging both the wall and vane responses, work on the protocol compiler, as well as C. Baffes, M.
and should improve the frequency recovery time after a Ball, J.-P. Carneiro, E. Cullerton, J. Czaikowski, B. Han-
trip. Automated RF recovery will also be added. na, R. Pasquinelli, D. Peterson, G. Romanov, A. Shemya-
kin, P. Varghese, and T. Zuchnik.

7: Accelerator Technology Main Systems

T06 - Room Temperature RF 111
 MOPOB17 Proceedings of NAPAC2016, Chicago, IL, USA ISBN 978-3-95450-180-9

REFERENCES
[1] D. Bowring et al., “Resonance Control for Fermilab’s
PXIE RFQ,” in Proc. IPAC’16, Busan, Korea, May
2016, paper MOPMW026.
[2] P. N. Ostruomov et al., “Development and beam test
of a continuous wave radio frequency quadrupole ac-
celerator,” PRST-AB 15, 110101, 2012.
[3] W. Jing et al., “Multi-physics analysis of the RFQ for
injections scheme II of C-ADS driver linac,” Chinese
Physics C 38, 10, 2014.
[4] A. L. Edelen et al., “Neural Networks for Modeling
and Control of Particle Accelerators,” IEEE Transac-
tions on Nuclear Science, vol. 63, no. 2, Apr. 2016.
[5] A. L. Edelen et al., “Neural Network Model of the
PXIE RFQ Cooling System and Resonant Frequency
Response,” in Proc. IPAC’16, Busan, Korea, May
2016, paper THPOY020.
[6] R. Neswold, C. King, “Generation of Simple, Type-
Safe Messages for Inter-task Communications,” in
Proc. ICALEPCS 2009, Kobe, Japan, Oct., 2009.
[7] R. Neswold, C. King, “Further Developments in
Generating Type-Safe Messaging,” in Proc.
ICALEPCS 2011, Grenoble, France, Oct. 2011.
[8] A. R. Lambert, “FNAL PXIE 162.5 MHz RFQ
GENERAL RF,” LBNL Eng. Note 10773, 11 Jan
2013.
[9] A. R. Lambert et al., “High-Intensity Proton RFQ
Accelerator Fabrication Status for PXIE,” in Proc.
IPAC‘15, Richmond, VA, May 2015, paper
WEPTY045.
[10] J. P. Edelen et al., “Low level RF control for the PIP-
II Injector Test RFQ”, NAPAC 2016, Chicago, IL,
this conference.
[11] User’s Guide, Model 18i/14i/M12i Cryogenic Tem-
perature Monitors, Revision 3c. Cryogenic Control
Systems, Inc., 2016.
Copyright © 2016 CC-BY-3.0 and by the respective authors

7: Accelerator Technology Main Systems

112 T06 - Room Temperature RF