Preamble

Development of an EPICS-based Automatic Superconducting Module Test System for SHINE Chen Luo, Kai Xu, Zheng Li, Ya-wei Huang, Yan-fei Zhai, Hai-long Wu, Hong Wu,Xu-ming Liu, Xiao-han Ou-yang, Zhen-yu Ma, Xue-fang Huang, Hong-ru Jiang, Zhi-gang Zhang, Wen-feng Yang,Qiang Chang,Xin-yu Li, Lei Gao, Xiang Zheng, Yu-bin Zhao, and Shen-jie Zhao 1 Shanghai Advanced Research Institute, Chinese Academy of Sciences, Shanghai 201210 ShanghaiTech University, Shanghai 201210 Shanghai Apactron Particle Equipment Co. Ltd, Shanghai 201815 The Shanghai High Repetition Rate XFEL and Extreme Light Facility (SHINE) is a high-frequency hard X-ray free-electron laser (FEL) facility with a maximum repetition rate of 1 MHz, delivering photon energy from 0.4 to 25 keV. A critical component of SHINE is the superconducting module. Given the complexity of its internal components and structure, rigorous and precise evaluation of its performance is essential prior to installation. To evaluate its performance, we developed an automated module test system based on the Experimental Physics and Industrial Control System (EPICS). This comprehensive test system integrates various key technical domains, including radio frequency (RF), cryogenics, vacuum, electronics, radiation, and control.

The test control framework employs simple network management protocol(SNMP) for software architecture, StreamDevice for standardized device driver access, PyDM for user interface, network time protocol(NTP) for timing synchronization, and Python scripting for test automation. The test system has been successfully applied to test multiple superconducting modules at our institute. This automation significantly reduces testing time and simplifies procedures, standardizes procedural execution, ultimately enhances the overall test consistency and efficiency, while also providing data acquisition that beam operation data analysis.

Keywords

SHINE· EPICS·Module Test·RF Test·Automation

INTRODUCTION

The SHINE is currently under construction and is designed to be a high repetition rate hard X-ray facility. The SHINE consists of a linear accelerator section, two undulators, two

beamlines, four cryogenic stations, and other infrastructures. 5

The linear accelerator is divided into five sections (L0–L4).

The beamline FEL-1 provides photo energy from 3.0-15.0 keV, and the beamline FEL-2 delivers photo energy from 0.2-

3.0 keV. The cryogenic station includes three 4KW cooling 9

capacities at 2k for the SHINE tunnel and 1KW cooling ca- pacities at 2K for the test halls. The general layout of SHINE is shown in Fig. . The injector section (L0) comprises one

twin-FPC cavity module (i1CM) and one nine-cell cavity 13

module (i8CM). The L1 section includes two 1.3 GHz mod- ules and two 3.9 GHz modules, while L2-L4 sections com- prise 52 1.3 GHz modules [ ]. Central to its design is the accelerator based on superconducting radio-frequency (SRF) technology, requiring the mass production and precise evalu- ation of numerous superconducting cryomodules. The mod- ule includes eight superconducting cavities, eight fundamen-

tal power couplers(FPC), eight tuning devices, sixteen HOM 21

couplers, one cryomodule system (including thermostat, vac- uum pipe system and vacuum vessel), a beam position mon- itor (BPM), a superconducting magnet and support system ]. The layout of the module is presented in Fig. . Before assembly, cavities are tested in both bare and dressed states through vertical tests (VT), and only qualified cavities are as- sembled to modules. One module is tested over 20 days, and

∗ This work was supported by Shanghai Municipal Science and Technology Major Project (No. 2017SHZDZX02).The SHINE project, and the Na- tional Natural Science Foundation of China (No. 12125508).

only those meet the specifications are installed in the SHINE tunnel. The assembly and testing of modules are conducted in two halls on the campus of the Shanghai Advanced Research Institute. The two halls,which have a combined area of 3,000 square meters, are equipped with four vertical test platforms (VTF1/2 in ATF1 and VTF3/4 in ATF2) and four horizontal test platforms (HTF1/2 in ATF1 and HTF3/4 in ATF2). The layout of the halls is shown in Fig. . It can assemble and test three modules per month in the halls. It has already completed more than twenty modules assembly and test. The module test platforms include radiation protection, vacuum, cryogen- ics, power sources, low-level RF(LLRF), RF interlocks, RF test, BPM, and SC magnets. The module test objectives are to verify: (1) whether all components of the module meet the design requirements; (2) whether the assembly process meets the specifications; (3) whether the module meets beam oper- ational requirements for installation in the tunnel.

EPICS is a control system developed by Los Alamos National Laboratory (LANL) and Argonne National Labo- ratory (ANL) [ EPICS is a standard model known for portability, expansion, and reusability. It has been ap- plied in hundreds of physics experimental facilities, such as Beijing Electron-Positron Collider (BEPC), Shanghai Syn- chrotron Radiation Facility (SSRF), Linac Coherent Light Source (LCLS), and others.

The stringent operational requirements of SHINE, particu- larly its high repetition rate, require that each superconduct-

ing module operate within extremely tight tolerances. Manual module testing procedures are time-consuming, prone to hu- man error, and struggle to ensure high repeatability across a large number of modules.

To mitigate these challenges and ensure quality assurance, an integrated and fully auto- mated test solution is crucial. This paper describes the design and implementation of an EPICS-based control system devel- oped specifically for module test. The system adopts reliable scripting to achieve standardized device access, test execu- tion, and real-time data archiving.

MODULE TEST REQUIREMENTS AND IMPLEMENTATION Module test requirements The module test objectives are defined by SHINE accep- tance criteria, covering accelerating gradient, heat load, high order mode(HOM), tuner performance, quadrupole magnet, BPM, cryomodule, and vacuum [ ]. The acceptance crite- ria are illustrated in the following table Specific requirements are as follows: •Measurement of the superconducting cavity spectrum and mode frequency; •Characterization of FPC conditions at room temperature and •Determination of the external quality factor of the FPC (Qe); •Cavity performance tests,Eacc vs Q0, Eacc vs X ray, Eacc vs current; •Automatic generation of test reports from raw data.

Item Parameter Acceptance criteria Gradient Eacc ≥ 20 MV/m Vc ≥ 166 MV Dc < 30nA Heat load Q0@Vc=166MV ≥ 3 . 0 E + 10 Stability Test 10h or FPC Temp. achieve equilibrium 2K load Static < 20W dynamic(166MV) < 120W Total < 140 W Temp.

T FPC cold < 150K T infra < 80 ◦ C Residual Magnetic Magnetic flux < 0.2 mGs@300K High oder mode Qe > 3.0E+11 Phom < 1.4W@20MV/m Tuner Slow tuner range 1300MHz ± 0 . 02MHz Piezo range 0-500Hz BPM impedance 50Ω ± 3Ω S21 < − 55 db @0 − 2 GHz Vacuum Beam vacuum <1.3E-7 Pa at 2K <1.3E-5 Pa at room temp.

Coupler vacuum <6.5E-6 Pa at 2K <1.3E-5 Pa at room temp.

Isolation vacuum <5E-4 Pa at 2K <5E-1 Pa at room temp.

Radiation GM <1mSv@Eu 2m far away from cavity centre

Implementation of the module test Platform There are many systems involved in the module test plat- form. The module test diagram is illustrated in Fig. . The module test platform integrates: •The solid-state power amplifier (SSA) provides cavity power with a maximum output of 5.2 kW, which is connected to the FPC through the waveguide.

•The LLRF controls the forward signals and feedback signals of the cavity via an FPGA-based loop, allowing the cavity to achieve resonance through the Self-Excited Loop (SEL)[

•The RF interlock protects the cavity and FPC by monitoring 98

quench, temperature, electronics, and other protection signals 99

to shut down the SSA during faults.

•Radiation is monitored using a Geiger–Müller counter posi- 101

tioned two meters from the center of the cavity. •Dark current is measured with Faraday cups at both ends of the module (feedcap and endcap), capable of detecting nanoampere-scale currents generated by radiation. •The cryomodule is regulated by valves and heaters to con- trol the levels of liquid helium and pressure. The static load is calculated from heater power and flowmeter rate, while the dynamic load is derived from applied RF power and flowme- ter measurements[ •During the module test, RF test is the Primary tasks. It takes a whole process during the test period. The cavity has four ports: one input power port Pf, one extraction power port Pt, two HOM ports . It measures power val- ues from eight cavities: PF, PR, PT, , total- ing 40 values. The cavity acceleration gradient at the resonant frequency and the loaded quality factor are also indicated. For long-term operation, the heating effect of the FPC, we use for- mula ( ) to calculate based on the PF and PT values in this test system. It has to use to calibrate Qt.

Qt can be calculated according to formulas ( ) and ( ).The following equations are obtained for ( ) and (

E acc =

E acc =

L is the effective acceleration length of the cavity, r/Q is the cavity geometric factor, while the loaded quality factor. Ac- cording to Eq. ( hompu

1 Q

For the strongly coupled cavity, according to Eq. ( hompu

1 Q

Q L = 2 ∗ π ∗ f 0 ∗ τ (6) 135

is the cavity resonace frequency, is the cavity decay The RF test schematic diagram is illustrated in Fig. . The frequency of each cavity is measured by a frequency counter.

The power value of each cavity is measured by the power me- ters. Tao is obtained from the LLRF sel loop. It is calculated from the PT decay time. To reduce the number of power me- ters and frequency counters, a seven one-eighths RF switch controller is used to switch the power of the hom port, as well as the cavity PF/PT frequency, spectrum, and oscilloscope.

The 3.9 GHz module in HTF4 is shown in Fig.

SOFTWARE IMPLEMENTATION Software architecture In this system, an SNMP agent is applied on each device to collect data and respond to requests from the SNMP manager that runs on both the device and the IOC server[ ]. Time

synchronization is achieved using the Network Time Proto- 154

col (NTP), while Management Information Bases (MIB)[ are used to query device status information and returned to the SNMP manager. The SNMP acquires these response data and maps them to process variables (PV) in the real-time database (db) file. These PVs are accessed and visualized through the operator interface (OPI), developed with PyDM and connected via the Channel Access (CA) protocol[ ]. In addition, PV data can be retrieved from the Archiver Appli- ance through its HTTP interface for historical data storage and retrieval. The software architecture of the RF test system is illustrated in Fig. . The SHINE archiver is shown in Fig.

Device Control StreamDevice is an EPICS support module designed for

devices with byte-stream communication interfaces. It incor- 170

porates an asynDriver interface to handle underlying commu-

nication protocols [ 20 , 21 ] and can be extended to support 172

additional bus drivers. In the RF test system, multiple de- vices are integrated, including: •ROHDE&SCHWARZ N8PRS power meter (USB 3.0 inter- face) •KEYSIGHT 53230A frequency counter (Ethernet interface) •RF switch controller (Ethernet interface) •KEYSIGHT EXA spectrum analyzer (VNC server interface) Except for the spectrum analyzer, which uses VNC for

communication, the other devices are controlled through the 181

StreamDevice. The StreamDevice protocol uses plain ASCII text to execute commands during IOC startup. Protocol files are written based on hardware control commands. An exam- ple of a power meter driver is as follows:

terminator = CRLF; 186

ReplyTimeout =1000; 187

ReadTimeout = 1000; 188

Separator = ","; 189

reset{ out "*rst"; start{ out "INIT: CONT %d"; getpower{

out "FETCH?"; in "%f"; Data Acquisition and Storage Data acquisition is carried out using the EPICS real-time

database, which communicates with the device layer over the 203

network to enable data collection and storage. Upon request from the OPI client, real-time data are transmitted and dis- played. The data are stored in the database file [ ]. An example of a real-time database record file is shown below: record (ao "(DEV) powermeter ") field (DESC, "set off value(dB)") field (EGU, "W") field (SCAN, "Passive") field (DTYP, "stream") field (VAL, "1") In this configuration:

The DESC field describes the signal name. The EGU field

specifies the engineering unit. The DTYP field declares the 219

device type. The SCAN field determines the scanning in- 220

terval. The minimum period is set to 0.1 seconds to ensure 221

accurate data acquisition and display. To improve reusabil- ity, the database file employs macro replacement. For exam- ple, (DEV) identifies the device, and (POW-port) loads spe- cific parameters through the dbLoadRecords command in the IOC startup script. Historical data storage is implemented us- ing the Archiver Appliance, which archives millions of PV from all systems. During module tests, real-time PV data are archived by PV name, ensuring long-term accessibility and enabling test reports to be generated from stored data.

User Interface PyDM is a PyQt-based framework for developing user in- terfaces in control systems. It supports both a drag-and-drop screen builder for simple applications and a Python-based framework for building complex interfaces[ ]. The com- bination of Qt’s cross-platform features and Python’s flexibil- ity makes PyDM effective for operator interfaces in EPICS environments. The OPI of test HALL1 is presented in Fig.

The user interface of the RF test system is divided into four sub-interfaces based on their functions for operation and dis- play. The RF test OPI is divided into four functional sub- interfaces:

•Sub-interface 1: Displays cryogenic parameters, RF param- 243

eters, radiation parameters, vacuum parameters, dark current parameters. •Sub-interface 2: Provides tools for superconducting cavity calibration. •Sub-interface 3: Supports channel selection on the 1-in-8 RF

switch, oscilloscope monitoring, data saving for cavities and 249

couplers, and calibration of the power meter. •Sub-interface 4: Configures measurement settings such as power meter functions, time and pulse points, startup/shut- down control of devices. The OPI of RF test sub-1 is shown in Fig. , sub-2 is shown in Fig. .sub-3 is shown in Fig.

E. Time synchronization 257

The time synchronization requirements depend on the 258

experiment[ ]. High precision applications require nanosec- ond accuracy, whereas millisecond precision is sufficient for module testing.

The NTP[ ], a TCP/IP application

layer protocol, is widely used to synchronize clocks between 262

clients and servers. A dedicated NTP server provides syn-

chronized system time. Timesyncd is adopted instead of 264

NTP in the RF test system. Timesyncd connects to the same time server within the Ubuntu operating system, offering a lightweight and well-integrated solution for millisecond-level

synchronization. 268

The NTP server configuration for timesyncd is located at:

TEST AUTOMATION Automation scripts written in Python, with PyDM-based interfaces:

Cavity tuning automation: A two-step coarse-to-fine algo- 274

rithm aligns all eight cavities in -mode.

FPC conditioning automation: Power ramping in both 276

pulse and continuous Wave(CW) modes ensures coupler sta- bility at room temperature and 2 K.

During the module test, the automatic conditioning of the 279

cavity addresses three key aspects: (1) For the TESLA cavity type, the multipacting (MP) [ ]region between 17 and

24 MV/m requires CW conditioning of each cavity individ- 282

ually; however, automatic conditioning can be performed on 283

eight cavities simultaneously. (2) Sporadic quenches occur- ring within the MP range are quickly recovered, resulting in a stable accelerating gradient (Eu). (3) When MP reappears after warm-up and cooldown of the modules in the tunnel, an automatic condition is needed. Using an automatic program

for cavity conditioning could reduce module test time. 289

A. Cavity Tuning Automation 290

The program is used to automatically determine the phase

difference of the cavities, enabling the simultaneous tuning of 292

eight cavities in -mode. The program is written in Python, with the user interface developed with PyDM. The cavity tun-

ing procedure consists of two stages. First, coarse tuning is 295

performed in 10° increments to locate the maximum value of

PT. Then, fine-tuning is carried out in 0.1° increments within 297

3° range around the resonance point to accurately con- firm the resonance phase by identifying the peak value of PT.

The flowchart of the tuning process is shown. 13 . The user 300

interface is presented in Fig. FPC automatic condition system Each FPC must be conditioned both at room temperature

and at 2 K in the detuning cavity. Conditioning at room tem- 305

perature is divided into two modes: pulse and CW. It needs to be conditioned in CW mode at 2K. A Python script is devel- oped to automate this process, with the interface implemented

in PyDM. For pulse conditioning, the power is increased from 309

500 W to 4 kW in 500 W steps, with a repetition rate (RR) of

10 Hz and duty from 5% to 20%. For CW conditioning, the 311

power is raised from 500 W to 4 kW in 500 W increments,

with each step maintained for 30 minutes. Conditioning is 313

completed at 4 kW when either the temperature of the cou-

pler’s cold section reaches 323 K at room temperature, or the temperature rise of the cold section at 2 K is less than 0.1 K within 1 hour. The process flowchart is shown in Fig.

Cavity automatic condition system The process is carried out using a Python script, with the user interface built with PyDM. The cavity condition proce- dure starts at 16.5 MV/m and progresses to 26 MV/m in 0.1 MV/m increments, each step for 2 minutes.

When cavity quench[ ] appears, the cavity automation system reset the interlock and SSA. If another interlock occurs, it will quite the automatic programm, and an audible alert is triggered.

The automatic system increases Eacc for its useable only for less than two hours. The user interface is presented in Fig.

The result of cavity automation conditioning for CAV2 is il- 330

lustrated in Fig.

CONCLUSION

In this work, we presented the development of an EPICS- based test system for the SHINE module. To date, more than 20 superconducting modules have been successfully tested in this system. For the 3.9 GHz modules, the result is: Vc

36MV, Q0 >2E+9@36MV. For the 1.3 GHz modules, the 338

result is: Vc >166MV for all modules except CM10, Q0 339

3E+10@166MV for all modules except CM2 when MP 340

quenches occurred before Q0 test. All modules, including i1CM, i8CM, 3.9GHz and 1.3GHz modules, have already been tested in ATF1 and ATF2. The i1CM and i8CM have been installed in the injector section. The injector achieved 100 MeV on October 30, 2024 [ ]. The CM1, CM2, and 3.9GHz modules have been installed in the L1 section, where

beam commissioning is ongoing. The CM3–CM15 are 347

currently beening installed in the L2 section. 348

The automation dramatically enhanced the efficiency and standardization of the procedure. Cavity condition tests showed that a complete performance, which previously required 72 hours of continuous manual operation, is now executed autonomously in approximately 28 hours, repre-

senting a 61% reduction in testing time and a significant 354

decrease in operational labor costs. The upgraded automated test system has demonstrated that the automation yields a dramatic improvement in accuracy and repeatability, critical for the quality control of SHINE’s large-scale module deployment.

Future work will focus on integrating data analysis[ ] and predictive maintenance algorithms into the control loop to further optimize the operational efficiency and reliability of the test facility.

Acknowledgments

We thank the RF, cryogenics, LLRF, vac- 364

uum, mechanical, electronics, and radiation-safety teams for on-site 365

support during all module test, and the control team for maintaining 366

EPICS and Archiver Appliance. We also acknowledge the partner company for its efforts in testing time and collecting data.

C. Feng, H.X. Deng, Review of fully coherent free- electron lasers.

Nucl. Tech. (2018). Y.W. Gong, M. Zhang, W.J. Fan, et al., Beam performance of the SHINE dechirper. Nucl. Sci. Tech. 32(3), 29 (2021).

J.F Chen, B Liu, D Wang, et al. Cryomodules development for SHINE project, WEPS42 H.T. Hou, J.F. Chen, et al. Prototypes Fabrication of 1.3 GHz Superconducting RF Components for SHINE Project, StreamDevice Software Documentation.

Avail- able: (accessed Aug. 2025). Argonne National Laboratory, EPICS – Experimental Physics Industrial Control System.

Available: (accessed Aug. 2025). F. Schölz, C. Adolphsen, J. Carwardine, et al. Cryomod- ule testing for LCLS-II. Proc. SRF2017 [C], Lanzhou, China, 2017, pp. 747–751.

SRF2017-THPB030

[8] C. Pagani, P. Pierini. Testing of superconducting RF cav- 392

ities and cryomodules for particle accelerators. Supercond.

Sci. Technol. 30, 023001 (2017). 6668/30/2/023001 Zhao, Ultra-high quality factor and ultra-high accelerating gradient achievements in a 1.3 GHz continuous wave cryomodule. Nucl. Sci. Tech. 36, 2 (2025).

Z.Y. Ma, S.J. Zhao, et al. High RF power tests of first fundamental power coupler proto- types for the SHINE project. Nucl. Sci. Tech. 33, (2022).

Branlard J. Low-level radio-frequency systems for accelera- tors. Nucl Instrum Methods Phys Res A, 768: 98–108, 2014.

[12] Q. Ma, L. Wang, J. Yang, and Y. Li, "Cryogenic system for su- 406

perconducting radio frequency cavities," Nucl. Sci. Tech, vol. 31, no. 11, p. 111, 2020, L. Wang, Q. Ma, Y. Li, and J. Yang, "Dynamic heat load mea- surement of a superconducting cryomodule," Nucl. Sci. Tech, vol. 29, no. 9, p. 109, 2018,

[14] W. Pan, R. Ge, M. Li et al., Cryogenic system of the high per- 414

formance 1.3 GHz 9-cell superconducting radio frequency pro- totype cryomodule. Appl. Thermal Eng. 255, 124042 (2024).

Padamsee H, Knobloch J, Hays T,etal. RF superconductivity for accelerators[M]. New York: John Wiley, 1998.

Hasan Padamsee. Superconducting Radio-Frequency Cavities [M]. Weinheim: Wiley-VCH, 2008.

BIBLIOGRAPHY J. Case, M. Fedor, M. Schoffstall, J. Davin. A Simple Network Management Protocol (SNMP). RFC 1157, IETF, 1990. Avail- able:

K. McCloghrie, M.T. Rose. Management Information Base for Network Management of TCP/IP-based Internets: MIB- II. RFC 1213, IETF, 1991. Available:

M. Davidsaver, A. Johnson, E. Norum, et al. EPICS: Evolution and future directions. In: Proc. ICALEPCS, Barcelona, Spain, 2011, pp. 293–296 [C].

ICALEPCS2011-MOAOB01 StreamDevice: General purpose device support for de-

vices with a byte stream based communication interface. 434

Scherrer Institute, Switzerland. Available online: (accessed 20 Aug 2025).

Zimoch D. StreamDevice – a device support module for com-

municating with devices via ASCII-based protocols. In: Proc 439

Int Conf Accelerator and Large Experimental Physics Control Systems (ICALEPCS 2003), Gyeongju, Korea, 2003.

L.R. Dalesio, M.R. Kraimer, A.J. Kozubal. EPICS architecture.

In: Proc. ICALEPCS, Berlin, Germany, 1991, pp. 278–282 Z.L. Zhao, H.C. Xu, P.R. Gong. Beamline data manage- ment system based on Archiver Appliance. Nucl. Tech. 41(3), (2018). (in Chinese).

Dalesio, Davidsaver, Clausen, PyDM: Python Display Manager for Control Systems. In:

Proc. ICALEPCS, Barcelona, Spain, 2017, TUPHA093 SLAC Laboratory. PyDM – Python Display Manager. 2022- 11-17. Available:

M.Q. Zeng, L. Yin, C.C. Yin, et al. Development and im- plementation of an EPICS-based timing equipment control system for SHINE beamlines and endstations. Nucl. Sci.

Tech. 46(7), 070102 (2023). (in Chinese). D.L. Mills. Network Time Protocol (NTP), Version 3: Specifi- cation, Implementation and Analysis. RFC 1305, IETF, 1992.

Available: Ubuntu. Network

chronization Configuration [EB/OL]. Avail- 465

able: (accessed Aug. 2025) H. Weise. European XFEL superconducting RF gun and cry- omodule tests. Nucl. Instrum. Methods Phys. Res. A 768, 1–6 (2014).

A. Gurevich, “Multipacting, RF heating and other mech-

anisms of surface resistance in superconducting radio fre- 472

quency cavities,” Accel. Tech. (2012).

[31] TESLA Collaboration. TESLA Technical Design Report, Part 475

II: The Accelerator. DESY 2001-011, Hamburg: DESY, 2001.

Hard X-ray free-electron laser device injector achieves 100 acceleration.

Posen,A. Cravatta,et “High gradient perfor- mance quench behavior verification omodule energy continuous MySQL. (accessed Aug. 2025). accelerator,” Phys.

Accel. Beams