LLRF Topical Workshop on Advanced Feedback Control (TWoAFC)

Oct 13, 2026, 8:00 AM → Oct 15, 2026, 6:30 PM Europe/Rome

Nova Gorica, Slovenia

The topical series of LLRF workshops is meant to bring together a diverse group of experts spanning disciplines beyond the traditional LLRF field in particle accelerators. The intent is to provide a basis for understanding the technologies and constraints which drive critical engineering decisions in multiple disciplines.

This year's event will focus on Advanced Feedback Control: Control systems are used across a wide range of applications, from simple proportional–integral (PI) loops to more advanced control architectures. In many cases, achieving the best performance requires a systematic, control-theory-based optimization approach. A careful and rigorous design and tuning process can substantially improve performance and robustness, often translating directly into better efficiency and overall system reliability. Bringing together the LLRF, orbit feedback, instabilities control, and collision point stabilization with other communities such as quantum computing, astrophysics or autonomous driving to share experiences and discuss new ideas and trends.

This event is the third topical workshop organized as a spin-off of the bi-annual LLRF workshop series. The LLRF workshop series was established in 2001 and starting in 2018, the scientific program committee began organizing a more focused, topical, event in the year between the main workshops.

Topics we will dive into: 

• Control Architecture and Strategies
• Control Hardware Platforms
• System Identification
• Control Loop Modeling, Tuning and Optimization
• Data driven control
• Sensors and actuators
  
  

Workshop timeline

The workshop will begin on the morning of Tuesday, 13 October, and conclude at 18:00 on Thursday, 15 October.

The programme will include a welcome reception on Tuesday evening and the workshop dinner on Wednesday evening. On Thursday morning, participants will visit the FERMI facility before returning to Nova Gorica for the final sessions and conclusion of the workshop programme.

This edition of LLRF Topical Workshop is sponsored by: 

LLRF-Workshop-poster-A2

Tue, October 13

8:00 AM Registration

9:00 AM Welcome & LOC Announcement

Welcome talk

1 Elettra Sincrotrone Trieste and its advanced light sources

Elettra Sincrotrone Trieste is a multidisciplinary research center of excellence, open to the international research community, specializing in generating high quality synchrotron and free-electron laser light and applying it in materials and life sciences. This talk will introduce the Laboratory and provide an overview of its two advanced light sources, FERMI, the seeded free-electron laser facility, and the ongoing Elettra 2.0 Project, the major upgrade of the Elettra synchrotron light source aimed at delivering significantly enhanced photon brightness and performance.
Further to an overview of the scientific mission of the facilities, the presentation will highlight the role of RF technologies in delivering reliable, stable and high-performance beams to the user community.

Speaker: Alessandro Fabris (Elettra Sincrotrone Trieste S.C.p.A.)

Control Loop Modeling, Tuning and Optimization

2 Minimizing the RF Beam Component During Bunch Splitting in the EIC

The Hadron Storage Ring (HSR) of the Electron-Ion Collider (EIC) will perform a 1 to 4 bunch split to reach its luminosity goals. After the bunch splitting, the beam is compressed by 197 MHz and 591 MHz RF systems. It is important that, during the splitting process, the voltage on the compression cavities stays close to 0 V to prevent any detrimental effects on the beam emittance and bunch-by-bunch charge variation. During this bunch splitting process, the beam current at 197 MHz reaches a peak of 1 A while the 197 MHz cavity voltage is still at 0 volts. Feedback is used to keep the cavity voltage near 0 V, however, there is not enough installed power to cancel the RF component of the beam. To fight this beam component, a “defocusing cavity” at 197 MHz is introduced to provide bunch shaping such that the beam component at 197 MHz is minimized. Simulations are performed to determine the voltage and phase program on the defocusing cavity throughout the splitting process. Regulation in the presence of transient-beam loading is also studied.

Speaker: Arshdeep Singh (Brookhaven National Laboratory)

3 A new scheme for beam-loading compensation at HL-LHC injection

Capturing the high-intensity proton beams in the HL-LHC era will be challenging for the RF system. The increase of the bunch intensity to $2.3 \times 10^{11}$ protons per bunch will require significantly more RF power for beam-loading compensation. At the same time, a higher RF voltage must be provided to capture and retain the beam at the injection plateau. Together, these two factors will push the peak RF power beyond the limit of the present klystrons. Capture of equidistant bunches from the injectors currently relies on forcing a constant RF voltage in amplitude and phase, for which only the half-detuning compensation scheme is applicable. This detuning leaves the beam-induced voltage only partially compensated. However, by modulating the RF phase along the turn, one can fully counteract the detuning caused by beam loading. We propose a novel technique that applies RF phase modulation at the bunch-to-bucket transfer to reduce the RF power during the injection process by up to a factor of two. We present simulations combining longitudinal beam dynamics and detailed time-domain models of the SPS and LHC RF feedback systems. Results from first beam tests in the LHC will also be shown.

Speaker: Birk Emil Karlsen-Baeck (CERN)

10:45 AM Coffee break

Control Loop Modeling, Tuning and Optimization

4 RF cavity feedback for high-intensity counter-rotating beams in the proposed muon collider RCS chain

The proposed muon collider complex is designed to accelerate high-intensity muon bunches to energies of several TeV. Due to the muon's short rest-frame lifetime, acceleration must be performed quickly. Thousands of superconducting cavities are required for the chain of rapid-cycling synchrotrons. Such large RF systems in this high-energy acceleration section pose the challenge of a very high fundamental-mode impedance. The resulting beam loading must be mitigated to preserve beam stability. The high-energy chain features a very short acceleration time of a few milliseconds and a high bunch population of $2.7\times10^{12}$. The feedback therefore faces particular challenges, as the RF system operates in a regime between those typically encountered in linear accelerators and synchrotrons. Additionally, both counter-rotating bunches are accelerated by the same RF system, causing the fields to interact in the cavities. This contribution presents the initial modelling of the cavity feedback system using the longitudinal beam dynamics code BLonD.

Speaker: Leonard Thiele (University of Rostock, CERN)

Tutorial

12:30 PM Lunch

5 Vendor no. 1

Control Loop Modeling, Tuning and Optimization

6 Analysis and Suppression of Ponderomotive Instabilities in SRF Cavities

Superconducting radio-frequency (SRF) cavities with high loaded quality factors are particularly susceptible to ponderomotive instabilities caused by the interaction between RF fields and mechanical vibrations. Predicting and suppressing these instabilities is essential for reliable operation of high-gradient continuous-wave accelerators. To address this issue, we have developed a practical analysis framework for SRF cavities under low-level RF feedback control. It combines electromechanical system identification, stability modeling, simulation, and experimental verification. Measurements on a 162.5 MHz SRF linac identified dominant mechanical modes near 118 and 280 Hz, which were included in open- and closed-loop models. Tests on two cavities with four PI gain settings showed good agreement between predicted and measured instability thresholds. The model also guided controller retuning after improper feedback gains caused frequent faults, reducing the daily fault rates from approximately 17 and 5 to below 0.4. The proposed approach provides practical guidance for LLRF tuning and improves the reliability of high-gradient SRF accelerators.

Speaker: Jiayi Peng (Institute of Modern Physics)

7 Simulation Study of Three-bunch Operation Methods for SwissFEL Upgrade

PSI is investigating a SwissFEL extension with a new hard X-ray beamline, Porthos, in addition to the existing hard X-ray (Aramis) and soft X-ray (Athos) beamlines. To allow all three beam lines to operate simultaneously at the maximum RF pulse repetition rate (100 Hz), three bunches must be accelerated in each RF pulse. The three bunches will be separated by 21 ns. The accelerating voltages and phases for the three bunches need to be independently adjustable to tune the FEL properties. This poses significant challenges for the LLRF system, especially at RF stations equipped with pulse compressors. To make the three-bunch operation feasible, the machine layout has been updated by introducing a dedicated bunch compressor for the Athos beamline. In this work, an acceleration model for the traveling-wave structures is developed, allowing the accelerating voltages and phases for all bunches to be simulated for arbitrary pulse shapes. We performed a start-to-end co-simulation including all RF stations and the tracking of all bunches, serving as a virtual machine based on the actual RF operation parameters of SwissFEL. This model allows us to simulate the control methods, including the tuning knobs for all bunches and the beam-based longitudinal feedback loops. This contribution presents the modeling and simulation results mentioned above.

Speaker: Zheqiao Geng (PSI - Paul Scherrer Institut)

8 Multivariable Integration of RF Phase Actuation into Fast Orbit Feedback

Next-generation storage-ring light sources combine higher-bandwidth fast orbit feedback (FOFB) systems with synchrotron frequencies that may lie within the FOFB bandwidth. Coupled-bunch mode-zero oscillations driven by RF phase and amplitude noise can produce horizontal orbit motion within the controllable frequency range. Existing work exploits the near orthogonality between the dispersion orbit associated with this motion and the betatron orbit space, allowing the disturbance to be extracted from beam-position measurements and corrected independently through the main low-level-RF phase set-point. This work instead investigates a fully integrated formulation in which magnetic and RF actuators are coordinated within a single multivariable FOFB design, and highlights advantages of this integrated approach. Next-generation systems already combine slow, high-authority corrector magnets with fast, lower-authority correctors through multimodal factorisations. We extend this architecture to RF phase actuation by treating it as an additional high-bandwidth actuator that accounts for its distinct spatial response.

Speaker: Idris Kempf (University of Oxford)

3:35 PM Coffee break

9 Vendor no. 2

Control Architecture and Strategies

10 Closed-loop Feedback Control of Coupled-bunch Instabilities

Bunch-by-bunch feedback control of coupled-bunch instabilities in the last 30 years became a ubiquitous feature of storage rings, both light sources and colliders. In my talk I will introduce the formalism of coupled-bunch instabilities and bunch-by-bunch feedback. Next, I will describe in detail the technology that these feedback systems rely upon. Using measurements from a number of real world machines, I will describe the commonalities between different applications as well as illustrate some of the unique challenges.

Speaker: Dmitry Teytelman (Dimtel, Inc.)

11 Model-Driven Control for High-Stability Magnetron RF Sources

Magnetrons offer high energy efficiency, making them attractive power sources; however, their RF output lacks the phase and amplitude stability required for high-precision applications like particle accelerators. While techniques like RF signal injection have been explored to stabilize magnetron outputs, this research introduces a Low-Level RF (LLRF) control framework designed to enhance signal quality further. This is achieved by combining injection locking with parallel control mechanisms, such as DC power supply modulation. To ensure efficient implementation, the proposed architecture leverages optimized digital processing, specifically utilizing 4-IQ sampling techniques on novel hardware architectures. Additionally, mathematical models are developed to enable controller synthesis following a Model-Based Design (MBD) methodology. By integrating injection locking, multi-variable parallel control, efficient digital processing, and model-driven design, this work seeds a robust and stable RF power source tailored to particle accelerators.

This ongoing work has been partially supported by the PID2023-148792OB-I00 project funded by MICIU/AEI/10.13039/501100011033 and by FEDER, UE.

Speaker: Jon Vivas (University of the Basque Country EHU)

12 Updates on an S-band Klystron Fast Feedback Loop for Plasma Wakefield Acceleration

Beam-driven plasma wakefield accelerators impose extremely stringent requirements on RF phase stability to ensure reliable beam generation and synchronization. At SPARC_LAB, a fast RF feedback system has been developed for the S-band RF power plant to satisfy the facility's synchronization requirement of less than 20 fs RMS phase jitter with respect to the reference master oscillator (RMO). The feedback system has been in operation since 2008 and has undergone several generations of upgrades to improve its performance and robustness. This work presents the latest developments of the fast feedback system together with experimental results demonstrating its phase noise suppression capability on the S-band klystron. Measurements show that the residual RF phase jitter can be reduced to below 15 fs RMS with respect to the RMO. These results demonstrate that the upgraded S-band fast feedback system is capable of meeting the femtosecond-level synchronization requirements of advanced beam-driven plasma wakefield accelerator at SPARC_LAB.

Speaker: Xianghe Fang (INFN-LNF)

13 Operational Experience with the Adaptive Feed-Forward at CERN Linac4

The CERN Linac4 low-level RF relies on an adaptive feed-forward (AFF) to correct the beam-loading transient that the field feedback (PI or LQG) cannot remove within a single pulse. The AFF learns a correction from the field error of previous pulses and reduces the beam loading at the batch start to below 0.4%, a factor of two better than the specification. Over several years of operation we have identified conditions in which the AFF learns a wrong correction: intermittent beam transmission through the cavity, pulses taken with the beam stopper inserted, breakdowns or faults that interrupt the beam, and outdated corrections left after a long pause or a change of settings. In such cases the beam quality degrades or the RF trips. We present these conditions together with the measures taken to keep the AFF running reliably without operator intervention. We added a beam-present input, restricted learning to pulses with the beam stopper out and no active interlock, introduced an automatic reset after a long pause or a change of settings, refined the correction for longitudinal painting, and tuned the parameters for greater robustness.

Speaker: Bartosz Bielawski (CERN)

SOCIAL EVENT: WELCOME RECEPTION

Wed, October 14

9:00 AM LOC Announcement

Control Loop Modeling, Tuning and Optimization

14 A Systematic Methodology for equivalent Circuit Modelling of Particle Accelerator Resonant Cavities for Control Applications

Resonant cavity models traditionally used for control design suffer from oversimplification, failing to capture spatial actuation effects or geometric imperfections. To address this limitation, we present a methodology for deriving cavity models via equivalent circuit representations. Most accelerator cavities (such as Tesla structures, RFQs, or DTLs) can be decomposed into elemental units (cavities, cells, or stems). The proposed framework models each unit and its
inter-component coupling using electronic circuits, leveraging electromagnetic software (e.g., CST, ADS) for parameter extraction. The resulting model improves the simulation efficiency due to its circuital representation. Furthermore, because the controller typically shares this circuit-based nature, the entire closed-loop system can be co-simulated within a single software environment. This enhanced model enables precise LLRF control that accounts for cross-coupling effects, laying the foundation for advanced control strategies.
This work was supported in part by the project PID2023-148792OB-I00
funded by MICIU/AEI/10.13039/501100011033 and by ERDF, EU

Speaker: Beñat González (University of the Basque Country UPV/EHU)

15 Comparative Analysis Of Varying Feedback Configurations For An RF System

The dynamics of an RF cavity control system can be modeled with a great deal of accuracy. This includes the dynamics of the cavity, beam+cavity interactions, microphonics, Lorentz-Force Detuning, the field control loop, and the tuning control loop. A time-domain simulator has been developed in Python to study the behavior of a complete RF control system with respect to varying feedback configurations, different controller types and control algorithms, and different operational conditions. Regarding field control: a comparison between traditional PID controllers, Active Disturbance Rejection Controllers (ADRCs), and Linear-Quadratic Regulators (LQRs) is performed. Metrics such as disturbance rejection, stability, peak power, and cavity field amplitude and phase transients are used to evaluate the performance of the controllers. A similar process is carried out for cavity tuning control loops.

Speaker: Arshdeep Singh (Brookhaven National Laboratory)

Control Hardware Platforms

16 When the Bare-Metal Returns: A Centralized Data Plane for Elettra 2.0 Control and Simulation

Elettra 2.0 control system will consolidate time-critical acquisition, processing, and actuation onto a dual HPC 512 HT core server. Using the DPDK kernel bypass, this data plane is designed to acquire Turn-by-Turn streams from 168 BPMs, 4 LLRFs and 120 BLMs, and drive about 1200 magnet power supplies at 100 kHz for RT feedback purposes, with a total of 160 Mpps data rate with a few microseconds jitter level over Ethernet. Moreover, the TbT streaming will enable the extraction of machine optics parameters continuously at up to 100 Hz rep-rate.
In order to validate the processing chain a test facility couples the real-time acquisition system connected to the real BPM electronics with a simulator running on the same machine to generate synthetic TbT data and feed the RT processing pipeline that is cross-checked with the simulator ground truth.
Commodity hardware ensures upgradability, while decades of machine physics algorithms remain in C code, accelerated through efficient, hardware-level execution.
When the bare-metal returns, the system radically simplifies and accelerates at the same time: commodity hardware below, optimized physics above, and no unearned layers in between.

Speaker: Giulio Gaio (Elettra Sincrotrone Trieste)

10:45 AM Coffee break

Control Hardware Platforms

17 RF Beam Noise Feedback System for the High-Luminosity LHC Crab Cavities

The High-Luminosity Large Hadron Collider (HL-LHC) project at CERN is set to introduce a series
of major upgrades across the LHC complex, including the implementation of crab cavities,
applied for the first time in a hadron machine. These cavities are designed to increase beam
luminosity by compensating for the geometric luminosity reduction caused by the crossing
angle at the interaction points.

However, the crab cavity controller and RF power stage introduce additional phase noise into
the system, which can lead to beam emittance growth if not properly mitigated. A new
solution, known as the beam noise feedback system, is therefore being developed to suppress
the impact of this noise.

This presentation will outline the main challenges of the project and introduce the mitigation strategy based on AMD RFSoC technology as well as some preliminary results obtained from the Super Proton Synchrotron.

Speaker: Dimitar Marinov (CERN)

Tutorial

12:30 PM Group Photo

12:35 PM Lunch

Control Hardware Platforms

18 The Development of Feedback Systems at SSRF and SHINE

Shanghai Synchrotron Radiation Facility (SSRF), a third-generation light source in operation since 2009, is upgrading its fast orbit feedback system with new MPSoC FPGA-based digital electronics, alongside developing an in-house bunch-by-bunch feedback system based on RFSoC FPGA. Meanwhile, Shanghai High Repetition Rate XFEL and Extreme Light Facility (SHINE), a 3-km-long continuous-wave superconducting linear accelerator currently under construction, is also implementing multiple fast orbit feedback loops using the same MPSoC FPGA platform. This workshop will share and discuss the latest progress on these feedback systems.

Speaker: longwei lai (Shanghai Advanced Research Institute, Chinese Academy of Sciences)

19 Obsidian: a compact platform for scalable and intelligent acoustic-band active stabilization control

Optimal beam parameters in high-power lasers and accelerators require active stabilization against environmental perturbations. Typically residing in the acoustic frequency band, these perturbations are corrected via piezo actuators or magnets. Common applications include coherent laser combining, laser pointing stabilization, resonance control in superconducting radio-frequency (SRF) cavities, and beam orbit regulation in storage rings. Recent experimental demonstrations of machine-learning-based stabilization in laser pointing and coherent laser combining have validated the concept of AI-enhanced active noise control. We present the Obsidian board, a general-purpose, open-source hardware platform and FPGA carrier designed for real-time control implementation. Supporting both Micro-Research Finland (MRF) timing and White Rabbit protocols, the Obsidian board interfaces with Low-Level Radio Frequency (LLRF) controllers or Beam Position Monitors (BPMs) through dedicated timing and Gigabit transceiver links. It executes hard real-time control algorithms while bridging MHz-to-kHz sampling rates for acoustic-band I/O. Key application demonstrations include MRF Event Generator (EVG) / Event Receiver (EVR) timing nodes, synchronous Machine Vision via a GeniCam camera interface, and laser pointing and coherence stabilization via FPGA-based ML pattern recognition.

Speaker: Qiang Du (Lawrence Berkeley National Laboratory)

Control Architecture and Strategies

20 Intra-Pulse RF Phase Stabilization in LLRF Systems

Achieving high RF field stability in accelerating cavities driven by microsecond-scale pulses remains a challenge for modern LLRF systems. The short pulse duration limits the implementation of conventional FPGA-based intra-pulse feedback, while pulse-to-pulse correction alone cannot fully compensate for phase perturbations introduced by the high-power RF chain.
An implementation of this novel feedback concept was developed and deployed at the INFN SPARC_LAB facility, representing the first realization of its kind. It is now routinely operated with state-of-the-art performance, demonstrating significantly improved phase stability compared with conventional pulse-to-pulse control.
We present a new approach for short-pulse operation, based on a different architecture that simplifies system integration while enabling intra-pulse feedback capabilities. The design and experimental validation are presented, including measurements performed at the INFN SPARC_LAB facility. This implementation extends the Libera LLRF platform with a dedicated solution for advanced feedback applications, showing strong potential for accelerators requiring improved field stability and pulse control.

Speaker: Borut Baricevic (Instrumentation Technologies d.o.o)

21 Lowpass Differentiator Controllers for the Longitudinal Bunch-by-bunch Feedback

Modern diffraction-limited storage ring light sources run at high beam currents and use harmonic cavities to lengthen the bunches for increased lifetime. These features make feedback control of the longitudinal coupled-bunch instabilities very challenging due to the bunches' wideband and variable dynamics. In the last few years, the long-ignored idea of using a differentiator as the feedback controller has made a comeback. I will outline the inherent control challenges, discuss the feedback controller design methodology, and present experimental results from several machines.

Speaker: Dmitry Teytelman (Dimtel, Inc.)

3:45 PM Round table

Poster Session & Coffee Break

22 Modernization of the SPEAR3 RF Control System: From PEP-II Heritage Analog LLRF to a Distributed FPGA-Based Architecture

The Stanford Synchrotron Radiation Lightsource (SSRL) is completing a comprehensive modernization of the SPEAR3 RF control system. The legacy system, originally designed for the PEP-II B-Factory circa 1997, has operated nearly unchanged for three decades. Its core is a custom analog RF Processor (RFP) in a VXI chassis performing I/Q feedback at approximately ±90 kHz bandwidth, supplemented by dedicated Comb Filter (CFM), Gap Voltage Feedback (GVF), and Gap Feed-Forward (GFF) analog modules. Supervisory control runs as six SNL programs on a VxWorks IOC, including a few thousand-line analog calibration sequence requiring ~20 minutes to execute. Inter-subsystem communication relies on the Allen-Bradley Data Highway+ serial bus at ~1 Hz. Interlock coordination is distributed across analog modules with no central first-fault detection and a non-functional arc detection system. All hardware platforms — VXI, CAMAC, PLC-5, SLC-500, and the obsolete Slo-Syn stepper drive chain — are past end-of-life.
The upgrade replaces the entire control electronics chain while retaining the operational RF plant: one klystron at ~800 kW driving four single-cell cavities at 476.3 MHz. Two Dimtel LLRF9/476 digital controllers replace the analog RFP: Unit 1 executes the primary vector-sum field control loop within 300 ns direct loop delay and four independent cavity tuner feedback loops; Unit 2 provides reflected-power monitoring and fast RF interlock generation. A new centralized Interface Chassis introduces first-fault detection and microsecond-scale hardware interlock coordination.
I will present the system architecture, key engineering decisions driving the transition from a monolithic analog control framework to a distributed hardware-accelerated FPGA architecture, and early commissioning experience at an operating synchrotron facility

Speaker: Faya WANG (SLAC)

23 Advanced LLRF control and automatic RF conditioning for the SSRIP facility

The Scalable System for Radioisotope Production (SSRIP) is a new compact linear accelerator developed at IFIN-HH (Măgurele, Romania), coordinated by INFN-LNF within the ELI-NP infrastructure. The accelerator is designed to produce high-brightness electron beams for radioisotope production applications. The SSRIP RF system consists of four independent RF stations, three operating at 2856 MHz (S-band) and one at 5712 MHz (C-band). Each station is equipped with a solid-state pulsed modulator and a high-power klystron amplifier. The RF control system is based on digital, temperature-stabilized and FPGA-controlled Low Level RF (LLRF) units, providing high-resolution amplitude and phase feedback for stable accelerator operation. An automatic RF conditioning procedure was developed to progressively increase the RF power up to the nominal operating level. The procedure continuously monitors vacuum levels and RF waveforms, automatically increasing or reducing the RF power according to the machine status, with minimal operator intervention. The subsequent beam commissioning, completed in July 2026, achieved an electron beam energy of 109 MeV with a bunch charge of 100 pC.

Speaker: Beatrice Serenellini (INFN - LNF)

24 New FPGA Architectures and the opportunities for implementing advanced control algorithms in LLRF systems

FPGA architectures have evolved dramatically in recent years with vast increases in logic and compute resources as well as fundamental architectural changes such as integration of microprocessors and data converters on the same chip. This makes it possible to develop new LLRF system hardware configurations as well as provide the resources to implement advanced control algorithms such as system identification and adaptive control in real time. The new hardware architectures and examples of advanced control and signal processing algorithms for LLRF systems will be presented.

Speaker: Dr Philip Varghese (Fermilab)

25 Development of DAC timing scan program for RAON Low Energy LINAC LLRF System

RAON is a heavy ion accelerator complex in Daejeon, Republic of Korea. Its beam commissioning of low energy superconducting linear accelerator (SCL3) have been finished by Institute for Rare Isotope Science (IRIS) in Institute of Basic Science (IBS). The purpose of this accelerator is the generation of rare isotope by ISOL (Isotope Separation On-Line) and its acceleration for the nuclear physics experiment. SCL3 consists of 22 quarter wave resonator (QWR) type cavities and 102 half wave resonator (HWR) type cavities and their operating RF frequencies are 81.25 MHz and 162.5 MHz. Every cavity can be controlled independently for the flexibility to accelerate the various A/q ions. An FPGA-based digital LLRF system has been in operation in this section since 2022. Because this system uses an LVDS interface to drive the DAC, a timing scan is required to determine the optimal interface timing and ensure reliable data transfer. An algorithm for automatically scanning the interface timing and applying the optimal timing parameters has recently been developed and tested. This presentation describes the algorithm and presents preliminary test results.

This work was supported by the Institute for Basic Science(IBS-I001-D1)

Speaker: Hyojae Jang (Institute for Basic Science)

26 Active Optical Delay Stabilization of 1.3 GHz RF-over-Fiber Reference Distribution for LLRF Systems

Precise phase regulation in LLRF systems relies on a phase-stable RF reference, making reference distribution an integral part of achievable field-control performance*. We present an actively stabilized RF-over-fiber link for distribution of a 1.3 GHz reference.
The system measures optical round-trip delay and compensates fiber-induced fluctuations through feedback to a fast optical delay actuator with 2.4 ns control range. By directly stabilizing the optical propagation delay rather than correcting the RF phase, the approach provides reciprocal delay compensation and avoids residual errors associated with non-reciprocal RF phase correction. The optical delay actuator enables fast cold start and response without the long warm-up times and limited control bandwidth of thermally controlled fiber spools.
Out-of-loop measurements demonstrate 6.6 fs RMS jitter for [1Hz – 1MHz] and 19.2 fs RMS drift over 8 h with 2 Hz sampling, corresponding to only ~20 fs RMS over the full bandwidth of [35uHz – 1MHz]. Residual frequency stability is 8.6E-15 at 1 s, reaching 1E-17 at 2000 s. These results demonstrate femtosecond-stable RF reference distribution for demanding accelerator LLRF systems.

Speaker: Kemal Shafak (Cycle GmbH)

27 A digital controller for Q0 vs Eacc measurement of SRF cavities at STFC Daresbury Laboratory

In a vertical test facility (VTF) a cavity is required to be driven on resonance so that performance characteristics such as Q0 and coupling factors can be calculated from measurements. This contrasts with the usual requirements of a controller in an RF station of a particle accelerator. In UKRI-STFC's VTF at Daresbury Laboratory the RF control system is presently distributed across an interdependent combination of hardware, firmware and software which range from custom designs to commercial off-the-shelf solutions. This can present issues when needing to upgrade the system to meet new customer requirements. We present a discussion of the system requirements and a proposal to meet them with a single digital controller.

Speakers: Conor Jenkins (ASTeC, STFC Daresbury Laboratory), Matthew Jones (ASTeC, STFC Daresbury Laboratory)

28 Development and Experimental Validation of an X-Band LLRF Prototype for High-Gradient Linear Accelerators

Compact, high-gradient linear accelerators increasingly employ X-band RF technology, placing stringent demands on amplitude and phase stability especially for short RF pulses. Meeting these demands requires low-level RF (LLRF) systems capable of precise signal generation and acquisition, with careful management of phase noise, timing jitter, and thermal effects. An X-band LLRF prototype has been developed within the EuPRAXIA Doctoral Network to address these requirements for next-generation linear accelerators. The prototype integrates a high-speed analog RF front-end with an FPGA-based digital platform for pulsed RF generation, acquisition, and signal processing. Its performance was characterized through laboratory measurements and subsequently evaluated on an accelerator RF test bench. This poster presents the system architecture, validation methodology, and experimental results, focusing on pulse-to-pulse amplitude stability and phase noise performance. The measurements demonstrate the feasibility of the proposed approach and provide a foundation for further development and deployment at high-gradient X-band accelerator facilities.

Speaker: Phani Deep Meruga (Instrumentation Technologies)

29 FPGA Control Platform for the Elettra 2.0 DLLRF Feedback System

A custom Intel Arria 10 SoC FPGA platform, developed for Beam Position Monitors, has been adapted as the core control hardware for the 500 MHz DLLRF system of the Elettra 2.0 upgrade. The architecture features an external RF Front-End down-converting signals to an 18.5 MHz IF, coupled with two FMC boards hosting four ADC channels (16 bits, 120 MS/s) and a DAC output stage with quadrature modulators. The feedback control chain is fully implemented inside the FPGA fabric. The pipeline utilizes a Digital Down Converter with non-IQ demodulation and CIC moving-average filters operating at the 1.1566 MHz Turn-by-Turn rate. CORDIC blocks extract amplitude and phase, which are processed by independent PI controllers equipped with anti-windup and anti-saturation algorithms. A Digital Up Converter converts the values back to I/Q components, driving two DAC channels for quadrature modulation. The entire processing loop achieves a total latency under 2 µs. High-speed SFP+ optical interfaces transfer real-time data at 10 Gb/s, while the SoC HPS manages configuration and diagnostics. The hardware platform design, real-time control implementation, and closed-loop performance will be presented.

Speaker: Gabriele Brajnik (Elettra-Sincrotrone Trieste)

30 High-Frequency MicroTCA Pulsed Cavity Controller at CERN

High-frequency pulsed cavity control is emerging as a common requirement across several projects at CERN, including AWAKE, FCC-ee, and a proposed upgrade to CLEAR. This poster presents the design and laboratory commissioning of a pulsed cavity controller built on the MicroTCA platform, using a SIS8300-KU FPGA carrier paired with a DWC8VM1 rear transition module.

RF signals are down-mixed and digitised, with baseband I/Q values recovered by digital down-conversion on the FPGA. Amplitude and phase stability during the pulse are maintained by a proportional-integral controller with pulse-to-pulse feedforward correction, which compensates for repeatable disturbances that fall outside the intra-pulse loop bandwidth. The controller supports arbitrary, tuneable reference waveforms, accommodating complex profiles such as those required for pulse compressor systems. Correction signals are output via a pair of baseband DACs, which drive the DWC8VM1's vector modulator to up-convert to the target RF frequency.

The system has been designed and validated for pulses of a few microseconds at 3 GHz, with the architecture supporting reconfiguration across 350–6000 MHz by changing the rear transition module, allowing reuse of the same platform across projects with different RF frequency requirements. Integration with CERN infrastructure includes direct FESA access to the device's register memory map, enabling operational control and monitoring consistent with existing accelerator control system conventions.

Speaker: Daniel Harryman (CERN)

31 Initial Development of a Resonant Kicker (RK) LLRF for PAL-XFEL

PAL-XFEL has been operating two beamlines simultaneously since 2021. The simultaneous operation method involves generating and accelerating single bunch of electrons at 60 Hz to provide 30 Hz beams to the hard X-ray(HX) and soft X-ray(SX) beamlines, respectively. It is intended to switch to a method developed by SwissFEL, which is capable of providing 60 Hz beams to both the HX and SX beamlines. To implement this, the two adjacent electron bunches are spaced 25ns apart, and the operating frequency is designed so that the RK operates at 20 MHz. We chose a method of driving the RK using a combination of LLRF and SSA. The initial LLRF prototype was developed and showed the RMS stability 0.007 %(Amplitude) and 0.02 °(Phase), which needs improvement nearly ten times. We are cosidering and seeking for solutions to improve the RMS stability.

Speaker: Jinyul Hu (Pohang Accelerator Laboratory)

32 Integration of White Rabbit into an FPGA-Based Low-Level Radio Frequency (LLRF) System for Deterministic Triggering and Low-Latency Interlock Communication

Low-Level Radio Frequency (LLRF) systems are critical components of modern particle accelerators, providing precise control of the amplitude and phase of RF fields in accelerating cavities. Advanced control and feedback architectures increasingly rely on deterministic timing and fast communication to coordinate RF control, diagnostics and protection functions across distributed equipment. Traditionally, synchronization, trigger distribution and time-critical communications rely on independent infrastructures, requiring dedicated cabling and additional hardware.

This work presents the integration of White Rabbit (WR) technology into an FPGA-based LLRF platform through the incorporation of the HATI White Rabbit IP core. The integration provides deterministic sub-nanosecond synchronization, hardware timestamping capabilities and the generation of internally synchronized triggers directly within the LLRF system. These capabilities provide a common timing reference for the coordination of control and feedback functions across distributed LLRF systems while reducing the need for dedicated timing and trigger distribution lines.

Furthermore, the original WR communication architecture has been extended to support the deterministic transmission of ultra-low-latency data packets over the existing WR fiber network. This mechanism enables the exchange of time-critical information between distributed accelerator systems. Potential applications include the fast transmission of machine interlocks as well as the exchange of control-related data between distributed LLRF controllers, such as I/Q components for coordinated or distributed feedback schemes. By combining precise synchronization and deterministic low-latency data exchange over the same optical network, the proposed architecture provides a common infrastructure for timing, protection and advanced distributed control applications.

The proposed implementation has been integrated and experimentally validated in a laboratory White Rabbit network. High-rate packet traffic was transmitted between two WR nodes to characterize communication latency, determinism and reliability. The presented results demonstrate the feasibility of using White Rabbit not only as a precision timing network, but also as a deterministic ultra-low-latency communication infrastructure, enabling applications ranging from fast machine protection to time-critical data exchange between distributed LLRF controllers.

Speaker: Juan Salvador Fernández Prat (Safran Electronic & Defense Spain S. L. U.)

33 Minimizing the RF power requirements at injection in the Large Hadron Collider

In the context of the High-Luminosity Large Hadron Collider (HL-LHC), high-intensity beams of 2.3 · 10¹¹ protons per bunch are planned to be injected at the LHC after Long Shutdown 3 (LS3) at CERN. To cope with these beam intensities, several strategies have been put in place to reduce the power requirements of the 2 x 8 accelerating cavities installed in the machine. Among them, one of the most challenging is the injection process. In this work, we present a scheme where we modulate the phase bunch-by-bunch in the LHC to minimize the peak RF power at injection. The ejection bunches at the Super Proton Synchrotron (SPS) are likewise phase-modulated for optimal capture. Implementing this approach required complex modifications to both the SPS and LHC RF systems, since the optimal modulated phases depend on several parameters such as beam intensity, number of circulating bunches, cavity voltages, etc. In this contribution, the necessary changes implemented in the SPS and LHC LLRF systems will be presented.

Speaker: Dr Diego Barrientos (CERN)

34 Operational Adjustments to the SELAP Controller for Robust Cavity Control Under Adverse RF Conditions

Since 2021, the Self-Excited Loop with Amplitude and Phase (SELAP) control algorithm has regulated CEBAF’s high-Q superconducting radio-frequency cavities, demonstrating robust field stability. Sustaining this performance in routine operations, however, required targeted controller adaptations to mitigate microphonics, cavity-to-cavity mechanical coupling, signal crosstalk, high-power amplifier (HPA) saturation, and heavy beam loading. This paper details the firmware and software implementations developed to address these operational constraints and presents beam-run performance data on cavity availability and trip-rate reduction.

Speaker: Dr Tomasz Plawski (Jefferson Lab)

35 Photonic RF Phase Monitor

We describe a timing-stabilized RF amplitude and phase monitor, using optical interferometry. Unlike RF-over-fiber using AM, RF voltage drives an EO phase modulator at the sensor. No bias is needed, and light is provided by the receiving system. The optical signal goes back through fiber and is interfered with an optical LO at the receiver, a coherent telecom component with optical I and Q channels for X and Y polarizations. This produces RF signals to be demodulated, with a result linearly proportional to the sensed voltage. The roundtrip optical phase change can be used to time correct the RF signal, providing data for feedback control while the fiber delayvaries due to thermal and acoustic effects. One can get polarization state data from the receiver to drive a polarization controller, aligning light to the axis of the EO phase modulator. We show simulations using ANSYS, compared with early experimental results using a single-channel interferometer and our FPGA-based DSP. We demonstrate RF transmission and reception, detecting amplitude and phase with correction for optical phase instability. With a fast digitizer,this sensing system can be used for temporally stable BPM and BAM signals. The scheme is capable of extension to multiple channels mutually synchronized.

Speaker: Russell Wilcox (Lawrence Berkeley National Laboratory)

36 QubiC: Enabling Full-Stack Hardware-Software Co-Design for Superconducting Quantum Processors

The performance of superconducting quantum processors is critically limited by the control stack, where traditional systems often rely on black-box abstractions that hinder full-stack co-design. Notably, superconducting qubit control and particle accelerator radio-frequency (RF) control systems can share the same fundamental FPGA-based, software-defined radio architecture. We present QubiC, an open-source quantum control system designed to bridge this gap. QubiC offers a vertically integrated framework that exposes low-level hardware control while maintaining high-level programming abstractions. Its architecture features native, high-bandwidth links to GPUs, enabling advanced capabilities such as AI-driven automated calibration and enhanced quantum state discrimination. We have experimentally validated QubiC by demonstrating high-fidelity control on superconducting processors. Overall, QubiC serves as a powerful platform to accelerate progress toward robust, high-performance quantum computing through hardware-software co-design.

Speaker: Gang Huang (Lawrence Berkeley National Laboratory)

37 Quench Detection and Trip Classification Developments at DESY

Automating quench detection and trip classification has been an ongoing research project at DESY for a number of years. We have recently developed firmware and close-to-hardware implementations of our model-based quench detection and trip classification system and are in the process of testing these in emulation and on real systems. In addition to several software implementations, we are testing Versal, native HDL, and hybrid processing solutions. We will report on the status of these projects, upcoming development plans, and obstacles to performance in such deployments.

Speaker: Joshua Einstein-Curtis (Deutsches Elektronen-Synchrotron DESY)

38 RF Conditioning IP Core with vacuum feedback

RF conditioning of accelerating cavities and power couplers is required to reach nominal operating voltage. High field gradients trigger multipacting and outgassing, which raise the vacuum pressure and limit the RF power that can be applied before arcing occurs. For efficient and safe conditioning, a feedback loop keeps the drive inside the optimal zone where vacuum activity is provoked but stays within safe limits. Existing systems at CERN are based on legacy PCs from the 1990s or VME hardware developed 20 years ago, both critically obsolete.

A new generic FPGA IP core has been developed to integrate RF conditioning directly into the Low Level RF (LLRF) cavity-controller, eliminating dedicated hardware as well as reducing maintenance and system complexity. Developed for the HL-LHC crab cavities on the MicroTCA platform, it implements configurable RF pulse generation, automatic RF power ramping, frequency modulation and a vacuum feedback loop. Software supervision monitors the process and optimises the operating parameters.

The architecture of the IP core, its main functionalities, and the first experimental results obtained during validation will be presented.

Speaker: Bruno Valentin Seifert (CERN)

39 Simple and Reliable Digital Electronics for Beam Monitoring and Control

Modern digital electronics chips available on the market have reached incredible levels of performance. They enable highly advanced monitoring and control of accelerated beams, with bunch-by-bunch resolution and even beyond. However, this performance comes at a cost: increasing device capabilities also leads to greater complexity, making these technologies less accessible to beginners and often increasing development time.

On the other hand, many applications do not require state-of-the-art performance. In this contribution, we present a project that aims to remain relatively simple while providing intermediate levels of signal-processing and computational performance without compromising reliability.

This project uses an FPGA to achieve precise control at the clock-cycle level. The purpose is to make a beam phase-lock loop, which would then provide a RF signal in phase with the beam for other diagnostics devices. Other applications based on the same hardware platform could later be realised, whether for feedback systems or simply for acquisition and monitoring.

A key requirement for this project was the integration with the control system through TCP/IP, using a generic solution. An additional objective was to favour open-source components and solutions whenever possible, in order to facilitate reuse, adaptation, and long-term maintainability while reducing dependence on proprietary solutions.

Speaker: Benoit Roche (ESRF)

SOCIAL EVENT: Social Dinner

Thu, October 15

FERMI TOUR

12:30 PM Lunch

40 Vendor no. 3

Tutorial

Control Loop Modeling, Tuning and Optimization

41 Active Noise Cancellation Measurements on LCLS-II SRF Cavities

Various advanced resonance control techniques have been developed and deployed at accelerator facilities to mitigate mechanical detuning in high-Q Superconducting Radio-Frequency (SRF) cavities. During the LCLS-II project, several of these approaches were evaluated and a real-time adaptive Narrowband Active Noise Cancellation (NANC) algorithm, developed through a collaboration between SLAC, DESY, and LBNL, emerged as a particularly promising method. The NANC algorithm has since been tested on LCLS-II-style cryomodule cavities at multiple facilities and has consistently demonstrated its ability to reduce microphonic detuning.
The LCLS-II implementation of NANC has matured significantly through successive stages of development and is now approaching readiness for larger-scale deployment. This capability is expected to provide benefits at SLAC as the laboratory prepares to commission 23 new SRF cryomodules for the LCLS-II-HE project while also restarting 37 LCLS-II cryomodules, some of whose cavities have experienced microphonic detuning that limited achievable operating gradients. This contribution describes the NANC approach as implemented on the LCLS-II Low-Level RF (LLRF) platform and presents results from recent testing campaigns.

Speaker: Andy Benwell (SLAC National Accelerator Laboratory)

System Identification

42 Environmental Noise and Disturbance Effects for high-Q SRF Systems

Noise and systematic disturbances are important to characterize when working with low-bandwidth SRF cavities that require tight RF regulation. For a future upgrade at XFEL, we have been investigating the microphonics and noise performance of the tunnel, cryomodule, and RF environment in order to better characterize the expected performance, control, and mitigations needed for the design.

Speaker: Joshua Einstein-Curtis (Deutsches Elektronen-Synchrotron DESY)

3:45 PM Coffee break

43 Vendor no. 4

System Identification

44 Mechanical-Mode System Identification for Superconducting Cavities: Grey-Box and Closed-Loop Chirp Methods

Mechanical-mode characterization is essential for understanding Lorentz-force detuning (LFD) and improving field stability in superconducting RF cavities. This work presents two complementary approaches for identifying LFD-induced mechanical dynamics under LLRF control. First, a physics-informed grey-box model is estimated from LFD transients. It extracts dominant mechanical modes when stepped-sine measurements are time-consuming or compromised by ponderomotive instability, and yields low-order models that reproduce measured detuning transients. Second, a closed-loop broadband chirp method injects a perturbation through the cavity-field amplitude setpoint while amplitude and phase feedback remain active. The mechanical transfer function is reconstructed from the squared-field perturbation and detuning. Measurements on a half-wave resonator show that one chirp sweep identifies dominant modes consistent with closed-loop stepped-sine measurements and grey-box identification. The effects of excitation strength, chirp duration, feedback gains, and microphonics processing are evaluated. Together, the methods provide tools for robust mechanical-mode identification and LLRF optimization.

Speaker: Yilin Miao (Institute of Modern Physics, Chinese Academy of Sciences)

45 Loop Delay Identification in LLRF Systems via Self-Excited Loop Control

LLRF systems typically employ the classical PI control algorithm to maintain constant cavity voltage and phase. The performance of PI control depends on the design and optimization of the proportional (kp) and integral (ki) parameters; therefore, accurate modeling of the closed-loop path is of great importance. Loop delay is a critical parameter of the closed-loop system, as it determines the ultimate limit of the closed-loop bandwidth. The common method for measuring delay is to compare the time difference between a step input and the corresponding step output; however, this approach struggles to cover the entire loop and inevitably includes the cavity's step rise time, leading to inaccurate loop delay measurements. This paper adopts the Self-Excited Loop (SEL) control technique to identify the loop delay, based on the relationship among the SEL resonant frequency, loop phase shift, and loop delay. The loop delay identification is tested on the buncher of the Dalian Advanced Light Source (DALS). The obtained results are then used to design the PI parameters for the buncher's closed-loop control. Finally, the amplitude and phase noise of the buncher under open-loop and closed-loop conditions are compared using these PI parameters.

Speaker: Zhiyuan Zhang (Institute of Advanced Light Source Facilities, Shenzhen)

46 Performance Comparison of Linear vs. Unscented Kalman Filter-Based Gun Detuning Estimation with a Focus on Parameter Identification

The interpulse control laws for the European XFEL's normal-conducting 1.3 GHz gun (rev. 5.2) require detuning measurements to modulate pulse width and phase. However, due to thermal inertia and the inability to place temperature sensors directly on the gun's inner wall, such measurements cannot be obtained directly. This paper compares the tracking performance of two RF-based detuning estimation methods to address this challenge: the established linear, stationary approach versus an Unscented Kalman filter using a nonlinear, parameter-identified model.

Speaker: Max Herrmann (DESY)

5:55 PM Feedback and open discussion

6:20 PM Closing remarks - SPC updates