We are developing a low-level RF (LLRF) controller based on RF System on Chip (RFSoC) for an electron linac. The AMD Zynq RFSoC was employed for this controller, which has a large-scale high-speed FPGA together with high-speed ADCs and DACs (8 channels each). The RFSoC also has an application CPU for Linux and a real-time CPU for time-critical tasks, capable of a 1 kHz repetition rate. A general-purpose pizza-box module with an RFSoC was designed and manufactured, and firmware for LLRF control was developed. This LLRF module will be first utilized for an X-band (11.424 GHz) transverse deflector system* for SACLA. A pulsed X-band RF signal is generated by upconverting a 476 MHz IF signal from the DAC and RF signals from the X-band high-power components are converted to 476 MHz IF signals and digitized by ADCs. The IF signal is converted to a baseband IQ signal and the phase and amplitude are obtained. Since the latency of ADC, DAC, and FPGA is as short as several 100 ns, the intra-pulse feedback control is anticipated to stabilize the phase and amplitude of the acceleration RF field. This presentation will give the design and basic performance of the LLRF controller.

The 3rd-generation synchrotron light source BESSY II is undergoing a series of modernization measures to maintain its leadership role until BESSY III starts its operation, planned for in 2035. The modernization of the LLRF control systems is one of these measures in the so-called BESSY-II+ project. Prior to the deployment of the new mTCA-based digital systems to control the fundamental frequency cavities at the booster and storage ring, a test stand was set up provide the opportunity for extensive offline testing and debugging. The test stand is consists of a HOM-damped 500-MHz cavity, a 80-kW Solid State Amplifier (SSA), a mTCA crate and all the ancillary systems required for high-power RF cavity operation. In this paper we discuss the results of the high power tests in continuous-wave operation, as required for the BESSY II storage ring, and in ramped mode, as required for the BESSY II booster during beam injection.

KEK Accelerator Test Facility (ATF) conducts beam instrumentation R&D for International Linear Collider (ILC) project. ATF includes 1.3 GeV normal conductivity S-band Linac and Damping Ring (DR). There are 9 S-band pulsed klystrons at Linac, which supply High-Power RF to accelerate electron beam up to 1.3 GeV, 1 CW klystron at DR. The electron beam is generated by a photocathode irradiation by a laser pulse. The laser pulse generation is synchronized with the accelerating fields by the laser system oscillator Piezo feedback. These Linac, DR High-Power RF field and laser pulse arrival time jitter and/or drift define the stability of the electron beam parameters, such as average energy, energy spread (RMS), emittance, bunch charge etc. This study demonstrates KEK ATF Linac and DR High-Power RF field phase and amplitude, as well as the laser system oscillator laser pulse arrival stability measurement results. Also, FPGA board based digital Low-Level RF phase&amplitude feedback system is described in this report.

Superconducting cavities with high Q-factor require precise tuning to match the RF frequency, ensuring stable electromagnetic fields and minimizing RF power consumption. At the XFEL accelerator, TESLA cavities are tuned using slow tuners (step motors) for coarse adjustments and fast tuners (piezoelectric actuators) for fine-tuning and compensating disturbances such as Lorentz Force Detuning (LFD) and microphonics*. Critical to this system, Piezo actuators require high-voltage (up to 100V) and high-current (up to 1A) driving signals for effective LFD compensation. However, they are vulnerable to overvoltage, overcurrent, and overheating**, and their protection is crucial since replacing damaged piezo in fully assembled modules is unfeasible. Additionally, piezo induced vibrations can affect the machine's stability. Optimizing piezo excitation—by reducing voltage, current, and current slope while ensuring effective LFD compensation—improves both reliability and machine stability. This paper explores the optimization of piezo excitation at XFEL, detailing methods and results applicable to other facilities with superconducting cavities.

The Proton radiation effects facility (PREF) was designed and constructed by the Institute of Modern Physics, which can provide high-quality proton beams with continuous and accurate tunable energy range, high current intensity, high duty cycle and large scanning area of 10-60MeV energy range. which consists of a proton source, RFQ linac injector, synchrotron and irradiation terminals. The RFQ works at 162.5 MHz, providing 1.2Mev proton beam for synchrotron. The RF system consists of a RFQ cavity, two 50 kW solid state amplifiers and digital low level RF control system (LLRF). The amplitude and phase stability requirements for the LLRF are 1% and ±1°separately. To meet requirements and to ensure reliability, a digital LLRF system was designed. The new digital LLRF is based on Virtex5 FPGA, fast ADCs and DACs, and CPCI bus. The progress and plans for future are presented.

EuPRAXIA, the "European Plasma Research Accelerator with eXcellence In Applications," represents the next generation of free-electron lasers (FEL). It aims to develop a compact, cost-efficient particle accelerator using innovative wake-field accelerator technology. High-energy physics often demands higher acceleration voltages, and X-band technology offers high gradients in compact structures. The EuPRAXIA@SPARC_LAB LINAC injector, featuring an S-band RF gun, four S-band structures, and sixteen X-band structures, achieves a maximum beam energy of 1 GeV. For femtosecond-level synchronization and stability, Low-Level Radio Frequency (LLRF) systems are essential. However, commercial X-band LLRF solutions are unavailable. This project, in context of the EuPRAXIA - Doctoral Network, develops an X-band LLRF prototype tailored to meet the EuPRAXIA@SPARC_LAB LINAC's stringent requirements. After validation on a testbench, the prototype will enable industrial production and commercialization. This paper presents the Front-End, Back-End analysis, and further evaluation of the prototype.

A Low-Level RF and Power Control system based on EPICS has been developed for the new H¯ RF Ion Source on the Pre-Injector Test Stand at ISIS Spallation Neutron and Muon source, UKRI-STFC Rutherford Appleton Laboratory. The Ion Source LLRF system provides a 2 MHz signal to a Solid-State 100 kW RF Amplifier that drives the Ion Source Plasma, the changing Plasma load requires fast Frequency agility and closed loop Power Control. This paper will detail the design and performance of the LLRF system.

The TRIUMF ISAC-1 and ISAC-2 accelerator chains uses multiple fixed frequencies in their RF cavities. These include 5.8933 MHz, 11.7866 MHz, 35.36 MHz, 106.08 MHz and 141.44 MHz. These need to be synchronized in phase with respect to each other’s. The new frequency sources use x2, x3 and x4 low phase noise multipliers to generate these frequencies from a single low phase noise 5.8933 MHz frequency synthesizer. Bench tests have shown that the frequency multipliers do not generate additional phase noises, except those that are theoretically produced due to frequency multiplication. With an average performance frequency source as a reference which has -85dBc/Hz at 10 Hz offset, the integrated rms phase noise of 141.44 MHz multiplied output is less than 0.5°.

The Radio Frequency Protection Interlock (RFPI) system main responsibility is to collect predefined set of signals and to protect each RF station. In case of safety limits violations from any of this input signals the RFPI has to instantenously drop permits for the LLRF or RF amplifier (eq. Solid State Amplifier - SSA or klystron) operation. This paper presents an overview of the final design of the RFPI system dedicated for Proton Improvement Plan II (PIP-II) at Fermilab.

The installation phase of the European Spallation Source (ESS) linear accelerator is nearly complete. As with other superconducting linacs operating in pulse mode, LLRF systems play a crucial role in controlling accelerating beam parameters. Modern LLRF systems go beyond providing fast and reliable feedback for RF signal regulation; they also ensure precise, dynamic cavity tuning. Additionally, they enhance machine availability by monitoring various signals to identify potential issues and implementing fast and slow algorithms to optimize cavity performance within safety limits, tailored to specific accelerator conditions. Preparation for these tasks begins during cryomodule and cavity testing, prior to tunnel installation. Key parameters such as Lorentz force detuning coefficients, piezotuner range and polarity, main mechanical cavity modes, Pi-mode frequencies, slow tuner sensitivity, and backlash must be accurately determined to enable peak LLRF performance. This paper outlines the development, implementation, and application of software tools designed to determine these parameters for cavities tested at ESS Test Stand 2 (TS2) and those installed in the accelerator tunnel.

Hefei Infrared Free-Electron Laser device (IR-FEL) is a user experimental device dedicated to energy chemistry research that can generate high brightness mid/far infrared lasers. It is driven by an S-band linear accelerator with a maximum electron energy of 60 MeV. The stability of the final laser output is determined by the quality of the electron beam, and optimizing the Low-Level RF (LLRF) Controlsystem can elevate the beam's ultimate quality. The IR-FEL linear accelerator boasts a beam length of 13μs, exhibiting a pronounced beam loading effect. The leading edge of the beam interacts with the RF field, absorbing energy, thereby influencing the acceleration process at the beam's tail. This interaction leads to an increase in beam emittance, impacting the final laser quality. However, by incorporating a feedforward algorithm to modulate the microwave field amplitude upon the beam's arrival, we can mitigate the beam loading effect and improve beam quality. Details regarding this upgrade, along with the experimental outcomes, will be elaborated upon in the main text.

Previous work has shown the efficacy of using machine learning for removal of noise in LLRF signals when operating in an industrial environment. Here we extend the analysis to include different noise power spectra. Specifically we analyze the impact on denoisig when correlated noise power spectra are used. Four different noise spectra are analyzed including red, pink, violet, and blue noise. We demonstrate the ability to remove the noise when trained on only white noise and compare this to results when retraining on different color spectra.

The FRANZ linac, consisting of a coupled RFQ-IH cavity and a subsequent CH rebuncher, requires an LLRF system with moderate performance demands. These include amplitude control to maintain a constant field in the cavity, constant phase synchronization between the accelerator and rebuncher, and plunger control to stabilize the cavities frequency at 175 MHz. Given the dead time from LLRF RF output to probe input is approximately 150 ns and the system operates in cw or 1 ms pulsed mode, a decision was made to design a system with a reaction time of 1 µs. To ensure flexibility, the system was designed with digital control. Consequently, an analog-digital hybrid system was implemented. The RF signal processing is performed using classical analog components, while the control and readout of the analog signals are managed by a ZYNQ SoC, which combines FPGA and ARM processors. The first proof-of-concept prototype for amplitude control, including reflection and vacuum monitoring, has been successfully operational with the RFQ since late 2023. Development of the next version, which will include phase and plunger control, is underway and is expected to undergo beam testing in 2025.

Hefei Advanced Light Facility (HALF) is a fourth-generation synchrotron radiation source based on diffraction limited storage ring. It comprises a 180-meter injector and a 480-meter storage ring. The injector incorporates a digital low-level radio frequency (LLRF) control system based on MTCA.4, ensuring a stable and adjustable microwave field for the acceleration structure. This article outlines the structure of the LLRF system, encompassing both hardware and software components. Within the software, we have mitigated signal drift induced by environmental temperature fluctuations by adding a reference tracking module. Building upon the existing IQ closed-loop functionality, we have successfully implemented separate amplitude and phase closed-loop functions. In high-power online testing, the IQ closed-loop demonstrated amplitude and phase stabilities of 0.0411% (RMS)/0.0638° (RMS), respectively. Furthermore, the phase stability achieved by the phase-independent closed-loop function reached 0.0646° (RMS). Currently, the LLRF system has fulfilled the design requirements of HALF.

ALBA Low-Level RF (LLRF) system has provided over a decade of reliable operation and has been adopted by other synchrotron facilities. To meet the evolving requirements of ALBA and ALBA-II, a new LLRF system has been developed. This system features FPGA and ADC/DAC MTCA boards designed by SAFRAN, enabling direct 500 MHz signal sampling without down/up-conversion. These enhancements reduce system complexity, minimize noise, and simplify maintenance. SAFRAN also supplies peripheral modules and the Tango device server generator, while ALBA implemented it and developed a new GUI. Upgraded GPIO and RF signal patch panels complement the new hardware. The legacy VHDL code has been updated to improve readability and functionality, incorporating advanced features such as octant selection and a harmonic direct feedback selection method. The latter, based on IIR filtering, isolates positive and negative revolution harmonics in the I/Q domain, feeding them back to amplifiers to effectively mitigate transient beam loading caused by the storage ring bunch train gaps. This upgraded LLRF system delivers enhanced performance and greater flexibility to address the future needs of ALBA and ALBA-II.

Detecting anomalies in superconducting cavities at the EuXFEL is essential for reliable operation. We began with a model-based anomaly detection approach focused on residual analysis. To improve fault discrimination, particularly for quench events, we augmented the detection with a machine learning-based classification. Key challenges are posed by the transition to real-time operation, requiring computational and integration adjustments. For the online application, we deployed two servers at one of the 25 stations to detect and log anomalies with a software implementation. In parallel, we pushed the development of a firmware solution that will counteract critical faults in real-time. At the current stage only the anomaly detection is in online operation, which is planned to be augmented with the online fault classification in the future. The resulting detection system delivers reports across various timescales, supporting both immediate responses and long-term maintenance.

Superconducting radio frequency (SRF) cavities in particle accelerators rely on accurately calibrated RF signals to assess cavity bandwidth and detuning, ensuring optimal performance. In practice, however, calibration drift due to humidity and temperature fluctuations over time poses a significant challenge, potentially resulting in suboptimal operation and reduced efficiency. This study explores how environmental variables such as humidity and temperature affect this phenomenon. Relative humidity, in particular, is difficult to control and has been shown to impact calibration drift strongly. Building on these insights, we introduce machine learning-based approaches to forecast both relative humidity and calibration drift in SRF cavities. By leveraging advanced algorithms and historical data on cavity operation and performance, we develop predictive models that identify patterns and trends indicative of relative humidity and calibration drift. Two approaches are presented in this work, including a polynomial NARMAX model and an attention-based deep neural network. These models enable real-time compensation and automated recalibration, improving system stability and efficiency.

The digital LLRF controller for the Facility for Rare Isotope (FRIB) project was designed to accommodate various cavity types with six distinct frequencies ranging from 40.25 to 322 MHz. The cavities also adopt different types of tuners, e.g. stepper motor, pneumatic, water flow, etc. A common hardware platform with design choices such as direct sampling of RF, compatible footprint for RF components (e.g. filters), same form factor PCBs, spare channels (RF, analog and digital) made it a reality. The design later turned out to be very adaptive to unforeseen new requirements as the project moved on. Those include adding an interface to enable and monitor the bias tee high voltage power supply, adding a serial interface to communicate with the tuner servo controller and monitoring a cold cathode gauge for faster interlock response. Most recently FRIB LLRF controller use case is expanded to support the testing of e-gun for the SLAC LCLS-II project which runs at a different RF frequency and uses a piezo tuner. Furthermore we are exploring a solution with this versatile platform to support the upgrade of the K500 cyclotron RF control with a continuous frequency range from 10 to 27 MHz.

The KEK e-/e+ LINAC delivers the beams to four storage rings with the top-up injections by switching the beam mode in 50 Hz repetition rate. The beam charge, energy, and number of bunches (one or two) are different for each ring. Therefore, the RF timing and phase are adjusted to each beam mode independently. To stabilize the RF phase drifts caused by the klystron high voltage, the cooling water and accelerating structure temperature, the RF phase feedback was introduced. The correction phase quantity is obtained by feedback calculation using non-injection mode without beam acceleration, and the value is added to set phase value in each mode. As a result, the RF phase in each beam mode has been stabilized.

In the SXFEL injector, the beam stability achieved superior performance, maintaining fluctuations below 0.01% after passing through four S-band accelerating units. However, the stability deteriorated to 0.02% upon exiting the X-band linearizer. To mitigate this degradation, a series of targeted enhancements were implemented, including an extensive upgrade of the low-level RF system’s front-end electronics and the integration of adaptive modulation techniques for input signal optimization. These measures effectively restored and improved the beam stability, achieving precision levels within 0.01%.

RF measurements are crucial for stabilizing the power source output and extracting beam data. As digital systems evolve, the analog-to-digital converter (ADC) now commonly reaches 16 bits and 100 MHz, enabling multi-channel low-level radio frequency (LLRF) systems to generate several gigabytes of data per second, overwhelming data storage and processing capabilities. This paper proposes a pre-processing method using Field Programmable Gate Arrays (FPGAs), which dynamically adjusts timing intervals based on operator requirements. For detailed waveform analysis, the LLRF can upload data over short time intervals with high precision. Conversely, for applications concerned with slow drift, long-time-range, low-precision data is transmitted. Thus, the total amount of uploaded data remains constant. A multi-order filter is applied to the raw data, with desired precision achieved at specific orders. The time precision ranges from 10 ns to 20 µs, while the time range spans from 20 µs to 40 ms.

In various of particle accelerator designs, amplitude and phase modulation methods are commonly applied to shape the RF pulses for implementing pulse compressors or compensating for the fluctuations introduced by the high-power RF components and beam loading effects. The phase modulations are typically implemented with additional phase shifters that requires drive or control electronics. With our recent next generation LLRF (NG-LLRF) platform developed based on the direct RF sampling technology of RF system-on-chip (RFSoC) devices, the RF pulse shaping can be realized without the analogue phase shifters, which can significantly simplify the system architecture. We performed a range of high-power experiments in C-band for evaluating the RF pulse shaping capabilities of the NG-LLRF system at different stages of the RF circuits. In this paper, the high-power characterization results with the Cool Copper Collider structure driven by RF pulses with different modulation schemes will be described. With the pulse modulation and demodulation completely implemented in digital domain, the RF pulse shaping schemes can be rapidly adapted for X-band structures just by adding analogue mixers.

The low-level RF (LLRF) systems for S-band linear accelerating structures are typically implemented with heterodyne base architectures. We have developed and characterized the next generation LLRF (NG-LLRF) based on the RF system-on-chip (RFSoC) for C-band accelerating structures and the platform delivered the pulse-to-pulse fluctuation levels considerably better than the requirement of the targeted applications. The NG-LLRF system uses the direct RF sampling technique of the RFSoC, which significantly simplified the architecture compared with the conventional LLRF. We have extended the frequency range of the NG-LLRF to S-band and experimented with different RFSoC devices and system designs to meet the more stringent requirement for S-band LLRF applications. In this paper, the characterization results of the platform with different system architectures will be summarized and the high-power test results of the NG-LLRF with the S-band accelerating structure in the Next Linear Collider Test Accelerator (NLCTA) test facility at SLAC National Accelerator Laboratory will be presented and analyzed.