Real-Time Beam Integrity Monitor

• The topic is defined as the continuous, real-time monitoring of beam parameters to provide immediate diagnostics for accelerator operations, machine protection, and experiment normalization.
• Instrumentation architectures incorporate transduction sensors, low-latency readout, and digital post-processing to convert raw signals into actionable beam-state variables.
• Real-time monitors leverage precise timing, synchronization, and advanced signal processing to detect beam anomalies reliably in high radiation and dynamic operational environments.

A real-time beam integrity monitor is an instrumentation or detector system that measures beam-related observables while the beam is being delivered and makes those observables available quickly enough to support accelerator operation, machine protection, detector protection, or experiment normalization. Reported implementations span bunch-by-bunch luminosity and beam-induced background monitoring at the HL-LHC, per-pulse proton charge and timestamp distribution at spallation sources, profile and rate imaging of ion and photon beams, beam permit fibre surveillance, and combined phase-position diagnostics based on direct digitization (Shibin et al., 2024, Zhuang et al., 2018, Levin et al., 2022, García-Argos et al., 2015, Zhao et al., 2015).

1. Scope of the concept

The meaning of “beam integrity” is facility-dependent. In the CMS Fast Beam Condition Monitor, it is tied to bunch-by-bunch collision rates, beam-induced background levels, and fast changes in the rate and time structure of signals that are symptomatic of abnormal beam conditions; the detector provides per-bunch histograms of time-of-arrival and time-over-threshold in real time and is explicitly described as a fast beam integrity monitor (Shibin et al., 2024). In the CSNS proton beam real-time monitor, beam integrity means per-pulse proton charge, continuity of the pulse sequence, time interval between pulses at the 25 Hz nominal repetition rate, and detection of abnormally low or spurious pulses (Zhuang et al., 2018).

In beam-profile systems, integrity is defined through transverse observables. The FRIB Scintillator Ion Beam Monitor is designed to verify that the beam is centered where expected, that the transverse profile is stable and within tolerance, that the intensity is in the desired range, and that there are no unexpected drifts, steering errors, or severe degradation of the beam shape (Levin et al., 2022). The BM4 photon-beam monitor at Tohoku University defines integrity through 1D profiles in horizontal and vertical, beam position, beam size, stability versus intensity and time, and halo or background discrimination (Kino et al., 2024). In hadron-therapy instrumentation based on Topmetal-II$^{-}$ sensors, the monitored variables are beam position, incidence angle, and intensity, measured online with a 3.3 ms frame time (Wang et al., 2016).

A further extension appears in machine-protection infrastructure. In the LHC Beam Interlock System context, “beam integrity” refers to the integrity of the beam permit signal path itself: optical-link attenuation, power margin, and the distinction between genuine unsafe beam conditions and false dumps caused by fibre or transceiver degradation (García-Argos et al., 2015). This suggests that the term does not denote a single observable, but a class of continuously acquired measurements that certify whether beam delivery, beam timing, beam trajectory, or the beam-permit chain remain within expected operational limits.

2. Instrumentation architectures

Real-time beam integrity monitors use heterogeneous sensing media, but their architectures are consistently organized as a transduction stage, a low-latency readout stage, and an analysis stage. At CMS, FBCM uses CO$_2$-cooled silicon pad sensors read out by the six-channel FBCM23 ASIC in 65 nm CMOS. Each channel contains a transimpedance front-end amplifier, a booster amplifier, and a leading-edge discriminator; the output is a binary high-speed asynchronous signal whose leading edge encodes time-of-arrival and whose width encodes time-over-threshold (Shibin et al., 2024). The CMS Beam Halo Monitor uses quartz Cherenkov radiators, UV-sensitive PMTs, charge-integrating QIE electronics, and $\mu$HTR back-end histogramming to provide per-bunch machine-induced background measurements (Stifter, 2015).

At CSNS, the architecture is based on three beam current transformers in the RTBT line, an NI PXI oscilloscope/ADC card for digitizing the CT waveform, and a self-developed TDC card for timestamping. The TDC emits a trigger message on the White Rabbit network; the PXI controller computes proton charge in software and transmits a measurement message carrying sequence number, trigger time, and proton charge values (Zhuang et al., 2018). In digitizer-centric systems, the beam pickup, analog front-end, ADC, and FPGA form a tightly integrated IQ chain. The beam-measurement systems reported for SSRF, CSNS, and ADS use BPMs or FCTs, band-pass filtering, optional analog down-conversion, direct or under-sampled digitization, and FPGA-based digital down-conversion to compute amplitude, phase, position, and energy in real time (Zhao et al., 2015).

Profile monitors for low- and medium-energy beamlines often replace direct electrical pickup with optical readout. The FRIB scintillator monitor uses movable scintillator targets, a high-resolution CMOS machine-vision camera, and analysis software that operates frame by frame (Levin et al., 2022). The CERN XBPF/XBTF family uses 1 mm scintillating fibres, SiPMs for fine-granularity position readout, PMTs for fast trigger and TOF planes, and White Rabbit-compatible back-end electronics (Ortega et al., 2019). The BM4 photon monitor combines a thin aluminum converter, scintillating-fibre x/y layers, a charged VETO, a trigger counter, and a streaming TDC readout chain (Kino et al., 2024). The supersonic gas curtain monitor forms a thin gas sheet in vacuum, lets the hadron beam ionize the curtain, extracts the ions onto an MCP and phosphor screen, and images the profile with a CMOS camera (Kumar et al., 18 Sep 2025).

These architectures differ in sensor physics, but they share a common systems pattern: minimally perturbative sensing, deterministic timing or triggerless acquisition, and digital post-processing that converts raw signals into operational beam-state variables.

3. Timing, synchronization, and triggerless operation

The defining characteristic of many beam integrity monitors is not merely sensitivity but timing discipline. The HL-LHC FBCM is asynchronous with respect to the LHC clock at the ASIC level and has no Level-1 trigger; the lpGBT continuously samples discriminator outputs with 0.78 ns time resolution, the back-end reconstructs hit timing and ToT at 0.78 ns granularity, and the firmware maps each pulse to a $(\mathrm{BCID}, \text{intra-BCID time})$ pair for bunch-by-bunch analysis in a 25 ns bunch structure (Shibin et al., 2024). The CMS Beam Halo Monitor is fully synchronous with the LHC clock, digitizes every 25 ns bunch crossing, retains four sub-BC timing bins, and publishes per-bunch MIB fluxes every $\sim 23$ s (Stifter, 2015).

Facility-wide time distribution is central at neutron and test-beam facilities. CSNS uses a White Rabbit timing network with GPS receiver, rubidium clock, and WR Grandmaster; proton-pulse timestamps use an LXI / IEEE-1588-style structure with TAI reference, and timestamp precision is reported as about 1 ns (Zhuang et al., 2018). XBPF/XBTF at CERN also relies on White Rabbit: VFC-HD adds WR timestamps with 8 ns precision, while the FMC-TDC used with the fast trigger planes provides 81 ps time resolution and a TOF resolution below 900 ps (Ortega et al., 2019).

Optical or scintillation profile systems operate on a slower but still operationally real-time cadence. The FRIB scintillator monitor demonstrated 1 Hz imaging with 1 s exposures for single-ion visualization (Levin et al., 2022). The BM4 photon-beam monitor uses a streaming TDC with 0.96 ns binning and reaches a 1 s beam-profiling accuracy of $10\,\mu\text{m}$ when sufficient statistics are accumulated (Kino et al., 2024). The scintillator FLASH monitor reports 20 kfps, corresponding to $50\,\mu\text{s}$ per frame, while real-time FPGA-based computation of beam position, beam shape, and beam dose takes $< 1\,\mu\text{s}$ (Levin et al., 2023). The gas-curtain monitor demonstrated 2D profile measurements within 100 ms to 1 s and an ion drift time of about $20\,\mu\text{s}$ (Kumar et al., 18 Sep 2025).

A plausible implication is that “real time” in this domain is application-specific: 40 MHz bunch resolution is required for collider background monitoring, $\sim 1$ ns timestamps are sufficient for spallation-source pulse alignment, and $_2$0 to $_2$1 profile refresh is operationally adequate for many beamline and therapy contexts.

4. Signal processing and inferred beam-state variables

The digital core of a beam integrity monitor is the mapping from raw pickup or image data to physically meaningful observables. In direct-digitization systems, the principal quantities are amplitude and phase extracted from IQ components: $_2$2 and, for a four-electrode BPM, the transverse coordinates are obtained with the normalized difference-over-sum algorithm

$_2$3

These relations are the basis of the SSRF, CSNS, and ADS digital beam-measurement instruments (Zhao et al., 2015).

For current-transformer systems, the processing is waveform-based. At CSNS, the digitized CT signal is passed through a Butterworth filter, baseline deduction, Savitzky–Golay filtering, careful time-window selection, and numerical integration,

$_2$4

to obtain the proton bunch charge (Zhuang et al., 2018). In FBCM, the signal model is binary rather than analog: the FPGA performs edge detection on each channel’s sampled waveform, computes ToA from the leading edge and ToT from the pulse width, assigns the hit to a BCID, and fills per-channel and per-BCID histograms of hit count, ToA, and ToT (Shibin et al., 2024). The same timing information is used to distinguish prompt collision products from beam-gas or halo backgrounds through shifted or out-of-time ToA distributions (Shibin et al., 2024).

Profile monitors use image-domain processing. The BM4 photon-beam monitor identifies photon events by a coincidence of trigger, x-layer, and y-layer hits within a 48 ns window with no VETO hit, then constructs x and y histograms and fits each projection with a double Gaussian having a common mean $_2$5 and separate core and halo widths $_2$6 and $_2$7 (Kino et al., 2024). The FRIB scintillator monitor performs background subtraction, centroiding, RMS beam-size extraction, and perspective and rotation corrections to express image coordinates in the beam frame (Levin et al., 2022). The hadron-therapy Topmetal monitor computes the center of gravity row by row and performs a linear fit to obtain the beam-track projection and incidence angle, while intensity is the summed charge in the active pixels (Wang et al., 2016).

For fluence-monitoring systems, the derived observable may be radiation damage rather than beam motion. In p-i-n diode arrays, the forward voltage at constant current is approximately linear with hadron fluence over a broad regime; in 3D diode arrays, the leakage current under reverse bias follows the linear damage relation

$_2$8

which allows real-time fluence and 2D profile reconstruction in high-fluence hadron beams (Hoeferkamp et al., 2020). This suggests that beam integrity monitoring can be formulated either as fast kinematic state estimation or as continuous dosimetric state estimation, depending on the facility.

5. Representative operational deployments

The literature contains several mature or prototype deployments that illustrate the breadth of the field.

System	Facility	Real-time function
FBCM	CMS HL-LHC	Bunch-by-bunch luminosity and beam-induced background
Proton monitor	CSNS	Per-pulse charge, timestamp, broadcast, archive
BHM	CMS cavern	Per-bunch machine-induced background flux
XBPF/XBTF	CERN Experimental Areas	Spill-by-spill profile, trigger, TOF, momentum
SGC-IPM	Pelletron / hadron-therapy context	Non-perturbative 2D profile in 100 ms–1 s

At CMS, FBCM operates as a stand-alone luminometer, always on and independent of the CMS central trigger, with per-BCID histograms integrated over about 1 s into “lumi words” (Shibin et al., 2024). Its prototype beam tests with hadron, muon, and electron beams validated the front-end board design, led to the choice of direct sensor–ASIC bonding, and identified an operational threshold region around $_2$9 to $\mu$0 fC for noise suppression and efficiency (Auzinger et al., 7 Jul 2025). The CMS Beam Halo Monitor provides a complementary detector-side view: directional Cherenkov counters at large radius isolate machine-induced background from collision products and publish per-bunch background rates to CMS and the LHC (Stifter, 2015).

At CSNS, the monitor is part of the experimental timing fabric itself: each proton pulse is measured before the target, marked with a high-precision timestamp, broadcast to the control room and neutron instruments, forwarded by redundant agents, and stored in a MySQL history database for offline use and integrity checks (Zhuang et al., 2018). In the CERN Experimental Areas, XBPF/XBTF combines profile tracking, trigger generation, TOF, and momentum spectrometry using the same scintillating-fibre technology; the system was commissioned in H4-VLE with efficiencies above the 90% requirement and TOF below 900 ps (Ortega et al., 2019).

Real-time profile imaging has also been demonstrated for photon, ion, and therapy beams. The BM4 monitor reconstructs GeV-photon beam position and beam size at three longitudinal locations using scintillating fibres, SiPMs, and streaming TDCs (Kino et al., 2024). The FRIB scintillator monitor spans more than five orders of magnitude in rate, from isolated single ions to $\mu$1 particles/s, with linear response within uncertainties and direct visualization of single-ion hits (Levin et al., 2022). In radiotherapy development, the scintillator FLASH monitor reproduces nearly identical beam profiles relative to commercial Gafchromic film and supports an IEC-compliant fast beam-interrupt signal (Levin et al., 2023). The gas-curtain monitor extends the concept to a non-perturbative vacuum-compatible 2D ionization profile monitor for carbon beams and explicitly analyzes its sensitivity under an example FLASH parameter set (Kumar et al., 18 Sep 2025).

6. Reliability, radiation hardness, limitations, and design trends

Reliability requirements are severe because many monitors operate in high radiation or in protection-relevant loops. FBCM is designed for a sensor environment up to $\mu$2 total ionizing dose and fluence up to $\mu$3 (1 MeV neq), uses radiation-tolerant 65 nm CMOS, lpGBT, VTRx+, and bPOL12V, and protects internal logic and configuration registers against SEUs with triple modular redundancy (Shibin et al., 2024). The BHM relies on quartz radiators, UV-sensitive PMTs, and a calibration system to track gain drift and aging (Stifter, 2015). The FLASH scintillator monitor showed a small $\mu$4/kGy signal decrease after a $\mu$5 kGy cumulative dose (Levin et al., 2023). In ion-beam imaging, PM scintillator degradation was measured as $\mu$6 per kGy, with partial recovery in air over a few hours (Levin et al., 2022). For fluence-monitoring diode arrays, p-i-n devices remain linear up to about $\mu$7, while 3D diodes were demonstrated to at least $\mu$8 (Hoeferkamp et al., 2020).

The literature also shows that “real time” is constrained by different failure modes. At CSNS, software processing after ADC capture is acceptable at 25 Hz, but the paper notes that much higher repetition rates or tighter feedback loops would motivate moving more processing into FPGAs or dedicated hardware (Zhuang et al., 2018). The BM4 photon monitor explicitly states that it is not a beam loss monitor system and does not replace standard electron-beam BPMs in the ring (Kino et al., 2024). In the PIP-II BPM study, horizontal position signal linearity showed irregularities around $\mu$9 to $(\mathrm{BCID}, \text{intra-BCID time})$0 mm that persisted under mesh refinement, indicating that calibration and geometric validation are indispensable before such a BPM can be used as a protection-grade diagnostic (Rouzky et al., 2024). For the gas-curtain monitor, single-pulse FLASH monitoring of broad fields requires higher gain and/or higher curtain density than the present proof-of-concept provides, although pencil-beam conditions are much closer to the demonstrated sensitivity envelope (Kumar et al., 18 Sep 2025). Earlier diode-array fluence monitors reported a total fluence uncertainty of about $(\mathrm{BCID}, \text{intra-BCID time})$1, with cable effects and temperature dependence as major contributors (Palni et al., 2013).

Across these implementations, several design patterns recur. Always-on triggerless operation, fine timing relative to the machine structure, radiation-hard or minimally invasive sensing, modular segmentation, and rich diagnostic observables such as ToA/ToT or 2D profile moments appear repeatedly as the preferred architecture for modern systems (Shibin et al., 2024). This suggests that the contemporary real-time beam integrity monitor is not merely a scalar current readout, but an integrated measurement stack in which timing, geometry, background rejection, and reliability engineering are treated as coequal design constraints.