TABLE 3.

Beam control mechanisms.

Reference (chronological order)	LINAC	Energy (MeV)	Detector(s)	Methods	Accuracy	Issues	Improvement Suggestions
Lempart et al. 2019	Elekta Precise	10	EDD 2–3G Diode (IBA)	Developed an in‐house control system circuit connected to the LINAC. The diode detects pulses of radiation, and the signal is fed into a two‐stage amplifier and conditioning circuit. Essentially, the photocurrent is converted to 5 V, which is input into the interrupt pin of an Atmega238 microcontroller unit (MCU). This MCU counts pulses by measuring the rising flanks of the signal using a Timer and Interrupt Service Route triggered by the rising flanks. Once the number of pulses counted matches the user‐defined expected, the MCU sends a logic signal to an optocoupler circuit. This circuit prevents trigger pulses from reaching the thyratron of the LINAC, which effectively prevents electron injection into the gun and the magnetron from producing RF for acceleration.	No additional pulses were delivered than expected. This control approach has very little latency.	LINAC stability issues require a warm‐up every 10 min due to the cooling of the machine as a result of the short beam on times in FLASH. It can be fixed with additional tuning. Not monitoring by dose and pulse height variation was observed, limiting accuracy.	Monitor by dose instead. Suggest using the transmission ion chamber (see Konradsson et al. 2024 row). Lower or modulate the gun current to achieve a desired dose with a fixed number of pulses
Szpala et al. 2021	Varian iX	18	Transmission Ion Chamber	Entered MUs in the treatment console, just as in CONV mode.	The stability of the dose for a given MU was satisfactory (standard deviation of charge ∼ 0.1‐0.2 nC). MU chamber was linear with dose and dose rate.	The smallest MU allowed by iX LINAC is 1. Output dependence on the repetition rate (PRF) due to the MU chamber not fully resetting before the next pulse, allowing the same delivery time for all PRF > 200 MU/min (and thus more dose). Samples were placed in the LINAC head. To increase DR further, must move samples farther toward the source. Eventually, samples would partially block the MU chamber	Use an external monitor chamber with recombination correction factors to re‐enable dose rate and steering servos to prevent output issues.
Rahman et al. 2021	Varian Clinac 2100C	10	Remote Trigger Unit (DoseOptics) (coincidence scattered radiation detector)	Developed a control circuit with an Arduino Mega 2560. Control was based on counting pulses (interrupt‐based) up to a preset amount from the RTU or a preset time. RTU (two scintillators and PMTs) detects pulses from outside of the field. The circuit would turn on a gating reed relay until the preset pulses or time was reached. The reed relay output was connected to pins 32 and 34 of the gating switchbox (Varian), effectively asserting an MLC hold‐off signal when the output was on.	Consistent within 1–2 Gy. No other information	Variations in dose per pulse make the system inaccurate. Long execution time of ancillary routine in Arduino firmware (display/keyboard). Interrupt‐based control factored this out.	Need a dose‐based monitor.
Ashraf et al. 2022	Varian C‐Series	10	W1 Scintillator (Exradin) Remote Trigger Unit	Utilized field programmable gate array (FPGA) controller (consisting of a real‐time input/output controller, detectors, trigger signals, and display) programmed with LabView. FPGA performed the following: Synced with the ‘Sync’ signal from the LINAC and generated signals Integrated the signal from the W1 to calculate DPP (dual channel gated integrator) Counted pulses and measure the pulse width from RTU Send a gating signal to the RPM switchbox when the desired dose or pulses are reached	Dose accurate within 1.5 Gy (more accurate for lower doses) & sometimes an extra pulse was delivered at high PRF.	Spurious particles from RTU due to decay time and positional dependence. Relay and gating system lag, causing delivery errors. No pulse width modulation to meet specific dose goals. Damage issues with W1	Modulate the last pulse. Use the gun current pulse for gating (can apply to any LINAC). Use a different scintillation detector with better radiation hardness (liquid?)
Garty et al. 2022	Varian Clinac 2100C	9	N/A	Modified the timer interface card. Replaced GDLY CNT signal (controls delay between electron gun/klystron, which in turn causes beam on or off) with their own. If voltage is applied, they are out of phase (beam off). Used a USB‐CTR08 card to generate control pulses. This accepts the KLY1 signal (klystron pulses) from the controller and turns on output after predetermined pulses (of KLY1). The output pulls up a test point of equal voltage to cause the gun/klystron to be out of phase. Otherwise, this is grounded, and the beam is on.	Pulse height fluctuated 1–2% for lower dose rates but up to 15% for higher. This was out of their control by using a pulse counter.	Pulse height drifted due to room temperature effecting the gun and accelerator. Burn‐in of diodes in the circuit that drives the klystron—replaced. Activation in the LINAC head. Gun and klystron are asynchronous in most irradiations and must wait a certain amount of time for stability.	Develop an alternative beam monitor: The current system uses a monitor chamber and film to measure the pulse height.
Dal Bello et al. 2023	Varian Truebeam	16	N/A	Varian performed the modifications and set up the control system. They set up a prototype patch that modifies the beam generation and monitoring firmware to read the pulse trigger signal and shut off the beam after a preset amount. The dose in a pulse was stabilized by changing the automatic frequency control. An upper limit on the number of pulses allowed was set up. This had two modes where the limit could be increased for tuning purposes. The target could also be inserted, and the target current could be used to characterize pulses.	No differences between the desired and delivered number of pulses	N/A	Real‐time dose monitoring is needed. Independent system
Cetnar et al. 2024	Varian Clinac iX	16	Beam Pulse Counter (Varian)	Varian offers their own LINAC conversion package, FLEX. The package comes with the Varian Beam Pulse Counter, where users can input desired # pulses, and the device controls delivery while providing audio/visual information. No specifics on what the counter is.	N/A (output was reproducible)	Machine head activation following long beam on times. Need to wait to enter the room safely.	Real‐time dose monitoring as opposed to PC. Suggest alternating current transformer placed after the scattering foil to measure a dose.
Oh et al. 2024	Varian Clinac 213EX	16	Beam Pulse Counter (Varian)	Also used the Varian FLEX package but provided more specifics. The package was the same as a pulse counter, which allows the user to deliver a preset number of pulses. The package allows 1–99 pulses to be delivered in one irradiation. The user can choose between 18–180 pulses/s. A required time gap between irradiations is set up as well. CONV beams remain available, but the vendors restrict any clinical use after modification	Out of 165 output measurements, seven deviated from the mean by > 5% and 11 by > 3%. Since using a pulse counter, this was out of their control. This improved after additional modifications to < 1% (see issues)	No active dose control Operation in service mode requires extra QA. One energy configuration is lost to set up the FLASH beam. Output fluctuations existed until tuning the PFN and increasing the mandatory time gap. Fluctuations during warm‐up	Real‐time dosimetry monitoring for FLEX OR Ability to adjust the gun current and RF from the klystron in combination with the existing system to achieve more dose modulation
Konradsson et al. 2024 (upgrade to Lempart et al. 2019)	Elekta Precise	10	External Transmission Ion Chamber (Elekta) mounted on the upper part of an applicator (55  cm from source) EDD 2–3G Diode (IBA)	Developed an in‐house control system circuit. Uses an Arduino Due microcontroller unit (MCU) to send a signal to an optocoupler to disable thyratron trigger signals and thus halt delivery. Signals are sent when one of the following criteria is met: The desired dose is reached as measured by the external MU chamber. The chamber is corrected for temperature and pressure now that is is external, for recombination in FLASH with a logistic model from a previous study of theirs, and normalization to a target dose. A preset number of pulses is reached, as measured by the diode at the field edge (Lempart et al. 2019). The diode signal is digitized and counted by the MCU. This can also be done by connecting an oscilloscope to the external MU chamber. Preset time is reached, t = (n‐1)/PRF, where n = number of pulses desired. Additionally, the system was synced to the pulse‐forming network (PFN) for dose modulation. The PFN has a ramp‐up time, and a user‐selected delay in the beam on signal is introduced.	Each component was examined separately for control and successfully interrupted every time. The MU chamber was linear with dose and DPP. The results of the expected delivered dose were within 5%. This was improved with PFN synchronization to 0.8% Number of pulses between the diode and MU chamber with oscilloscope agreement observed	Transmission MU chamber correction factors introduce uncertainty Cannot fully rule out the possibility of underdosing due to premature beam interruptions Reporting of DPP with modulation using PFN is difficult Results were initial tests only, not long‐term	Need a different beam monitor that is dose rate independent. They are currently investigating a beam current transform (BCT) paced at the exit window. Want to modulate the last pulse instead of the first. Requires software developments and new transmitter components
Deut et al. 2024	Elekta SL 18 MV	10	Unbiased Silicon Diode Sensor	A PC circuit was developed in‐house. The silicon diode measures radiation pulses. The diode signal is converted to voltage with a trans‐impedance amplifier. The resulting voltage signal is then altered by feeding it to a Sallen‐Key filter and amplification. This signal is then passed to a Schmitt Trigger, effectively producing a square 5 V pulse for each radiation pulse measured. These square signals are counted with an Arduino NANO board until a desired amount is reached. It is unclear from here how the LINAC is connected to this circuit to be shut off.	Compare the pulse counter to readings from an oscilloscope relating to the number of pulses delivered. They observed no pulse loss.	N/A	Desire to have a dose‐based monitor as opposed to counting pulses. The group is working on characterizing silicon sensors for online pulse monitoring. This study illustrated the success of these devices. In the future, they will correlate the sensor reading to dose from film measurements for an online monitor.

Abbreviations: BC, beam characteristics; DPP, dose per pulse; PC, pulse counting; PDD, percent depth dose; TAD, total absorbed dose.