High-voltage output of accelerator power supplies in free-electron laser facilities and compact devices
The linac modulator in the facility hall feeds the klystron that drives the accelerating structure, and the modulator output runs 115 kV at 75 A pulses, 5 microseconds flat-top, repeated at 50 Hz. The pulse forming network sits in a 3000-liter oil tank, and the oil temperature holds at 45 degrees Celsius under continuous operation. Flat-top droop on the pulse spec sits below 0.05 percent, measured on the last 2 microseconds of the pulse. The beam energy at the linac exit tracks the modulator voltage squared, so a 0.1 percent voltage drift moves the beam energy by 0.2 percent, enough to push the output wavelength off the operation band. The modulator control loop updates every pulse, sampling the flat-top voltage and adjusting the charge setpoint for the next pulse. Pulse-to-pulse amplitude stability on the installed units reads plus or minus 0.03 percent on a 10-minute capture, which meets the operation spec without headroom for aging drift. The charging supply for the PFN, a 5 kV DC unit on the tank lid, draws 18 kW average from the mains and regulates the charge within 5 V.
Magnet supplies along the beamline take the largest share of the cabinet count. Quadrupole magnets run at 120 A across a 12 V load, and the current ripple spec sits at 50 ppm peak to peak. Dipole supplies hold 350 A at 28 V with the same ripple band. Each magnet supply occupies one rack slot, and a typical beamline cabinet row holds 18 slots. Current drift over a shift measures 150 ppm on the original generation supplies, which forced the operation group to re-reference the beam tunes between shifts. The newer supplies, installed during the last upgrade, hold 60 ppm across a shift, and the re-reference routine now runs weekly instead of daily. The magnet coils water-cool at 22 degrees inlet, and the deionized loop holds resistivity above 1 megaohm-centimeter. A failed coolant flow switch on one quad supply in 2024 shut down the row for 3 hours, and the flow switches were put on a quarterly test schedule after that event. The magnet supply output current gets a weekly check against a 50 A shunt standard, and the check takes 15 minutes per cabinet row.
Tank oil maintenance drives the calendar. The oil in the modulator tank gets sampled every 6 months for dielectric strength and dissolved gas analysis. Dielectric breakdown voltage spec sits at 30 kV per 2.5 mm gap, and the last two samples read 27 and 28 kV, just under the band. The group changed the oil at the 4-year mark, replacing 2500 liters and topping with fresh oil after the sediment drain. Partial discharge inside the tank, measured with an ultrasonic sensor mounted on the tank wall, read 18 pC after 8 years of service, climbing to 45 pC before the oil change. The crowbar spark gap inside the tank trips at 120 percent of modulator voltage, and the gap erosion shows as a slow rise in trip threshold. Gap realignment happens annually during the winter shutdown, taking 8 hours per modulator. The oil level sight glass, checked monthly, shows a 20 mm drop per year from evaporation, and the sight glass refill uses the same oil grade as the bulk stock. The tank breather filter, a desiccant canister on the tank lid, changes color when saturated, and the canister gets replaced at the annual service visit.
Klystron collector heating sets the duty cycle limit. The collector dissipates 8 kW average at full pulse duty, and the water cooling loop runs 12 liters per minute at 20 degrees. A blocked flow alarm on the collector loop once let the tube run hot for 20 minutes, and the collector vacuum seal softened, causing a slow vacuum rise that took the klystron offline for 6 weeks of replacement. Flow alarms now interlock the modulator HV enable, and a second flow sensor was added in series after the incident. The klystron, a 220000 euro line item, and the facility keeps one spare on site. Tube lifetime on the installed fleet averages 35000 operating hours, and the replacement schedule blocks the calendar 2 months ahead of the predicted end. The klystron cathode current, 0.5 A at full beam, drifts 5 percent over the tube life, and the modulator charge voltage compensates the drift to hold the beam power flat. The cathode heater supply, 12 V at 10 A, draws power continuously, and the heater current check happens at the weekly tune session.
Compact devices in the adjacent hall use a different supply topology. The compact linac modulator stacks solid-state switches instead of a thyratron, and the output runs 80 kV at 40 A pulses of 4 microseconds at 100 Hz. The stack contains 24 switch modules in series, and a failed module triggers a module-level bypass that lets the machine run at 90 percent voltage until the next maintenance window. Module replacement takes 1 hour per unit, and the spare modules sit on the shelf in packs of four. The compact units run a smaller oil tank, 800 liters, and the oil change interval shrinks to 2 years because the smaller volume degrades faster. Cooling on the compact units uses forced air on the switch stack instead of water, and the intake filters clean monthly. The compact hall logs 2 module failures per quarter, against zero thyratron failures in the main hall over the same period. The switch module gate drive board, a 12-channel unit on the stack ring, gets a visual inspection at the quarterly visit, and the board capacitors get an ESR check at the annual shutdown.
Interlock logic decides what an operator can and cannot start. The beamline runs a hardwired interlock chain: door switches, radiation monitor contacts, vacuum gauge contacts, water flow switches, and modulator fault relays all wire into the safety PLC. The PLC refuses the beam enable if any contact opens, and the operator sees a plain English fault message on the control screen. Bypass keys for maintenance sit in a locked cabinet, and the key logbook records every removal. One bypass left engaged after a maintenance window caused an uncontrolled modulator ramp in 2022, and the interlock logic was rewritten to require a deliberate reset sequence after every maintenance contact. The reset takes 90 seconds, and the operators complain about the delay on busy days, but no second incident has occurred since. The radiation monitor in the accelerator tunnel trips at 10 microsieverts per hour, and the trip log shows one spurious event per year, traced to a loose detector cable. The cable routing was re-done after the second spurious event.
Beam energy measurement calibrates the output. The beam energy probe, a toroidal current monitor in the beamline, samples the beam current at 100 MHz bandwidth, and the energy estimate integrates the current against the known accelerating voltage. The probe calibration uses a reference resistor standard on the bench, and the calibration check happens at the monthly tune session. The beam position monitor, a 4-button pick-up near the linac exit, reads the beam centroid to 0.1 mm resolution, and the position data feeds the magnet trim loop. Position drift over a shift measures 0.3 mm, and the trim loop corrects the drift within 10 seconds. The beam loss monitor, an ion chamber near the beamline, trips at 5 percent loss, and the trip log shows one event per quarter, usually from a mistuned magnet on recipe switch.
Shift operations split the hall into day and night crews. The day crew runs beam tuning, witness runs, and sample exchanges; the night crew holds steady output and watches trends. The modulator voltage trend file logs every pulse, and a nightly script flags any 10-minute window where the mean drifts more than 0.05 percent. The magnet supplies get a weekly current check against a calibrated shunt, and the checks record in the equipment logbook. Scheduled downtime blocks 2 weeks per year per hall, split between winter and autumn, and the blocks cover oil sampling, gap realignment, module swaps, and cable inspection. Unplanned downtime on the main hall averaged 18 hours per year before the last upgrade, and 11 hours after. The compact hall averaged 30 hours per year, mostly on switch modules and filter changes. The nightly trend review, done by the night crew lead, takes 20 minutes and includes the modulator voltage, magnet current, and oil temperature plots for the shift.
Cable and connector maintenance hides the real wear. The HV cable between modulator and klystron runs in oil-filled conduit, and the conduit joints get a visual inspection quarterly. One joint in the main hall showed oil seepage in 2023, and the group re-banded the joint before dielectric loss turned into a fault. The klystron socket at the tube end sees thermal cycling every beam start, and the socket contacts oxidize after 500 thermal cycles. Socket replacement happens at 600 cycles, taking 4 hours. The grounding straps between modulator tank and hall floor get a torque check annually, because loose straps raised the cabinet ground impedance enough to couple noise into the magnet supply reference lines. The reference line noise showed up on the tune quality plots as a 20 ppm ripple, and the torque fix brought the ripple back under 50. Nobody on the operations floor calls the supply chain the bottleneck any more, and the monthly maintenance hour budget per modulator sits at 6 hours. The spare cable stock, two lengths of 10 m HV cable, sits in the hall rack room, and the cable connectors get a monthly visual check for oxidation.
The modulator trigger chain uses a thyratron grid supply, a 5 V pulse unit on the tank lid, and the grid pulse width is 2 microseconds. The thyratron grid current, 2 A peak, drifts 10 percent over the tube life, and the modulator charge voltage compensates the drift to hold the beam power flat. The grid supply capacitor bank gets replaced at the 6e8 shot mark, along with the thyratron swap. The spare grid supply costs 1800 dollars, and the spare sits on the shelf in the hall rack room. The trigger cable between the grid supply and the thyratron, a 1 m coax, gets a quarterly visual check, and the cable connector gets a torque check at the annual shutdown.
The beam dump, a water-cooled copper block at the beamline end, absorbs the unused beam power, and the dump cooling runs at 8 liters per minute. The dump temperature holds 35 degrees under full beam, and the dump water flow alarm trips at 6 liters per minute. The dump has never tripped, but the flow gauge reading trends down 0.3 liters per minute per year as the filter cartridges clog, and the cartridges get changed on the annual shutdown. The dump radiation shield, a 200 mm lead wall, gets a visual inspection at the quarterly visit, and the shield joints get a torque check at the annual shutdown.
The modulator oil pump, a 200 liter per minute unit on the tank bottom, circulates the oil through a cooler and a filter, and the filter pressure drop reads 150 kPa on a clean unit. The pump vibration, measured at the quarterly visit, trends up from 1.0 mm per second to 1.8 mm per second over 2 years, and the pump bearing gets replaced at the 2.5 mm per second mark. The pump motor, a 3 kW unit, draws 6 A at full flow, and the motor current trend shows no drift over 3 years. The spare pump costs 4200 dollars, and the spare sits on the shelf in the hall rack room.
The beamline vacuum gauges, an ion gauge unit on the chamber wall, get a quarterly reading check, and the gauge drift measures 5 percent per year. The gauge filament, a 0.1 mm tungsten wire, lasts 5000 hours, and the filament replacement happens at the annual shutdown. The spare gauge costs 1200 dollars, and the spare sits on the shelf in the hall rack room.
The hall ventilation, a 4000 cubic meter per hour make-up air unit, holds the hall temperature at 22 degrees, and the air filter gets replaced at the quarterly visit. The ventilation duct, a 300 mm spiral pipe, gets a visual inspection at the quarterly visit, and the duct joints get a torque check at the annual shutdown. The hall floor drainage, a 100 mm trench drain along the rack room wall, gets a weekly visual check, and the drain strainer gets a cleaning at the monthly visit. The hall lighting, a 54 W LED fixture row, gets a brightness check at the monthly visit, and the fixture driver gets an ESR check at the annual shutdown. The hall door, a 1.5 meter wide sliding unit, gets a visual inspection at the quarterly visit, and the door track gets a cleaning at the annual shutdown.
