准分子激光器高压电源在激光微加工与诱导沉积中的能量
An excimer laser does not amplify light continuously; it stores energy in a pulse-forming network and discharges it into a gas volume in tens of nanoseconds. The supply that charges that network therefore sets the energy of every pulse, one charge at a time.
Charging is a resonant process in most modern tools. The supply charges a capacitor bank through an inductor, the voltage on the bank follows a half sine, and a switch stops the charge at the peak. The precision of the stop is the precision of the pulse energy.
A stop that is late by one percent gives a pulse energy that is high by one percent, and in micromachining that one percent is a depth difference of a few hundred nanometres per pulse, which accumulates into a visible step at the end of a scribe.
Multiple pulse energy control is what the process actually needs. A row of holes drilled with two hundred pulses requires the sum of two hundred energies to fall within a band, and the band is often two percent, so the per-pulse error must be well under one percent and must not have a systematic trend.
The dominant error source is the measurement of the bank voltage. A divider with a temperature coefficient of fifty parts per million per degree will drift by a quarter of a percent over the warm-up of a tool, and that drift appears as a slow increase in pulse energy through the first hour of a shift.
Calibration against the discharge is the practical answer. The bank voltage is measured with a divider, and the divider is trimmed against the actual energy measured by the tool's energy meter at the output, so the systematic errors of both are absorbed into one calibration constant.
Jitter is the second error source. The switch that stops the charge has a delay that varies with temperature and with the voltage at which the switch is triggered, and that delay translates into a voltage error because the bank voltage is still rising when the command arrives.
A switch with a fixed delay is not enough; the delay must be predictable. Triggering the switch from a comparator that watches the bank voltage directly, rather than from a timer that predicts the peak, removes most of the jitter because the comparison happens at the actual voltage.
The comparator has its own delay, which is short and constant, and the residual is a fixed offset that the calibration absorbs. This architecture is why resonant chargers with a voltage comparator reach charging accuracies of a few parts in ten thousand.
Pulse repetition adds thermal stress. A tool running at a few hundred hertz charges and discharges the bank continuously, and the losses in the charging inductor and the switch accumulate in a small volume. Water cooling of the inductor is common, and the supply must be specified with the repetition rate in the cooling calculation.
Degassing of the gas is the process that fills the gaps between bursts. The laser fires in bursts, the gas between bursts is stagnant, and the supply idles. A charger that idles at a fixed voltage consumes little, and a charger that idles by cycling the bank wastes energy and heats the switch.
Ripple on the bank at idle is not a problem until the next burst begins. If the burst starts while the bank is still settling from the idle regulation, the first pulse is off by a fraction of a percent, and the first pulse is often the one that determines whether the process starts cleanly.
The remedy is a defined pre-charge state. Before a burst, the supply brings the bank to a standby voltage below the firing voltage and holds it there for a defined period, so the first charge in the burst starts from a known state rather than from whatever the idle loop left behind.
Energy measurement is the third pillar. A tool that measures pulse energy at the output can correct the charging voltage pulse by pulse, and the correction loop has a bandwidth of a few pulses. That loop is what keeps a long drilling job stable as the gas ages and the discharge impedance changes.
The loop must be slow enough not to fight the charger and fast enough to follow the gas. A bandwidth of ten hertz suits a tool running at two hundred hertz, and it corrects a drift over tens of pulses without amplifying the pulse-to-pulse variation that is inherent in the discharge.
Gas aging produces a monotonic drift. As the halogen concentration falls, the discharge voltage rises for the same energy, and the correction loop raises the charging voltage to compensate; when the charging voltage reaches its ceiling, the tool must stop and refill, and that ceiling is a design parameter of the supply.
A fabrication case illustrates the accumulation. A scribing tool produced a step at the end of each line that grew with line length, and the step matched a slow increase in pulse energy through the burst. The cause was the idle regulation state, which left the bank at a voltage that drifted as the switch warmed. Fixing the standby state removed the step.
A second case involved a first-pulse anomaly after a gas change. The energy of the first pulse in each burst was high by two percent for the first ten bursts after the refill, then settled. The cause was the correction loop's initial condition, which assumed a discharge voltage that no longer applied, and resetting the loop's initial value to the measured value removed the anomaly.
A third case involved jitter at high repetition. The charging switch overheated, its delay grew, and the charging accuracy fell from three parts in ten thousand to twenty. Adding a water-cooled heatsink and derating the switch restored the accuracy.
Maintenance in this role is about the divider and the switch. The divider is checked annually against a reference, the switch is checked for delay drift quarterly, and the cooling circuit is checked for flow monthly. A charger that loses accuracy usually loses it at one of those three points.
Spares are few: one charging switch, one divider, one comparator board and one inductor. The inductor is the expensive item, and it is rarely spared, because a failure in the inductor is usually a failure of the cooling rather than of the winding.
Cost comparisons should include the energy meter. A charger with voltage-comparator control and a defined standby state costs more than a timer-based charger, and the difference is recovered by the reduction in scrapped work, which in micromachining is measured in wafers rather than in percentage points.
Limits are real. A charger cannot stabilise energy if the gas is exhausted, and no specification should promise stability across a refill transient. The supply can only hold the electrical state; the gas state belongs to the tool.
The intent of the design is compact. Charge a capacitor to a voltage that is known to a few parts in ten thousand, stop the charge at the same point every time, and make the state at the start of a burst identical to the state at the start of the last one.
Everything else, from the cooling to the calibration, exists to keep that sentence true over a shift, a week and a year of pulses.
Laser-induced deposition puts a different demand on the same hardware. A deposition head that writes a metal line by decomposing a gas with focused pulses needs a stable energy per pulse but tolerates a wider band, so the charger's specification relaxes on accuracy and tightens on availability.
Availability in that role means the charger must run for hours without a stop for calibration. A charger that drifts by a fraction of a percent per hour is acceptable if the tool's correction loop can follow it, and unacceptable if the loop's bandwidth is a few hertz and the drift is faster than the loop.
Electromagnetic layout completes the electrical design. The charging inductor radiates a field that couples into the energy meter's signal cable, and the effect is a small offset in the measured energy that appears only while the charger is switching. A screened cable and a twisted pair inside the enclosure remove it.
The discharge itself is a radio-frequency event. The current loop through the electrodes and the peaking capacitor forms a circuit that rings at tens of megahertz, and the charger's control board must be placed outside that loop's near field, or the comparator will trigger on noise.
A practical arrangement puts the charger, the control board and the energy meter in three separate zones, with the only connection between them being screened cables that enter the enclosure at defined points.
The final judgement is economic and dull. A charger that holds a few parts in ten thousand costs perhaps twice a timer-based unit, and a single scrapped workpiece justifies the difference, which is a calculation that a process engineer makes in a minute and a purchasing department makes in a week.

