4 Kilowatts Lost in the Diode Stack: Inside the Brutal Tolerances of Industrial CO? Laser Power Supplies
Published on 10/09/2026 at 05:12 | Editorial boerse-global.de
Engineers specifying power electronics for pulsed CO? lasers are working against margins that leave almost no room for error. A body of detailed technical analysis now circulating in specialist circles lays out just how unforgiving those margins are — from nanosecond-scale switching windows to temperature swings that can quietly unravel a resonance design.
The pulse profile leaves little slack
For helium-free LE-CO? lasers, the required parameters are strict: pulse rise times of no more than 300 ns, pulse widths between 100 and 300 ns, and charging accuracy of ±0.5%. The projected service life exceeds 5 × 10? shots.
At an operating voltage of 30 kV, a 300 ns rise translates to at least 100 kV/µs. With a capacitance of 10 nF, the discharge loop inductance should stay below 1.86 µH.
Running at a repetition rate of 700 Hz gives a pulse interval of 1.43 ms. A 10-nF capacitor charged to 30 kV stores 4.5 J, meaning the recharge within that interval demands an average charging power of roughly 3.15 kW.
To keep those peaks under control, the analysis recommends laminated busbars with a connection inductance below 10 nH, an equivalent series inductance (ESL) of less than 20 nH for the storage component, and charging repeatability of ±0.5%.
Where the losses pile up: multiplier chains and Marx generators
Diode chains in CO? lasers see switching frequencies between 20,000 and 200,000 operations per second. Reverse-current effects combined with parasitic loop inductance can produce voltage spikes reaching twice the output voltage.
Take a concrete case: 10 kV blocking voltage, 1 A recovery current and a 500 ns recovery time at 100 kHz yield 250 W of loss per diode. Across an 8-stage chain with 16 diodes, that adds up to 4 kW — roughly 20% of total output power.
The recommended countermeasures are diodes with a reverse recovery time (t_rr) under 100 ns, reverse recovery charge (Q_rr) below 1 µC and leakage currents under 1 µA, supported by conductor loops under 20 mm and a junction temperature margin above 30%.
TEA-CO? lasers, which call for voltages from 30 to 100 kV, frequently rely on Marx generators. A 10-stage design with 5 kV per stage works out to 50 kV. Delivering 40 J of pulse energy requires 0.8 J per stage, which at 5 kV means a minimum capacitance of 64 nF.
A parasitic loop of 50 nH already produces an inductive voltage drop of 1,000 V when current rises by 1 kA in 50 ns. The design limits cited are stage delays below 10 ns, loop inductances under 50 nH per stage, and component matching of ±2%.
High-frequency feed and the thermal trap
At excitation frequencies of 13.56 MHz, voltage along a 60-cm electrode can vary by 30 to 50%. Achieving a homogeneous discharge calls for at least 4 feed points, 1-nF coupling capacitors, series resistors of 200 ? with ±1% tolerance, connection inductances below 5 nH, and 1 M? bleeder resistors per coupling capacitor.
In a 10 kW system with 16 branches, roughly 600 W of high-frequency power lands on each path — a duty that argues for Class I capacitors with a dissipation factor (DF) below 0.1%.
Thermal stability turns out to be the decisive weakness. As little as 10 V of residual ripple on a 10 kV bus can cause significant fluctuations in laser output. Measurements show that power supply efficiency can fall from 92% to 75% when the unit heats up by 50 °C.
If a Class II resonant capacitor loses 10% of its capacitance over a 50 °C temperature rise — components in this class can shed 10 to 15% — the first series resonance frequency f_r1 shifts by 5.4%.
The technical papers therefore steer designers toward N4700 Class I capacitors with a dissipation factor below 0.1% and capacitance tolerance of ±2%, backed by partial discharge testing at 5 pC and 1.5 times the rated voltage.
Leakage inductance does double duty in LLC designs
In LLC resonant power supplies, the transformer's leakage inductance serves as the series resonant inductor. A leakage spread of ±10% keeps the resonant frequency within a ±5% band. Push manufacturing tolerance to ±30%, however, and the frequency drifts by about 15% — enough to jeopardise the soft-switching range.
The advice here: match the capacitor value to the measured leakage inductance, keep the ratio of magnetising inductance to series inductance between 3 and 7, and aim for winding capacitances below 100 pF. One worked example puts 50 µH and 4700 pF at 328 kHz; with a main inductance of 300 µH, the second resonance point drops to 124 kHz.
Modern supplies using inverters in the 20 to 100 kHz range and turns ratios of 1:80 to 1:150 deliver 20 to 50 kV DC and reach efficiencies of 92 to 96%.
Classic 50 Hz transformers, by contrast, manage only 80 to 85% while weighing 15 to 30 kg. High-frequency transformers are built 5 to 8 times smaller and show load-step settling times under 0.5 ms, compared with 10 to 40 ms at mains frequency.
To safeguard these systems, the benchmarks applied are filter dissipation factors below 0.1% at 50 kHz and rectifier recovery times under 100 ns — all within a switching frequency range for industrial topologies that spans 20 to 200 kHz.
