Simulate first
Modelled the full loop in LTspice before touching a soldering iron — checking that the feedback settled instead of oscillating, and that the current held steady across a range of target voltages.
personal project · 2025–2026
A 450nm diode laser needed a current source stable enough not to destroy itself, so I designed and built one: a constant-current driver, simulated in LTspice and soldered by hand, together with Anja Schubeck.
A laser diode's brightness depends on current, not voltage — and its resistance drops as it heats up. Drive it from a plain voltage source and that turns into a feedback loop with only one ending: thermal runaway and a dead diode.
The build itself started simple: a 450nm laser diode in a 12 mm × 30 mm copper module, chosen to cut paper cleanly, mounted on a heatsink to carry the heat away, and a driver to keep the current inside it constant no matter how warm it gets.
The driver is the part that actually took the work. It had to hold current steady well enough to protect a diode with essentially no tolerance for overshoot — while staying simple enough to solder onto a single perfboard.
The driver, soldered on perfboard. The blue potentiometer on the left sets the target current by hand.
Simulated in LTspice before a single joint was soldered — a constant-current regulator built around a precision op-amp, a Zener reference, and a MOSFET doing the actual work.
Reference. A Zener diode (1N750) fed from a resistor sets a stable few volts, independent of small drifts on the supply rail — the number everything downstream is measured against.
Target. A resistor divider steps that reference down to a small target voltage, filtered by an RC stage so the comparison sees a clean, drift-free number rather than supply noise.
Control loop. An LT1006 precision op-amp compares that target against the voltage across a small shunt resistor in series with the laser diode — which is just Ohm's law read backwards, a direct measurement of the current actually flowing.
Actuation. Whatever the op-amp sees, it corrects for by driving the gate of an NMOS in series with the diode — throttling it down the instant current tries to rise, which is exactly the runaway a diode laser needs protecting from.
Modelled the full loop in LTspice before touching a soldering iron — checking that the feedback settled instead of oscillating, and that the current held steady across a range of target voltages.
Built by hand on perfboard: op-amp, shunt, MOSFET, and a potentiometer wired in so the target current can be tuned after the fact rather than fixed at design time.
The 450nm diode sits in its copper module on a finned heatsink, since the module conducts heat well but has almost nowhere to put it on its own.
Dialled the potentiometer in against a measured beam, well inside the diode's rated current — steady output, no drift as the module warmed up.
The diode and beam path stay inside a closed housing — nothing you'd walk into by accident.
Laser safety glasses rated for 450nm come on before the driver is powered, every time.
The whole point of the driver: current is capped in hardware, not just in intent — the diode is protected even if a setting is wrong.
Run against a beam block, never an open room, so a stray reflection has nowhere uncontrolled to go.