Driving a 905 nm pulsed laser diode: current, edge rates and what actually kills diodes

The driver writes the pulse the diode actually sees

A datasheet pulse width is a specification, not a guarantee. What reaches the junction is a current waveform with finite edges, overshoot and ringing, and it is that waveform - not the number on the dial - that sets the pulse energy the target sees.

This article uses our own LZ-905-PD01 as the worked example, because the numbers are published. Its headline parameters:

The first thing to notice is the mismatch between the driver's ranges and the diode's ratings. The driver will happily run 100 ns at 30 kHz. The diode will not.

Edge rates are a fixed tax on every pulse

The current pulse is a trapezoid: a 7.5 ns rise, a flat top, and a 12 ns fall. That is about 20 ns of transition on every pulse, which is 40 % of a 50 ns pulse and 20 % of a 100 ns pulse.

For a trapezoidal pulse, measured at half maximum, the arithmetic works out cleanly: the charge delivered is the peak current multiplied by the full width at half maximum, so

is exact for an ideal trapezoid. The problem is not the arithmetic, it is reaching the flat top at all. If the pulse is too short for the current to settle, the diode never sees its rated peak current, and you get less energy than the calculation promises. Two mechanisms cause that: the driver's own current regulation loop, and the inductance of the wiring between driver and diode.

Loop inductance decides whether you reach peak current

A 905 nm diode that runs at 25 A needs the driver to push 25 A into it in a few nanoseconds. That is a rate of change of current, and inductance resists it:

To reach 30 A in 7.5 ns you need di/dt of about 4 A/ns. With 20 nH of loop inductance - which is roughly what a pair of 3 cm leads gives you, about 6-8 nH per centimetre of wire in air - the inductive drop is:

Eighty volts of drop, on a driver whose output range ends at 50 V and whose MOSFET is rated at 80 V. The current never gets there, or the switch dies, or both. Practical consequences:

• Keep the driver-to-diode loop as short and as flat as physically possible. Every centimetre counts against you.
• Run the return path directly under or beside the forward path. A twisted pair is better than two separate wires.
• Use a low-inductance bypass capacitor bank of a few microfarads with a low ESR, placed at the diode, not at the driver.
• Do not put the diode on a long cable to a remote head and expect full peak current. Either put the driver at the diode, or accept a longer pulse with less peak current.

The storage capacitor matters more than the supply

The energy per pulse is small - 18 µJ for a 180 W / 100 ns part - but it has to be delivered in 100 ns. That is a charge of

A 100 nF capacitor charged to 22 V stores about 24 µJ, which is barely one pulse and it will sag badly during it. Size the local bank in microfarads, and let the power supply refill it between pulses: at 10 kHz the average power is only 180 W × 0.1 % = 0.18 W, so the supply is trivial and the decoupling is everything.

Duty cycle: the driver and the diode have different limits

The diode's limit is 0.1 % duty cycle. The driver's frequency range goes to 30 kHz. Combine them carelessly and you get 100 ns at 30 kHz = 0.3 %, three times the diode's rating. The diode will work, get hot, and fail months later in the field.

Enforce the limit in the product's firmware, not in a note on a test report. The table is in the 905 nm selection article: 100 ns caps you at 10 kHz, 50 ns at 20 kHz, 33 ns at 30 kHz.

If you need a high repetition rate with a long pulse, the solution is to split the energy across emitters driven in sequence, or to move to a different laser technology. It is not to run the part over its rating.

Timing jitter is range error

In a time-of-flight system, 1 ns of timing uncertainty is 15 cm of range error, because the light has to travel out and back. A driver with a 7.5 ns rise time and a few nanoseconds of jitter puts that error straight into your measurement.

• Use the driver's external TTL input when the system has a master clock, and keep the trigger path short and impedance-matched.
• Specify the start reference from the optical pulse, not from the electrical trigger, if the accuracy budget is tight. A monitor photodiode or a pick-off inside the transmitter is the usual answer.
• Check the temperature drift of both the driver's delay and the diode's own turn-on delay.

What kills diodes

Listed roughly in order of how often we see them:

1. Over-current on a short pulse. More current than the datasheet allows, for a few nanoseconds, is still over-current. The junction fails at the facet first.
2. Duty cycle violation. Slow, thermal, and it shows up as degraded power rather than a dead part.
3. Inductive spikes above the reverse rating. Our 905 nm parts are rated for 20 V reverse. Ringing on the falling edge goes negative, so a reverse-clamping diode across the laser is cheap insurance.
4. ESD. These are Class 1A static-sensitive devices. Grounded wrist strap, grounded iron, conductive packaging, and no handling by the pins.
5. Hot plugging. Connecting a diode to a live driver means the decoupling capacitors discharge through it.
6. No current limiting into the diode. A constant-voltage supply with no regulation will deliver whatever the diode will take. Always regulate current, never voltage.
7. Condensation. Operating below the dew point with a cold package destroys the facet. Dry air or nitrogen in the housing, and moisture-barrier packaging in storage.

A commissioning checklist

• [ ] Measure the optical pulse, not the electrical one: a photodiode with at least 1 GHz of bandwidth into an oscilloscope with matched input impedance.
• [ ] Measure pulse energy with a calibrated energy meter as well, because a fast scope can underestimate the area of a pulse with slow tails.
• [ ] Confirm the flat top is flat: current should reach its plateau well before the pulse ends, with no more than a few percent of overshoot.
• [ ] Verify the duty cycle at the maximum repetition rate the product will ever command, including any fault modes.
• [ ] Check the diode's case temperature at the worst-case ambient, after the product has been running for an hour.
• [ ] Confirm the reverse clamp is present at the diode, not on the far end of a cable.
• [ ] Record the measured energy against the calculated value. If the measurement is more than 10 % below the calculation, the loop or the compliance voltage is the reason.