VCSEL vs edge-emitting laser: choosing a source for 3D sensing and LiDAR

Two ways to build a laser

An edge-emitting laser (EEL) is a Fabry-Perot cavity built into the plane of the wafer. Light leaves from a cleaved facet, and the emitting aperture is a stripe - about 110 µm by 30 µm in our 905 nm parts - so the beam is elliptical and astigmatic: 9° parallel to the junction, 25° perpendicular to it.

A VCSEL (vertical-cavity surface-emitting laser) has its cavity perpendicular to the wafer, so light leaves the surface. The aperture is round, so the beam is round: our 940 nm part is 19° at 1/e², symmetrical, with a doughnut far field.

Everything else follows from that geometry.

The numbers, side by side

Read the third row and the first row together. The EEL has thirty times the peak power, and the VCSEL has thirty times the pulse energy. Both statements are true, and neither is the whole answer.

Peak power is not energy

This is the same identity that governs every pulsed source:

A 6 W VCSEL driven for 100 ns produces 0.6 µJ. The same VCSEL driven for 1 µs produces 6 µJ - as much as a 60 W edge emitter at 100 ns. The EEL reaches 18 µJ in 100 ns; to match that energy, the VCSEL would need a 3 µs pulse.

So the honest summary is:

• The EEL delivers energy fast. High peak power, short pulses, high instantaneous return.
• The VCSEL delivers energy slowly. Low peak power, long pulses, low instantaneous return.

Whether "fast" or "slow" is better depends entirely on your receiver and your resolution requirement, not on the laser.

Where the long pulse helps, and where it hurts

It helps when the receiver integrates: a photon-counting or integrating receiver accumulates the whole return, so what matters is total energy, not its shape. The VCSEL's 3 % duty cycle also means far more pulses per second at the same average power, which is exactly what averaging-based 3D sensing wants.

It hurts if you need to separate two targets at different distances. A 100 ns pulse smears the return over about 15 m, because light travels 0.3 m per nanosecond and has to make the round trip. A 1 µs pulse smears it over 150 m. In a direct time-of-flight system where you time the leading edge, the precision can still be good - but the ability to resolve a person standing in front of a wall disappears.

For a gesture sensor at 1 m and a vehicle at 200 m, that does not matter. For a rangefinder that has to see a deer in front of trees, it is fatal.

Range, honestly

Using the published model in the range equation article, and treating the 940 nm VCSEL as close enough to 905 nm for a silicon APD:

The 905 nm VCSEL array is the interesting entry: 70 W from a surface emitter is enough to compete with a mid-range edge emitter, with a circular beam and no FAC. Arrays scale by adding emitters on the same die, which is why VCSEL power has been climbing so quickly.

What else changes

Wavelength stability. A VCSEL's emission wavelength moves much less with temperature than an edge emitter's 0.10 nm/K, which matters when a narrow optical filter is doing the heavy lifting against sunlight. The trade is a wider spectral width and a lower coherence length - which is good news for speckle and bad news for interferometry.

Speckle. A VCSEL array is many independent emitters lasing at slightly different wavelengths, so the speckle pattern averages out. A single-mode edge emitter produces strongly coherent light and a visible speckle pattern on the target. For imaging, this is often the deciding factor. The same mechanism is the cure in visible and RGB laser illumination, where speckle has to be removed rather than averaged away.

Elliptical beam vs round. A 25° fast axis is not just a collimation problem. At 1 km, a 25° fan is a stripe 440 m tall, most of which never reaches your receiver - see FAC and fibre coupling. VCSELs are born round.

Failure modes. An edge emitter fails at the facet, usually from catastrophic optical damage caused by over-current or contamination. A VCSEL fails by thermal rollover, and individual emitters in an array can fail without killing the device, which degrades gracefully but is hard to detect.

Cost at volume. Edge emitters are made by cleaving and coating facets, one wafer into many bars. VCSELs are made and tested entirely on-wafer, which is why they became the default for high-volume consumer sensing.

A decision rule

Choose an edge emitter when: the requirement is beyond a few hundred metres, the pulse must be under 100 ns, the target is not cooperative, or you need the highest peak power per dollar. Range-finding, vehicle LiDAR, long-range industrial sensing.

Choose a VCSEL when: the requirement is a few metres to a few hundred metres, the receiver integrates, you want a round beam and no FAC, you want surface-mount integration and wafer-scale cost, and you can live with a long pulse. Consumer 3D sensing, gesture and proximity, machine vision, short-range automotive flash.

Choose both when they are not competing: a 905 nm edge emitter for the long-range channel and a 940 nm VCSEL array for the wide-field short-range channel is a common and sensible architecture in automotive sensing.