Two independent design axes: ranging vs steering
| Concept | What it addresses |
|---|---|
| FMCW vs ToF (pulsed) | How the laser measures distance (FMCW also provides velocity via Doppler) |
| Spinning vs solid‑state | How the beam is steered across a scene to cover field‑of‑view |
Scanning still exists — even for FMCW
To see a wide scene, a LiDAR has to direct photons into many directions. That can be done by a spinning motor, by micro‑mechanics, by wave‑interference, or by “flash” illumination — but the coverage problem remains.
Three steering families dominate today’s solid‑state designs
In practice, the industry mixes methods to reach a wide 2D field of view: one approach for one axis, and another for the perpendicular axis.
| Method | Moving parts | Steering range | Speed | Key challenge |
|---|---|---|---|---|
| MEMS mirror | Yes (micro) | ±25° | kHz | Mirror fatigue, limited angle |
| OPA (phased array) | None | ±15° | MHz potential | Wavelength‑steering conflict, sidelobes |
| None | ~10–20° | Tied to chirp rate | Range and steering coupled together |
Why “no spinning motor” doesn’t mean “no scanning problem”
Mechanical scanners solve coverage by brute force: they physically sweep. Solid‑state FMCW shifts that burden onto micro‑mechanics (MEMS), wave‑interference (OPA), or physics‑coupled steering (wavelength + grating).
Fast and compact, but it’s still a moving structure. The engineering question becomes long‑term reliability versus field‑of‑view.
The “fully solid‑state” dream: electronic steering, no moving parts. The trade is steering range and artifacts (sidelobes), plus the wavelength‑sweep coupling problem.
Exploits the FMCW chirp to steer “for free” — the approach Voyant Photonics etches onto a fingernail-sized silicon chip. The catch: it couples two functions, so changing range also changes angle.
Today’s approach is tiling multiple apertures around the vehicle, then fusing point clouds. That creates seams, complexity, and vertical‑FOV constraints.
Why seamless 360° stays hard
A single flat aperture — lens, chip, or mirror — only sees the hemisphere in front of it. Physics caps that near 180°, and real optics hold good beam quality to roughly 120° before the edges degrade. Full surround coverage therefore needs emitters pointing in many directions at once.
Today’s answer is multi-aperture tiling: three or four chips mounted at different angles, each covering ~120°, overlapped to reach about 340° horizontally. It works, but it adds seams where objects can slip between sensors, and it forces point clouds from several units to be synchronized and fused in software — extra latency and complexity.
The vertical dimension is the quieter limit. Solid-state chips typically span only 20–30° vertically, so even a fully tiled ring sees a horizontal band rather than a sphere — and for a low object near the ground, that band matters. The pragmatic path the field is taking is sensor fusion: pairing LiDAR with camera and radar so no single device has to deliver the full surround, the approach robotaxi fleets already rely on.