Automotive

Radar

Automotive radar measures how far away something is, how fast it's moving relative to you, and roughly where it is angularly — by transmitting radio waves and analyzing what bounces back. Radio waves shrug off rain, fog, direct sun glare, and total darkness in a way cameras simply can't, which is exactly why radar has been in cars since long before "ADAS" was a marketing term.

Quick reference

What it is: An RF sensor that measures range, relative velocity, and angle of surrounding objects from a reflected radio signal.
Mental model: Shouting into a canyon and timing the echo — except the echo's pitch also tells you whether the canyon wall is moving toward or away from you.
Key numbers: 77–81 GHz is the modern automotive band; range/velocity resolution comes down to centimeters and works identically in fog, rain, or pitch dark.
Reach for it when: You need reliable range and closing-speed data regardless of weather or lighting — adaptive cruise control, automatic emergency braking, blind-spot monitoring.

How FMCW radar measures range and velocity

Almost every automotive radar today is FMCW (Frequency-Modulated Continuous Wave): instead of a single fixed frequency, it transmits a chirp — a signal whose frequency sweeps linearly upward over a short interval, then resets and sweeps again. The reflected chirp comes back delayed by however long the round trip to the object and back took.

Both chirps sweep at the same rate, so the frequency gap between them (f_beat) stays constant while both are ramping — that gap, not the raw time delay, is what the radar actually measures.

The radar mixes the transmitted and received signals together, which produces a much lower "beat" frequency equal to the difference between them — easy to sample and process compared to the original multi-GHz signal. Range and velocity fall out of two different properties of that beat signal:

Range: R = (c × T_chirp × f_beat) / (2 × B) — worked example: bandwidth B = 1 GHz, chirp duration T_chirp = 40 µs, measured f_beat = 2 MHz gives R = (3×10⁸ × 40×10^-6 × 2×10⁶) / (2 × 10⁹) = 12 meters.
Velocity: Compare the beat frequency across consecutive chirps rather than within one — a moving object adds a small Doppler shift chirp-to-chirp. v = (λ × f_doppler) / 2, where λ is the radar's wavelength (about 3.9 mm at 77 GHz).

One chirp alone gives range; a whole sequence of chirps, compared against each other, gives velocity — which is why automotive radar always transmits chirps in rapid bursts ("frames") rather than one at a time.

Angle of arrival: knowing where, not just how far

Range and velocity alone don't say where an object is angularly — that comes from comparing the same reflected signal arriving at two or more receive antennas, spaced a known distance apart. An object off to one side is very slightly closer to one antenna than the other, so its reflection arrives with a measurable phase difference between them.

The wave reaches RX1 after traveling an extra d·sinθ compared to RX2 — that extra distance shows up as a phase difference Δφ = 2π·d·sinθ / λ, which DSP inverts to solve for θ.

Production radars use several receive antennas (and increasingly several transmit antennas too, in a MIMO arrangement) specifically to get a sharper angle estimate than two antennas alone could provide — more antennas means more phase-difference measurements to combine.

Frequency bands and radar classes

Band	Status	Typical class	Notes
24 GHz	Legacy, being phased out	SRR (short-range)	Narrow allowed bandwidth limits range resolution; many regions now restrict or sunset this band.
76–77 GHz	Current mainstream standard	LRR (long-range)	The forward-facing radar behind the front grille/bumper for ACC and AEB.
77–81 GHz	Modern, wider bandwidth	SRR / MRR (corner radars)	Wider allowed bandwidth (up to ~4 GHz) gives finer range resolution — used for blind spot, cross-traffic, and newer "imaging radar."

A typical modern car carries one long-range radar facing forward plus four corner (short/medium-range) radars — enough overlapping coverage that there's no fully blind angle around the vehicle.

The radar processing chain

The physical bounce off the object (top) and the electronic signal-processing chain (bottom) are two different things happening in sequence — the dashed arc is the RF path, the solid chain is what happens to the signal once it's back.

Radar vs. LiDAR vs. camera, briefly

Each sensor gets its own full page (LiDAR and camera sensors are planned next), but the short version explains why cars carry all three instead of picking one:

	Radar	LiDAR	Camera
Best at	Range + velocity, any weather	Precise 3D shape/position	Classification (what is it?)
Weak in	Angular resolution, object classification	Heavy rain/fog/snow, cost	Darkness, glare, fog, rain
Works in the dark?	Yes, unaffected	Yes, unaffected	Degraded without added illumination

This is exactly why ADAS relies on sensor fusion rather than any single sensor — radar confidently says something is there and how fast it's closing; camera says what it probably is; combining them covers each one's blind spot.

Real-world applications

Adaptive Cruise Control (ACC) — the forward long-range radar's range and closing-velocity measurements are exactly what a car needs to hold a set following distance automatically.
Automatic Emergency Braking (AEB) — a rapidly closing object the driver hasn't reacted to triggers braking, using the same range/velocity data at a much shorter reaction budget.
Blind-spot monitoring — corner radars watch the zones a mirror can't fully cover, unaffected by darkness or a dirty mirror.
Rear cross-traffic alert — rear corner radars detect approaching traffic while reversing out of a parking spot, before it's visible to the driver.

Gotchas

Field notes

Radar has genuinely poor angular resolution compared to LiDAR or camera — it can tell you confidently that "something" is at a range and angle, but not reliably what it is. Classification is usually left to camera or fusion logic, not radar alone.
Stationary metal objects (manhole covers, overhead signs, bridge structures) can produce strong reflections that look like real obstacles — a classic source of "ghost" detections that fusion and tracking logic have to filter out.
Multiple vehicles' radars operating in the same band on a busy road can mutually interfere — interference mitigation is real, ongoing automotive radar engineering, not a solved problem.
24 GHz designs are being phased out of new vehicles in many regions as spectrum regulations tighten — a design choice worth checking before starting new automotive radar work.