Field note · Selection
mmWave vs PIR vs Ultrasonic vs LiDAR: Where Each One Fails
Feature tables list what a sensor can do. They never list what it cannot, and that is where projects fail.
The whole comparison in one table
| Technology | In one line | Afraid of |
|---|---|---|
| PIR | Only sees a moving heat source | Still people, heat sources, high ambient temperature |
| Ultrasonic | A tape measure in still air | Wind, temperature gradients, soft or angled surfaces |
| mmWave radar | The only one that does still targets + rain + sealed enclosure | Metal multipath, under 30 cm, small apertures |
| LiDAR | Indoor precision champion | Rain, dirty windows, price |
If you are in a hurry, that table is enough to decide. The rest is detail.
PIR: only sees a moving heat source
How it works: it measures changes in incoming infrared. Passive, no emitter, very cheap.
What it is afraid of
- Still people — it only sees change, so someone reading at a desk disappears within seconds
- Heat sources — sunlight, a radiator, an AC vent all look like someone walking past
- High ambient temperature — when the room approaches body temperature, contrast collapses and range drops sharply
Pick it when: you only need "did someone walk through", the budget is minimal, or the device runs on a battery.
Ultrasonic: a tape measure in still air
How it works: it emits a burst of inaudible sound and times the echo. Distance comes from time of flight.
What it is afraid of
- Wind — moving air shifts the medium carrying the sound and creates echoes that are not there
- Temperature gradients — the speed of sound changes with temperature; without compensation you get a systematic error
- Soft or angled surfaces — fabric, curtains, sloped walls absorb or scatter the echo, so it reports nothing
- Distance — practical range is a few metres, and the beam is too wide to localise anything
Pick it when: level, tank fill, or parking — the target is directly below, flat, and the air is still.
mmWave radar: the only one that does still targets, rain and sealed enclosures
How it works: it transmits an FMCW chirp and measures the beat frequency of the return. Range from the beat, velocity from Doppler, angle from phase across elements.
What it is afraid of
- Metal multipath — metal reflects almost everything, so one target appears at several ranges. Cheap modules cannot suppress it
- Very close range — under about half a metre the transmit signal leaks into the receive path and blinds the sensor
- Doppler-only designs — they measure velocity, not presence, so a stationary person vanishes. Architecture matters more than the marketing
- Small apertures — angular resolution comes from aperture. A matchbox cannot separate two people a metre apart at ten metres
Pick it when: you must detect people who are not moving, work through rain, fog, dust or darkness, seal the sensor behind plastic, or measure range and velocity together.
LiDAR: indoor precision champion
How it works: it fires laser pulses and times the return, building a point cloud. High end scans; low end is a single-point rangefinder.
What it is afraid of
- Rain, fog, snow, dust — optical wavelengths scatter badly on droplets; 100 m in clear air can become 20 m in heavy rain
- A dirty window — loading bays and roadside poles get dirty, and performance degrades silently until someone cleans it
- Price — a single-point unit is affordable; a scanning unit producing a usable point cloud is not, and draws accordingly
- Eye safety — transmit power is capped by laser class, which caps range
Pick it when: indoor precision, mapping, profiling, volume measurement — anything where you need a point cloud rather than a presence flag.
How to read your own site
Five questions about the installation, not the sensor:
- Will the target ever be still? Yes → PIR is out immediately
- Is it outdoors, and does it rain? Yes → LiDAR is out
- Is there wind? Yes → ultrasonic is out
- Is there metal structure nearby? Yes → budget for real clutter rejection, not the cheapest module
- What is the shortest distance you must detect? Under 30 cm → mmWave radar struggles, everything else is fine
Three misjudgements I keep seeing
"PIR is cheaper, start there"
Cheaper per unit, far more expensive per outcome when people sit still. An occupancy system that reports an empty room is not a bargain — it gets switched off within a month.
"Radar is radar"
A Doppler module and an FMCW module are both sold as radar and behave completely differently. Doppler measures velocity and cannot see a stationary target. FMCW measures range and can. Ask which architecture before discussing price.
"Higher frequency is better"
77 GHz resolves angle better than 24 GHz at the same aperture, but has higher path loss, tighter tolerances, and may not be legal in your market. Frequency is a trade, not an upgrade. Check the band compliance lookup first.
Next
Still choosing? Run the technology selector — four
questions, one answer.
Already know the technology and just need to size it?
Run the range calculator.
Not sure either way? Describe the site and I will tell you.
Tell me about the site, not the sensor
What you detect, how far, indoors or outdoors, and what the environment does to it — rain, wind, heat, metal, dirt. That is enough for a straight answer.