mmWave Occupancy Sensors: Engineering for Smart Buildings

Why PIR Is No Longer Enough

The Passive Infrared sensor is the workhorse of the building automation industry. Hundreds of millions are installed. They detect movement of a warm body across the lens, which makes them excellent for hallway lights and corridor security. They are also fundamentally limited in ways that matter for modern smart-building products.

PIR cannot detect a stationary person. A meeting room with three people sitting at a table reads as empty to a PIR. Lights time out, HVAC ventilation drops, and the meeting becomes uncomfortable. To compensate, PIR-based products typically extend their hold time to 15 or 30 minutes after the last motion. The result is energy wasted on empty rooms after people leave, plus discomfort while they sit still inside.

The second failure mode is counting. PIR cannot count. A room with one person and a room with ten people look identical. Modern building automation use cases (demand-controlled ventilation, dynamic space planning, real-time analytics) need accurate counts, which PIR cannot provide.

60 GHz mmWave radar resolves both failures.

What mmWave Radar Adds

Three capabilities together close the PIR gap for modern smart-building products.

• Stationary occupant detection. The breathing of a seated person produces a periodic micro-motion at 0.1 to 0.5 Hz, easily resolved by the radar. The room is correctly reported as occupied.
• Counts up to ten or twenty people. Each person is tracked as an independent target, so the count drives demand-controlled ventilation, lighting groups, and meeting-room analytics.
• Zone awareness. A single radar above the doorway distinguishes between occupied and unoccupied halves of a long room, which a PIR cannot do without multiple devices.

The technology is mature. The TI IWR6843ODSEVM is purpose-tuned for occupancy detection with elevation, and the wider IWR6843 family covers most other building geometries. See our EVM comparison for the platform decision in detail.

The Architectural Choice: Sensor-Per-Room or Centralised

Two architectural patterns dominate radar occupancy products.

Sensor-per-room (most common). One ceiling-mounted sensor per enclosed space, between 9 and 25 square metres. The sensor talks directly to the building management system via Matter, BACnet, KNX, or REST. Mechanical envelope is small, retrofit is easy, the failure of one sensor affects one room. This is the right pattern for office, hospitality, and education.

Centralised multi-sensor head. Two or three radar sensors connected to a single in-ceiling processing unit, which serves a larger open space such as a co-working floor or an exhibition hall. The processing unit handles cross-sensor fusion, which gives better counting accuracy in spaces above 50 square metres. Used in higher-end installations.

The decision usually follows the space density and the existing wiring topology. A retrofit into a 1970s office building benefits from per-room sensors using wireless backhaul. A new-build with a structured cabling plan can afford centralised heads.

Building Integration: Matter, BACnet, KNX, REST, MQTT

A radar occupancy sensor is useless on its own. The value is created when the building automation system reads the sensor's data and acts on it. Five integration patterns cover almost every installation:

• Matter is the modern open standard pushed by Apple, Google, Amazon, and the broader Connectivity Standards Alliance. New consumer-tier and commercial buildings should default to Matter for new installations.
• BACnet is the dominant protocol in commercial HVAC. Native BACnet/IP support unlocks direct integration with variable-air-volume controllers and chillers, without a gateway.
• KNX dominates European premium residential and commercial buildings. A KNX-native radar occupancy sensor wins specification decisions in Germany, Austria, Switzerland, and the Benelux.
• OPC-UA for industrial buildings and integration with manufacturing systems.
• REST and MQTT for cloud-connected products with their own analytics platform. The radar sends JSON occupancy events to a cloud broker, which feeds dashboards and AI optimisation.

Most enterprise radar products implement two or three of these in parallel. The choice for any given installation is made at commissioning time. This means the firmware architecture must keep the protocols modular and the certification scope manageable.

The Privacy Argument: Why Radar Wins Over Cameras for HVAC

Camera-based occupancy works technically. It also fails commercially in many markets, because the data-protection conversation derails the project. Hospital workplaces, school cafeterias, restrooms, and bedroom monitoring in care homes are all settings where camera surveillance is either legally restricted or commercially untenable.

Radar at 60 GHz produces a range-Doppler map. There is no image. There is no identifiable feature of any individual. Under Article 4 of the GDPR, this is not personal data. Under HIPAA in the United States, it is not Protected Health Information. The conversation with the data-protection officer is short.

In practice, a complete privacy story for an enterprise installation includes:

• A Data Protection Impact Assessment specific to the installation, even though the sensor itself does not capture personal data.
• A clear statement of what data the sensor outputs (count, occupancy state, motion magnitude) and where it is sent.
• Configurable retention for any aggregate data stored downstream, with a default of zero personal-identifying inference.
• A documented sensor decommissioning policy for end-of-life or change of building tenant.

For installations covering corridors, common areas, or single-occupant offices, radar is one of the few technologies that can meet 2026 European privacy expectations and still provide useful occupancy data.

Performance and Sensor Placement Guide

Placement is half the engineering. A perfect radar mounted in the wrong position underperforms a mediocre PIR mounted correctly.

For ceiling-mounted occupancy sensors using the IWR6843ODSEVM-class platform:

• Ceiling height 2.4 to 3.5 metres is the sweet spot. Below 2.4 m and the field of view becomes too small to cover a useful area; above 3.5 m and the signal-to-noise on small motion degrades.
• Mounting position central in the room, not in a corner. Off-centre mounting works but requires per-installation calibration.
• Distance to large reflective surfaces (whiteboards, windows, metal cabinets) above 1.5 m to limit multipath ghost targets.
• Distance to HVAC vents above 1 m. Air movement near the sensor creates spurious low-amplitude returns that the algorithm has to learn to discount.

Performance under correct placement:

• Time-to-detect a person walking in: under 500 ms.
• Time-to-confirm a stationary occupant: 2 to 5 seconds (one breathing cycle minimum).
• Time-to-confirm room vacated: 20 to 60 seconds, configurable.
• False-occupied rate: below 0.05 events per hour after tuning.
• False-unoccupied rate: below 0.02 events per hour after tuning.

Energy ROI: Quantified HVAC and Lighting Savings

The business case for premium radar occupancy sensors rests on energy. A typical office building spends 35 to 50 percent of its energy on HVAC and lighting. Accurate occupancy data lets the BMS turn things off when they are not needed, which is impossible with PIR's stationary blind spot.

Independent measurements across pilot installations we have engineered:

• Lighting energy reduction versus time-clocked or PIR-controlled lighting: 15 to 25 percent.
• HVAC energy reduction versus PIR-driven demand-controlled ventilation: 12 to 20 percent.
• Combined building energy reduction versus a 2010-vintage automation baseline: 18 to 28 percent.

At European 2026 electricity prices, the energy saving on a single mid-size office building (3000 square metres, 200 occupants) reaches 35,000 to 65,000 EUR per year. The radar sensor BOM premium over PIR amortises in 18 to 24 months. After that the savings flow to the operator.

Productisation: From Reference to Sellable Sensor

Reference designs are not products. A sellable enterprise occupancy sensor needs nine engineering investments beyond the reference firmware:

1. Custom PCB with the IWR6843AOP package, mechanical envelope sized for ceiling tile mounting (typically 65 by 65 mm).
2. RF radome design that does not perturb the antenna pattern more than 2 dB at 60 GHz.
3. Tamper detection and provisioning UX for the building installer.
4. Wireless backhaul (Matter over Thread, BLE, Wi-SUN, LoRaWAN), depending on building type.
5. Over-the-air firmware update with rollback, secure boot, signed images.
6. BMS protocol stack (BACnet, KNX, or OPC-UA) certified by the relevant body.
7. Localisation for installer apps in the markets you sell into.
8. Long-term support commitment, typically ten years for commercial real estate.
9. End-of-line calibration that takes under 60 seconds per unit at the contract manufacturer.

Engineering for High-Volume Manufacturing

Occupancy sensors are sold in thousands of units per building, tens of thousands across a portfolio operator. The unit-economics differ substantially from low-volume industrial radar products. Three engineering priorities specific to this segment:

• BOM cost down to single-digit EUR for non-RF components. The radar SoC dominates the BOM, so squeezing the supporting circuitry is where cost savings happen.
• Test fixtures at the CM that handle hundreds of units per hour. A 60-second calibration step that takes 90 seconds with bench equipment becomes a bottleneck at volume.
• Reliability budget targeting field MTBF above 15 years. The sensor lives behind a ceiling tile in an office for 15+ years. Electrolytic capacitors and similar wear-out components do not belong in this BOM.

Frequently Asked Questions

Lukas Hofstätter

Founder and Managing Director, HALready GmbH

Lukas leads embedded engineering at HALready, with focused experience in 60 GHz mmWave radar applications across smart building, healthcare, and industrial markets. His work includes BMS protocol integration, GDPR-aware sensor architecture, and the productisation of high-volume building automation hardware.

LinkedIn profile