World Sleep Industry ENCYCLOPEDIA

Millimeter-Wave Monitoring Modules

1.2.1

Name, Definition, and Delivery Boundary

A millimeter-wave monitoring module is a radar-sensing subassembly intended for integration into another product. It may combine some or all of a millimeter-wave transceiver, antennas, reference clock, power regulation, sampling chain, signal-processing hardware, memory, and embedded control on a circuit board, in a package, or in a system-on-module. It can support detection of presence, range, motion, posture, or mechanical motion associated with breathing and cardiac activity, but it normally still depends on a host for power, mechanical placement, configuration, interpretation, and the user interface. A device with its own deployable enclosure, finished software, and defined end use is a complete radar product even if it contains the same sensing components.

Integration depth varies substantially. At one end is a radar IC with antennas in package and only enough support circuitry to expose intermediate-frequency or raw digital samples. At the other end is a board with a processor, memory, regulator, and application firmware that emits a target list, presence state, or vital-rate estimate over a serial link. A development board, evaluation kit, and reference design are engineering tools; they do not automatically have the size, connector durability, environmental rating, production controls, or authorization status of a production module. Selection therefore requires an exact part number, hardware revision, antenna type, processing level, licensed firmware, and controlled document set rather than comparison of labels such as “60 GHz,” “AI,” or “vital signs.”

Module capability is not finished-product capability. Detection range, field of view, range and angle resolution, sensitivity to a nearly stationary person, and coverage of physiological estimates all change with the host cover, mounting angle, power quality, temperature, computing resources, configuration, and algorithm. A supplier demonstration applies to its named board, firmware, target, and scene; it cannot simply become a claim for a bedside unit, smart bed, air conditioner, light fixture, or care terminal. The integrator must maintain a responsibility chain from module inputs, processing, and status codes through the finished product's display, alert logic, and data retention.

Composition, Interfaces, and Data Semantics

A typical module may contain an RF monolithic integrated circuit, one or more transmit and receive channels, antennas on the package or board, a reference oscillator, analog-to-digital conversion, a digital front end, hardware accelerator, control processor, flash memory, and power management. The combination determines how much work remains in the host. An RF front end needs external acquisition and substantial computation; a radar system-on-chip can perform range, velocity, and angle processing; a module with application firmware may export only high-level events. Its bill of materials and block diagram should distinguish fitted functions from options and state whether debug connectors, test points, and devices visible on an evaluation board exist on the production item.

Electrical documentation should state supply range, average and peak current, power sequencing, reset and boot timing, logic levels, pin tolerances, link rate, connector definition, and sleep and wake behavior. A label such as SPI, UART, I²C, USB, CAN, or proprietary high-speed link does not by itself establish interoperability. The protocol also needs framing, byte order, units, scaling, checks, sequence numbers, timestamps, buffer depth, timeout behavior, and rules for dropped packets. The host must not interpret a duplicated frame, stale data, incomplete boot, or broken link as a stable vital sign or as proof that a space is unoccupied.

The output layer must be explicit. Raw ADC or complex samples, a radar cube, range-Doppler map, point cloud, track, bed-occupancy state, respiratory rate, heart-rate estimate, and alert are different objects produced at successive processing stages. They have different bandwidth, latency, auditability, and risk. Each output needs a validity indication, confidence or quality measure, target identity, configuration version, and defined sampling time base. A numerical zero must distinguish a true measured zero from no target, out of range, algorithm not locked, and communication failure. If filtering or classification occurs inside the module, the host also needs enough diagnostic information to audit invalid and degraded states.

Antenna, Mechanical, and RF Integration

An antenna being inside the package does not make mechanical integration arbitrary. Module axes, main beam, polarization, near-field clearance, ground and copper keep-out, mounting holes, connectors, and neighboring components all affect the usable field of view. A version with antennas on its circuit board is additionally sensitive to substrate, stack-up, fabrication tolerance, and distance from the board edge. The radiation pattern of a different evaluation board is not valid evidence for the selected module. Installation drawings should express azimuth, elevation, and obscured regions in host coordinates, and coverage must be checked with the intended bed position, sleeping postures, and room geometry.

The enclosure or decorative panel is part of the RF path. Dielectric properties, thickness, layer count, curvature, antenna-to-cover spacing, coating, adhesive, moisture, and manufacturing variation can cause attenuation, standing waves, pattern distortion, angular bias, or sidelobe changes. Displays, metal brackets, batteries, heat spreaders, wiring, screws, and conductive coatings may block or redirect energy. “Concealed installation” means that selected nonmetallic materials can pass the operating band under specified construction; it does not mean that every wall, textile, wood panel, or plastic is transparent. The final material stack must be measured rather than accepted from appearance alone.

Movement of the host, a resin or coating substitution, dust or condensation in front of the antenna, and changes in a bed frame or furniture can alter clutter and multipath. A product offering multiple mounting orientations or interchangeable covers needs configuration and calibration for each supported combination, not one universal correction. Antenna gain, field of view, and maximum distance must also be tied to the actual chirp, transmit power, and processing settings. If a lens or cover is changed to increase gain, the beam, blind regions, human RF exposure, and applicable regional emission limits require fresh assessment.

Power, Clock, Thermal Behavior, and Electromagnetic Coexistence

Radar transmission, sampling, and processing create dynamic loads. The host supply must stay within voltage and ripple limits when boot, calibration, transmit bursts, peak computation, and wireless backhaul coincide. Harmonics from switching regulators, digital clocks, display drivers, motors, and high-speed buses can enter the intermediate-frequency or reference-clock path, creating fixed clutter, a higher noise floor, or false targets. Design verification should cover supply corners, ground return paths, decoupling, shielding, and cable routing, and should use radar-output performance—not merely successful boot—as the acceptance criterion.

Module self-heating, enclosure heat flow, and ambient temperature can change transmit power, receiver gain, noise, oscillator behavior, and phase stability. Sleep duty cycle affects both thermal equilibrium and the latency and continuity of low-frequency physiological observations. Runtime calibration can correct drift, but it may also introduce a step in gain or phase; an algorithm that relies on inter-frame coherence or a historical static-clutter estimate must rebuild that state after calibration. Specifications should identify where temperature is measured, derating rules, over-temperature behavior, the calibration schedule, and how outputs are marked during recovery.

Radars sharing a band can raise one another's noise floor, create missed detections or blind regions, and sometimes produce ghost targets. Nearby radios and host digital electronics can also cause in-band or out-of-band disturbance. Products with several modules, two-bed rooms, showrooms, and dense housing cannot be qualified only with a single radar in a quiet room. Frequency planning, staggered frame timing, synchronization, randomization, interference detection, or rejection of corrupted chirps may help, but each method has capacity and performance costs. The system should expose degraded or invalid operation during interference instead of hiding it behind a smoothed output.

Firmware, Configuration, and Failure States

Sweep start and stop frequencies, bandwidth, slope, sampling rate, chirps per frame, frame period, active transmit and receive channels, gain, and power mode jointly determine unambiguous range and velocity, resolution, data rate, heat, and energy use. Changing parameters for more range can reduce resolution or frame rate or increase energy. A configuration file is therefore not a generic text template that can safely be copied between boards. The module should reject combinations outside validated limits, record the values actually applied, and allow the host to read complete identifiers for hardware, RF firmware, application firmware, algorithm model, and configuration.

Boot self-test, RF monitoring, watchdog events, temperature status, memory faults, calibration failure, interface overflow, and loss of algorithm lock require distinct codes and host actions. Continued numerical output does not prove that the chain is healthy, and a host must not retain the previous result through restart, upgrade, target reassignment, or configuration writes without a clear stale indication. For breathing- or cardiac-motion outputs, large movement, inadequate signal quality, unresolved multiple people, and departure of the target should produce an unavailable or limited state. A default, interpolated, or last-known value must never masquerade as a current observation.

A firmware update can alter RF scheduling, data formats, classification thresholds, and the validated performance envelope. Updates need compatibility checks, verification of integrity and origin, recovery from power loss, controlled rollback, and safe migration of configuration. Debug ports, factory commands, and unused services should be access-restricted in production. Interface, RF, algorithm, and finished-product tests must be repeated according to the impact of a version change, with traceable release notes. If old host software cannot understand a new field or status code, the update should fail safely rather than continue silently.

Calibration, Production, and Finished-Product Validation

Internal IC self-calibration, board-level range-bias and receiver gain/phase compensation, antenna beam-tilt measurement, and final-product calibration address different errors and do not replace one another. Some coefficients are unique to a board, antenna, and RF path. A change in the cover, bracket, or host board may invalidate coefficients obtained earlier. The manufacturer should define the calibration target, known distance and angle, environment, equipment, acceptance limits, coefficient storage, and response to failure, and it must prevent a serial number from being paired with another unit's calibration record.

Production screening should at least cover boot current, digital interfaces, transmit and receive paths, range and signal quality for a known reflector, required channel consistency, and retrieval of calibration data. A golden unit can reveal drift in the test station, but it does not replace screening of each manufactured product. Test fixtures, shielded or anechoic conditions, target reflectivity, and software versions must be controlled. An IC self-test alone can miss an antenna solder defect, wrong cover assembly, connector problem, supply noise, or blockage introduced by the host.

Finished-product validation should use production enclosure, power supply, host processor, display and radios, final firmware, and mounting accessories. It should span temperature, voltage, orientation, bedding, body size, movement, multiple people, pets, and interference. Raw or intermediate signal performance and the final user-facing result should be evaluated separately, with coverage, latency, packet loss, reacquisition time, false results, missed events, and invalid fraction reported. If the module outputs vital estimates, each output needs an appropriate synchronized reference and error reporting by population and scene. A brief stationary bench test cannot establish all-night validity.

Compliance, Safety, Privacy, and Intended-Use Limits

The module's frequency band, occupied bandwidth, transmit power, duty cycle, and antenna must comply with the spectrum and equipment-authorization rules where the product is sold and used. An existing module grant or supplier report normally carries conditions for specified antennas, integration, labeling, and host instructions. It does not automatically cover a different antenna, altered cover, higher gain, simultaneous transmitters, or another jurisdiction. The finished-product manufacturer remains responsible for applicable radio, electromagnetic-compatibility, electrical-safety, and human RF-exposure assessments and for keeping the shipping configuration aligned with its technical file.

Radar does not produce an ordinary camera image, yet raw echoes, tracks, occupancy, sleep timing, and vital trends can reveal sensitive behavior. The module-to-host link, storage, and cloud path should implement data minimization, access control, authentication, encryption, update controls, logging, and deletion. Diagnostic captures do not justify indefinite retention. The integrator should state which data are visible to the module vendor, product manufacturer, platform operator, and user, and should provide workable behavior for offline use, account removal, resale, and end of support.

A module is not an independent medical conclusion. A supplier reference algorithm, laboratory result, or statement that breathing can be detected does not prove that an integrated product can diagnose sleep apnea, recognize arrhythmia, provide life support, replace a nurse call, prevent falls, or monitor an infant. Intended use, risk controls, alert timing, clinical evidence, and regulatory responsibility attach to the finished product and its claims. Difficulty breathing, chest pain, fainting, repeated gasping awakenings, or other danger signs require professional assessment; module output must not be used to exclude disease or delay help.