A battery-free RAIN RFID temperature sensor has been used to measure bearing temperature inside a sealed, rotating Formula 1 electric motor, a measurement point that conventional wired and optical instrumentation cannot reach. The work brought together Ferrari Sports Management as the end user, Radio6ense as the wireless sensing specialist, Axzon as the supplier of the Magnus S3 RAIN IC, and Dewesoft Italy for data acquisition integration. The result was the early detection of a defective bearing whose thermal signature was completely invisible to the wired PT1000 probe sitting nearby.
The measurement problem inside an MGU-K
The target was the bearing of the MGU-K, the Motor Generator Unit Kinetic that recovers energy under braking in an F1 hybrid power unit. Bearing temperature is, in the words of the case study, a critical predictive indicator of frictional increase and incipient degradation. It is also one of the hardest things in the powertrain to actually instrument.
The measurement point is on a rotating component inside a sealed metallic housing. Running wires to it means slip rings or a rotating connector, both of which add failure modes and packaging problems in an assembly where there is no space to give. Infrared pyrometry needs line of sight, which a closed motor case does not offer. A wired probe mounted on the stationary structure nearby can be fitted easily enough, but it only sees heat that has already conducted out through the surrounding metal, by which time the signal has been smoothed and delayed.
That leaves a passive wireless sensor as the only realistic route, and it has to survive an environment that is deliberately hostile to radio: a metal enclosure, high rotational speed, and an inverter throwing electromagnetic noise around at close range.
Passive RAIN RFID as the sensing layer
The sensor is built around Axzon’s Magnus S3 RAIN IC, which carries an on-chip temperature sensor and operates in the UHF band at 860 to 960 MHz under EPC Gen2 and ISO 18000-63. Quoted accuracy is plus or minus 0.5 degrees C typical, over a range of minus 30 to plus 120 degrees C.
Crucially, there is no battery and no local power source of any kind. The tag harvests a few microwatts of RF energy from the interrogator field, takes the reading, and returns it by backscatter modulation. Nothing on the rotating assembly needs servicing, recharging or wiring, which is what makes the whole approach viable on a component that spins inside a sealed case.
Radio6ense designed the RF front end around that constraint. The tag uses two capacitively coupled C-shaped dipole antennas, co-designed with the chip using full-wave electromagnetic simulation in CST Solver. The coupling to the reader antenna is near-field and E-field dominant, and that choice is doing real work: an E-field dominant link is far less susceptible to the magnetic noise the inverter generates, which is the interference that would otherwise swamp a conventional magnetically coupled arrangement at these power levels.
On the interrogation side, a Kathrein RRU1440 reader sits outside the motor housing, with its antenna embedded inside the MGU-K and fed through a coaxial feedthrough in the case wall. That keeps the reader electronics out of the thermal and vibration environment while placing the RF aperture where it needs to be, inside the enclosure with the tag.
Turning tag reads into a measurement channel
Getting a temperature out of a RAIN tag is one thing. Getting it into a race engineer’s data set, time-aligned with torque and current, is another, and this is where the architecture is interesting. The system is built in three layers.
At the physical layer the passive RAIN sensor takes the reading and the UHF reader retrieves it via backscatter, with embedded reader firmware known as HEXA managing the Gen2 air interface transactions. Above that sits an openDAQ bridge, which exposes the entire RFID subsystem as a standard openDAQ device so that, as the case study puts it, each passive sensor becomes a well-defined measurement channel. At the top, DewesoftX imports those temperature channels as fully native signals, synchronised alongside motor variables such as torque, rotational speed and inverter currents.
The practical consequence is that the RFID sensor stops being a separate telemetry curiosity with its own log file and becomes just another channel in the DAQ system, on the same time base as everything else. For a diagnostic application that depends on correlating a thermal excursion with a specific load event, that synchronisation is not a convenience, it is the point.
Benchmarked against wired and infrared references
The system was tested on a bench running on-track-equivalent operating profiles, including rapid load transitions, high-torque regimes and fast thermal excursions. Two reference sensors were run alongside the RAIN tag: a PT1000 wired probe on a stationary structural mount, and a non-contact infrared sensor requiring line of sight.
With a healthy bearing, the RFID measurement tracked the infrared reference closely, capturing both absolute temperature and rapid transients. The PT1000 trace, by contrast, stayed smooth and heavily filtered by heat conduction through the structure it was mounted to, exactly the behaviour you would expect from a sensor that is not on the part it is trying to measure.
The defective bearing is where the difference mattered. The RFID channel revealed localised and repeatable thermal deviations under the same operating conditions that produced nothing unusual on the healthy unit. The stationary PT1000 probe remained largely unchanged throughout. Only the RAIN sensor had both the spatial selectivity, being physically on the bearing, and the temporal responsiveness to pick up the early thermal signature of the fault.
Why this matters beyond motorsport
The headline claim from Axzon is that this is not a lab demonstration but predictive diagnostics operating at the limit of motorsport engineering, and the case study supports the broader reading. What has been demonstrated is a passive in-motor telemetry layer that gives direct, synchronised access to bearing-level temperature inside a sealed electric machine, and that detects anomalies conventional sensors miss entirely.
The source describes the approach as extensible toward higher-temperature applications and on-vehicle deployment. That points fairly obviously at electric powertrains more generally, where the constraint is rarely the sensing element itself and almost always physical access to the measurement point. Any rotating, sealed or otherwise unreachable component in an EV drive unit presents the same problem the MGU-K bearing did, and the same answer applies: if you cannot run a wire to it and cannot see it, a passive UHF tag harvesting RF energy through the housing is one of the very few options left.
More broadly, this is a good illustration of where RAIN RFID sensing has got to. Battery-free sensor ICs with on-chip temperature measurement have been available for some years, but the engineering that turns them into a trustworthy measurement, the co-designed tag antenna, the coupling mode chosen to reject a specific noise source, the reader aperture placement, and the integration into a synchronised DAQ environment, is what separates a tag that returns a number from an instrument an engineer will act on.
