Prototyping platform Nordic Thingy:53 at Rutronik
Based on Nordic's nRF5340 Dual-Core Arm Cortex M-33 multiprotocol system-on-chip (SoC), Thingy:53 features the nPM1100 power management IC (PMIC), the nRF21540 front end module (FEM), and a power amplifier/low noise amplifier (PA/LNA) that allows embedded machine learning (ML) models to be run directly on the device.
The prototyping platform is equipped with multiple motion and environmental sensors and a 1350 mAh Li-Poly battery. It supports Bluetooth LE, Thread, Matter, Zigbee, IEEE 802.15.4, NFC, and Bluetooth Mesh RF protocols.
Rutronik supported Nordic in the realisation of the product with specialised know-how in the fields of sensor technology, PoCs, and ML as well as interdisciplinary system competence. The result is an usable product for building advanced wireless proofs-of-concept and prototypes with ML capabilities.
Thingy:53 is a multi-sensor prototyping platform with multi-protocol short-range wireless connectivity that provides reduced time-to-market for embedded applications with machine learning.
The integrated ML functionality enables developers to use the Thingy:53's sensors in applications such as speech recognition or motion pattern identification. For example, the accelerometer and a microphone with Pulse Density Modulation wake the nRF5340 SoC from standby mode when triggered by motion or sound. By keeping the platform in sleep mode for as long as possible, power-saving ML applications can be realised. In addition, this efficient use extends battery life.
Thingy:53 includes multiple sensors that measure temperature, humidity, air quality and pressure, as well as ambient light and colour. The platform also features a low-power accelerometer and a six-axis inertial measurement unit. A MEMS microphone and buzzer, as well as individually programmable buttons and RGB LEDs, complete the product.
Components from Bosch SE, Rohm Semiconductor, Toshiba, and Murata, among others, are used.
Stable wireless connection the nRF5340 SoC has both a network processor and a dedicated application processor. By using dedicated compute resources, the processor enables a trouble-free, stable wireless connection. The 128MHz Arm Cortex-M33 application processor ensures that the prototyping platform can process complex machine learning computing tasks and complex algorithms. Sufficient memory provides a 1MB flash and 512kB RAM.
ML training and sensor data analysis in the cloud Nordic's nRF Edge Impulse app allows sensor data to be transferred via Bluetooth LE to the linked mobile device, uploaded to the cloud (Edge Impulse Studio account) for training, and trained ML models to be downloaded to the Thingy:53.