Nordic Thingy:53 Provides Rapid Prototyping for Embedded Machine Learning and IoT

Nordic Semiconductor has a family of rapid prototyping platforms called Thingy. These devices go beyond traditional development boards by adding sensors and interfaces to their popular system-on-chip (SoC) products. Today, they have announced a new member of that family, the Nordic Thingy:53.

As the name implies, Nordic built this new Thingy around their flagship nRF5340 SoC. It also includes an excellent assortment of sensors, over-the-air update capabilities, and out-of-the-box integration to the embedded machine learning platform Edge Impulse.

The nRF5340 is complimented by an onboard nPM1100 power management IC (PMIC) and an nRF21540 front-end RF module (FEM.) Also on the board are a multitude of sensors. And while wireless is a significant part of Thingy:53, there are physical connections like USB-C, a Qwiic/Stemma/Grove style connector, a programming header, a power switch, and a port for a debug board. That board contains additional connections for measuring current consumption with tools such as the Power Profiler Kit.

Let's start by looking at the SoCs.

SoCs

At its introduction, the nRF5340 was the first SoC with two separate Arm processors and wireless connectivity. Similar to dual "core," this concept is two slightly different processors that split edge processing tasks. One Arm Cortex-M33 runs at 128 MHz with 1 megabyte of Flash and 512 kilobytes of RAM for high performance. In addition, it has a floating-point unit (FPU) and supports digital signal processing (DSP) instructions. Another Arm Cortex-M33 targets network processing. It runs up to 64 MHz with 256 kilobytes of Flash and 64 kilobytes of RAM.

The nRF5340 has a 2.4 GHz radio that supports Bluetooth LE, Bluetooth mesh, and proprietary protocols. It also supports standards like Zigbee, Thread, Matter, and even NFC. In addition, for increased range and robustness with wireless communication, Thingy:53 includes an nRF21540 FEM. This RF front-end IC contains a power amplifier (PA) and low-noise amplifier (LNA).

If you have not heard of Matter, you might recognize its previous name: Project Connected Home over IP. This working group is part of the Connectivity Standards Alliance (CSA), formerly known as the Zigbee Alliance. Any device that supports Matter can communicate with smart home devices like Apple HomeKit, Amazon Alexa, Google Assistant, Samsung's SmartThings, and many others. While not a wireless protocol, Matter operates as an application layer on top of a physical layer like Bluetooth.

If you have not heard of Matter, you might recognize its previous name: Project Connected Home over IP. This working group is part of the Connectivity Standards Alliance (CSA), formerly known as the Zigbee Alliance. Any device that supports Matter can communicate with smart home devices like Apple HomeKit, Amazon Alexa, Google Assistant, Samsung's SmartThings, and many others. While not a wireless protocol, Matter operates as an application layer on top of a physical layer like Bluetooth.

An nPM1100 PMIC handles charging the internal 1350 mAh LiPo battery and the buck converter when powered from the USB-C connection.

Sensors

Thingy:53 has an extensive array of sensors for embedded machine learning applications. These sensors include:

• Temperature
• Humidity
• Air quality
• Air pressure
• Color/light sensor
• 6-axis IMU (accelerometer + gyroscope)
• Magnetometer
• PDM (pulse density modulation) microphone

Two sensors, the microphone and accelerometer, can wake the SoC when actions occur. For example, the microphone might detect a sound, or the accelerometer could activate when it detects motion. These capabilities aid in low-power operations.

There are also user-definable push buttons, an RGB LED, and an SWF-style port for RF measurements with a spectrum analyzer.

In the unlikely event that you need a sensor the Thingy:53 does not already have, you can extend its capabilities using the Grove, Stemma, and Qwiic connector. This connector taps into several ecosystems of easy-to-connect sensors, actuators, and displays.

Collecting data from all of these sensors is fantastic, but that data needs to go somewhere to be practical.

Machine learning and firmware development

You can tell that Thingy:53 targets embedded machine learning (ML) applications. First, Nordic collaborated with Edge Impulse on the default firmware. Second, that firmware is ready to collect sensor data and transmit it via Bluetooth LE to a mobile device running the Edge Impulse Mobile app.

This combination shortens the time to get data into Edge Impulse Studio, where you can develop (and test) an embedded ML model. Then, when ready, you can deploy that model back to the Thingy:53 to run on the nRF5340's high-performance application processor.

Edge Impulse is a development platform for embedded (or edge) machine learning. It accelerates the implementation of a ML solution by accepting data from nearly any sensor type, developing a model, testing it (simulation or live), and deploying it to edge targets, like the Thingy:53.

For more involved firmware development, Nordic's nRF Connect SDK supports the nRF53 and Thingy:53. This Zephyr RTOS-based SDK is the same one used for other Nordic families like the nRF52 and nRF91. In addition, the nRF Programming mobile app enables you to load new firmware over the air.

For more information

Nordic says distributors, including Farnell, will have stock as of today. To learn more about this new IoT prototyping device, visit the Thingy:53 product page.