Silicon Labs intros SoCs for edge ML – with Matter, Zigbee, OpenThread, Bluetooth
Chip-maker Silicon Labs has introduced two families of 2.4 GHz wireless system-on-chips (SoCs), featuring integrated AI/ML accelerators and support, variously, for short-range protocols including Matter, Zigbee, OpenThread, Bluetooth Low Energy (BLE), and Bluetooth mesh.
The Austin-based firm said integrated (tinyML) AI/ML hardware acceleration in the new BG24 and MG24 lines provide “up to” four-times faster inferencing and six-times lower power consumption, compared with cloud-based data processing. A new AI/ML software development kit ptimises TensorFlow Lite for all its SoCs, it said.
The BG24 SoC is for Bluetooth operations, including with BLE and Bluetooth mesh, and is targeted for smart home, smart lighting, and medical products. The MG24 line supports other 02.15.4 connectivity technologies, including Matter, Zigbee, OpenThread, and proprietary versions. It is being pitched also for building automation products.
Matter, it should be noted, is the new Zigbee-based standard from the old Zigbee Alliance, itself rebranded last May as the Connectivity Standards Alliance (CSA), for smart home products, developed as part of the key Project Connected Home over IP (CHIP) project, led by Swedish flatpack furniture maker IKEA.
The Matter standard has gathered momentum. CSA claimed Matter-representation at the Consumer Electronics Show (CES) in Last Vegas earlier this month was strong, with a number making announcements about Matter support and deployments, including from Amazon, Belkin, Comcast, GE Lighting, Google, Infineon, LG, Nordic Semiconductor, NXP Semiconductors, Samsung, and Texas Instruments, among others.
CSA claims 200-odd companies are involved with the Matter standard. Infineon and Oppo have just joined the CSA board. Schneider Electric has integrated Matter into its home and buildings portfolio. “Make no mistake, this is an inflection point for technology.
Silicon Labs suggests Matter will find a home in lighting systems in office buildings, and might be used in conjunction with tinyML running on edge-based SoCs such as the MG24 – to leverage a combination of motion and audio sensors (to detect even the sound of typing) to monitor room occupancy and turn the lights on/off.
It stated: “ML computing at the edge enables other intelligent industrial and home applications, including sensor-data processing for anomaly detection, predictive maintenance, audio pattern recognition for improved glass-break detection, simple-command word recognition, and vision use cases like presence detection or people counting with low-resolution cameras.”
The potential to bring greater intelligence to edge applications is only just being realised, said Silicon Labs. “Today, those considering deploying AI or machine learning at the edge are faced with steep penalties in performance and energy use that may outweigh the benefits. The BG24 and MG24 alleviate those penalties as the first ultra-low powered devices with dedicated AI/ML accelerators built-in,” it said in a statement.
The BG24 and MG24 have the largest Flash and RAM capacities in the Silicon Labs portfolio. “This means that the device can evolve for multi-protocol support, Matter, and trained ML algorithms for large datasets,” it said. The new SoCs also provide PSA Level 3-Certified Secure Vault, the highest level of security certification for IoT devices, which is important in devices like door locks, medical equipment, and other sensitive deployments.
Silicon Labs has partnered with tinyML specialists SensiML and Edge Impulse to provide an “end-to-end toolchain that simplifies the development of ML models optimized for embedded deployments of wireless applications,” it said. Around 40 companies are testing its BG24 and MG24 systems in a “closed Alpha program”, it said, spurred in particular by the tinyML functions in both and the Matter support in the MG24.
The single-die BG24 and MG24 SoCs combine a 78 MHz Cortex-M33 processor, 2.4 GHz radio, 20-bit ADC, combined Flash (up to 1536 kB) and RAM (up to 256 kB), and a hardware accelerator for processing ML algorithms while offloading the M33 – “so applications have more cycles to do other work”. The units will be generally available from April.
Matt Johnson, chief executive at Silicon Labs, said: “The BG24 and MG24 wireless SoCs represent an awesome combination of industry capabilities including broad wireless multiprotocol support, battery life, machine learning, and security for IoT edge applications.”