By Steve Bush 24th January 2022
Silicon Labs has introduced a wireless microcontroller with a hardware AI accelerator that can work with Matter, Zigbee, OpenThread, Bluetooth Low Energy and Bluetooth mesh. It comes along with a software tool kit for machine learning.
Called EFR32MG24, “this new hardware and software platform will help bring AI and ML applications and wireless to battery-powered edge devices,” according to the company. “The family supports multiple wireless protocols and incorporates PSA Level 3 protection, ideal for smart home, medical and industrial applications.”
The main processor is a 78MHz Arm Cortex-M33 with DSP instruction and floating-point unit, with up to 1.5Mbyte of flash and 256kbyte of ram.
Regarding on-board AI/ML acceleration, the company has so far said little. It told Electronics Weekly: “The AI/ML accelerator is an in-house-designed co-processor focused on the complex math and calculations associated with machine learning workloads. It is separate from the ARM M33 processor, which offloads AI/ML workloads to the accelerator to improve performance and battery life.”
The radio supports 2 (G)FSK (with configurable shaping), OQPSK DSSS and (G)MSK modulation.
-104.5dBm sensitivity is claimed for 250kbit/s O-QPSK DSSS, -104.9dBm for 125kbit/s GFSK, -97.5 Bm for 1Mbit/s GFSK and -94.4 dBm for Mbit/s GFSK. Transmit power is up to 19.5dBm.
Operation is across 1.71 to 3.8 V and -40 to 125°C, and packaging is 5 x 5mm QFN40 or 6 x 6mm QFN48.
Consumption examples include: transmit at 5.1mA for 0dBm output, 20mA for 10dBm and 155mA for 19.5dBm, 4.4mA receive for 1Mbit/s GFSK, and 32.2μA/MHz (at 39MHz) active processor consumption.
Hardware cryptographic acceleration is included for: AES128/192/256, ChaCha20-Poly1305, SHA-1, SHA-2/256/384/512, ECDSA+ECDH (P-192, P-256, P-384, P-521), Ed25519 and Curve25519, J-PAKE, and PBKDF2.
Other security features are Arm’s TrustZone, true random number generation, secure boot, secure debug unlock, physically uncloneable function (PUF for secure key management), anti-tamper and secure attestation.
Peripherals include: a 12bit 1Msample/s or 16bit 77ksample/s ADC, a DAC, analogue comparators, serial interfaces, a keypad scanner and timers including PWM generation.
A related sub-set family, EFR32BG24, looses OQPSK DSSS modulation and supports only Bluetooth Low Energy (5.2) and Bluetooth Mesh. Both the MG24 and BG24 families can also work with proprietary protocols.
Applications are foreseen in gateways, hubs, sensors, switches, door locks, LED bulbs and luminaires, implementing location services, predictive maintenance, glass break detection and wake-word detection.
“For example,” said SiLabs, “in a commercial office building, many lights are controlled by motion detectors that monitor occupancy. When typing at a desk with motion limited to hands and fingers, workers may be left in the dark. By connecting audio sensors with motion detectors through the Matter application layer, the sound of typing can be run through a machine-learning algorithm to allow the lighting system to make a more informed decision”.
Parts will be broadly available in April.
Top: Pro Kit for MG24 and BG24
Lower: Development board with 20bit ADC and sensors
Tagged with: AI Bluetooth IIoT IoT machine learning Silicon Labs wireless
Your email address will not be published. Required fields are marked *
Download the Elektra Awards 2022 media pack and book your sponsorship package to be part of the most prestigious awards event celebrating its 20th anniversary this year!
Get our news, blogs and comments straight to your inbox! Sign up for the Electronics Weekly newsletters: Mannerisms, Gadget Master and the Daily and Weekly roundups.
Read our special supplement celebrating 60 years of Electronics Weekly and looking ahead to the future of the industry.
Read the Electronics Weekly @ 60 supplement »
Read the first ever Electronics Weekly online: 7th September 1960. We've scanned the very first edition so you can enjoy it.
Read the very first edition »
Electronics Weekly teams up with RS Grass Roots to highlight the brightest young electronic engineers in the UK today.
Read our special supplement celebrating 60 years of Electronics Weekly and looking ahead to the future of the industry.
Read the Electronics Weekly @ 60 supplement »
Read the first ever Electronics Weekly online: 7th September 1960. We've scanned the very first edition so you can enjoy it.
Read the very first edition »
Tune into this Xilinx interview: Responding to platform-based embedded design
Tune into this podcast to hear from Chetan Khona (Director Industrial, Vision, Healthcare & Sciences at Xilinx) about how Xilinx and the semiconductor industry is responding to customer demands.
By using this website you are consenting to the use of cookies. Electronics Weekly is owned by Metropolis International Group Limited, a member of the Metropolis Group; you can view our privacy and cookies policy here.