Alif Semiconductor has launched a Bluetooth Low Energy wireless microcontroller for Matter that features a neural co-processor for AI and machine learning workloads. The company claims the Balletto family is the first Matter wireless microcontroller range to feature hardware optimisation for neural network processing.

Supplied in an ultra-small WLCSP package, the Balletto MCU allows smart home and consumer device manufacturers to implement AI/ML functions such as speech recognition, adaptive noise cancellation, vocal targeting, and beam forming in true wireless stereo (TWS) earbuds, and sensor fusion in lifestyle wristbands and other types of space-constrained devices.

According to Alif, Balletto is optimised for audio and sensor functions due to its DSP and AI/ML capabilities and large on-die memory. The device’s Arm Cortex-M55 core, which achieves an EEMBC CoreMark score of 704 at 160MHz, includes the Arm Helium M-profile vector extension (MVE), providing a 500% improvement in DSP performance.

Alongside the Cortex-M55 CPU, Balletto includes an Arm Ethos-U55 Neural Processing Unit (NPU), which can perform up to 46 GOPS and is backed by a large 2MB tightly coupled memory (TCM), extends the audio capabilities with hardware acceleration of AI/ML models. Neural network processing performance is up to 15x better than a Cortex-M4 processor can achieve. This produces outstanding results in audio encoding/decoding functions, including for the high-bit rate, low-latency LC3 codec which underpins the BLE Auracast broadcast audio technology.

Balletto features Alif Semiconductor’s aiPM technology, which can dynamically power only the logic and associated memory that are in use at any given time, thus achieving very low system power consumption. The aiPM power management unit implements four system level power modes including a Stop mode which draws just 700nA.

With its own separate processor and memory, Balletto’s radio supports concurrent Bluetooth LE 5.3 and Thread operations via a single on-chip antenna, making it suitable for use in home automation networks and systems that support the Matter protocol. Receive sensitivity of -101dBm and algorithms help with interference avoidance from other devices, and coexistence with the other protocols sharing the 2.4GHz frequency band.

Balletto has a dual-PA architecture: a high-power amplifier (HPA) provides a +10dBm output for maximum range and signal strength; via the low-power amplifier (LPA), output is +4dBm for optimised power consumption.

“Device makers and developers require more on-device intelligence if they are to deliver on the promise of edge AI across consumer, enterprise and industrial markets,” said Paul Williamson, Senior Vice President and General Manager, IoT Line of Business at Arm. “Alif’s new range of Arm-powered SoCs provide a significant performance uplift in ML applications enabling smaller, cost-sensitive devices to be smarter and more capable.”

“Alif Semiconductor’s Balletto family is targeting BLE challenges faced by the wireless audio and smart home industry today with a platform adapted to high bit-rate audio and the Matter ecosystem,” said Reza Kazerounian, President and Co-Founder of Alif Semiconductor. “With the launch of multiprotocol support, coupled with our industry leading AI/ML powered MCU, it has never been easier to build battery life-friendly connected intelligent edge devices.”