Interest in artificial intelligence (AI) technology seems to be increasing by the day since the unveiling of OpenAI’s ChatGPT in November of last year. Under the umbrella of AI technology is machine learning (ML) which is a type of AI. ML uses large quantities of data collected from machines and sensors in conjunction with algorithms to analyze vast amounts of data. In turn, the data is used to make improvements in the algorithms as well as execute it’s the intended function of the algorithm. One example is ML implemented to spot patterns or anomalies in a manufacturing line. Edge ML is a technique in which the data collection, ML algorithm, and execution all occur in the end device or smart device. This offers the added advantage of improved security, latency, and efficiency.
Recently, a company from India focusing on IoT devices, Eoxys, is releasing a new module designed for edge ML applications known as the Xeno+ Nano ML Module. Two key components make the module suitable for edge ML applications. First, the module features Syntiant’s NDP120 Neural Decision Processor. The processor has been built specifically with battery powered deep learning applications in mind. As a result, it can run multiple applications concurrently with reduced power consumption. Secondly, the module also includes InnoPhase’s INP1014 Talaria TWO certified Wi-Fi and BLE module. This is also a low-power module that supports 2.4GHz 802.11 b/g/n Wi-Fi protocols and 2Mbps PHY BLE 5.0.
The module is available in two different versions, differing mainly on the processor used. One version uses a STM32L4 series Arm Cortex-M4 while the other utilizes a Nuvoton Cortex-M32 Trust Zone series MCU. Both come on a PCB sized at 60 x 38mm with 20 castellated pins on either side of the board to support serial interfaces and GPIOs. The NDP120 supports I2S, PDM, TDM, and I2C. On the other hand, the MCU supports UART, SPI, I2C, PWM, and GPIOs. A USB-C connection is available on the boards with a serial debug port as well 3 pins supporting Arm's Serial Wire Debug (SWD) interface. For power, the module supports 2.5V to 5.0V battery power from rechargeable or non-rechargeable batteries. When an active USB connection is provided an onboard MOSFET switch will automatically cut power provided from the battery and utilize the 5.0V USB provided power.
Also included in the module are key security features. These include tamper-resistant key storage in memory (Flash and SRAM), Arm Cortex TrustZone support, hardware crypto accelerators, CRC calculation unit, six tamper detection pins, and Arm Platform Security Architecture accommodation. The module is available for pre-order with a price expected to be around $70. Overall, the module can provide a good platform for users looking to experiment with Syntiant’s neural processing chip or users looking to implement edge ML applications utilizing microphones or sensors.