The Xylo™-Audio board is based on SynSense‘s Xylo™ family of devices, which brings the flexibility of NN-based audio processing to microwatt energy budgets. It provides the capabilities for real-time audio detection. A high performance MEMS mic, power consumption measuring unit, external audio input port and a USB-C port are also provided by Xylo™–Audio development kit.

Xylo™Audio is not limited to keyword detection, but is capable of detecting almost any audio feature. Application development is made easy with SynSense’s open-source Python library, Rockpool.

Requirements

Key features

• AFE integrated, works directly with kinds of MEMS microphones. Supportsboth single end and differential mode

• Event-based analog feature extraction with neural technology for power saving

• With programmable AFE gain, the amount of information extractedfromtheincoming signal can scale in complexity, which supports large rangeof input sound intensity

• 4-wire SPI slave interface for register and RAM configuration, support bothsingle and burst mode access

• Support up to 1000 reservoir neurons and up to 8 classification

• 1-bit configurable interrupt indicating classification / detection done• 4-bit monitor pins for debug purpose, internal debug registers available• Up to 100 MHz internal operating frequency

• Extremely low memory footprint (~150 KB), memory power control granularity down to 2 KB

• Ultra-low average working power consumption (~550 uW)