The Aspinity
AML100
Near-Zero Power AnalogML™ Processor for Always-on Detection, Classification, Notification, and Prevention
The AML100 accurately detects and classifies sensor-driven events from raw, analog sensor data, allowing developers to reduce always-on system power by 95% or more.
AML100 Features
- World’s first pure analog AI processor
- Consumes <20µA when always-sensing
- Enables always-on system power <50uA for most applications >20 TOPS/W equivalent compute efficiency
- Easy to use programming kit compatible with standard ML tools
- Supports up to 4 analog sensors
- Easy to integrate into current digital architectures
- 7mm x 7mm 48-pin QFN package
AML100 Functional Block Diagram
Always-on AI System Implementation
Aspinity’s AML100 chip controls the entire system, handles the majority of the ML workload
- Digitizing data before determining relevance drains system power
- Use analog AI inference to identify relevant data because ML workloads compute more efficiently in the analog domain
- Keep most system components dormant until relevant data is curated
Supported Sensors
MEMS Microphone
ECM
Hydrophone
Image Sensor
PPG
PIR
Piezo
Accelerometer
Gyroscope
Electric/Magnetic
Capacitance
ECG
Radar
Applications
Aspinity AML100 integrates with any embedded system to achieve an order-of-magnitude reduction in system power consumption
