| name | a-32ch-event-based-bio-signal-frontend-neuromorphic |
| description | Skill for understanding and applying the 32-channel event-based bio-signal acquisition front-end for adaptive neuromorphic processing (arXiv:2607.12901v1). This skill outlines the dual-mode encoding (PFM and aADM) approach for low-power neural signal acquisition and its compatibility with spiking neural network processors. |
| tags | ["neuroscience","brain-network","neural-dynamics","spiking-neural-network","computational-neuroscience","hardware","bio-signal-acquisition"] |
| related_skills | [] |
32-Channel Event-Based Bio-Signal Acquisition Front-End for Adaptive Neuromorphic Processing
Overview
This skill is based on the arXiv paper: "32-channel event-based bio-signal acquisition front-end for adaptive neuromorphic processing" (arXiv:2607.12901v1) by Narayanan Shyam, Saptarshi Ghosh, and Giacomo Indiveri.
The paper presents a 32-channel application-specific integrated circuit (ASIC) designed for acquiring bio-signals (e.g., neural signals) and encoding them in an event-based format suitable for neuromorphic processing. The key innovation is the dual-mode encoding per channel:
- Pulse Frequency Modulation (PFM): Converts signal amplitude to pulse frequency.
- Adaptive Asynchronous Delta Modulator (aADM): An auto-scaling delta modulator that adjusts the encoding data-rate based on the input signal envelope, enabling high data compression for low-power transmission.
The system is fabricated in a 180 nm CMOS process and provides a configurable interface for integration with state-of-the-art Spiking Neural Network (SNN) neuromorphic processors.
Core Concepts
Event-Based Sensing
Event-based sensors only transmit information when a change occurs, reducing data redundancy and power consumption compared to traditional sampled systems.
Dual-Mode Encoding
- PFM: Simple, linear frequency encoding suitable for moderate signal dynamics.
- aADM: Adaptive delta modulation that tracks the signal envelope, providing high compression for slowly varying signals and higher fidelity for rapid changes.
Adaptive Data Rate
The aADM dynamically adjusts its sampling rate based on the signal's amplitude variations, optimizing the trade-off between fidelity and bandwidth.
Neuromorphic Compatibility
The output format (asynchronous spike-like events) is directly compatible with Spiking Neural Network (SNN) processors, enabling end-to-end low-power neural signal processing pipelines.
Application Scenarios
- Brain-Computer Interfaces (BCIs): Wireless acquisition and processing of neural signals for assistive technology.
- Implantable Medical Devices: Long-term monitoring of neural or other bio-signals with minimal power consumption.
- Wireless Sensor Networks: Event-based transmission reduces bandwidth requirements in distributed sensing systems.
Implementation Steps
-
Understand the Dual-Mode Architecture:
- Study the PFM and aADM circuit designs (refer to the paper's schematic diagrams).
- Note the configurability of each channel via bias currents and reference voltages.
-
Simulate the System:
- Use transistor-level simulators (e.g., Spectre, HSPICE) to verify the PFM and aADM responses to various input signals (sine waves, spikes, noise).
- Validate the adaptive behavior of the aADM under varying signal envelopes.
-
Integrate with SNN Processor:
- Map the event output (pulses or spikes) to the input spikes of an SNN.
- Ensure voltage levels and timing compatibility between the ASIC and the neuromorphic chip (e.g., Loihi, SpiNNaker, or custom ASNN).
-
Power Optimization:
- Tune the bias currents to minimize static power while maintaining sufficient signal-to-noise ratio.
- Leverage the event-driven nature to achieve ultra-low average power consumption.
-
System Validation:
- Test with real bio-signals (e.g., EEG, EMG, or neural spikes) to evaluate fidelity and compression ratios.
- Compare performance against traditional ADC-based acquisition systems.
Key Parameters
- Number of Channels: 32
- Encoding Modes: PFM and aADM (selectable per channel)
- Process: 180 nm CMOS
- Interface: Configurable bias currents for power/performance trade-off
- Output: Asynchronous event streams (voltage pulses)
References
- Shyam, N., Ghosh, S., & Indiveri, G. (2026). 32-channel event-based bio-signal acquisition front-end for adaptive neuromorphic processing. arXiv preprint arXiv:2607.12901.
- Indiveri, G., et al. (2011). Neuromorphic silicon neuron circuits. Frontiers in Neuroscience, 5, 73.
- Essaifi, A., et al. (2020). Asynchronous delta modulation for neural signal compression. IEEE Transactions on Biomedical Circuits and Systems.
Activation
Trigger this skill when working on:
- Designing low-power bio-signal acquisition systems
- Developing brain-computer interfaces
- Exploring event-based sensing for neuromorphic computing
- Integrating analog front-ends with spiking neural networks
Notes
- The chip is specifically designed for compatibility with SNN processors, making it a valuable component for neuromorphic engineering pipelines.
- The dual-mode approach allows flexibility: PFM for deterministic latency requirements, aADM for maximum compression.