LIF Spiking Neural Network ASIC: A 4-Neuron Neuromorphic Chip Design on TinyTapeout

This project is a 4-neuron Spiking Neural Network (SNN) ASIC chip design based on the Leaky Integrate-and-Fire (LIF) model, specifically built for the TinyTapeout platform and manufactured using GlobalFoundries' 180nm process. The project explores the implementation of neuromorphic computing on edge devices, while leveraging the TinyTapeout platform to lower the barrier to chip design, providing practical references for the open-source chip and edge AI fields.

As the third generation of neural networks, Spiking Neural Networks (SNNs) transmit information via discrete spike signals, which are closer to biological nervous systems and have advantages in energy efficiency and real-time processing. The LIF model is a commonly used neuron model, with core processes including: integration (accumulating membrane potential from input currents), leakage (membrane potential decaying to resting potential over time), and firing (generating a spike and resetting when membrane potential exceeds the threshold). This model is simplified yet efficient, making it suitable for hardware implementation.

TinyTapeout is an open-source chip design platform that enables individuals and small teams to manufacture ASICs at low cost (hundreds of dollars) via Multi-Project Wafer (MPW) services. This project chose the GF180nm process due to its maturity, stability, and low cost, making it suitable for verifying neuromorphic computing concepts and educational purposes.

The 4-neuron design includes three core components: 1. Neuron core (membrane potential integrator, leakage circuit, threshold comparator, spike generator); 2. Synaptic connections (weights determine network behavior, and different topologies enable multiple computing modes); 3. Event-driven architecture (significant energy is consumed only when neurons fire, making it suitable for edge computing).

SNN ASICs have potential in multiple fields: edge AI (natural sparsity is suitable for resource-constrained devices), real-time signal processing (audio/vibration analysis, brain-computer interfaces), low-power sensing (combining with event cameras to achieve microwatt-level intelligent perception), and brain science research (physical simulation platform for verifying neuroscience theories).

The project shares technical achievements through open source and provides learning resources, allowing anyone to research and improve it. The TinyTapeout model proves that chip design is no longer limited to large companies; the popularization of open-source EDA tools and MPW services makes "garage chip designers" a possibility.

Although this project is small in scale, it represents an important step towards the democratization of neuromorphic computing. From algorithm to silicon implementation, from high-cost to low-cost tape-out, the open-source community is redefining the boundaries of chip innovation, providing an entry path and practical reference for edge AI and low-power computing developers.