This article demonstrates the first functional neuromorphic spiking neural network (SNN) that processes the time-to-first-spike (TTFS) encoded analog spiking signals with the second-order leaky integrate-and-fire (SOLIF) neuron model to achieve superior biological plausibility. An 8-kb SRAM macro is used to implement the synapses of the neurons to enable analog computing in memory (ACIM) operation and produce current-type dendrite signals of the neurons. A novel low-leakage 8T (LL8T) SRAM cell is proposed for implementing the SRAM macro to reduce the read leakage currents on the read bitlines (RBLs) when performing ACIM. Each neuron’s soma is implemented with low-power analog circuits to realize the SOLIF model for processing the dendrite signals and generating the final analog output spikes. No data converters are required in our design by virtue of analog computing’s nature. A test chip implementing the complete output layer of the proposed SNN was fabricated in 90-nm CMOS. The active area is $553.4\,\, \times 118.6 \mu \text{m}^{2}$ . The measurement results show that our SNN implementation achieves an average inference latency of 196 ns and an inference accuracy of 81.4%. It consumes $242 ~\mu \text{W}$ with an energy efficiency of 4.74 pJ/inference/neuron.

中文翻译：

---

使用首次尖峰时间编码方案和低泄漏 8T SRAM 中的模拟计算的神经形态尖峰神经网络

本文演示了第一个功能性神经形态尖峰神经网络 (SNN)，它使用二阶泄漏积分激发 (SOLIF) 神经元模型处理首次尖峰时间 (TTFS) 编码的模拟尖峰信号，以实现卓越的生物学特性合理性。 8 kb SRAM 宏用于实现神经元突触，以实现内存模拟计算 (ACIM) 操作并产生神经元的电流型树突信号。提出了一种新颖的低泄漏 8T (LL8T) SRAM 单元，用于实现 SRAM 宏，以减少执行 ACIM 时读取位线 (RBL) 上的读取泄漏电流。每个神经元的体细胞都采用低功耗模拟电路来实现 SOLIF 模型，用于处理树突信号并生成最终的模拟输出尖峰。由于模拟计算的性质，我们的设计中不需要数据转换器。实现所提出的 SNN 完整输出层的测试芯片是用 90 nm CMOS 制造的。活动区域为 $553.4\,\, \乘以118.6 \mu \text{m}^{2}$ 。测量结果表明，我们的 SNN 实现实现了 196 ns 的平均推理延迟和 81.4% 的推理精度。它消耗 $242 ~\mu \text{W}$能量效率为 4.74 pJ/推理/神经元。