RESUMO
Stochastic computing (SC) is widely known for its high error tolerance and efficient computing ability of complex functions with remarkably simple logic gates. The noise of electronic devices is widely used to be the entropy source due to its randomness. Compared with thermal noise and random telegraph noise (RTN), flicker noise is favored by researchers because of its high noise density. Meanwhile, unlike using RRAM, PCRAM and other emerging memory devices as the entropy source, using logic devices does not require any additional process steps, which is significant for industrialization. In this work, we systematically and statistically studied the 1/f noise characteristics of 14 nm FinFET, and found that miniaturizing the channel area of the device or lowering the ambient temperature can effectively increase the 1/f noise density of the device. This is of great importance to improve the accuracy of the SC system and simplify the complexity of the stochastic number generator (SNG) circuit. At the same time, these rules of 1/f noise characteristics in FinFET devices can provide good guidance for our device selection in circuit design.
RESUMO
In embedded neuromorphic Internet of Things (IoT) systems, it is critical to improve the efficiency of neural network (NN) edge devices in inferring a pretrained NN. Meanwhile, in the paradigm of edge computing, device integration, data retention characteristics and power consumption are particularly important. In this paper, the self-selected device (SSD), which is the base cell for building the densest three-dimensional (3D) architecture, is used to store non-volatile weights in binary neural networks (BNN) for embedded NN applications. Considering that the prevailing issues in written data retention on the device can affect the energy efficiency of the system's operation, the data loss mechanism of the self-selected cell is elucidated. On this basis, we introduce an optimized method to retain oxygen ions and prevent their diffusion toward the switching layer by introducing a titanium interfacial layer. By using this optimization, the recombination probability of Vo and oxygen ions is reduced, effectively improving the retention characteristics of the device. The optimization effect is verified using a simulation after mapping the BNN weights to the 3D VRRAM array constructed by the SSD before and after optimization. The simulation results showed that the long-term recognition accuracy (greater than 105 s) of the pre-trained BNN was improved by 24% and that the energy consumption of the system during training can be reduced 25,000-fold while ensuring the same accuracy. This work provides high storage density and a non-volatile solution to meet the low power consumption and miniaturization requirements of embedded neuromorphic applications.