5 September 2023 in IEEE Journal on Emerging and Selected Topics in Circuits and Systems
The synapse is a key element of neuromorphic computing in terms of efficiency and accuracy. In this paper, an optimized current-controlled memristive synapse circuit is proposed. Our proposed synapse demonstrates reliability in the face of process variation and the inherent stochastic behavior of memristors. Up to an 82% energy optimization can be seen during the SET operation over prior work. In addition, the READ process shows up to 54% energy savings. Our current-controlled approach also provides more reliable programming over traditional programming methods. This design is demonstrated with a 4-bit memory precision configuration. Using a spiking neural network (SNN), a neuromorphic application analysis was performed with this precision configuration. Our optimized design showed up to a 82% improvement in control applications and a 2.7x improvement in classification applications compared with other design cases.
10 August 2023 in IEEE Transactions on Circuits and Systems I: Regular Papers
Memristors provide a tempting solution for weighted synapse connections in neuromorphic computing due to their size and non-volatile nature. However, memristors are unreliable in the commonly used voltage-pulse-based programming approaches and require precisely shaped pulses to avoid programming failure. In this paper, we demonstrate a current-limiting-based solution that provides a more predictable analog memory behavior when reading and writing memristive synapses. With our proposed design READ current can be optimized by $\sim$ 19x compared to the 1T1R design. Moreover, our proposed design saves $\sim$ 9x energy compared to the 1T1R design. Our 3T1R design also shows promising write operation which is less affected by the process variation in MOSFETs and the inherent stochastic behavior of memristors. Memristors used for testing are hafnium oxide based and were fabricated in a 65hybrid CMOS-memristor process. The proposed design also shows linear characteristics between the voltage applied and the resulting resistance for the writing operation. The simulation and measured data show similar patterns with respect to voltage pulse based programming and current compliance based programming. We further observed the impact of this behavior on neuromorphic-specific applications such as a spiking neural network.
17 April 2024 in Scientific Reports
The synapse is a key element circuit in any memristor-based neuromorphic computing system. A memristor is a two-terminal analog memory device. Memristive synapses suffer from various challenges including high voltage,SETorRESETfailure, andREADmargin issues that can degrade the distinguishability of stored weights. EnhancingREADresolution is very important to improving the reliability of memristive synapses. Usually, theREADresolution is very small for a memristive synapse with a 4-bit data precision. This work considers a step-by-step analysis to enhance theREADcurrent resolution or the read current difference between two resistance levels for a current-controlled memristor-based synapse. An empirical model is used to characterize the $${\hbox {HfO}}_{2}$$ based memristive device. $$1\textrm{st}$$ and $$2\textrm{nd}$$ stage device of our proposed synapse design can be scaled to enhance theREADcurrent margin up to $$\sim$$ 4.3 $$\times$$ and $$\sim$$ 21%, respectively. Moreover,READcurrent resolution can be enhanced with run-time adaptation techniques such asREADvoltage scaling and body biasing. TheREADvoltage scaling and body biasing can improve theREADcurrent resolution by about 46% and 15%, respectively. TENNLab’s neuromorphic computing framework is leveraged to evaluate the effect ofREADcurrent resolution on classification, control, and reservoir computing applications. HigherREADcurrent resolution shows better accuracy than lower resolution even when facing different levels of read noise.