On the non-STDP behavior and its remedy in a floating-gate synapse

Area of Science:

• Neuromorphic Engineering
• Solid-State Circuits
• Artificial Intelligence Hardware

Background:

• Floating-gate (FG) transistors offer nonvolatile weight storage for neuromorphic systems.
• Traditional FG synapses exhibit anomalous non-STDP behavior, deviating from biological learning rules.

Purpose of the Study:

• To theoretically analyze the cause of non-STDP behavior in FG synapses.
• To propose and experimentally validate a solution for achieving biologically accurate spike-timing-dependent plasticity (STDP).

Main Methods:

• Theoretical analysis of FG transistor operation under varying gate waveforms.
• Fabrication and experimental testing of a modified FG synapse in a 0.35-μm CMOS process.
• Circuit simulation for generating optimized control gate waveforms.

Main Results:

• Identified the theoretical cause of non-STDP behavior in traditional FG synapses.
• Demonstrated that modified control gate waveforms achieve biologically consistent STDP over a wide parameter range.
• Experimental validation of the proposed solution using a fabricated FG synapse.

Conclusions:

• The proposed method effectively corrects non-STDP behavior in FG synapses.
• This advancement enables more biologically plausible learning in neuromorphic hardware.
• Optimized gate waveform generation circuits are feasible for practical implementation.