Double Bias Memristive Dot Product Engine Signal Degradation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing memristive crossbar arrays suffer from signal degradation due to resistance in wires and electrodes, which affects the accuracy of vector-matrix processing operations.
Innovation Solution
A double bias memristive dot-product engine is implemented, where additional voltage inputs are applied to both ends of each row electrode to minimize sneak path currents and reduce signal degradation, using a crossbar array with memristive elements at each junction to perform weighted sum operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a crossbar array with memristive elements is used to perform vector-matrix processing, then processing speed and efficiency are improved, but signal degradation due to wire and electrode resistance reduces accuracy
Solution Approach 1:
The patent applies preliminary anti-action by introducing compensation voltages at both ends of row electrodes before the actual computation occurs. These pre-applied voltages counteract the voltage drops caused by wire resistance during the computation process, thereby preventing signal degradation before it affects accuracy. The compensation voltages are calculated based on the known resistance values and expected current flows, allowing the system to proactively cancel out harmful effects.
Solution Approach 2:
The patent changes the voltage parameter distribution across the crossbar array by applying different voltages at opposite ends of row electrodes. Instead of using a single reference potential, the system dynamically adjusts voltage levels at multiple points to compensate for resistance-induced variations. This parameter change approach allows the system to maintain accurate voltage differences across memristive elements despite resistive losses in interconnect wires.
2Measurement precision
If additional voltage inputs are applied to both ends of row electrodes to compensate for resistance, then signal degradation is reduced, but device complexity increases
Solution Approach 1:
The patent achieves universality by making the column electrodes serve dual functions: they act as both computation input lines (receiving voltage inputs for matrix multiplication) and as reference potential lines (providing stable voltage references for compensation). This multi-functionality reduces the need for separate dedicated reference lines, thereby limiting the increase in device complexity while still enabling accurate compensation for resistive effects.
Solution Approach 2:
The patent applies equipotentiality by using column electrodes maintained at a common reference potential to provide stable voltage references across the array. By ensuring that all column electrodes share the same reference potential, the system creates equipotential regions that simplify the calculation and application of compensation voltages, reducing control complexity while maintaining signal accuracy.
3Measurement precision
If compensation voltages are applied to minimize sneak path currents, then processing accuracy is improved, but power consumption increases
Solution Approach 1:
The patent applies partial action by selectively compensating only for the resistive effects that significantly impact computation accuracy, rather than attempting to compensate for all possible error sources. The compensation voltages are applied only to the extent necessary to counteract wire resistance effects, avoiding excessive compensation that would waste energy. This selective approach balances accuracy improvement with energy efficiency.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The double bias scheme significantly reduces signal degradation, improves accuracy, and allows for the operation of denser crossbar arrays by minimizing errors and ensuring consistent voltage distribution across the array.
Implementation Method 1
Resistive memory elements referred to as memristors are devices that may be programmed to different resistive states by applying electrical voltage or currents to the memristors. After programming, the state of the memristors may be read. The state of the memristors remains stable over a specified time period long enough to regard the device as non-volatile.
Implementation Method 2
The resistance attributable to the wires can result in signal degradation—e.g., decrease in voltage—along each row or column. The disclosure provided herein describes a vector-matrix processing system and method utilizing memristor-based crossbar arrays that reduces signal degradation caused by resistance in the wires, lines, or electrodes connecting the memristors along each row or column.
Data Source
AI summary
A double bias dot-product engine for vector processing is described. The dot product engine includes a crossbar array having N×M memory elements to store information corresponding to values contained in an N×M matrix, each memory element being a memristive storage device. First and second vector input registers including N voltage inputs, each voltage input corresponding to a value contained in a vector having N×1 values. The vector input registers are connected to the crossbar array to supply voltage inputs to each of N row electrodes at two locations along the electrode. A vector output register is also included to receive voltage outputs from each of M column electrodes.