Neuromorphic Device IR Drop Compensation via Reference Cells
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In semiconductor neuromorphic devices, the increasing number of synaptic weights and biases in artificial neural networks leads to challenges in storing and implementing them in hardware, particularly due to high-intensity technology and low-power requirements, and the issue of IR drop (voltage drop) in column lines during read operations.
Innovation Solution
A neuromorphic device and method that utilize reference synaptic cells to compensate for IR drop in column lines by outputting reference read currents, allowing for digital data compensation and reliable operation through a Computation-in-Memory (CiM) device configuration with first and second synaptic cells and a column control circuit.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the number of synaptic weights and biases is increased to support more layers and neurons in artificial neural networks, then the computational capability and accuracy of the neural network is improved, but the IR drop in column lines increases and power consumption rises
Solution Approach 1:
Reference column lines are introduced as intermediary elements to measure and compensate for IR drop in data column lines. The reference synaptic cells connected to these reference column lines provide a baseline measurement that allows the system to correct for voltage drops without affecting the actual computational data paths.
Solution Approach 2:
A feedback mechanism is implemented where the currents measured from reference column lines are used to adjust and compensate for IR drop in the data column lines. The column control circuit continuously monitors the reference currents and applies corrections to maintain voltage stability across all column lines during read operations.
2Quantity of substance
If more synaptic cells are added to increase storage capacity for synaptic weights, then the density and functionality of the neuromorphic device is improved, but the IR drop and power consumption increase
Solution Approach 1:
The synaptic cell array is segmented into functional groups with dedicated reference synaptic cells interspersed among the data synaptic cells. This segmentation allows the reference cells to be activated separately for measurement purposes without requiring all cells to be simultaneously active, reducing overall power consumption while maintaining measurement accuracy.
3Reliability
If reference column lines are added to measure and compensate for IR drop, then the voltage stability and reliability are improved, but the device complexity increases
Solution Approach 1:
The reference synaptic cells are merged with the data synaptic cells in the same crossbar array structure, sharing common row lines and physical space. This integration allows the reference measurement function to be added without requiring a completely separate measurement system, thereby reducing overall device complexity while maintaining voltage stability.
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 solution ensures reliable operation by compensating for voltage drops in column lines, enhancing the reliability and efficiency of neuromorphic device operations, particularly in high-density synaptic arrays, thereby addressing the challenges of IR drop and power consumption.
Implementation Method 1
a plurality of first synaptic cells disposed between a plurality of row lines and a plurality of column lines and configured to output a plurality of read currents to the column lines, the read currents corresponding to pulse signals applied through the row lines
Implementation Method 2
a column control circuit configured to output digital data by compensating for the read currents output through the column lines, based on the reference read currents output through the reference column lines
Data Source
AI summary
A Computation-in-Memory (CiM) device includes a plurality of first synaptic cells disposed between row lines and column lines and configured to output read currents to the column lines, the read currents corresponding to pulse signals applied through the row lines; a plurality of second synaptic cells disposed between the row lines and two or more reference column lines and configured to output two or more reference read currents to the reference column lines, the reference read currents corresponding to the pulse signals applied through the row lines; and a column control circuit configured to output digital data by compensating for the read currents output through the column lines, based on the reference read currents output through the reference column lines.