Neural Network Cell Array Compensation for Bit Line Voltage Drop

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Artificial neural networks using analog-to-digital converters experience significant voltage drops due to high output currents, leading to decreased computation accuracy.

Innovation Solution

A neural network device incorporating a digital-to-analog converter, a cell array with memory cells, and an analog-to-digital converter to manage input and output voltages, along with dummy conductances to compensate for computational discrepancies.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If high output current is used for analog computation, then computation speed is improved, but voltage drop increases leading to decreased accuracy

Engineering Contradiction:
Improvecomputation speedVSAvoidcomputation accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent introduces a bit line voltage compensation circuit as an intermediary component that detects voltage drops on bit lines and applies compensating voltages to counteract them. This mediator circuit allows the system to maintain high output currents for fast computation while correcting the resulting voltage drops to preserve accuracy, thus resolving the contradiction between computation speed and accuracy.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent implements a feedback mechanism where the actual output current is measured and used to adjust the input voltage dynamically. The compensation circuit receives feedback about voltage drops and automatically adjusts compensating voltages to maintain accurate computation results, enabling the system to operate at high speeds without sacrificing precision.

Inventive Principle:
Principle #23Feedback

2Productivity

If analog-to-digital converter is used to detect current output, then computation speed is improved, but voltage drop due to high output current decreases accuracy

Engineering Contradiction:
Improvecomputation speedVSAvoidcomputation accuracy
Core Design Contradiction:
ProductivityVSManufacturing precision

Solution Approach 1:

The compensation circuit acts as an intermediary between the analog computation core and the analog-to-digital converter. It corrects voltage drops in the bit lines before the ADC detects the output current, ensuring that the ADC receives accurate voltage signals even when operating at high speeds with high output currents.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent applies voltage compensation in advance before the ADC performs detection. By pre-correcting voltage drops on bit lines through the compensation circuit, the system ensures that when the high-speed ADC detects the output, the voltage levels are already accurate, thus maintaining both speed and precision.

Inventive Principle:
Principle #10Preliminary action

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 effectively mitigates voltage drops, maintaining computation accuracy by balancing currents and voltages within the neural network device.

Implementation Method 1

a digital-to-analog converter configured to convert a digital signal into input voltages

Methodology Applied
Scientific EffectElectrostatic field: Electric Field

Implementation Method 2

a cell array including a plurality of memory cells that are arranged in a plurality of bit lines and a plurality of word lines and has weights of a neural network transferred thereto, wherein the cell array is configured to output, through the plurality of bit lines, output voltages obtained by performing computation on the input voltages

Methodology Applied
Scientific EffectElectrical conduction: Conduction (electrical)

Implementation Method 3

an analog-to-digital converter configured to detect the output voltages and convert the output voltages into a digital signal

Methodology Applied
Scientific EffectVoltage detection: Ohm's Law

Data Source

PatentEP4700656A1Neural network device considering voltage drop and method of implementing the same
Publication Date: 2026.02.25 PEBBLE SQUARE INC
  • EP4700656A1 patent drawingFigure 1
  • EP4700656A1 patent drawingFigure 2
  • EP4700656A1 patent drawingFigure 3

AI summary

Provided is a neural network device including a digital-to-analog converter configured to convert a digital signal into input voltages, a cell array including a plurality of memory cells that are arranged in a plurality of bit lines and a plurality of word lines and has weights of a neural network transferred thereto, wherein the cell array is configured to output, through the plurality of bit lines, output voltages obtained by performing computation on the input voltages that are input through the plurality of word lines, and an analog-to-digital converter configured to detect the output voltages and convert the output voltages into a digital signal.