Neuromorphic ADC Reference Circuit for Drift-Stable Readout

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Neuromorphic computing devices face challenges with temperature and time dependency in resistive memory cells, leading to inconsistencies in read currents and signal voltages, which affect the accuracy and reliability of analog-to-digital conversions.

Innovation Solution

Incorporating an analog-to-digital converter with a voltage generator using a resistive memory element of the same material as the crossbar array, generating a reference voltage independent of temperature and time, to produce divided voltages for comparing signal voltages and generating digital signals that are consistent across temperature and time variations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If resistive memory cells are used in neuromorphic computing devices, then computational efficiency is improved, but temperature and time dependency causes inconsistencies in read currents and signal voltages

Engineering Contradiction:

Improvecomputational efficiencyVSAvoidconsistency of read currents and signal voltages

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent uses resistive memory elements with the same resistive material in both the crossbar array and the voltage generator. This homogeneity ensures that both the signal voltage from the crossbar array and the reference voltage from the voltage generator exhibit identical temperature and time dependency characteristics, allowing the dependencies to cancel out during comparison operations.

Inventive Principle:

Principle #33Homogeneity

Solution Approach 2:

The voltage generator continuously generates a reference voltage that mirrors the temperature and time dependency of the signal voltage through using the same resistive material. This creates a feedback mechanism where the reference voltage automatically compensates for drift in the signal voltage, maintaining consistent comparison results without external intervention.

Inventive Principle:

Principle #23Feedback

2Device complexity

If temperature and time dependency is present in resistive memory cells, then device operation is simplified, but analog-to-digital conversion accuracy deteriorates

Engineering Contradiction:

Improvedevice operation simplicityVSAvoidanalog-to-digital conversion accuracy

Core Design Contradiction:

Device complexityVSMeasurement precision

Solution Approach 1:

The voltage generator acts as an intermediary that produces a reference voltage with matching temperature and time dependency characteristics. This intermediary reference voltage serves as a mediator between the signal voltage and the comparison operation, enabling accurate analog-to-digital conversion despite the presence of environmental dependencies in the resistive memory cells.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Reliability

If resistive memory elements with same material are used in voltage generator, then temperature and time independence is achieved, but device complexity increases

Engineering Contradiction:

Improvetemperature and time independenceVSAvoidanalog-to-digital converter structure

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent merges the voltage generation functionality with the resistive memory elements already present in the crossbar array by using the same resistive material. This combining approach generates temperature and time independent reference voltages without introducing entirely new component types, thereby limiting the increase in device complexity to the necessary voltage generator circuitry while leveraging existing material properties.

Inventive Principle:

Principle #5Merging (Combining)

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

This approach ensures digital signals remain stable and consistent, reducing errors and the need for error-detection mechanisms, enabling faster and more energy-efficient operations in neuromorphic computing tasks like classification and regression.

Implementation Method 1

generates a first voltage based on a reference voltage and the at least one resistive memory element

Methodology Applied

Scientific EffectElectrical conduction: Conduction (electrical)

Implementation Method 2

divides the first voltage to generate at least one divided voltage

Methodology Applied

Scientific EffectElectrical resistance: Electrical Resistance

Implementation Method 3

compare a signal voltage from the crossbar array with at least one divided voltage to generate at least one comparison signal

Methodology Applied

Scientific EffectVoltage comparison: Ohmmeter

Data Source

PatentUS11302392B2Analog-to-digital converter and neuromorphic computing device including the same

Publication Date: 2022.04.12 SAMSUNG ELECTRONICS CO LTD

AI summary

An analog-to-digital converter is connected to a crossbar array including a plurality of resistive memory cells. Each of the plurality of resistive memory cells includes a resistive element. The analog-to-digital converter includes a voltage generator and processing circuitry. The voltage generator includes at least one resistive memory element including a same resistive material as the resistive element included in the crossbar array, and is configured to generate a first voltage based on a reference voltage and the at least one resistive memory element and to divide the first voltage to generate at least one divided voltage. The processing circuitry is configured to compare a signal voltage generated from the crossbar array with the at least one divided voltage to generate at least one comparison signal and generate at least one digital signal corresponding to the signal voltage based on the at least one comparison signal.