Capacitor-Weighted Segmentation Buffer for Stable Low-Power DAC Integration

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional buffer circuits face challenges with high power consumption, large chip area, and low accuracy, particularly in AI applications like convolutional neural networks, where existing solutions such as CPU-based convolution operations or bit-wise current-based compute-in-memory circuits are not power-efficient or accurate.

Innovation Solution

A capacitor weighted segmentation buffer is introduced, comprising a push-pull buffer circuit and multiple capacitors with different capacitance values, integrated with a digital-to-analog converter, which operates in discrete-time modes to provide stable buffering across process, voltage, and temperature variations, reducing chip area and power consumption while maintaining high accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If conventional open-loop buffers are used, then the circuit structure is simple, but the performance suffers from PVT variations and requires separate DAC components resulting in large chip area

Engineering Contradiction:
Improvebuffer circuit structureVSAvoidPVT variation performance
Core Design Contradiction:
Device complexityVSReliability

Solution Approach 1:

The patent merges the DAC and buffer functions into a single integrated circuit block. The buffer circuit shares capacitive elements with the DAC, eliminating the need for separate DAC components and reducing chip area while maintaining performance stability against PVT variations through the unified design

Inventive Principle:
Principle #5Merging (Combining)

2Reliability

If closed-loop buffers are used, then stability and phase margin performance are improved, but speed and energy consumption deteriorate

Engineering Contradiction:
Improvestability performanceVSAvoidpower consumption
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent employs dynamic element matching (DEM) technique where capacitive elements are dynamically switched and redistributed during operation. This dynamic approach maintains stability through controlled charge redistribution while avoiding the continuous power consumption associated with traditional closed-loop feedback mechanisms

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The buffer operates in discrete-time modes with periodic reset and buffer phases. During the reset phase, capacitors are discharged; during the buffer phase, they store and transfer charge. This periodic operation achieves stability through controlled charge management while maintaining low power consumption by avoiding continuous feedback current

Inventive Principle:
Principle #19Periodic action

3Adaptability or versatility

If CPU-based convolution operations are used in AI applications, then computational flexibility is maintained, but power consumption increases significantly

Engineering Contradiction:
Improvecomputational flexibilityVSAvoidpower consumption
Core Design Contradiction:
Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The patent replaces traditional CPU-based computational operations with an analog compute-in-memory circuit that performs convolution operations directly in the analog domain. The capacitor-weighted buffer circuit enables weighted sums to be computed through passive charge redistribution rather than active computational processing, dramatically reducing power consumption while maintaining computational functionality for AI applications

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

4Device complexity

If bit-wise current-based compute-in-memory circuits are used, then integration is improved, but accuracy deteriorates

Engineering Contradiction:
Improveintegration levelVSAvoidcomputational accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent uses capacitor-based charge storage and redistribution to accurately represent and process analog weights and inputs. By copying charge representations through capacitive coupling rather than current-based bit-wise operations, the circuit maintains high accuracy in analog computations while achieving full integration with the DAC buffer structure

Inventive Principle:
Principle #26Copying

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 capacitor weighted segmentation buffer achieves low power consumption, ultra-small area, and high accuracy, enabling efficient integration with digital-to-analog converters and analog compute-in-memory circuits, thereby addressing the limitations of conventional buffer designs in AI applications.

Implementation Method 1

a first capacitor having a first terminal coupled to the control terminal of the first transistor and a second terminal arranged to receive a first input signal

Methodology Applied
Scientific EffectCapacitance: Capacitance

Data Source

PatentEP4312367A1Capacitor weighted segmentation buffer
Publication Date: 2024.01.31 MEDIATEK INC
  • EP4312367A1 patent drawingFigure 1
  • EP4312367A1 patent drawingFigure 2
  • EP4312367A1 patent drawingFigure 3

AI summary

A capacitor weighted segmentation buffer includes a push-pull buffer circuit and a plurality of capacitors. The capacitors include a first capacitor having a first terminal coupled to a control terminal of the first transistor and a second terminal arranged to receive a first input signal; a second capacitor having a first terminal coupled to a control terminal of the second transistor and a second terminal arranged to receive the first input signal; a third capacitor having a first terminal coupled to the control terminal of the first transistor and a second terminal arranged to receive a second input signal; and a fourth capacitor having a first terminal coupled to the control terminal of the second transistor and a second terminal arranged to receive the second input signal.