Adaptive Filter Gradient Sharing for Massive MIMO DPD Convergence

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Massive Multiple-Input-Multiple-Output (MIMO) systems face challenges in computation load and convergence speed due to the large number of antennae, particularly in digital pre-distortion (DPD) algorithms, which struggle to efficiently track changes in power amplifier characteristics across numerous branches.

Innovation Solution

The proposed method transfers gradient-based information among all branches of the adaptive filter system, allowing uninterrupted parameter updates and improving convergence speed and tracking capability, by sharing gradient results and adjusting update rates based on branch similarity.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If the number of antennae is increased in massive MIMO systems, then beamforming capability is improved, but computation load and convergence speed of DPD algorithms deteriorate

Engineering Contradiction:
Improvebeamforming capabilityVSAvoidcomputation load
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent merges the parameter update processes of multiple antenna branches by transferring gradient-based information across all branches. Instead of independently updating each branch's parameters, the system combines gradient information from selected branches and uses it to update parameters across all branches simultaneously, reducing the overall computation load while maintaining beamforming capability

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent creates a universal parameter update mechanism that serves all antenna branches. By computing gradient-based information for selected branches and transferring it to update parameters across all branches, the system achieves multi-functionality where a single computation process benefits the entire massive MIMO system, reducing redundant calculations

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Reliability

If the number of antennae is increased in massive MIMO systems, then beamforming capability is improved, but convergence speed of DPD algorithms deteriorates

Engineering Contradiction:
Improvebeamforming capabilityVSAvoidconvergence speed
Core Design Contradiction:
ReliabilityVSSpeed

Solution Approach 1:

The patent ensures continuous parameter updates across all branches by transferring gradient-based information. Instead of interrupting updates in non-selected branches, the system continuously updates all branches using transferred gradient information, maintaining uninterrupted convergence progress across the entire system

Inventive Principle:
Principle #20Continuity of useful action

Solution Approach 2:

The patent implements a feedback mechanism where gradient-based information from selected branches is transferred and used to update parameters in all branches. This feedback loop allows the system to continuously refine parameters across all antenna branches, accelerating convergence while maintaining beamforming performance

Inventive Principle:
Principle #23Feedback

3Speed

If gradient-based information is transferred among all branches, then convergence speed is improved, but communication overhead between branches increases

Engineering Contradiction:
Improveconvergence speedVSAvoidcommunication overhead
Core Design Contradiction:
SpeedVSQuantity of substance

Solution Approach 1:

The patent applies local quality by selectively transferring gradient-based information only from and to specific branches rather than all-to-all communication. The system identifies selected branches for gradient computation and transfers information efficiently, reducing unnecessary communication overhead while maintaining convergence acceleration benefits

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP3791481B1System and method for adaptive filtering
Publication Date: 2023.10.04 TELEFONAKTIEBOLAGET LM ERICSSON (PUBL)
  • EP3791481B1 patent drawingFigure 1~2
  • EP3791481B1 patent drawingFigure 3~4
  • EP3791481B1 patent drawingFigure 5

AI summary

A method in an adaptive filter system is provided. The method comprises obtaining parameters for a plurality of branches of the adaptive filter system (S310). The method further comprises computing gradient-based information for a selected one of the plurality of branches (S320). The method further comprises updating the parameters for the plurality of branches based on the gradient-based information for the selected branch (S330). An adaptive filter system is also provided.