Pipelined MIMO-OFDM Detector for Reducing Computational Complexity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The complexity and computational intensity of tree search operations in MIMO-OFDM communication systems, particularly due to the need for efficient sorting and indexing functions, lead to processing inefficiencies and increased costs as the number of antennas, OFDM subcarriers, and modulation order increase.
Innovation Solution
The implementation of a pipelined architecture with multiple atomic sorting and indexing units, allowing concurrent processing and reduced idle time and memory requirements, enhances the throughput of the MIMO-OFDM detector by parallelizing the computation and sorting of node metrics across tree levels and subcarriers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If tree search operations are implemented with increased number of antennas, OFDM subcarriers, and modulation order, then detection accuracy and system capacity are improved, but computational complexity and processing cost increase significantly
Solution Approach 1:
The patent divides the tree search operation into multiple pipeline stages, where each stage processes a specific portion of the search tree. The search tree is segmented by tree levels, with each level processed by a dedicated processing unit. This segmentation allows parallel processing of different tree levels simultaneously, reducing the overall computational complexity while maintaining detection accuracy.
Solution Approach 2:
The patent introduces a time dimension to the tree search operation by implementing a pipelined architecture. Instead of processing the entire search tree sequentially, the system processes multiple tree levels concurrently across different time stages. This dimensional transformation enables parallel computation of node metrics across different levels, significantly reducing processing time and complexity.
2Productivity
If sequential processing is used for tree search operations, then implementation is simpler, but processing time and throughput are reduced
Solution Approach 1:
The patent implements a continuous pipelined processing architecture where data flows continuously through multiple processing stages. Each stage processes node metrics for a specific tree level and passes results to the next stage without interruption. This continuous action eliminates idle time between processing stages, maximizing throughput and reducing overall processing time compared to sequential batch processing.
Solution Approach 2:
The patent performs preliminary computation of node metrics for each tree level before proceeding to the next level. By pre-computing metrics for multiple levels in parallel and storing them in buffer memory, the system eliminates waiting time during sequential processing, thereby increasing throughput and reducing total processing time.
3Measurement precision
If comprehensive sorting and indexing is performed for all node metrics, then detection accuracy is improved, but memory requirements and processing overhead increase
Solution Approach 1:
The patent extracts only the essential information from node metrics - specifically the indices of the lowest metric values - and stores these in buffer memory for the next processing stage. Instead of storing complete node metric data structures, the system extracts and retains only the critical index information needed for continuing the tree search, significantly reducing memory requirements while maintaining detection accuracy.
Solution Approach 2:
The patent performs partial sorting and indexing operations, focusing only on identifying and retaining the indices of the lowest metric values at each tree level rather than completely sorting all node metrics. This partial action approach provides sufficient detection accuracy by preserving the most relevant paths while reducing the computational overhead and memory usage associated with complete sorting operations.
Data Source
AI summary
A nonlinear MIMO-OFDM detector includes a vector arithmetic unit (VAU) that sequentially computes first metrics corresponding to a first current tree level of a first search tree and second metrics corresponding to a second current tree level of a second search tree. A sorting and indexing unit (SIU) that sorts the first metrics and the second metrics sequentially received from the VAU and that sequentially provides first indices of lowest first metrics and second indices of lowest second metrics to the vector arithmetic unit. The lowest first metrics are first inputs to the VAU for a first next tree level of the first search tree and the lowest second metrics are second inputs to the VAU for a second next tree level of the second search tree. The VAU and the SIU are pipelined to compute the second metrics concurrently with sorting and indexing of the first metrics.