High-Rank MIMO Symbol Detection via Virtual Channel Combining
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High-rank MIMO symbol detectors, such as those used in LTE and 5G NR, face complexity issues due to increased layers, leading to degraded performance in existing linear detectors and excessive complexity in non-linear detectors.
Innovation Solution
Implementing virtual channel combining using multiple lower rank symbol detectors to decompose high-rank channels into virtual channels, followed by preprocessing and combining Euclidean distance or log-likelihood-ratio values to achieve high-rank symbol detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the number of layers (rank) used for MIMO transmission increases, then the data rate and capacity are improved, but the complexity of symbol detection increases significantly
Solution Approach 1:
The high-rank MIMO detection problem is segmented into multiple lower-rank sub-problems. The channel matrix is decomposed into multiple virtual channels, and separate symbol detectors process each virtual channel independently. This segmentation reduces the computational complexity of each detector while maintaining the overall data rate through parallel processing of multiple codewords.
2Device complexity
If linear symbol detectors (ZF, MMSE) are used for high-rank scenarios, then the device complexity is reduced, but the detection performance degrades
Solution Approach 1:
By segmenting the high-rank channel into multiple virtual channels with lower ranks, the system can employ simple linear detectors (ZF or MMSE) on each segment while achieving performance comparable to complex non-linear detectors. The segmentation transforms an intractable high-complexity problem into multiple tractable low-complexity problems.
Solution Approach 2:
The outputs from multiple low-rank symbol detectors are merged/combined to produce the final detection result. This combining process integrates the results from parallel virtual channel detections, achieving high-rank detection performance with the computational efficiency of low-rank detectors.
3Reliability
If non-linear symbol detectors (ML, SD) are used for high-rank scenarios, then the detection performance is improved, but the device complexity becomes excessive
Solution Approach 1:
The patent applies segmentation by decomposing the high-rank channel into multiple virtual channels, allowing the use of simple linear detectors instead of complex non-linear detectors. This segmentation principle reduces complexity from exponential (required for ML/SD at high rank) to polynomial complexity.
Solution Approach 2:
Instead of implementing a single complex non-linear detector for the entire high-rank channel, the system creates multiple copies of simpler low-rank detectors that process virtual channels. These detector copies operate in parallel, providing near-ML performance with much lower complexity.
4Productivity
If multiple codewords are detected simultaneously in high-rank MIMO, then the spectral efficiency is improved, but the interference management and detection complexity increase
Solution Approach 1:
The patent segments the detection of multiple codewords by assigning different sets of virtual channels to different codewords. This segmentation allows independent detection of each codeword with reduced interference, as each detector only needs to process a subset of layers rather than all layers simultaneously.
Solution Approach 2:
The patent introduces a new dimension of virtual channel decomposition to manage multiple codewords. By transforming the detection problem from a single high-rank space into multiple lower-rank virtual channel spaces, the system achieves better interference management through dimensional decomposition.
Data Source
AI summary
A method may include receiving a signal over a channel, decomposing the channel into a plurality of virtual channels, and performing symbol detection by two or more symbol detectors on the plurality of virtual channels. At least one of the two or more symbol detectors may include less than all of the plurality of virtual channels, and the method may include combining outputs from the two or more symbol detectors to obtain log likelihood ratio (LLR) values for layers of the channel.