High-Speed Ultra-Energy-Efficient Memristor-Based Massive MIMO SIC Detector Circuit With Hybrid Analog-Digital Computing Architecture

Jia-Hui Bi, Shaoshi Yang, Sheng Chen, Ping ZhangIEEE Trans. Vehicular Technology 2025 · DOI

TL;DR

This paper introduces a novel circuit design for massive MIMO (Multiple-Input Multiple-Output) systems that leverages memristor crossbar arrays for ultra-energy-efficient and high-speed signal detection. By combining analog matrix computation with hybrid analog-digital slicing, their proposed circuit significantly outperforms traditional digital processors, FPGAs, and GPUs in terms of speed and energy efficiency for Successive Interference Cancellation (SIC) detection.

What Problem Does This Paper Address?

Massive MIMO technology, crucial for enhancing wireless communication rates and spectral efficiency, faces a significant challenge: the substantial power consumption and computational complexity associated with its large number of antennas and sophisticated baseband signal processing. Specifically, algorithms like Successive Interference Cancellation (SIC), which offer better detection performance than simpler linear methods, involve iterative matrix computations and non-linear "slicing" operations. While emerging memristor-based computing offers impressive speed and energy efficiency for matrix operations, efficiently implementing the slicing part of SIC detection has remained an open problem. This paper aims to bridge this gap by proposing a memristor-based circuit that can handle both the matrix computations and the slicing operations required for SIC detection in massive MIMO systems.

Methods

The core innovation lies in a hybrid analog-digital architecture for the SIC detector. The proposed circuit is structured into K stages, where each stage comprises:

  • Memristor-based Matrix Computing Module: This module utilizes memristor crossbar arrays to perform the complex matrix computations inherent in the SIC algorithm. The analog nature of memristors allows for these operations to be executed at high speeds and with low energy consumption, bypassing the traditional von Neumann bottleneck.
  • Hybrid Analog-Digital Slicers: These novel slicers are designed to handle the non-linear "slicing" operation. They first convert analog voltages (representing quantized signal estimates) into a digital (binary) vector. Based on this binary representation, they then output the corresponding quantized voltage levels. Two slicer structures are proposed: a "directly select" structure and an "indirectly select" structure, with the latter offering better multiplexer channel utilization for higher-order modulation schemes.

The paper also details how complex-valued vectors and matrices are converted into their real-valued counterparts to be compatible with the analog matrix computation.

Datasets

Not discussed in the paper. The evaluation is based on simulations using a 32x64 massive MIMO system with 16-QAM modulation.

Key Insights and Ideas

  • Bridging the Analog-Digital Divide for SIC: The most significant insight is the successful integration of memristor-based analog matrix computation with a novel hybrid analog-digital slicer to implement the non-linear SIC detection algorithm. This overcomes a key limitation of previous memristor-based approaches that were primarily suited for linear operations.
  • Ultra-Low Power and High Speed: The inherent parallelism and in-memory computing nature of memristor crossbar arrays, combined with the efficient analog processing, leads to dramatic improvements in both speed and energy efficiency compared to conventional digital hardware.
  • Memristor Precision Matters: The performance of the proposed detector is sensitive to the precision of the memristors. Simulations show that at least 6 bits of precision are needed to achieve detection performance comparable to digital methods.
  • Scalability for Future Wireless: The proposed architecture is highlighted as particularly suitable for future massive MIMO systems, suggesting its potential to meet the ever-increasing demands for connectivity and data rates.

Key Results

The proposed memristor-based SIC detector circuit demonstrates significant performance gains:

  • Speed: Approximately 43 times faster than a traditional 8-core DSP.
  • Energy Efficiency: Approximately 110 times higher than a traditional 8-core DSP.
  • Comparison: Outperforms a high-performance FPGA and offers comparable speed to a powerful GPU, while significantly exceeding the GPU's energy efficiency.
  • Computational Speed: Achieves approximately 5.5 tera operations per second (TOPS).
  • Computational Energy Efficiency: Reaches approximately 1.41 TOPS/W.
  • Memristor Precision: At least 6 bits of memristor precision are required for performance close to digital benchmarks.

Limitations and Future Work

  • Memristor Precision: The paper acknowledges that the performance is directly tied to memristor precision, and achieving high precision can be challenging. Future work could explore error correction codes or more robust circuit designs to mitigate precision limitations.
  • Simulation-Based: The results are based on simulations (using LTspice®). Real-world hardware implementation and testing would be the next crucial step.
  • Specific Modulation Schemes: While 16-QAM is used for illustration, further investigation into the performance with other modulation schemes and their impact on slicer design would be beneficial.
  • Hardware Implementation Details: While components like comparators and multiplexers are specified, a full hardware implementation roadmap and detailed analysis of parasitic effects would be valuable.

Summary updated

BibTeX
@article{bi2025high,
  author = {Jia-Hui Bi and Shaoshi Yang and Sheng Chen and Ping Zhang},
  title = {{High-Speed Ultra-Energy-Efficient Memristor-Based Massive MIMO SIC Detector Circuit With Hybrid Analog-Digital Computing Architecture}},
  journal = {IEEE Trans. Vehicular Technology},
  year = {2025},
  doi = {10.1109/tvt.2025.3544093},
}
Figure