A Massive MIMO System for Power-Constrained Devices without Error-Correcting Memory

Due to hardware complexity and cost constraints, power-limited devices often avoid using error-correcting codes (ECC). The memory errors generated from these devices can propagate through the wireless channel during transmission, making the decoding at the receiver more challenging. In a massive MIMO system, which is equipped with a large number of antennas to support a large number of device users, this issue can be critical. In this paper, We propose a two-stage solution to address this challenge. The first stage employs an equalizer-based detection mechanism to mitigate communication channel impairments and interference while the second stage employs a Gradient Boosted Decision Tree (GBDT) algorithm to predict and correct memory errors originated at the devices. We demonstrated the efficacy of our approach through Python simulations, utilizing the GBDT model via the Scikit-Learn library. The results indicate a detection accuracy of over 99 % after approximate-inversion based equalization and GBDT-based error detection at the massive MIMO base station receiver. Finally, we design a static random access memory (SRAM) with Cadence Spectre to illustrate the correlation between voltage scaling and error rates, demonstrating that the proposed solution can be utilized to reduce the operating voltage of the user devices.