This paper presents an algorithm-adaptable, scalable, and platform-portable generator for massive multiple-input multiple-output (MIMO) baseband processing systems. This generator is written in Chisel hardware construction language, and produces instances that implement distributed massive MIMO base station (BS) processing, including channel estimation and beamforming. The generator can be reused for different MIMO systems and hardware datapath designs by changing the parameters. The generator is paired with a Python-based system simulator, which incorporated together can emulate a system testing various baseband signal processing algorithms. The field programmable gate array (FPGA) emulation is performed with generated instances using various parameter values. To demonstrate the algorithmic adaptability, a Golay-sequence-based channel estimation method, a beamspace calibration method, and a channel denoising algorithm are evaluated across a range of channel models. The performance of the generator, necessity of the algorithmic adaptability, and ease of hardware generation are evaluated and discussed. The emulated register-transfer level (RTL) implementation with different system parameters shows that with beamspace methods, the demodulation error vector magnitude is improved by up to 29.8%.

本文提出了一种适用于大规模多输入多输出 (MIMO) 基带处理系统的算法适应性强、可扩展且平台便携的发生器。该生成器使用 Chisel 硬件构建语言编写，并生成实现分布式大规模 MIMO 基站 (BS) 处理的实例，包括信道估计和波束成形。通过更改参数，发生器可以重复用于不同的 MIMO 系统和硬件数据路径设计。该生成器与基于 Python 的系统模拟器配对，结合在一起可以模拟测试各种基带信号处理算法的系统。现场可编程门阵列 (FPGA) 仿真是通过使用各种参数值生成的实例来执行的。为了证明算法的适应性，在一系列信道模型中评估了基于 Golay 序列的信道估计方法、波束空间校准方法和信道去噪算法。对生成器的性能、算法适应性的必要性以及硬件生成的难易程度进行了评估和讨论。具有不同系统参数的仿真寄存器传输级 (RTL) 实施表明，使用波束空间方法，解调误差矢量幅度提高了 29.8%。