Spectral Optimization and DPD for 5G Massive MIMO Systems

Overview

During my postdoctoral tenure at Tecnológico de Monterrey (2018–2026), supported by a CONACyT research fellowship, I led the development of DPD algorithms and spectral modeling techniques for 5G massive MIMO transmitter arrays, where per-antenna PA nonlinearity and crosstalk effects degrade system-level spectral efficiency.

Key Contributions

• Developed surrogate behavioral models for PA linearization under high sparse data conditions, comparing polynomial, spline, and machine learning approaches (regression trees, random forests, CNNs) for wideband nonlinear modeling.
• Designed DPD algorithms optimized for massive MIMO architectures, addressing the challenge of scaling linearization across large antenna arrays while managing computational complexity.
• Performed system-level spectral efficiency analysis quantifying the impact of PA nonlinearity on massive MIMO capacity, beam pattern distortion, and adjacent channel interference.
• Published results in Sensors (2022), IEEE ISCAS (2022), and contributed a book chapter to Machine Learning for Complex and Unmanned Systems (CRC Press, Taylor & Francis, 2024).
• Advised 3 M.Sc. students and served on Ph.D. thesis committees related to this research line.

Tools & Technologies

MATLAB, Python (scikit-learn, TensorFlow), Keysight ADS, RF measurement testbed, statistical modeling frameworks.

Impact

This work bridges classical RF linearization theory with modern machine learning, demonstrating that data-driven surrogate models can match or exceed conventional polynomial DPD approaches — particularly in scenarios with limited or irregularly sampled training data, which is common in multi-band and multi-antenna deployments.