Master Thesis: AI‑Enhanced Next‑Gen DPD for 1 GHz+ Instantaneous Bandwidth Power Amplifiers

Join our Team

About this opportunity:

Modern radio access networks increasingly rely on wideband power amplifiers (PAs) with instantaneous bandwidths of 1 GHz and beyond. These PAs exhibit strong nonlinearities and memory effects, which can cause spectral regrowth, degraded error-vector magnitude (EVM), and reduced energy efficiency.

Digital predistortion (DPD) is essential for linearizing PAs, but conventional approaches can become difficult to scale to very wide bandwidths. They may require complex models, extensive tuning, and significant computational and memory resources. Recent advances in AI and machine learning offer new opportunities for more efficient model structures, parameterization, and adaptation.

This thesis will investigate next-generation DPD models for 1 GHz+ bandwidth PAs, with a focus on hybrid approaches that combine domain knowledge of PA behavior with AI/ML techniques. The work will primarily involve simulation and offline processing of realistic PA models and/or measured data from Ericsson-relevant scenarios. Laboratory validation may be included where appropriate.

What you will do:

• Review state-of-the-art DPD methods for wideband power amplifiers, including polynomial, LUT-based, and AI/ML-based approaches.

• Design and implement one or more DPD model structures that efficiently capture wideband memory effects and frequency-dependent PA behavior.

• Explore AI/ML for applications such as feature learning, model selection, parameter initialization, or improved tuning—while maintaining practical and interpretable solutions.

• Develop an evaluation framework in MATLAB, Python, or a similar environment using realistic PA models and/or measured PA data.

• Evaluate and compare the proposed approach with a representative baseline in terms of linearization performance, robustness, stability, computational complexity, and memory requirements.

• Provide recommendations for future DPD development in Ericsson wideband radios, including potential follow-up work on real-time implementation, faster tuning, and live adaptation.

The skills you bring:

• Pursuing Master’s in Electrical Engineering, Applied Physics, Computer Science, or a related field.

• Strong interest in signal processing, communication systems, RF technologies and power amplifiers.

• Programming skills in MATLAB and Python.

• Knowledge of digital predistortion, RF systems, nonlinear systems, or machine learning is an advantage.

• You are analytical, structured, and able to work independently on an algorithm-development problem.

• You are motivated by working at the intersection of advanced algorithms, AI, and next-generation radio technology.