MSc thesis project proposal

[Taken] Recurrent Neural Networks for Digital Predistortion

Digital Predistortion (DPD) is used to counteract the nonlinearity in power amplifiers (PA) [1] to minimize signal distortion and enhance PA efficiency, which is critical to reducing the energy consumption of communication systems [2]. In the era of 5G, the wideband signal increases the memory effect in PAs, and we need better DPD algorithms that can better capture the temporal dependency of data. Recurrent Neural Networks (RNN) such as Long Short-Term Memory (LSTM) [3] and Gated Recurrent Unit (GRU) [4] are powerful tools for sequential modeling and are promising candidates to enhance the performance of DPD with lower cost compared to the conventional memory polynomial methods [5]. This is the first step toward our ASIC design for an NN-based DPD.


Background Material

  1. S. P. Yadav and S. C. Bera, "Nonlinearity effect of Power Amplifiers in wireless communication systems," 2014 International Conference on Electronics, Communication and Computational Engineering (ICECCE), 2014, pp. 12-17, doi: 10.1109/ICECCE.2014.7086613.

  2. How DPD improves power amplifier efficiency - 5G Technology World

  3. Understanding LSTM Networks -- colah's blog

  4. Understanding GRU Networks. In this article, I will try to give a… | by Simeon Kostadinov | Towards Data Science

  5. D. R. Morgan, Z. Ma, J. Kim, M. G. Zierdt and J. Pastalan, "A Generalized Memory Polynomial Model for Digital Predistortion of RF Power Amplifiers," in IEEE Transactions on Signal Processing, vol. 54, no. 10, pp. 3852-3860, Oct. 2006, doi: 10.1109/TSP.2006.879264.

  6. Changsoo Eun and E. J. Powers, "A new Volterra predistorter based on the indirect learning architecture," in IEEE Transactions on Signal Processing, vol. 45, no. 1, pp. 223-227, Jan. 1997, doi: 10.1109/78.552219.

  7. C. Tarver, L. Jiang, A. Sefidi and J. R. Cavallaro, "Neural Network DPD via Backpropagation through a Neural Network Model of the PA," 2019 53rd Asilomar Conference on Signals, Systems, and Computers, 2019, pp. 358-362, doi: 10.1109/IEEECONF44664.2019.9048910.

  8. C. Tarver, A. Balatsoukas-Stimming and J. R. Cavallaro, "Design and Implementation of a Neural Network Based Predistorter for Enhanced Mobile Broadband," 2019 IEEE International Workshop on Signal Processing Systems (SiPS), 2019, pp. 296-301, doi: 10.1109/SiPS47522.2019.9020606.

  9. Y. Wu, U. Gustavsson, A. G. i. Amat and H. Wymeersch, "Residual Neural Networks for Digital Predistortion," GLOBECOM 2020 - 2020 IEEE Global Communications Conference, 2020, pp. 01-06, doi: 10.1109/GLOBECOM42002.2020.9322327.

  10. MATLAB: Digital Predistortion to Compensate for Power Amplifier Nonlinearities


  1. Explore and understand the MATLAB Simulink DPD example [6].

  2. Design an RNN-DPD algorithm using the MATLAB PA model as a subject and train your algorithm in PyTorch. 

  3. Evaluate the performance of your DPD algorithm in a MATLAB testbench.

  4. (Optional) Evaluate your DPD algorithm on real PAs (e.g. RF WebLab).


  • Familiar with MATLAB & Python.

  • Knowledge of digital signal processing.

  • Previous experience with any deep learning framework (PyTorch/Tensorflow) is a plus.


dr. Chang Gao

Electronic Circuits and Architectures Group

Department of Microelectronics

Last modified: 2022-10-10