Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

NN-Based Joint Mitigation of IQ Imbalance and PA Nonlinearity With Multiple States

Yundi Zhang, Wendong Cheng, Li Chen

TL;DR

This work tackles the joint mitigation of IQ imbalance and PA nonlinearity in RF transmitters under multiple operating states by introducing HN-R2TDNN, a neural-network model that combines hard parameter sharing across states with a hypernetwork that dynamically generates the output-layer parameters conditioned on the current state. The approach builds on multi-task learning to share hidden representations while adapting the final mapping per state, enabling effective joint DPD without retraining when states change. Empirical results show significant improvements in ACPR and NMSE across nine states compared to SVDEN and HG-R2TDNN, and the model demonstrates stronger spectral regrowth suppression. The proposed method offers a practical path to robust, state-aware joint DPD in dynamic wireless environments.

Abstract

Joint mitigation of IQ imbalance and PA nonlinearity is important for improving the performance of radio frequency (RF) transmitters. In this paper, we propose a new neural network (NN) model, which can be used for joint digital pre-distortion (DPD) of non-ideal IQ modulators and PAs in a transmitter with multiple operating states. The model is based on the methodology of multi-task learning (MTL). In this model, the hidden layers of the main NN are shared by all signal states, and the output layer's weights and biases are dynamically generated by another NN. The experimental results show that the proposed model can effectively perform joint DPD for IQ-PA systems, and it achieves better overall performance within multiple signal states than the existing methods.

NN-Based Joint Mitigation of IQ Imbalance and PA Nonlinearity With Multiple States

TL;DR

This work tackles the joint mitigation of IQ imbalance and PA nonlinearity in RF transmitters under multiple operating states by introducing HN-R2TDNN, a neural-network model that combines hard parameter sharing across states with a hypernetwork that dynamically generates the output-layer parameters conditioned on the current state. The approach builds on multi-task learning to share hidden representations while adapting the final mapping per state, enabling effective joint DPD without retraining when states change. Empirical results show significant improvements in ACPR and NMSE across nine states compared to SVDEN and HG-R2TDNN, and the model demonstrates stronger spectral regrowth suppression. The proposed method offers a practical path to robust, state-aware joint DPD in dynamic wireless environments.

Abstract

Joint mitigation of IQ imbalance and PA nonlinearity is important for improving the performance of radio frequency (RF) transmitters. In this paper, we propose a new neural network (NN) model, which can be used for joint digital pre-distortion (DPD) of non-ideal IQ modulators and PAs in a transmitter with multiple operating states. The model is based on the methodology of multi-task learning (MTL). In this model, the hidden layers of the main NN are shared by all signal states, and the output layer's weights and biases are dynamically generated by another NN. The experimental results show that the proposed model can effectively perform joint DPD for IQ-PA systems, and it achieves better overall performance within multiple signal states than the existing methods.
Paper Structure (12 sections, 14 equations, 5 figures)

This paper contains 12 sections, 14 equations, 5 figures.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. IQ Imbalance and PA Nonlinearity
  4. The Problem of Joint DPD
  5. Existing Methods
  6. Proposed Joint DPD Model
  7. The Proposed HN-R2TDNN
  8. Offline Training Method
  9. Experimental Results
  10. Experimental Setups
  11. Results
  12. Conclusion

Figures (5)

  • Figure 1: Complex baseband equivalent model of the IQ-PA system and the basic idea of joint DPD
  • Figure 2: Existing models. (a) R2TDNN and SVDEN. (b) HG-R2TDNN
  • Figure 3: Proposed Method. (a) Proposed HN-R2TDNN for joint DPD with multiple states. (b) The parallel loading approach of training data when using ILA
  • Figure 4: The ACPR and NMSE performance of the 3 models within all 9 states. (a) ACPR. (b) NMSE.
  • Figure 5: PSDs of the signals within operating state 1 (20MHz and -19dBm)