A Sparse Bayesian Technique to Learn the Frequency-Domain Active Regressors in OFDM Wireless Systems

Area of Science:

• Electrical Engineering
• Signal Processing
• Wireless Communications

Background:

• Nonlinear behavioral modeling of power amplifiers (PAs) is crucial for wireless systems.
• Most research has focused on time-domain (TD) modeling, with limited exploration in the frequency domain (FD).
• Orthogonal frequency division multiplexing (OFDM) systems present an opportunity for FD modeling.

Purpose of the Study:

• To develop and demonstrate a frequency domain Sparse Bayesian Learning (FD-SBL) algorithm for modeling nonlinear distortion in wireless OFDM systems.
• To identify an efficient and accurate reduced set of regressors for PA behavioral models.
• To enable prediction of nonlinear distortion for successive OFDM symbols.

Main Methods:

• Proposed a novel FD-SBL algorithm for PA nonlinear distortion modeling.
• Utilized SBL to identify active FD regressors and estimate PA model coefficients.
• Applied estimated coefficients for predicting distortion in subsequent OFDM symbols.

Main Results:

• Achieved a validation Normalized Mean Squared Error (NMSE) of -47 dB for a 30 MHz bandwidth signal, comparable to TD-SBL (-46.6 dB).
• FD-SBL demonstrated superior performance for a 100 MHz bandwidth signal, yielding an NMSE of -38.6 dB.
• TD-SBL encountered excessive processing time and numerical issues at 100 MHz bandwidth, rendering it impractical.

Conclusions:

• The proposed FD-SBL algorithm provides an efficient and accurate method for nonlinear distortion modeling in OFDM systems.
• FD-SBL overcomes the computational limitations of TD-SBL for high-bandwidth signals.
• This approach is promising for enhancing the performance and efficiency of modern wireless communication systems.