Transmitter Characterization and Optimization

PhD Student 
Karl Freiberger

Characterization of wireless transmitters and their individual building blocks is of utter importance when optimizing the system performance, both in the design phase but is also online, during operation. By characterization, we subsume the tasks of system identification and measurement/estimation of figure of merits. Figure of merits characterize the performance of the transmitter in a condensed way, e.g., the error vector magnitude (EVM) is a single number characterizing the inband distortion and noise to signal ratio. System identification is about estimating structure and parameter values of transmitter (building block) models, e.g. a nonlinear model with memory characterizing the behavior of the power amplifier (PA). This is an optimization problem: We want to find an optimal model and parameters values with respect to some objectives and constraints. Even more challenging than characterizing a nonlinear system e.g. the PA, is to optimize a digital predistorter, i.e., a preceding nonlinear system, such that the overall cascade has advantageous properties.

The focus of this thesis is characterization and optimization of transmitters for wireless local area networks (WLANs), i.e., the IEEE 802.11ac standard. The first part of this thesis investigates a measurement method to estimate EVM with less stringent hardware requirements (bandwidth, synchronization) but reduced measurement floor compared to the traditional EVM measurement. The second part deals with system identification and optimization. We show that multi-objective optimization (MOO) of EVM and out-of-band error (e.g., spectral mask) leads to better or at least equal DPD solutions than traditional approaches employing a mean-squared error (MSE) objective. Another point addressed is to relax the requirements on the analog to digital converter (ADC) sampling the PA output, for obtaining the necessary information to adapt the DPD. Decoupling the ADC samplerate from the signal bandwith becomes increasingly important for handling the wide bandwidths of modern communication standards.

MOO_DPD.png
This thesis is supervised by Christian Vogel.