Implementation of DSP-based algorithms on USRP for mitigating non-linear distortions in the receiver
Florian Schlembach · Common Library Network (Der Gemeinsame Bibliotheksverbund) · 2013
In recent years, software-defined radio (SDR) has attracted increasingly more attention in regards to modern communication systems. The concept of SDR defines a radio device that is capable of flexibly reconfiguring its radio interface by software. This opens multiple fields of application and makes SDR an enormously adjustable and versatile radio technology. However, RF impairments induced by cheap and simple RF front-ends turn out to be a significant limitation in practice. Non-linear distortions emerge from non-linear components of the direct down-conversion chain that are driven into their saturation level. This is a result of a finite linearity and limited dynamic range of the RF frontend. The focus of this thesis are non-linear distortions in wideband receivers and a mitigation of them by means of digital signal processing. The idea is to artificially regenerate the non-linear distortions in the digital domain by applying a memoryless, polynomial model. An adaptive filter adjusts these reference distortions in their magnitude and phase and subtracts them from the distorted signal. A hardware implementation of a mitigation algorithm on a typical SDR-platform is presented. No prior implementation of this pure-digital approach is known. An implementation design flow is described following a top-down approach, starting from a fixed-point high-level implementation and ending up with a low-level hardware description language implementation. Both high-level and low-level simulations help to validate and evaluate the implementation. In conclusion, the implementation of the mitigation algorithm is a sophisticated mitigation technique for cleaning a down-converted baseband spectrum of non-linear distortions in real-time. Therefore, the effective linearity of the RF front-end is increased. This may lead to a significant improvement in the bit error rate performance of cleansed modulated signals, as well as to an enhanced sensing reliability in the context of cognitive radio.