|Titel:||Blind Estimation and Mitigation of Nonlinear Channels|
|Reihe:||Mobile Nachrichtenübertragung, Nr.: 75|
Abstrakt in Englisch
The worldwide demand for wireless data services is on the rise for several years now. With the introduction of smartphones and tablets, this trend has intensified. Many users now own multiple devices and data needs to be easily available across all of them. This has lead to an embrace of cloud services not only for documents, but also for photos, music and even video data, yielding another spike in traffic demand. Multicarrier modulation is the current answer to the ever rising traffic demand. It allows the efficient usage of large bandwidths at relatively low computational complexity. It has been in use in digital video broadcasting and wireless local area networks for a while and is now introduced in cellular communications with the advent of LTE. The downside of multicarrier modulation is its very high dynamic range which results in instantaneous power peaks. Amplifiers need to be driven with a large power reserve in order to cope with these peaks. This however reduces their energy efficiency dramatically which is especially bad for battery powered devices. Furthermore, development and production of these highly linear amplifiers is costly. Driving the amplifiers with less reserve causes nonlinear distortion of the power peaks and hence a significant reduction in signal quality. In the past, both transmitter and receiver based methods have been presented to mitigate these distortions. However, very often perfect knowledge of the nonlinearity characteristic is assumed which is not realistic especially in receiver based methods. The focus of this thesis is on methods that estimate the nonlinearity characteristic on the receiver side. Of special interest are so called blind algorithms because they don’t require special pilot signals and hence can be used with existing standards. The main focus is on the formal derivation of blind estimation methods and their low-complexity implementation. It is shown that there is virtually no performance gap between estimated and perfect nonlinearity knowledge. One of the methods has been implemented on a software defined radio platform. It is shown that significant performance gains can be reached for real nonlinear amplifiers. The system runs in real time on cheap off-the-shelf components proving the low complexity of the method which is ready for implementation on today’s hardware.