Doctor's Theses (authored and supervised):
"Low-Complexity Iterative Receiver for a Multi-User OFDM System in Time-Varying MIMO Channels";
Supervisor, Reviewer: G. Matz, M. Nicoli;
Institut für Nachrichtentechnik und Hochfrequenztechnik (E389),
oral examination: 04-30-2009.
Achieving high rates at high velocity is one of the ambitious goals of future wireless communications. Increasing demand for communications at vehicular speeds in turn requires highly complex processing at the base station to estimate the fast changing channel and decode large amounts of data. Real-time processing requirements or energy constraints limit the amount of signal processing aﬀordable at the terminals. Thus, sub-optimal methods and algorithms permitting reduction of computational complexity while still keeping sustainable signal quality are highly desirable.
The NP-complete optimum maximum a posteriori sequence detector can be approximated by an iterative receiver, performing channel estimation and parallel interference cancelation (PIC) followed by linear minimum mean square error (LMMSE) detection. The two LMMSE ﬁlters required for channel estimation and for multi-user detection are a common source of complexity in scientiﬁc computing. This work investigates two approaches, which aim at further reducing complexity while avoiding strong performance degradation.
The ﬁrst approach explores linear detection, using Krylov subspace methods to approximate the output of an LMMSE ﬁlter. An LMMSE ﬁlter is derived for detection in the user space, i.e. using matched ﬁltering prior to PIC. Combined with the Krylov subspace method, this allows joint antenna detection while achieving considerable complexity reduction.
A second approach investigates non-linear maximum likelihood multi-user detection. Focus is put on its low-complexity implementation by means of a sphere decoder. Making use of the channel basis expansion, an eﬃcient implementation of a sphere decoder which is more suitable to fast varying channels is developed. Applied to computations of log-likelihood ratios, the reduced-rank sphere decoder results in a receiver achieving one magnitude less complexity at no performance loss.
This thesis aims at presenting a fair comparison between these two approaches in order to facilitate determining a trade-oﬀ between complexity and an acceptable performance degradation, depending on the system parameters and the available hardware.
Electronic version of the publication:
Created from the Publication Database of the Vienna University of Technology.