Optimized Training Signal Design
Independent thesis Advanced level (professional degree), 20 credits / 30 HE creditsStudent thesis
In this thesis, the problem of finding an optimal training sequence for estimating a MIMO flat fading channel with spatially and temporally correlated Gaussian noise is considered.
The methods analyzed tailor the training sequence not only according to the known statistical CSI but also to the specific purpose the channel estimate will fulfill. The task of obtaining the optimal training sequence is formulated in two different ways, either guaranteeing a specific performance or setting a maximum training power budget. Two different applications are considered, the ZF precoder and the MMSE equalizer. The performance of the training sequence obtained by minimizing a metric that is representative for this applications is compared to using the training sequence that minimizes the channel estimate mean square error. Additionally, since some approximations are required to solve the optimization problem when using the application-oriented metrics, the impact of these is analyzed.
Two different approximations that allow convexification and lead to SDP formulations are considered for each problem. The two approximations and problem formulations are analyzed in terms of performance, training power and outage probability. The SDP formulations are then compared to the solutions provided by builtin functions in MATLAB that converge to a minimum in order to obtain information about how far from optimal the solutions obtained from the SDP are.
Place, publisher, year, edition, pages
EES Examensarbete / Master Thesis, XR-EE-SB 2014:002
Electrical Engineering, Electronic Engineering, Information Engineering Telecommunications
IdentifiersURN: urn:nbn:se:kth:diva-142371OAI: oai:DiVA.org:kth-142371DiVA: diva2:699996
Subject / course
Master of Science in Engineering - Electrical Engineering
Bengtsson, Mats, ASSOCIATE PROFESSOR