MASc seminar - Kartik Vamaraju

Monday, July 27, 2015 3:00 pm - 3:00 pm EDT (GMT -04:00)

Candidate

Kartik Vamaraju

Title

Single-Input Multiple-Output Media Based Modulation

Supervisor

Amir Khandani

Abstract

Modulation is traditionally based on the idea that signal constellations should be deterministically constructed and known by the transmitter and receiver. Communication involves the transmitter randomly selecting from this known finite constellation. Media Based Modulation (MBM) is a departure from this paradigm. Tunable mirrors at the transmitter are used to establish independent channel realizations, the selection of which is used to encode information. The constellation is a function of the fading statistics of the environment and is therefore unknown at the transmitter and must be estimated by the receiver.

The main topic of this seminar is communication involving a Single-Input Multiple-Output (SIMO) MBM system communicating over a static Rayleigh fading channel when there is perfect Channel State Information (CSI) at the receiver. The geometric properties of MBM constellations are such that the performance of MBM systems strongly depend on the number of receiving antennas. Simulation results are presented to demonstrate the substantial performance gains experienced when additional receiving antennas are added to an MBM system when the initial number of receiving is low. The potential for MBM systems communicating over static Rayleigh fading channels to outperform SBM systems communicating over Additive White Gaussian Noise (AWGN) channels is also shown. Moreover, with sufficient receiving antennas it is possible not only to employ constellation sizes that are much larger then what is used in practice with negligible energy loss, but also in a regime where the noise power is greater then the energy per bit.

The application of channel coding to MBM is then discussed. Generally, the lack of Channel State Information (CSI) at the transmitter significantly hinders the capability of the code designer. Simulation results involving the application of a Single Parity Check (SPC) symbol code applied to an MBM system are presented. The geometry of MBM constellations has a significant impact on coding gain, and in particular the coding gain experienced with MBM systems communicating over a static Rayleigh fading channel can be substantially larger then what is experienced with the same code applied to an SBM system communicating over an AWGN channel.

Finally, algorithms are presented to solve the Maximum Likelihood (ML) symbol detection problem for MBM constellations. The inherent randomness of MBM constellations prevents the use of lattice decoding algorithms which are used by many SBM systems. A novel ML symbol detection algorithm is developed using ideas from sphere decoding and shown empirically to be significantly faster then brute force search. An approximate ML symbol detection algorithm is also presented in which the search radius is determined by the L-infinity norm and the optimal candidate selection is determined by the L-2 norm. Simulation results demonstrate a noticeable reduction in search effort by using the approximate algorithm in exchange for manageable energy losses.