Date of Award

Winter 2007

Project Type


Program or Major

Electrical Engineering

Degree Name

Master of Science

First Advisor

Michael J Carter


Orthogonal Frequency Division Multiplexing (OFDM) has been widely adopted as a modulation format for reliable digital communication over multipath fading channels, e.g. IEEE 802.11g "WiFi" networks, as well as broadband wireline channels, e.g. DSL modems. However, its robustness to channel impairments comes at the cost of increased sensitivity to symbol timing and carrier frequency offset errors, and thus requires more complex synchronization methods than conventional single-carrier modulation formats.

In this thesis, a class of synchronization methods based upon the intrinsic autocorrelation structure of the OFDM signal is studied from a statistical perspective. In particular, the reasons for the existence of irreducible time and frequency offset estimation errors in the limit of increasing signal-to-noise ratio (SNR) are investigated for correlator-based synchronizers for the non-fading channel case and several fading channel models. It is demonstrated that the primary source of irreducible synchronization errors at high SNR is the natural random distribution of signal energy in the cyclic prefix of the OFDM symbol.

Comparisons of the distribution of correlator output magnitude between the non-fading and fading channel cases demonstrates that fading skews the distribution with respect to the non-fading case. A potential mechanism for reducing the effect of innate signal energy variability, correlator output windowed averaging, is studied from the perspective of its influence on the distribution of interpeak intervals in the temporal correlator output signal. While improved performance is realized through averaging for the non-fading channel case, this technique is not as effective for fading channels. In either instance, the windowed averaging method increases the latency of the synchronization process and thus introduces delay in the overall demodulation process.