|M.Sc Student||Avrashi Gilad|
|Subject||Time Varying Carrier Frequency Offset Estimation in|
Multicarrier Underwater Acoustic Communication
|Department||Department of Electrical Engineering||Supervisors||Professor Israel Cohen|
|Dr. Alon Amar|
|Full Thesis text|
In this work we consider a group of closed-form carrier frequency offset estimators for multicarrier underwater acoustic communication. Acoustic modems are used for various underwater platforms including environmental monitoring beacons, and autonomous underwater vehicles (AUVs). With a growing need for applications such as image and video streaming, a considerable amount of attention has been given to high bandwidth communication schemes which can handle the dynamic underwater acoustic channel. In the past decade, multicarrier communications have been extensively used for these applications.
Carrier frequency offset (CFO) in orthogonal frequency division multiplexing (OFDM) communication systems may cause inter-carrier interference and degrades the performance of OFDM decoders. Over the past decades, numerous techniques were proposed to estimate frequency offsets in OFDM, focusing mainly on radio communication channels which are quite different from the acoustic channel. The time variations of the underwater acoustic communication (UAC) channel are non-negligible with respect to the propagation speed and are subject to multipath effects, ultimately causing non-uniform Doppler shifts and short coherence time. These variations have encouraged UAC modem designers to use block-by-block frequency estimators which require a grid search in the frequency domain and thus suffer from high computational complexity.
Recently, low complexity CFO estimators for OFDM in the UAC channel have been
suggested that replace the need for an exhaustive grid search. Using equi-spaced pilot
tones in the frequency domain and setting them to be identical, results in a periodic
time-domain block with a period equal to the number of pilot tones. By utilizing this,
the problem of estimating the CFO is shown to have a closed-form and can be solved
effectively given that it is constant during one block. These estimators, however, suffer from two drawbacks making them hard to implement in practical modems: (1) The use of equi-spaced identical pilots results in high peak to average power ratio (PAPR); (2) The assumption of having a constant CFO limits the design of the OFDM block.
In this work, we propose a transmitter-receiver design that allows closed-form CFO
estimation. Instead of identical
pilots, we propose a method for designing the pilot tones, such that the time
periodicity feature is preserved while the e
ect on PAPR is negligible. By looking at pilot design as a phase retrieval problem with a time-domain envelope chosen to satisfy the low PAPR requirement, we are able to derive a tunable design algorithm for the transmitted signal.
For the receiver side, we develop a closed-form time-varying CFO estimator. The
method expands the previously developed closed-form CFO estimator, which utilizes only the pilot-independent samples of the OFDM block to a complete scheme which also includes the pilot-dependent part. We then develop estimators that can be adapted to time-varying channels. Numerical simulations indicate that the bit error rate (BER) performance of the OFDM modem is significantly improved by using the proposed method. Furthermore, we show that by applying the proposed pilot-design algorithm, PAPR values are within fractions of dB of the original random pilots design without compromising the CFO estimation scheme.