Blind Turbo Demodulation for Differentially Encoded OFDM with 2D Trellis Decomposition

TL;DR

The paper addresses pilot-free, DAB-like OFDM where blind estimation of channel phase, gain, and noise is essential for turbo-DE-PSK demodulation. It introduces a fully blind approach that combines a 2D trellis-based joint phase and symbol detection with power- and null-tone-based estimators for $G$ and $\sigma^2_w$, respectively, and frames the problem as a unified hidden Markov model. The key contributions are (i) a 2D trellis decomposition that separates phase estimation from symbol demodulation across time and frequency, (ii) a discretized phase hypothesis framework with $L$ levels embedded in the trellis, (iii) a 2D gain/variance estimator leveraging 2D blocks and null tones, and (iv) extensive simulations showing performance approaching that of receivers with perfect CSI under realistic DAB-like formats. The results demonstrate robustness to practical time-varying channels and provide insights into the trade-offs among inner-code length, phase quantization, and 2D block size, indicating strong potential for pilot-free OFDM broadcast systems.

Abstract

Digital Audio Broadcasting (DAB)-like systems employ differentially encoded (DE) phase-shift keying (PSK) for transmission. While turbo-DE-PSK receivers offer substantial performance gains through iterative decoding by making the DE-PSK an inner code, they rely on accurate channel estimation without pilots, which is a key challenge in DAB-like scenarios. This paper develops a fully blind turbo-DE-PSK scheme that jointly estimates channel phase, channel gain, and noise variance directly from the received signal. The design leverages a two-dimensional (2D) trellis decomposition for blind phase estimation, complemented by power-based estimators for channel gain and noise variance. We provide a comprehensive system assessment across practical system parameters, including inner code length, phase quantization, and 2D block size. Simulation results show that the blind 2D turbo demodulator approaches the performance of receivers with perfect channel knowledge and remains robust under realistic transmission conditions.

Figures (7)

• Figure 1: System block diagram of a convolutionally encoded, bit-interleaved PSK mapper with differential encoding and OFDM modulation at the transmitter side. The receiver comprises an OFDM demodulator, channel estimator, and an iterative demodulation/decoding module.
• Figure 2: DE-PSK symbol organization before IDFT. $\varnothing$ denotes the null tones used to avoid interference.
• Figure 3: Trellis constructions for the (a) optimal DQPSK symbol demodulator and the (b) joint channel phase and DQPSK symbol demodulator with $L=32$.
• Figure 4: Received signal after DFT, null tone removal, and cyclic prefix removal. The red rectangular box indicates the 2D blocks used for channel estimation and demodulation.
• Figure 5: BER performance over iterations (IT) of the optimal receiver design for inner code length $N=4$ (dotted lines) and $N=10$ (solid lines) over the AWGN channel.