Dft-s-ofdm Detection Technology

3GPP's long time evolution (LTE) is targeted toward the next decade's cellular communication standard. LTE's design requirements and system architecture have changed a lot compared with the current 3G system, especially for physical layer. LTE adopts OFDM as the downlink transmission scheme, whereas for uplink, it uses a single carrier scheme named DFT-S-OFDM for the consideration of peak to average power ratio (PAPR). Although DFT-S-OFDM has similarities with OFDM, its detection schemes, resource allocation methods are different from OFDM, of which the detection scheme is the focus in this thesis.With the increasing complexities of the modern wireless communication system, often the performance metric uses the whole system output instead of an intermediate module's output. Therefore, how to simulate and validate becomes a problem. We refer to LTE's specifications and proposals of 3GPP RAN4 WG to construct a link level simulator. And then use it as the platform to validate the proposed algorithms.Based on this platform, the problem of how to perform soft decision under receiver diversity is studied. The existed literature has not fully considered the receiver diversity and the equalization effect on the signal. Therefore, we propose a new DFT-S-OFDM soft decision scheme:firstly extending the results under SISO antenna configuration to receiver diversity case; secondly applying the statistics after equalization to the soft decision. Moreover, for the non-ideal channel estimation condition, the effectiveness of the proposed algorithm is validated by taking the channel estimation error into account when deriving the formulae. Simulation results show that for receiver diversity, receiver diversity with non-ideal channel estimation and SISO scenarios, with different modulation type and code rate combinations, the proposed scheme has 0.5～1.6dB performance gain.Also, the problem of how to detect the DFT-S-OFDM signal using the local ML method under the uncoded system is studied. A modified MMSE-ML algorithm is proposed, by finding the second probable constellation symbol position after MMSE-ML search, and then applying the ML search on these two positions. Simulations are carried out using QPSK and 16QAM modulation types, for the modified MMSE-ML, the 1st order MMSE-ML, the 2nd order MMSE-ML and the linear MMSE method to show their contrasts. Simulation results show that compared with the 2nd order MMSE-ML, the proposed method has little performance loss with huge amount of computational complexity reduction. And with little additional computational complexity, the proposed method has a 0.5-0.7dB gain over the 1st order MMSE-ML.