Discrete Cosine Transform for Pre-coded EGPRS

Abstract: Due to the rapid growth of wireless multimedia service, new features such as improved spectral efficiency, latency and increased peak throughput are standardized in the current GSM/EDGE system. Higher order modulations and symbol rates are defined in Enhanced General Packet Radio Service phase 2 (EGPRS2) to achieve better throughput. However, the computational complexity in the traditional receiver can be very high when increased modulation orders are applied; additionally, the system becomes more sensitive to the impairments at an increased symbol rate. It is therefore desirable to have a less complex and more robust demodulator. Recently, a new study item which introduces the multi-carrier technique was proposed in the 3rd Generation Partnership Project (3GPP) standardization. Based on the channel partitioning using the Discrete Fourier Transform (DFT), a simple equalizer can be used, which greatly reduces the computational complexity on receiver, meanwhile achieves good throughput and robustness against impairments. In this thesis, another channel partitioning method by means of the Discrete Cosine Transform (DCT) is studied. Transmitter and receiver algorithms are developed, including a pre-filter designed at the receiver to facilitate the channel diagonalization. The link level performance is evaluated by means of simulations, under different test scenarios. The system’s robustness against impairments and peak-to-average ratio (PAR) reduction are also evaluated, and compared with a system based on the DFT pre-coding. From the simulations, the conclusions can be drawn that in this implementation, the DFT precoded EGPRS2 outperforms the DCT pre-coded EGPRS2 in all scenarios. The DCT pre-coded system also shows worse robustness against impairments and higher peak-to-average ratio reduction loss in throughput. The impact of pre-filter design on the DCT pre-coded system has also been analyzed, and it shows that there is a tradeoff between achieving good symmetrization, and preserving channel information in the frequency domain.