Microsoft Word - ICC06.DOC

This paper presents a new algorithm for frequency offset estimation for Polynomial Cancellation Coded Orthogonal Frequency Division Multiplexing with symbols overlapped in the time domain (Overlap PCC-OFDM). The algorithm exploits the Subcarrier Pair Imbalance (SPI) caused by frequency offset. The estimation is performed in the frequency domain. No training symbols or pilot tones are required. Simulations show that this estimator is an approximately linear function of frequency offset. There are three ways to reduce the variance of the estimation: increasing the number of subcarrier pairs, using a two-dimensional Minimum Mean Square Error (MMSE) equalizer before the estimation and using PCC-OFDM pilot symbols with no overlapping. 1. Overlap PCC-OFDM Polynomial Cancellation Coded Orthogonal Frequency Division Multiplexing (PCC-OFDM) was designed to reduce the Interchannel Interference (ICI) caused by frequency offset [1]. In PCC-OFDM, each data value to be transmitted is mapped onto a group of subcarriers. In this paper, the case where each data value to be transmitted is mapped onto pairs of subcarriers is considered. Despite its advantages, PCC-OFDM is not bandwidth efficient in its simplest form [2]. One way to overcome this drawback is to overlap PCC-OFDM symbols in the time domain [3]. Figure 1 shows the procedure of symbol overlapping. With overlapping, each overlapped symbol consists of three parts, the current PCC-OFDM symbol, the second half of the preceding PCC-OFDM symbol and the first half of the following PCC-OFDM symbol. Fig. 1 PCC-OFDM symbols overlapped in the time domain Figure 2 shows a simplified block diagram of an Overlap PCC-OFDM communication system. The high speed data stream is fed into a serial to parallel converter and converted into n lower speed parallel substreams. The th i vector to be transmitted is represented by i n i d d , 1 , 0 , , − . They are mapped onto the values i N i a a , 1 , 0 − that modulate the N subcarriers in the th i symbol period. For conventional OFDM, which has N n = and i k i k d a , , = , there is a simple one-to-one mapping of data values onto the subcarriers. For PCC-OFDM, n and N are not equal. In this paper, we have 2 N n = . In the channel, the signal is filtered by the channel response ( ) t h and additive noise ( ) t n is injected. At the receiver the th i output vector of the DFT is i N i z z , 1 , 0 − . The demodulated subcarriers are then weighted and added to generate the data estimates i n i v v , 1 , 0 − . For the subcarrier pair i M z , 2 and i M z , 1 2 + , an estimate is calculated using ( ) 2 , 1 2 , 2 , i M i M i M z z v + − = . To recover the transmitted data values from the overlapped symbols, a twodimensional Minimum Mean Square Error (MMSE) frequency domain equalizer can be used [3]. Fig. 2. Block diagram of a PCC-OFDM system In [4], a blind frequency offset estimator has been presented for PCC-OFDM. As shown in [5], the frequency offset estimator for PCC-OFDM can also be used for PCC-OFDM with symbols overlapped in the time domain. Recently, a new MMSE frequency 2 T 0 2 T T 2 3T T 2 Time To MMSE Equalizer Transmitter Receiver Divide into low bit rate streams Parallel to Serial overlap DAC and filtering N point IDFT Mapping data onto subcarriers High speed data i b , 0