^{1}

^{*}

One of the limitations of using OFDM technique is its higher PAPR in the time domain signal. The higher PAPR OFDM signal would cause the fatal degradation of BER performance and undesirable spectrum regrowth in the nonlinear channel. One of the promising PAPR reduction methods for OFDM signal is the Partial Transmit Sequence (PTS) method which can achieve better PAPR performance with reasonable computation complexity. However the PTS method is required to inform the phase coefficients of PTS as the side information to the receiver for the correct demodulation of data information through the data or separate channels. To simplify the transceiver of OFDM system with the PTS method, the phase coefficients of PTS are usually embedded in the data information. However since the phase coefficients of PTS are obtained after the PTS processing only for the data information at each OFDM symbol, it is hard to embed the phase coefficients of PTS in the data information separately without degradation of PAPR performance. To solve this problem, this paper proposes a new PAPR reduction method based on the packet-switched transmission systems in which all the clusters within the certain number of OFDM symbols have the sequential cluster ID numbers embedded in the header of each cluster. The salient features of the proposed method are to reduce the PAPR performance by re-ordering of clusters (ROC) in the frequency domain at the transmitter and to reconstruct the original ordering of clusters by using the cluster ID number demodulated from each cluster at the receiver. This paper also proposes a reduction technique of computation complexity for the proposed ROC method by using the feature of IFFT processing. This paper presents various computer simulation results to verify the effectiveness of the proposed ROC method with the reduction technique of computation complexity.

The OFDM (Orthogonal Frequency Division Multiplexing) technique has received a lot of attention especially in the field of wireless communications because of its efficient usage of frequency bandwidth and robustness to the multi-path fading. From these advantages, the OFDM technique has already been adopted as the standard transmission technique in the wireless LAN [

To solve the PAPR problem in OFDM systems, various PAPR reduction methods have been proposed up to today including such as the Clipping and Filtering method [7,8], Selected Mapping method (SLM) [9,10], and Partial Transmit Sequence method (PTS) [11,12]. One of the promising methods among these methods is the PTS method which can achieve the better PAPR performance with reasonable computation complexity. In the PTS method, each transmission OFDM symbol is divided into several clusters in the frequency domain and all subcarriers in each cluster are weighted by one of the predetermined phase coefficients in the time domain so as to reduce the PAPR performance. Although the PTS method can achieve the better PAPR performance with reasonable computation complexity, this method is required to inform the phase coefficients of PTS to the receiver as the side information through the data or separate channels. To simplify the transceiver of OFDM system with the PTS method, the phase coefficients of PTS obtained after the PTS processing are usually embedded in the data information separately. However since the phase coefficients of PTS are obtained after the PTS processing only for the data information, the PTS method is unable to reduce the PAPR performance both for the data information and phase coefficients of PTS simultaneously. From this fact, the PAPR performance after embedding the information of phase coefficients into the data information would be degraded from the original PAPR performance.

To solve the problem of conventional PTS method as mentioned above, this paper proposes a new PAPR reduction method for the OFDM signal based on the packet-switched transmission systems. In the proposed OFDM system, one packet is composed of all the clusters over the certain number of OFDM symbols in which each cluster has the sequential cluster ID number embedded in the header of each cluster. The reduction of PAPR performance is performed for each packet by ReOrdering of Clusters (ROC) in the frequency domain. The proposed ROC method can reduce the PAPR performance over the certain number of OFDM symbols simultaneously which include both the data information and the cluster ID numbers. At the receiver, the proposed ROC method enables the correct demodulation of data information for each cluster independently, because the ROC method only changes the locations of clusters in the frequency domain over the certain number of OFDM symbols at the transmitter. From this fact, the cluster ID number embedded in the header of each cluster can be identified from the demodulated data information for each cluster. By using the demodulated cluster ID number for each cluster, the original ordering of clusters over the certain number of OFDM symbols can be reconstructed correctly at the receiver. The proposed ROC method is completely different from the PTS method which is required to inform the phase coefficients of PTS as the side information to the receiver separately. This paper also proposes a low computation complexity technique for the ROC method which enables the re-ordering of clusters in the time domain by using the feature of IFFT algorithm.

In this paper, Section 2 presents PAPR properties of OFDM signal. Section 3 presents the proposed ROC method of using the low computation complexity technique. Section 4 presents the various computer simulation results to verify the effectiveness of the proposed method as compared with the conventional OFDM, and finally Section 5 concludes this paper.

Let denotes the frequency domain OFDM signal, where N is the number of FFT/IFFT points and n is the frequency index. The discrete time domain OFDM signal is obtained by taking the N-point Inverse Discrete Fourier Transform (IDFT) for X_{n} as given by the following equation:

where k is the discrete time index, W_{N}=e^{−j}^{2π/N} (known as the twiddle factor of IFFT), and j^{2} = −1. From (1) it can be seen that the time domain signal is generated by summation of complex independent random information data of. From this fact, the distribution of time domain signal has the property of Gaussian which has larger peak amplitude as compared with the single carrier transmission signal. To evaluate the envelop variations of OFDM time domain signal, the peak to average power ratio (PAPR) is usually employed as given by:

Here it should be noted that the discrete time PAPR can be evaluated precisely by taking more than four times oversampling ratio for the OFDM signal [

This section proposes the ROC method [

In the proposed ROC method, one packet is constructed by all the clusters within the certain number of OFDM symbols which have the sequential cluster ID numbers based on the packet-switched transmission systems. The cluster ID number for assigning each cluster is ordered sequentially from the 1-st to the (P∙V)-th when the number of clusters in one OFDM symbol is V and the re-ordering of clusters is performed over P OFDM symbols. In the ROC method, the processing for the reduc-

tion of PAPR performance is conducted over the consecutive P OFDM symbols by changing the order of clusters in the frequency domain.

In the proposed ROC method, the re-ordering of clusters is conducted in the frequency domain as mentioned above. However, when employing the frequency domain approach, the IFFT processing is required for each reordering of clusters in the frequency domain because the PAPR performance is required to evaluate in the time domain. This means that the computation complexity for the proposed ROC method becomes larger to achieve the better PAPR performance, because the computation complexity increases proportionally to the required number of IFFT processing. To solve this problem, this paper proposes a low computation complexity technique for the ROC method which can perform the re-ordering of clusters in the time domain by using the feature of IFFT algorithm.

When the number of clusters in one OFDM symbol is 4 (V = 4) and the re-ordering of clusters is performed over two symbols (P = 2) as shown in Figures 1 and 2, the data information including the cluster ID number for each cluster in the frequency domain is given by the following equation:

where is the data information at n-th sub-carrier of -th symbol (1 ≤≤ 2) for the v-th cluster (1 ≤v ≤ 4). In the proposed method, the time domain signal for each cluster is firstly calculated by the following equation:

where n_{v} (1 ≤ v ≤ 4) is the first sub-carrier number for the cluster v as given in (3). The time domain signal given in (4) corresponds to that the data information for the cluster v is located from the sub-carrier number 0 to M/V − 1 in the frequency domain. By using (4), the time domain signal for all clusters (1 ≤ v ≤ 4) over two symbols (1 ≤≤ 2) can be obtained. By using these time domain signals, the re-ordering of clusters can be conducted in the time domain. When the i-th cluster is changed to the j-th cluster in the frequency domain, its time domain signal can be given by the following equation:

where is the time domain signal for the i-th cluster given in (4) and is the time domain coefficient for changing the location of cluster to the j-th cluster which is given by the following equation:

From (5), it is concluded that the re-ordering of clusters in the frequency domain can be conducted in the time domain by using (6). The required number of computation processing for the proposed method is N times multiplications as shown in (5) which is smaller computation complexity than Nlog_{2}N when using the IFFT processing. This means that the proposed method can reduce the computation complexity by log_{2}N times as compared with that for using the IFFT processing.

As mentioned above, the proposed ROC method can reduce the PAPR performance by using the time domain signal obtained after IFFT processing for each cluster as given in (4). From this fact, the computation complexity required in the proposed ROC method would increase as increasing the number of clusters V. To solve the above problem, this paper employs the sprit-radix DIF-IFFT technique [