PAPR Reduction of SC_FDMA Based on modified Tone Reservation Method

Neelam Dewangan
Chattrapati Shivaji Institute of Technology,Durg(C.G) CSVTU, Bhilai neelamdewangan@csitdurg.in

Suchita Chatterjee
Chattrapati Shivaji Institute of Technology,Durg(C.G) CSVTU, Bhilai Suchitachatterjee@csitdurg.in

Abstract—In wireless communication systems, peak–to-average power ratio (PAPR) is a performance measurement that is indicative of the power efficiency of the transmitter. In this paper, an improved method based on tone reservation method for PAPR reduction was proposed. To improve PAPR reduction performance, the more appropriate magnitude of the time domain PAPR reduction signals was found by the proposed method. Simulation in single carrier frequency division multiple access (SC-FDMA) system demonstrates that PAPR reduction performance of the proposed method is better than that of the original approach when several PAPR reduction subcarriers were used. Keywords-PAPR; Tone Reservation; SC-FDMA

II.

PAPR OF SC-FDMA SIGNALS AND TONE RESERVATION METHOD

A. PAPR of SC-FDMA Signals SC-FDMA can be regarded as discrete Fourier transform (DFT) spread OFDMA, where time domain data symbols are transformed to frequency domain by DFT before going through OFDMA modulation [3].

I.

INTRODUCTION

Peak-to-average power ratio (PAPR) is indicative of the power efficiency of the mobile terminal. High PAPR at the transmitter can cause the transmitter high power amplifier (HPA) to go into non-linear regions and degrade the transmitter power efficiency. And low PAPR means longer battery life and cheaper output power amplifier for the same average transmitted power. To get lower PAPR, many PAPR reduction approaches for orthogonal frequency division multiplexing (OFDM) have been proposed [1], which can also be used to reduce PAPR in orthogonal frequency division multiple access (OFDMA) system, such as amplitude clipping, clipping and filtering, tone reservation (TR) and so on. In these methods, TR method is based on adding PAPR reduction signals to the data signals to reduce its peaks. TR method does not only eliminate the need for side information but also prevents the bit error rate (BER) degradation. In this paper, an improved method based on TR method for PAPR reduction is proposed. To improve PAPR reduction performance, the more appropriate magnitude of the time signals used to reduce PAPR was found by the proposed method. Simulations in single carrier frequency division multiple access (SC-FDMA) system demonstrate that the proposed method has better PAPR reduction performance than the original approach in [2]. The rest of this paper is organized as follows. In Section II, PAPR of SC-FDMA signals and tone reservation method are reviewed. Section III introduces the approach for PAPR reduction in [2]. Then an improved method was proposed in Section IV. The simulation results and performance analysis are shown in Section V. Finally, conclusions are given in Section VI.

xm
Figure 1. Block diagram of SC-FDMA transmitter. x m ( m = 0,1,∀, M − 1 ) are time domain symbols after SC-FDMA modulation. M is the total number of subcarriers, and N ( N < M ) is the number of subcarriers occupied by the input data.

Fig. 1 shows the block diagram of an SC-FDMA transmitter, where M is the total number of subcarriers, xm ( m = 0,1,∀, M − 1 ) are time domain symbols after SCFDMA modulation, and N ( N < M ) is the number of subcarriers occupied by the input data. SC-FDMA transmitter groups the modulation symbols into blocks each containing N symbols. Then each time domain data block was transformed to a frequency domain data block by an N-point DFT. Next it maps each of the N-point DFT outputs to one of the M orthogonal subcarriers. Then an M-point IDFT transforms subcarrier mapping outputs into a complex time domain signal xm. The cyclic prefix (CP) is a guard time to prevent intersymbol interference (ISI) due to multipath propagation. PAPR was widely used to evaluate envelope fluctuations of signals. PAPR of SC-FDMA signals without pulse shaping with symbol rate sampling can be expresses as follows [4]. max xm
2 2

PAPR =

m = 0,1,∀, M −1 M −1

1 M

(1)

m= 0

xm

TR method does not need any side information and any receiver operation.∀. x[n2 ] ≤ ≤ ≤ IV. k is the subcarrier index.∀. So where N is the total number of subcarriers. The magnitude of c ( c ≥ c ) was calculated out by { } . ˆ because the magnitude of c is larger than that of c . (2) ∠c [ n] = = ∠c [ n ] + ( ∠x [ n1 ] + π − ∠c [ n1 ]) ∠c [ n] + ( ∠c [ n1 ] − ∠c [ n1 ] ) (6) Lastly. N − 1 . the magnitude ˆ (9) c= c of c was set as (5) and the phases of c [ n] were calculated 1 2π kn C k exp j N N out by (6). The time domain PAPR reduction vector c is computed out as follow. c [ n1 ] must be modified into an opposite phase of x [ n1 ] . c has a constant the improved method has better PAPR reduction magnitude and a time varying phase for all n. c = c [ n ] = c [ n1 ] = x [ n1 ] − x [ n2 ] 2 ˆ ˆ ˆ ˆ And let θ [ nl ] = θ [ n1 ] . respectively. the frequency domain PAPR reduction vector C has only one nonzero element Ck. where nl is a time index of the l-th maximum peak. and x is the corresponding time domain data vector of the frequency domain data vector X. When add c to x. III. θ [ n l ] can be expressed as follow . we use one reserved subcarrier to generate time magnitude and the phase of c must be modified.1. θ [ nN ] be a set of the ˆ ˆ (5) phase differences between c [ n l ] and x [ nl ] . let c be the modified c. Next. Both the transmitter and the receiver agree on reserving a small subset of tones for generating PAPR reduction signals. ♣ X k = 0. i. (4) time index of the l-th maximum peak. And the first maximum peak magnitude x [ n1 ] was suppressed as x [ n1 ] = x [ n1 ] − c [ n1 ] = x [ n1 ] + x [ n2 ] 2 < x [ n1 ] . Let c be the time domain PAPR reduction vector which was generated by the original method.e. The phase of domain PAPR reduction vector c by the original method. the new time domain vector x can be represented as x = x + c = IDFT ( X + C ) . where nl is a ˆ c = c [ n] = n = 0. (3) And the second maximum peak magnitude x ( n2 ) will be never larger than the first maximum peak magnitude x ( n1 ) . First.B. we propose an improved method based on the original method in [2]. k∈S k∉S . THE O RIGINAL APPROACH BASED ON TR METHOD ˆ the phase differences between c [ nl ] and x [ nl ] . the time domain PAPR reduction vector c above has been generated by the original method in [2]. whose magnitude and phases have been modified by (5) and (6). To reduce the first maximum peak magnitude x [ n1 ] . ˆ and c be the more appropriate time domain PAPR reduction vector which was generated by the proposed improved ˆ ˆ method. When TR method is used. Since c has a constant c magnitude and a time varying phase for all n. ˆ ˆ θ [ nl ] = ∠c [ nl ] − ∠x [ nl ] (10) The approach based on TR method in [2] uses one reserved subcarrier to generate time domain PAPR reduction signals for reducing the first maximum peak magnitude of time domain data vector x. THE PROPOSED IMPROVED METHOD In this section. ♦ ♥ C k = 0. By adding the c to x. Those reserved tones are not used for data transmission. Then normalize c as follow. X and C must lie in disjoint frequency subspaces. (7) where c is the corresponding time domain PAPR reduction vector of the frequency domain PAPR reduction vector C. the first maximum peak magnitude x [ n1 ] was suppressed further. performance. the First. θ [ n2 ] . Tone Reservation Method Tone reservation (TR) method is based on adding time domain PAPR reduction signals to the time domain data signals to reduce its peaks. generate the new time domain data vector x which has lower PAPR. x [ n2 ] + c [ n2 ] x [ n1 ] + x [ n2 ] 2 x[n1 ] (8) where k is the subcarrier index and S is a set of subcarrier indices for data transmission.

then ˆ ˆ x [ n1 ] = x [ n1 ] + c [ n1 ] − 2 x [ n1 ] c [ n1 ] . 3. SIMULATION RESULTS ˆ Then N − 1 candidate magnitudes of c will be found by (15). l = 2. since c has a constant magnitude. N can be derived from (11). (19) l = 2.(12) ˆ When l = 1 . Lastly. CCDF = 1 − Pr ( PAPR ≤ PAPR0 ) .3.e. θ [ n1 ] = π .∀. But the maximum PAPR reduction gain for the proposed method is the same as that for the original method. ˆ even though the phase ∠c [ nl ] is the same to ∠x [ nl ] for ˆ l = 2.∀.e. 10 4 uniformly random data points were generated to acquire the CCDF of PAPR which is calculated by Monte Carlo simulation. the magnitude ˆ of ( x [ n1 ] + c [ n1 ]) will be not larger than that of ˆ So the new vector x + c has lower PAPR than x + c . input data .3. x [ n1 ] − x [ nl ] 2 2 ˆ c l = 2.∀.∀. respectively.3.∀. 3.e. N (the worst case). (13) ˆ c= c ˆ ˆ c exp j ( ∠c [ n1 ] − ∠c [ n1 ]) c ˆ So the optimal magnitude of the PAPR reduction vector c ˆ can be set as (16) after derivation. c must be as large as possible. ˆ c = = min ˆ ( c[n ] ) l Complementary Cumulative Distribution Function (CCDF) of PAPR. (15) V.∀. The total number of subcarriers M was set to 256. 2 2 2 set to an opposite phase of x [ n1 ] . N be maximums.∀. N l = 2. x [ n1 ] − x [ nl ] 2 2 { } } . (14) where c and c is calculated out by (4) and (16). Then the appropriate time ˆ domain PAPR reduction vector c could be calculated out as follow. x [ n1 ] − x [ nl ] 2 2 ( x [ n ] + c [ n ]) . ˆ[n c l] = ˆ 2 x [ n1 ] + x [ nl ] cos θ [ nl ] { ( )} . θ [ nl ] = 0 .∀.3. x [ nl ] ≤ x [ n1 ] 2 2 ˆ And c calculated out by (16) will be never smaller than c . the optimal magnitude ˆ of c must be the minimum of N − 1 candidate magnitudes. And the phase of c [ n1 ] is ˆ Next. 3. Simulations . ˆ ˆ Because c has larger magnitude than c . i. ˆ[n So let c l ] for l = 2.02dB [2]. N . i. N .∀. Subcarrier mapping modes is localized modes. 1 1 ˆ To reduce x [ n1 ] further. (16) are according to SC-FDMA system model in [4]. (12) and (13).e. i. N min x [ n1 ] − x [ n2 ] 2 c ˆ ˆ ˆ = x [ nl ] + c [ nl ] + 2 cos θ [ nl ] x [ nl ] c [ nl ] 2 ( ) . N min 2 x [ n 1] + x [ n l] x [ n1 ] − x [ nl ] 2 { } . which is 6. which is defined by (19). is the probability that PAPR is higher than a certain PAPR value PAPR0. (11) ≥ l = 2. The CCDF curve which is closer to the vertical axis has the better PAPR characteristic.∀. (18) c ˆ c exp j ( ∠x [ n1 ] + π − ∠c [ n1 ] ) = c { ˆ[n c l] ≤ ˆ 2 x [ n1 ] + x [ nl ] cos θ [ nl ] { ( )} . l = 2. i. N . 3. i. then ranges of ˆ[n c l ] for l = 2. to make sure (11) for all l.Both original and proposed methods are to suppress x [ n1 ] for PAPR reduction and to prevent other peak magnitudes larger than x [ n1 ] after PAPR reduction. (17) where ˆ x [ nl ] = x [ nl ] + c [ nl ] 2 2 2 ≥ ≥ ≥ l = 2. So it is very important that x [ nl ] must be always smaller than the x [ n1 ] in both methods. 3.e. N min ˆ 2 x [ n 1] + x [ n l] cos θ [ n l] { x [ n1 ] − x [ nl ] 2 2 ( )} Let method I be the original method in [2] and method II be the proposed improved method in this paper. 3. since the phase of c [ n1 ] must be an opposite ˆ phase of x [ n1 ] .

REFERENCES [1] T. 2009 Myung. On Broadcasting. And 16QAM symbol constellations were considered. Jun. “An Approach for PAPR Reduction Based on Tone Reservation Method”. vol. Sept. the result of reducing PAPR is better. no. pp.30-38. 2 illustrates that the PAPR reduction performance comparison of two methods for 16QAM. “Single carrier FDMA for uplink wireless transmission”.21 dB reductions. vol. respectively. “Peak-to-average Power Ratio of Single Carrier FDMA Signals with Pulse Shaping”. Wu. 2008.70 dB and 7. 3. no. When both methods are used. respectively... The total number of subcarriers M was set to 256. IEEE Trans.50 dB reductions. Comparison of the PAPR reduction performance for one PAPR reduction subcarrier and three PAPR reduction subcarriers. [3] In Fig. resulting in 1. D. CONCLUSION In this paper. it is confirmed that the PAPR reduction performance of method II is better than that of method I. 3. 3 and 5. Yunmok Son. 2. 257-268. 10-13 Jan. Simulation results in SC-FDMA system shows that the PAPR reduction performance of the proposed improved method is better than that of the approach in [2] for using only one subcarrier and several subcarriers. IEEE Vehicular Technology Mag. 11-14 Sept. [4] . and Y.60 dB more PAPR reduction than method Iat the CCDF of 10-4.. 4 illustrates that the better PAPR reduction effect will be not obvious after several iterations because the magnitudes of peak signals will be closer and closer to each other. pp. Indoor and Mobile Radio Communications. and Goodman D. Jiang. IEEE Consumer Communications and Networking Conference.J. Chang Hyun Nam.G. an improved method for PAPR reduction based on the original approach in [2] was proposed. The proposed method improves PAPR reduction performance by finding out the optimal magnitude of PAPR reduction signals.block size N to 64 and the number of PAPR reduction subcarriers i to 1. 2. input data block size N to 64.. 54. PAPR for method I and that for method II reduce approximately to 7.. Figure 2. and Hwang Soo Lee.04dB and 1. “An Overview: Peak-to-Average Power Ratio Reduction Techniques for OFDM Signals”. Comparison of the PAPR reduction performance for method I and method II in SC-FDMA system using 16QAM modulation. Fig. H. When i = 3 . Fig. input data block size N to 64 and the number of PAPR reduction subcarriers i to 1. Comparison of the PAPR reduction performance for three PAPR reduction subcarrier and five PAPR reduction subcarriers in SC-FDMA system using 16QAM modulation. J. Junsung Lim. resulting in 0.32dB and 6. when i = 1 . IEEE International Symposium on Personal. method II achieves approximately 0. 2006 [2] Figure 3. and Goodman. PAPR for method I and that for method II reduce approximately to 7. From Fig.66 dB and 1.86dB at the CCDF of 10-3. The total number of subcarriers M was set to 256. respectively. Fig. Figure 4.15 dB at the CCDF of 10-3. G. 3 shows that if PAPR reduction subcarriers are more. 2006 Myung H. 1. Junsung Lim. VI.