Inter national J our nal of Electrical and Computer Engineering (IJECE) V ol. 8, No. 6, December 2018, pp. 5215 5226 ISSN: 2088-8708, DOI: 10.11591/ijece.v8i6.pp5215-5226 5215 New T echnique Combining the T one Reser v ation Method with Clipping T echnique to Reduce the P eak-to-A v erage P o wer Ratio Hajar Abdelali 1 , Smail Bachir 2 , and Mohamed Oumsis 3 1 Lims Laboratory , F aculty of Sciences Dhar Mahraz, Sidi Mohamed Ben Abdellah Uni v ersity of Fez, Morocco 2 Xlim Laboratory , UMR CNRS 7252, Uni v ersity of Poitiers, France 3 Lrit Laboratory , Unit Associated W ith CNRST , URA C 29, Uni v ersity Mohamed V , Rabat, Maroc Article Inf o Article history: Recei v ed Dec 26, 2017 Re vised Jul 3, 2018 Accepted Jul 29, 2018 K eyw ord: OFDM System P APR Reduction W iMAX IEEE 802.16e T one Reserv ation Clipping Conjug ate Gradient ABSTRA CT Nonlinear di stortions and impairment s appear in multicarrier signal with high fluctu- ations when amplified by a Radio Frequenc y Po wer Amplifier (RF P A). This article is a contrib ution to impro ving the T one Reserv ation (TR) method dedica ted to reduc- ing fluctuations in the Orthogonal Frequenc y Di vision Multiple xing (OFDM) mod- ulation. This method e xploits the null sub-carriers in communications standards in order to generate a correction signal used to reduce the Peak-to-A v erage Po wer Ratio (P APR) with ne v ertheless limited performances due to the reduced number of these sub-carriers. F or this purpose, our contrib ution is to combine the TR method with the Clipping (CL) method to im pro v e the P APR reduction without significantly impacting the quality of transmissions. W e also sho w the g ain pro vided by this strate gy compared to the use of classical methods. Experiments using a simulated e xample on a complete W iMax 802 : 16 e transmitter ha v e been made in order to in v estig ate the P APR reduc- tion performances on presence of the non-linear Po wer Amplifier model based on g ain compression response and phase distortion. Copyright c 2018 Institute of Advanced Engineering and Science . All rights r eserved. Corresponding A uthor: Hajar Abdelali, LIMS Laboratory , F aculty of Sciences Dhar Mahraz, Sidi Mohamed Ben Abdellah Uni v ersity of Fez, Morocco. abdelali.hajar@usmba.ac.ma 1. Intr oduction OFDM technique [1] is a v ery popular modulation method and tends to become the most widespread multicarrier modulation in recent radio communication systems, such as W ireless Local Area Netw ork (WLAN), W orldwide Interoperability for Micro w a v e Access (W iMax) and D VB. T oday , it w as chosen for the fourth Generation ( 4 G) mobile communication systems, where the Long T erm Ev olution Adv anced (L TE-A) stan- dard allo ws broadband services with a theore tical transmission rate up to 100 Mbps. The adv antages of OFDM modulation are: its optimal spectral congestion, its fle xibility in the allocation of the sub-carriers and its sim- plicity of implementation using F ourier transforms at the transcei v er . Ho we v er , it presents some disadv antages and dra wbacks such as; OFDM system generates high po wer peaks in t ime-domain signal, termed as Peak to A v erage Po wer Ratio (P APR) [2] results in nonlinear distortions when amplified by the Radio Frequenc y Po wer Amplifier (RF P A) [3],[4]. These distortions lead to a in- band spectral re gro wth and out-of-band spectral re gro wth that result in de graded communication metrics lik e J ournal Homepage: http://iaescor e .com/journals/inde x.php/IJECE Evaluation Warning : The document was created with Spire.PDF for Python.
5216 ISSN: 2088-8708 Adjacent Channel Po wer Ratio (A CPR), Error V ector Magnitude (EVM) and Bit Error Rate (BER). In literature, there are se v eral approaches for P APR reduction lik e CL [5],[6], Clipping-Filtering (CF) [7],[8],[9], Selecti v e Mapping (SLM) [9],[10],[11], P artial T ransmit Sequence (PTS) [9],[10],[12], Block Cod- ing [13], T one Injection [14] and TR [15],[16]. These methods dif fer in their principle and impl ementation, b ut also in their dra wbacks: some of them add side information decreasing data rate and/or bandwidth lik e PTS, others sacrifice BER performances for the benefit of P A ef ficienc y such as CL and CF while some others are ef ficient b ut too dif ficult to implement lik e SLM. In this paper , our contrib ution called (TRCL) is a ne w method based on the combination of tw o meth- ods to reduce P APR, TR [15],[16] with Conjug ate Gradient Optimization algorithm [17],[18] and CL [5],[6]. These tw o methods present a certain complementarity: the TR method based on the addition of signal i s one of the methods ha ving the property of do wnw ard compatibility and impro ving the performance of EVM and BER due to their lo w de gradation and their lo w intak e of supplementary computational comple xity , b ut its perfor - mance remains limited because of the reduced number of free subcarriers, while the CL is v ery simple of those kno wn this day and theoretically allo ws an unlimited clipping of the signal b ut unfortunately with a de gradation of all the communication criteria and a decrease in performance caused by a loss of some information. Despite its limitations, this is the TR method that is preferred in practice because it requires no mod- ification the recei v er . In [19], the author proposed the use of the TR method based on the conjug ate gradient algorithm in order to reduce the v alue of P APR and impro v e the speed of con v er gence. In this paper , we are going to describe and study our TRCL method and compare it with other e xisting methods TR and CL. W e will see belo w ho w our ne w method impro v es the performance in terms of P APR reduction and the communication criteria EVM and BER. The study of the performances in terms of reduction of the P APR is carried out on the standard W iMax - IEEE 802.16e [20]. T o study the impact of this association, simulations in the presence of a Solid State Po wer Amplifiers (SSP A) modeled by a nonlinearity of the g ain and phase, are carried out in order to e v aluate the quality of the transmissions. The paper will be or g anized as follo ws: section 2 will present the Problem position, while section 3 , will discuss tw o methods of P APR reduction: Clipping and T one Reserv ation based on the Conjug ate Gradient Optimization Algorithm, and present the proposed technique based on the conjunction of CL and TR (TRCL). In section 4 , we present the simulation results of our ne w method TRCL compared to the TR and the CL. Finally , we conclude, in section 5 . 2. Pr oblem P osition T o transmit a signal, the OFDM signal model (figure 1) di vides at first a frequenc y range into se v eral independent orthogonal sub-carriers which ha v e the same width of frequenc y band. The representation of an OFDM signal can then be written: c ( k ) = 1 p N N 1 X k =0 C k e j ( 2 nk N ) ; 1 n N (1) where C k is the data symbols with k = 0 ; ; N 1 is the inde x of the sub-carrier , C = [ C 0 ; C 1 ; ; C N 1 ] indicate the carrier v ector data and N is the number of sub-carriers. Subsequently , it applies the algorithm of the in v erse F ourier T ransform (IFFT) on this OFDM signal by the F ollo wing equation: x = Q X (2) where Q is the matrix of transformation IFFT of size N. From equation 1, when all the sub-carriers are weighted at high le v el, the constructi v e sum needs to appear peaks in the temporal domain. This establi shes the major incon v enience of the OFDM systems [1]. T o measure these fluctuations, the P APR, defined as a highest peak po wer to a v erage po wer of one OFDM symbol, such as: P AP R ( x ) dB = 10 l og 10 max 0 t T j x ( t ) j 2 E ( j x ( t ) j 2 ) (3) IJECE V ol. 8, No. 6, December 2018 : 5215 5226 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 5217 Channel co ding Sym b ol map Serial to P arallel IFFT to Serial P arallel Cyclic extension D A C RF tx ADC Timing and sync h Remo v e extension freq Cyclic Serial to P arallel FFT to Serial P arallel Sym b ol demap Deco ding Bits F requency corrected signal Bits OFDM T ransmitter RF rx OFDM Receiv er Figure 1. Model of OFDM transmitter and recei v er for WIMAX In literature, we can find se v eral techniques for P APR reduction [2] which present some i mpro v em ents b ut at the cost of man y incon v eniences. F or e xample, the CL method has in theoretical unlimited performances in term of P APR reduction b ut its application induces systematically errors of transmission (lik e increasing BER and/or Out-Of-Band distortion, etc.). These dra wbacks can be a v oided using the TR method, b ut the use of only nul subcarriers allo ws about 2 to 4 dB of i mpro v ement . Here, we propose ne w technique combining TR and CL methods to impro v e performances by a v oiding the disadv antages of each one. As we pre viously noted, we will cite the adv antages and the incon v eniences of the tw o techniques TR and CL: TR : this technique is one of the methods ha ving the property of do wnw ard compatibility and impro v es the performance of ”EVM” and ”BER” due to their lo w de gradation and their lo w intak e of supplemen- tary computational comple xity b ut it is limited by a restricted number of free subcarriers in the standard, comple xity of implementation, etc. CL : this technique is v ery simple and has an unlimited performances in term of P APR reduction b ut its application induces systematically de gradation of all transmission criteria (BER and/or Out-of-Band distortion) 3. Pr oposed T echnique Unuse d sub c arriers Data sub c arriers t t iFFT P APR Reducer x x + c c De dic ate d sub c arrie rs for c orr e ctive signal (PR T) + A A A t t Figure 2. The combination of both techniques TR and CL (TRCL). Here, we will in v estig ate the possibilities of combining both techniques TR and CL (figure 2). These tw o techniques can be complementary because the TR allo ws to reduce the majority of peaks in an OFDM symbol. Still a fe w samples of the s y m bol that cannot be reduced by this method and which will be af fected by the non-linearities of the circuits such as the po wer amplifier . In this case, we propose to apply the CL method to a v oid an y e xceeding the clipping v alue. Ne w T ec hnique Combining the T one Reservation Method... (Hajar Abdelali) Evaluation Warning : The document was created with Spire.PDF for Python.
5218 ISSN: 2088-8708 3.1. Clipping CL [5],[6] is the simplest method of those kno wn this day which allo ws reducing the v alue of P APR. The objecti v e of this method is to clip all points whose amplitude e xceeds a predetermined threshold A to reduce the po wer v ariation. This procedure is repeated until no point e xceeds this threshold. Equation 4 sho ws the principle of clipping. The simplicity of this method has an impact on the signal reception performance will be reduced because the information remo v ed by clipping are lost. f ( x ) = A:sig n ( x ( t )) if j x ( t ) j > A x ( t ) otherwise (4) 3.2. T one Reser v ation TR [15],[16] is the most recent method based on the addition of a signal to the original signal to reduce the P APR. The Useful sub-carriers are reserv ed for the broadcast of the original signal and the sub-carriers not used are e xploited to generate the signal of cancellation of peak without loss of flo w of data and Reduced the P APR. The generation of the temporal signal for the reduction of the P APR by process TR w as formulated as problem of con v e x optimization which requires a high computational cost of the order of O ( N r N 2 L ) , where N r is the number of reserv ed carriers, N the number of carriers of the OFDM and L is the o v er -sampling f actor . T o solv e this problem, there are four iterati v e optimization algorithms, b ut in this paper we are going to concentrate on one optimization algorithm: Conjug ate Gradient. It is an iterati v e method just lik e the method of the gradient, b ut it uses an algorithm so that the direction of research is optimal. This direction is determined by the gradient at the point X n and the gradient of the pre vious point X n 1 [17]. F or the first iteration, the chosen direction is the ne g ati v e gradient: d 0 = g 0 (5) At the k + 1 th iteration, the point X n +1 is chosen by finding the minimum of the function in the direction pre viously determined to reduce the number of iterations. X k +1 = X k + k :d k (6) k is the conjug ate gradient step, and d k represents the direction of conjug ate gradient. The ne xt direction is determined by the pre vious direction. This method in v olv es combining the pre vious direction d k 1 with the gradient at the point X k to calculate the ne w direction: d k = g k + k :d k 1 (7) There are se v eral v ersions of the method of the combined gradient, the y v ary by the w ay the constant k is calculated. Here is the method of calculation of Fletcher -Ree v es [21]: k = g T k :g k g T k 1 :g k 1 (8) k represents then the report between the standard to the square of the current gradient and the standard to the square of the pre vious gradient. Here is the method of calculation of Polak-Ribire [22]: k = ( g k g k 1 ) T :g k g T k 1 :g k 1 = g T k 1 :g k g T k 1 :g k 1 (9) IJECE V ol. 8, No. 6, December 2018 : 5215 5226 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 5219 3.3. New TRCL Method The number of null subcarriers in a gi v en standard is limited, the TR method reduces the maj ority of peaks in an OFDM symbol, b ut al w ays remains some samples whose amplitude is high and which will consequently be af fected by the non-linearities of the circuits lik e the po wer amplifier . In this case, we propose to apply , after the TR, the CL method in order to a v oid an y e xceeding of the A threshold. Therefore, the first phase is to reduce the P APR, using the TR method which calculates the correcti on signal c, from the null subcarriers, according to the Conjug ate Gradient algorithm which is used to reduce the con v er gence time of the optimization algorithm. The second phase consists in applying a clipping to the same threshold A as the pre vious phase in order to eliminate the remaining peaks 4. Simulation Results In this section, we present the e xperimental e v aluation results of our ne w method TRCL with the standard IEEE 802 : 16 e or “W iMAX” [20]. The follo wing table 1 represents the W iMAX param eters used in simulations to impro v e the P APR reduction of OFDM signals. T able 1. The IEEE 802.16 parameters used in simulation Name of the parameter v alue System OFDM Modulation 64QAM Num of Sub-carriers 512 Sub-carriers Datas 384 Sub-carriers Pilots 42 Sub-carriers GI 86 P AP R tar g et 2 : 17 This standard is a broadband wireless technology of the fourth generation allo wing data transmis sion on long distances (dozens of kilometers until 50km) at high speed 70 Mbits/s and with an optimal security . The systems based on the 802 : 16 e standard is called “Mobile W imax”. The communication by this standard can be realized on Line Of Sight (LOS) or Not Line Of Sight (NLOS). The number of carriers for WIMAX 802.16e can v ary between 2048 , 1024 , 512 and 128 , and there are three types of subcarriers for OFDM: Subcarrier data, Pilot subcarriers and Subcarriers care (or subcarriers zeros). The distrib ution of these sub-carriers v aries accordi ng to the mode of allocation FUSC (Fully Used Sub Channel ization) and PUSC (P artially Used Sub Channelization) and the nature of the link DL (Do wnlink) or UL (Uplink): these tw o modes dif fer by the f act of total or partial use of channels and depend on the need for the users. In this paper , the simulation is car ried out for 512 subcarriers and we consider the Mode sub-channel FUSC in the do wnlink DL by using a F ast F ourier T ransform (FFT) (figure 3). 4.1. P APR r eduction T o mak e an e v aluation on the mo v ement of a signal, there are se v eral criteria to mak e it, among t hese criteria we find the P APR and the CCDF . As we ha v e already seen pre viously that the P APR is the report between the maximum po wer and the a v erages po wer of a temporal signal. Ho we v er , for a better mo v ement e v aluation of a signal , we use the second parameter CCDF which is a Complementary Cumulati v e Distrib ution Function of the mo v ement on an interv al of fix ed time. Thus CCDF is the probability that the v alue of P APR is superior to a fix ed v alue. C C D F = P r ( P AP R P AP R 0 ) (10) Ne w T ec hnique Combining the T one Reservation Method... (Hajar Abdelali) Evaluation Warning : The document was created with Spire.PDF for Python.
5220 ISSN: 2088-8708 Sym b ol mapping Input data X 1 × 384 X 1 × 426 C 1 × 86 X 512 × 1 X 512 × 1 x 576 × 1 x 1 × 576 P APR reduction Serial to parallel S/P Flexible OFDM system P/S BER EVM Ev aluation blo c k Output data Q 512 × 512 512pt. iFFT t P APR Reducer c De dic ate d sub c arrie rs for c orr e ctive signal (PR T) +E -E Pilot insertion S/P , Remo v e GI , FFT , Equalizing and Demapping OFDM demo dulation Add guard in terv al White noise P A Figure 3. OFDM data symbol generation and TRCL scheme for 802.16e standard F or that purpose, for an y signal, the dynamics of this signal will al w ays be superior or equal to 0 dB thus the v alue of P APR will also ha v e a probability superior or equal to 0 dB. The more the CCDF parameter has a lo w v alue, the more the probability of ha ving a high P APR v alue tends to 0 . 0 2 4 6 8 10 12 10 −3 10 −2 10 −1 10 0 PAPR, PR0 dB Probabilite, PR > PR0 CCDF of PAPR with transmission IEEE 802.16e     Original with TR with CL with TRCL Figure 4. CCDF of P APR with dif ferent method to reduce the P APR in IEEE 802.16e In this part, the impact of the studied methods on CCDF will be discussed. The figure 4, sho ws the CCDF curv es in function of P AP R 0 obtained by the dif ferent methods allo wing the P APR v alue reduction (TR, CL, TRCL) and the CCDF of a OFDM classic signal (i.e. without reduction technique of the P APR ”Original”). W e can notice that the combinaison TRCL reduce the P APR from 10 : 6 dB to 4 : 8 dB with a probability of approximately 10 2 . It represents a reduction f actor of about 5 : 8 dB, while CL of fers a reduction f actor of approximately 5 : 5 dB and TR of fers approximately 3 dB. From that, we can conclude that the CCDF of the TRCL and CL methods are comparable and decrease rapidly compared to the others methods (TR and Original). The second results sho wn in figure 5, represents the temporal e v olution of the OFDM signal’ s en v elope for the dif ferent P APR reduction methods. These results are obtained from a unit mean po wer signal with a clipping of normalized amplitude A = 1 : 65 to ha v e a v alue of P AP R tar g et equal to 2 : 17 (see equation 11). A = 10 P AP R tar g et 10 :E ( j x ( t ) j 2 ) (11) It can be noted that the TR (figure 5.a) method lea v es a fe w peaks that e xceed the A , the CL method IJECE V ol. 8, No. 6, December 2018 : 5215 5226 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 5221 0 0.5 1 1.5 2 2.5 T × 10 4 0 0.5 1 1.5 2 2.5 3 3.5 |x| Original Avec TR Avec CL Avec TRCL (a). W ith the dif ferent methods of the P APR reduction. 0 0.5 1 1.5 2 2.5 T × 10 4 0 0.5 1 1.5 2 2.5 3 3.5 |x| Original Avec TR Avec CL Avec TRCL (b). W ithout TRCL. Figure 5. “The temporal e v olution of the en v elope for OFDM signal. (figure 5.b) clipped all the peaks that are higher than A , and for TRCL (figure 5.a) we can sho w that the results obtained are identical to those of CL, b ut with an impro v ement in BER and EVM transmission errors (as we will see later on). In the ne xt section, a nonlinear amplifier and a channel model will be introduced to observ e the de gradation of the CL method and the performance of our TRCL method in terms of BER and EVM. 4.2. P erf ormances in pr esence of nonlinear P A model The emission information in wireless netw orks o v er a great distance such as WIMAX requires the use of po wer amplifiers P A [3], to ensure a complete transmission. The role of P A is to increase the po wer of the RF signal without distorting the signal at its input and without ener gy dissipation when issued to its output. The follo wing equation 12 represents the relationship non-linear between the input signal to a P A and its output: P out = G ( P in ) :P in (12) where P in and P out are the P A po wer of the input and the output signal, respecti v ely , and G is the g ain of the P A depending on the amplifier’ s input po we r . The g ain v alue remains constant for the weak po wers i n the linear zone, it will be called also linear g ain, then it decreases when the po wer increased in the other areas. P A is characterized by tw o inputs-outputs relationships also called transfer characteristics of con v er - sions AM/AM (Amplitude Modulation) and AM/PM (Phase Modulation). The AM/AM represents the v aria- tion of the output signal po wer depending on the input signal po wer (as sho w in figure 6), while the AM/PM represents the phase dif ference between the P A output and input signals depending on the input signal po wer . Ne w T ec hnique Combining the T one Reservation Method... (Hajar Abdelali) Evaluation Warning : The document was created with Spire.PDF for Python.
5222 ISSN: 2088-8708 −20 −15 −10 −5 0 5 10 15 −10 −5 0 5 10 P in P out     P out,sat P in,sat 1 dB compression point Saturation point  Linear zone  Compression      zone  Saturation      zone P in,1dB IBO (dB) OBO (dB) P out,1dB P out mean P in mean Figure 6. AM/AM characteristics The mathematical model of the amplifier P A is based on the most wides pread models, namely , the Rapp model descri bed by the e xpression of con v ersion AM/AM (equation 13) and the Saleh model which is described by AM/PM (as sho w in equation 14) [23]: F AM AM ( j x ( t ) j ) = j x j 1 + j x j A sat 2 p 1 2 p (13) where j x ( t ) j is the instantaneous modulated en v elope of the input signal. W ith A sat = 2 : 1 and p = 3 F AM P M ( j x ( t ) j ) = p : j x j 2 1 + p : j x j 2 (14) with p = pi= 3 and p = 1 . The parameters p and p are used to describe the beha vior of the amplifier . The AM / AM curv e of the P A is di vided into three functioning zones (According to figure 6), the linear zone of the amplifier where the output po wer is proportional to the input po wer according to a g ain of the amplifier which remains constant. After a certain po wer input , the g ain starts to decrease slo wly , this zone is called compression zone. And finally , the saturation zone where the amplifier reaches its maximum output po wer . The point to 1 dB compression of g ain represents the po wer output where the dif ference between the g ain calculated in the linear zone and the real g ain of the P A is w orth than 1 dB. The saturation point is a point where the amplifier reaches the maximum v alue to issue an output signal and its po wer appointed po wer of saturation. Most used amplifiers are non-linear , since the linear amplifiers are v ery e xpensi v e. Ho we v er , the non-linearities of a P A has to impact the e xistence of distortions temporal and frequenc y signals to transmit which are reflected by a distortion of the signal. T o a v oid such distortions, we must tak e a step back in po wer , compared to the point of compression to 1 dB steps. This decline allo ws to w ork in the linear zone as well as the signal is closest to the non-l inear zone of the P A. It is defined by the IBO (Input Back-of f) and the OBO (Ouput Back-of f) which represents the decline in a v erage po wer input and output, respecti v ely , compared to the point of compression to 1 dB (figure 6). In our simulations, we ha v e added noise to simulate a Additi v e White Gaussian Noise channel (A WGN). This is a fundamental characteristic of the transmission channel. This channel noise is modeled by a random signal n ( t ) , whose probability distrib ution follo ws the Gaussian la w . The recei v ed signal: r ( t ) = s ( t ) + n ( t ) (15) IJECE V ol. 8, No. 6, December 2018 : 5215 5226 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 5223 2 3 4 5 6 7 8 9 10 11 10 −4 10 −3 10 −2 10 −1 10 0 EbNo (dB) BER a. BER as a function of SNR with IBO=5     Original with TR with CL with TRCL 2 3 4 5 6 7 8 9 10 11 10 15 20 25 30 35 EbNo (dB) EVM (%) b. EVM as a function of SNR with IBO=5     Original with TR with CL with TRCL Figure 7. EVM and BER as a function of SNR with IBO=5 2 4 6 8 10 10 −3 10 −2 10 −1 10 0 EbNo (dB) BER a. BER as a function of EbNo with IBO=0     Original with TR with CL with TRCL 2 3 4 5 6 7 8 9 10 11 15 20 25 30 35 40 EbNo (dB) EVM (%) b. EVM as a function of EbNo with IBO=0     Original with TR with CL with TRCL Figure 8. EVM and BER as a function of SNR with IBO=0 where s(t)represents the transmitted signal. After the addition of nois e, we first calculated the percentage of Error V ector Magnitude (EVM) which corresponds to t he error v ector’ s amplitude between the origin constellation and the recei v ed constellation. Second, we calculated the Bit Error Rate (BER) which is the ratio between the number of errors bits at reception (Ne) and the total number of transmitted bits (N). In the first figures (7.a and 7.b), we can sho w the e v olution curv es of the EVM and the BER according to the SNR (Eb/No) that v aries between 2 dB and 11 dB for a backw ard mo v ement of I B O = 5 dB, to mak e a comparison between the v arious methods for P APR reduction. F or the lo w SNR v alues ( S N R < 7 ) dB, we can see that all BER and EVM curv es ha v e almost the same beha vior . F or strong SNR ( S N R > 7 ) dB, a first remark concerns the Clipping method which de grades all communication criteria. W e also see that of TR which coincides with TRCL with a non-ne gligible impro v ement of BER and with a slight impro v ement of EVM compared to the other methods. After , the amplifier operate s in the non-linear zone at the point at 1 dB of compression corresponding to a decrease in po wer of zero i nput ( I B O = 0 dB). The curv es in Figure 8 sho w the comparison between the BER and the EVM according to SNR (Eb/No) v arying between 2 dB and 11 dB to mak e a comparison between v arious methods which allo ws t o reduce the v alue of P APR. From the results of these figures, W e can also see that the CL method de grades all the criteria and the TRCL method maintains a good impro v ement when the noise le v el e xceeds E b= N o = 8 dB. W e can notice that our TRCL method al w ays retains the best le v el of performance on both criterias. In the follo wing figure 9, we e v a luated the transmission quality in terms of BER and EVM for each method by v arying the v alue of the amplifier’ s input do wn IBO (IBO v aries between 3 dB and 2 dB in a step 1 dB) with fix ed SNR v alue = 12 dB. W e noted the impro v ements made by applying a clipping after the TR method. Ne w T ec hnique Combining the T one Reservation Method... (Hajar Abdelali) Evaluation Warning : The document was created with Spire.PDF for Python.
5224 ISSN: 2088-8708 −3 −2 −1 0 1 2 10 −4 10 −3 10 −2 10 −1 IBO (dB) BER a. BER as a function of IBO with SNR=12     Original with TR with CL with TRCL −3 −2 −1 0 1 2 10 15 20 25 30 IBO (dB) EVM (%) b. EVM as a function of the IBO with SNR=12     Original with TR with CL with TRCL Figure 9. EVM and BER as a function of the IBO −30 −20 −10 0 10 20 30 −80 −70 −60 −50 −40 −30 −20 −10 0 frequency, MHz power spectral density Transmit spectrum OFDM (based on 802.16e)     Original with TR with CL with TRCL mask Figure 10. T ransmit spectrum OFDM based on 802.16e without amplifier Simulation results in figure 9.a sho w that when the IBO increases, the BER tends to w ard zero with the TRCL method. It is the proof that there is less disruption. In ef fect, an IBO lo w engenders significant distortions in the signal. W e can also observ e in figure 9.b that TRCL impro v es EVM by about 3% when the IBO increases, that is to say that, when the amplification is done more and more in the linear area. Ho we v er , lo w v alues of the IBO mean that the po wer amplifier operates on the limit of its saturation zone and will cause distortions in the amplified signal. Figure 10 and 11 sho w the signal spectrum before and after the use of methods which reduce the P APR, for a bandwidth of 30 MHz. W e can see that the application of TR and TRCL method lead to a spectrum broadening of the signal, this is due to the use of the 64 free sub-carriers of the standard. It is v erified on the spectra input and output of the amplifier . In addit ion, at the output of the amplifier , we can see the spectral ascents in t he bands due to nonlinear ef fects. Ho we v er , these spectral lifts stay belo w the IEEE 802 : 16 standard mask. −30 −20 −10 0 10 20 30 −80 −70 −60 −50 −40 −30 −20 −10 0 frequency, MHz power spectral density Transmit spectrum OFDM (based on 802.16e)     Original with TR with CL with TRCL mask Figure 11. T ransmit spectrum OFDM based on 802.16e with amplifier IJECE V ol. 8, No. 6, December 2018 : 5215 5226 Evaluation Warning : The document was created with Spire.PDF for Python.