International Journal of Engineering Trends and Technology (IJETT) - Volume4Issue4- April 2013 A Novel Phase Model for Nakagami-m Fading Channels Jyotheeswar K M1, C.Subhas2 1 Student, 2 Professor Department of Electronics and Communication Engineering, Sree Vidyanikethan Engineering College, TIRUPATI – 517 102, A. P., INDIA Abstract—The Nakagami-m distribution approximates a Rayleigh for = , Rician for > and Hoyt for . < < distributions. Utilizing the characteristic function (c.f.) of a gamma Random Variable (RV) and by approximating it to that of a Rician and Hoyt RVs, a novel Probability Density Functions (p.d.f.) are obtained for the real and imaginary parts of the complex Nakagami-m fading gain. Keywords—Nakagami-m, Characteristic functions, distribution, Rician distribution, phase model. 1. Hoyt INTRODUCTION With the advent of mobile technology, there has been a rigorous and widespread study on the wireless channels. These studies include modeling the physical phenomena that are characteristic of those channels and mitigating the effects of the channel on the transmitted data. One such important characteristic of a ‘no-wire’ channel is ‘fading’. The fundamental cause for the fading effect is that the transmitting antenna, generally in a mobile handset, is at the ground level. While the receiving antenna may be at a few tens of meters, the presence of buildings, power scatters, and sharp edges, make the received signal fluctuate very rapidly, which is termed as fading. There has been a significant research to model this effect so that mitigating techniques can be easily described and further applied to the fast growing wireless technology. Physical modeling of such an effect offers the flexibility of easy analysis for any user. Modeling, generally, considers overall effect as a black box effect, where the end user, may not have to explicitly consider the whole phenomenon. Hence the user can easily expect what can be the output for a given input. Since, the received signals are totally random in nature, the study of random variables are very much needed. Specifically, probability density functions (p.d.f.s), which give the probability ‘that the random variable taking a particular value’, are required. In recent years, Nakagami-m p.d.f gained much importance as it models well many experimental data. ISSN: 2231-5381 In the literature, there exists p.d.f. for the Nakagami-m distribution [3] and also the phase distribution considered is uniform in nature. In this paper, the real and imaginary parts of a complex Nakagami-m fading gain are derived and from this the phase distribution is derived. The paper is organized as follows. After this introduction, in section 2, the actual characterization of the real and imaginary parts of the fading gain is given. In section 3, the required phase distribution is obtained. In section 4, numerical results are shown to validate the obtained results. 2. CHARACTERIZATION OF REAL AND IMAGINARY PARTS OF COMPLEX NAKAGAMIm FADING GAIN Let a signal transmitted through a wireless channel has undergone Nakagami-m fading with parameter value m and mean square value Ω and let its envelope be R, and in order to obtain the p.d.f. of phase, let Θ be the random phase. Then R ℐ = X + ℐY. (1) where X and Y are the in phase and quadrature components of R ℐ If R is a Nakagami-m random variable, then the c.f. of , is given as Ψ (ℐω) = ℐ =Ψ (ℐω) = ℐ Ω (2) In order to obtain the novel model, it is instructive to consider two necessary cases, viz., > 1 and ≤ < 1. From [2], it is evident that the fading parameter ‘m’ is inversely proportional to the fading severity i.e., for small m, the p.d.f. experiences a high fading level and =∞ corresponds to a deterministic signal. According to central limit theorem, the p.d.f. of a sum of (large number of) random variables approximates a Gaussian density function of zero mean. http://www.ijettjournal.org Page 1298 International Journal of Engineering Trends and Technology (IJETT) - Volume4Issue4- April 2013 For large m ( > 1), the p.d.f. of R has a non-zero mean, since a line of sight can be expected (in order to decrease the fading severity). This case is more predominant in rural and in some parts of sub urban areas. In order to get the p.d.f. of X and Y, without loss of generality, we assume that the variance, which gives the spread of the p.d.f. of R about the mean, gets equally divided into the in phase and quadrature components For small m ≤ < 1 , the p.d.f. of R has a zero mean and the p.d.f. of X and Y experiences unequal variance levels. Considering the above analysis, for > 1 the characteristic function (c.f.) of X can be approximated to that of a N Ω Ω (µ , ) distribution and the c.f. of Y to that of a N (µ , ) distribution. Further assuming independency of X and Y, the approximate c.f. of is given by +µ + , (10) Ω = Ω, (11) Ω = Ω, (12) (3) ℐ Ω +Ω = Ω, ℐ ℐ Ω ≈ Ω ℐ ℐ Ω ≈ Ω ℐ ℐ Ω( ) / ( ℐ Ω( )) / , is valid for the corresponding parameters. Taking square roots on both sides of (12), we obtain ≤ 1, (6) Ω ⎞ ⎠ (15) ℐωΩ ⎞ ℐω ⎝ − 1 , 0≤ ℐωΩ ⎝ ≈ µ ⎛ = µ ℐω (5) where (14) ℐ Ω ⎛ (ℐω) ≈ (13) ℐ Ω The relation (4) For ≤ < 1, the c.f. of X can be approximated to that of a N (0, Ω (1+ )/2) distribution and the c.f. of Y to that of a N (0, Ω (1− )/2) distribution. Further assuming the independency of X and Y, the approximate c.f. of is given by = ℐ ℐ Ω Case when m > 1 (m is large) Since in this case, we can expect a direct LOS component, the mean of the received signal will be non-zero. This can be concluded by considering the Central limit theorem, which derives that the p.d.f. of sum of the random variables will be of Gaussian with zero mean. So, the below equation is valid ℐ (ℐω) ≈ µ We put the condition that + must have the same c.f. as that of + given by (2), the conditions must hold. ⎠ Substituting implying that R approximates a Hoyt (or Nakagami-q) distribution with = µ , Ω =Ω , = (16) in (15), we obtain = ,0≤ ≤1, (7) Let U and V be independent Nakagami distributed random variables with fading parameters and , respectively, and mean square values Ω andΩ , respectively. This implies and are gamma distributed with means Ω and Ω , respectively, and parameters and , respectively, so, we get (ℐω) = ℐ (ℐω) = ℐ Ω , ℐ ( ℐ , (17) ) ℐ Similarly, substitution of = µ ,Ω = Ω , = (16) (8) in (15) results in Ω , (9) ℐ ℐ ≈ ℐ ISSN: 2231-5381 ℐ ≈ http://www.ijettjournal.org ( ℐ , (18) ) Page 1299 International Journal of Engineering Trends and Technology (IJETT) - Volume4Issue4- April 2013 If the means and of the random variables respectively (see (8) and (9)), satisfy = + , = + and , , (19) = , (20) Then we conclude the following If we choose the distribution of |X| to be the same as the distribution of U and the distribution of |Y| to be the same as the distribution of V, then, for > 1/2, > 1/2, and condition (18), the c.f.s of and are given, respectively, by the approximations ℐωµ ℐωΩ (ℐω) ≈ ( ( ) = (1 − )g( ; , (21) , (22) ( )= + ℐωΩ ) ( ( =2 , ( ) − ) − ) , , (24) )= | |( ) =2 ( ) ( ) − , x > 0 )= | | (− − 0,x ≥ 0 ISSN: 2231-5381 (30) ( − ) 1 + 2 ( ; ) ) , (− ; ) , (31a) − , x < 0 ( ) ) ( ; ) , (− ; , ) (31b) µ = 1− √Ω cos φ, (32) µ = 1− √Ω sin φ, (33) (26) Ω = ) ( + The parametersµ , µ , Ω , Ω , ,and in are expressed in terms of the parameters m,Ω , and φ as (25) On the other hand (− ; , ) denotes the p.d.f. of a negative-sided Nakagami random variable with parameter and mean square value , and is given by , ) (23) 0,x ≤ 0 (− ; ( 1 + 2 2 ( )= Let ( ; , ) denote the p.d.f. of a positive-sided Nakagami random variable with parameter and mean square value , given by ( ; (29) Therefore, the p.d.f. of X is given as + )=2 ( = − It is clear that if eq. (21) and (22) are satisfied, then the p.d.f. of | | and | | are given by | |( ) , A similar procedure on Y gives the following p.d.f. ℐωµ ℐωΩ ( | g(− ; ℐωΩ ) (ℐω) ≈ |( ) = 2 )+ , To find , we use the method of mean matching. Since the distribution of X approximates a Gaussian distribution with mean , we obtain Such that, from (10), (11) and (12) = In order to obtain, the p.d.f. of X and Y, we linearly combine the above two functions as follows. Ω = Ω Ω 1+ 1− cos 2 , (34) 1− 1− cos 2 , (35) (27) = 1+ 1− cos 2 , (36) = 1− 1− cos 2 , (37) (28) http://www.ijettjournal.org Page 1300 International Journal of Engineering Trends and Technology (IJETT) - Volume4Issue4- April 2013 Case when 1⁄ 2 ≤ < 1 (m is small) In this case, the sum of sinusoids obtained at the receiver has the p.d.f. of mean zero and of different variances. So, we can approximate the corresponding c.f.s of Hoyt and gamma RVs and further substituting Ω = Ω and m = m we obtain = , (38) Ω ℐ ( ℐ Ω = Ω Ω , (46a) < ∞, ≠ 0, A similar procedure on Y gives the following p.d.f. 1 ( ) = [ ( ; 2 , ( ℐ Ω ) 1+ − 1 , (40) 1− − 1 , (41) 1+ −1 , (42) ) + (− ; , | | = (39) −∞ < ) < ∞, − )] , , (46b) ≠0 0, = 0 = Further in order to extend the obtained results, we try to obtain the magnitude distribution from the joint distribution function, so that the results can be validated with that of the original Nakagami –m distribution. We have that = − = 0, = 0, If the means Ω and Ω and the parameters m and m of the random variables and , respectively satisfy Ω = ) ℐ Ω ) = ( ( −∞ < Similar substituting of Ω = Ω and m = m gives Ω | | = ,Θ ( , )= ( cos ) ( sin ) (47) In order to get the magnitude marginal distribution of R, we integrate the above joint function w.r.t to i.e., = 1− − 1 , (43) Then we conclude the following: If we choose the distribution of | | to be the same as the distribution of U and the distribution of | | to be same as the distribution of V, then the c.f.s of and are given, respectively, by the approximations ∞ ( ) =∫ ( cos ) ( sin )d By substituting, the required equations into eq. (51), we finally arrive at ( )=2 ( , ( ) ) ( ) (1 − (ℐω) ≈ )(1 − ) + (1 − ) + ℐ Ω ] (49) 3. DISTRIBUTION OF PHASE OF COMPLEX NAKAGAMI-m FADING GAIN (45) Similar to the case of > 1, the p.d.f.s of | | and | | are given by g (x, ;,) and g (-x,:,) functions By mean matching, the value of is obtained as = Therefore, the p.d.f. of X is given as We now find the distribution of the phase Θ corresponding to the novel model given by the p.d.f.s (31) and (46) for the cases of > 1 and1⁄2 ≤ < 1, respectively. Phase distribution when > 1 From (28) and the fact that = cos Θ , = joint p.d.f. of R and Θ is expressed as , )+ (ℐω) ≈ ISSN: 2231-5381 × [(1 − (44) ℐ Ω ( ) = [ ( ; (48) ) + (− ; , )] ,Θ ( sin Θ , the , ) = (1 − )(1 − ) ( cos ; ( sin ; , ) http://www.ijettjournal.org , )× Page 1301 International Journal of Engineering Trends and Technology (IJETT) - Volume4Issue4- April 2013 + (1 − (− cos ; + +(1 − ) (− cos ; ) × ( sin ; , ) × (− sin ; , )( ) ( cos ; ) , ) × ( sin ; , ) , The phase distribution given by [3] can be obtained from (49) by putting µ = µ ≪ Ω ⁄2 and = = ⁄2 , which results in ) , , ⁄ 0 < < ∞, − < < Θ( ∞ ) =∫ ( cos ) ( sin )d + − − ( ) ( ) ( ) ( + × + × + × − ) × − ( ) ( ) ( ) ( ( )+ ( )+ ( )+ ) ( ) (52) where 2 ( ( ) = ( ) | cos | ) + ) ( | sin | <1 The p.d.f. of phase Θ is given by ( ) = ∫ which results in ( )= ( ( ) ) ( ) |cos | − < 4. |sin | ,Θ ( , ) (53) < For the case when > 1 , the original p.d.f.s of the real part X and the imaginary part of Y of the complex Nakagami-m fading gain R ℐ , are simply the Gaussian functions with the corresponding means and variances. The original phase p.d.f. is given by 1 ( ) ( )≈ 1 + 2√ cos( − ) × 1 2 − √2 cos( − ) ISSN: 2231-5381 < (55) ( )≈ ( ( (56) )) Fig.1. illustrates the probability density functions of X and Y for the case of large m along with that of the original Gaussian approximation. Fig.2. shows the p.d.f. of phase Θ of the complex Nakagami-m fading gain. Also shown are the Rician approximation and the approximation. ⁄ Also, it is clear that the has four modes and ⁄ only one mode matches well with the Rician approximation, which clearly states that ⁄ is only applicable 3 for = ± 4 and = ± 4 Fig.4. gives the plots of phase Θ along with the Hoyt approximation and the ⁄ approximation. We find that there are singularities at = 0 = ± .These singularities corresponds to the high probability concentrations of the phase around = 0 and = ± , implying that the real and imaginary parts of the fading gain have high probability of being either equal or antipodal. Fig.5 shows the magnitude distribution for = 2 (for large m case) of the novel model along with the original p.d.f. given below NUMERICAL RESULTS − < |cos sin | Fig.3. shows the p.d.f of the novel model for X and Y for the case of small m. Since the p.d.f. of X and Y are obtained by linearly combining the positive sided and negative sided p.d.f. this novel model deviates from its Hoyt approximation at zero. The same discrepancy can be seen also for the case of large m. i=1 corresponding to first quadrant and similar cases Phase distribution when 1⁄ 2 ≤ ( ⁄ ) And we call this the ⁄ characterization. For the case1⁄2 ≤ < 1, similar plots are obtained and the original phase function is given as (51) Now it can be easily seen that ( )= ( ) (50) by (30), The p.d.f. of Θ is where g(.) is given by (26), and given by ( )= ( ; , Ω) = 2 Fig.6. shows the same for Ω ( ) − (57) Ω = 0.9 (for small m case). (54) http://www.ijettjournal.org Page 1302 International Journal of Engineering Trends and Technology (IJETT) - Volume4Issue4- April 2013 Fig.1. PDFs of real part X and imaginary part Y of complex Nakagami-m fading gain and of corresponding Rician approximation for m=5, φ = π/6, and Ω=10. Fig.3. PDF of real part X and imaginary part Y of complex Nakagami-m fading gain and of corresponding Hoyt approximation for m=0.9 and Ω=10. Fig.2. PDF of phase Θ of complex Nakagami-m fading gain and of corresponding Rician approximation for m=5, φ = π/6 and its comparison with PDF of Θ obtained from Nakagami-m/2 approximation. Fig.4. PDF of phase Θ of complex Nakagami-m fading gain and of corresponding Hoyt approximation for m=0,9 and its comparison with PDF of Θ obtained from Nakagami-m/2 characterization. ISSN: 2231-5381 http://www.ijettjournal.org Page 1303 International Journal of Engineering Trends and Technology (IJETT) - Volume4Issue4- April 2013 5. CONCLUSION In this paper, a new model for Nakagami-m fading channels has been proposed. This is obtained by approximating the model with the distributions of Hoyt and Rician distributed envelopes. The phase distribution here derived is non uniform in nature and the results confirm the validity of the model for the two cases > 1 and0.5 ≤ < 1. REFERENCES Fig.5. Magnitude distribution for m=2 [1].Ranjan K. Mallik,”A New Statistical Model of the complex Nakagami-m Fading Gain”, IEEE Trans. Commun vol. 58, no. 9, september 2010. [2].M. A. Taneda, J. Takada, and K. Araki, ”The problem of the fading model selection”, IEICE Trans. Commun, vol. E84-B, no.3. pp. 660.-666, Mar. 2001. Jyotheeswar completed his Bachelor B.Tech [3]M. Nakagami, S. Wada, and S. Fujimura, “Some degree considerations on random phase problems from the standpoint fading”, Inst. (ECE) from Sreenivasa Institute of Technology and Electrical Commun. Engineers Proc., vol. 36, no. 11, pp. 595-602, Nov. 1953 Mangagement Studies, Chittoor and is currently [4].M.D. Yacoub, G. Fraidenraich, and J.C.S. Santos Filho, pursuing phase-envelope his master’s degree at Sree Vidyanikethan “Nakagami-m joint distribution”, Electronn. Lett., vol. 41, no.5, pp.259-261, Mar. 3, 2005. Engineering College, Tirupathi. [5].N. C. Sagias and G. K. Karagiannidis, “Effects of carrier phase error on EGC receivers in correlated Nakagami-m fading”, IEEE Commun. Lett., vol.9, no.7, pp. 580-582, Jul 2005. [6].G. Fraidenraich, C. Subhas M. didD.his Yacoub, Ph.D.J.inR. wireless Mendes, and communications J. C. S. Santos at Filho, “Second order statistics for diversity combining of nonJawaharlal Nehru IEEE Technological University Anantapur, identical Hoyt signals”, Trans. Commun., vol. 56, no. 2, pp,183-188, Feb, 2008. Anantapur, A.P. He has over 26 years of Industry, Research, Academic and Administrative experience. Presently he is with Sree Vidyanikethan Engineering College, Tirupati, A.P. as Professor of Electronics and Communication Engineering and Dean (Academic). He has eight publications in international journal with good impact factors and ten presentations in IEEE international conferences. His interests of research are wireless communications and signal processing. He is a member of IEEE and IEICE of Japan, and Fellow of IETE. Fig.6. Magnitude distribution for m=0.6 ISSN: 2231-5381 http://www.ijettjournal.org Page 1304