RADIO SCIENCE Journal of Research, NBSjUSNC-URSI

Vol. 68D, No.4, April 1964

Some Statistical Parameters Related to the

Nakagami-Rice Probability Distribution
William R. Burns
Contribution from Central Radio Propagation LaboratorYt National Bureau of Standardst
Bouldert Colo .
(R eceived N ovemb er 22, 1963)
Formulas and tables a re given for t he m ean and stand a rd dev iat ion of R = 20 loglo r
wh er e t h e r a ndom va ria ble 1 h as t h e N akagami-Ri ce distribu t ion. This d ist ri b ut ion is of
inter est in conn ection wit h t he s hort -te rm fadin g cha racteristics of som c recci ved rad io fi eld s.
A p a rt icula rl y simple formu la f or t he m ean of R is ob tained in te rm s of th e well- kn own
ex po nential in tegral function - Ei( - x) . Addi t ion al inform at ion co ncernin g t he m ed ia ll a nd
in te rd ecil e r a nge of If is a lso gi ve n.

1. Introduction
Tn certa in r adio propagation problems t he fi eld stretlg t.h at a r eceiver may be app roxim ated
by t he vector s um of a co nstant vector a nd a Ra~Tlei g h-dis Lribu te d vector [Nor ton , V ogler ,
:\1ansfLCld , a nd S hort, 1955]. Jf i t is ~tssuilled t ll<) t t he R ayleig h-dis Lri bu ted v ector has a n rill s
a illpli t ud e Ie a nd t ltHt Lh e consta nt vector has ~Ul nn s ampli tude of un ity, Lh en the proba bili t~T
distribuLion of t he r es ul tan t P of Lheir sum as give n by N Hlmgami [1940] a nd Ri ce [1944, 1945]
lll ay be put in t he form
(l.1 )
wh ere I o(x) is t be m odified B essel fun ction of first kind H,nd order zer o. Jf t he R a,y leig h di stribu ted vector h as a n l"I11 S a mpli tud e lei , t he co nstan t vector has an rill s ampli tud e le z, and t he
resul ta nt of t he ir urn is PI, Lhen ( l.l ) gives t he dis tribu t ion of r = r l/1e 2 as a function of k = k l/ kz.
Th e primary purpose of t his p a per is t he d etermin aLion of the mean R and sLand ard deviat ion (Til or R = 20 loglo 1'. Addi t ional inform ation co ncernin g t he m ed ia n R (0.5 ) and in terdecile
nw ge R (0. 1)- H (0. 9) of R is t~),ke n from th e p ap er b y Norton et al. [1 955]. T he ph ase cf> of th e
vector sum of ~t R a~'lei gh-di stribu ted vector and a co nsta n t vector is di scussed by Nor ton ,
Shultz, and Yarbrough [1952] .

2. Calculation of Rand
W e will determine

(T il

(JR

from t he r ela tion

wb er e R,2 is the m ean of R2.

:\1aking use of (1.1 ), we find that

(2 .1)

and
(2.2)

J n app endix 1 of t hi s pap er , t hese two in tegr als a r e evalu a ted and th e three equations wh ich
follow are d erived .
(2. 3)
R =( 10 loglO e)[-Ei(- l W)] .
429

For

k;::: 1, use
2 _

( 'i!..

O'/l - (lOlog lQe) ..(6+2~

l
n=l
F or le< l , use

0'~ =( 10 logl o e)2

t(

00

(- 1) n+ 1(1!F)n+1( n ~) _ [ oo (- 1) n- I(1!F)nJ 2 "')

(10 , ' 1) .' (10 + 1)
L:.
L:;
.n.n
,
}= 1 J
n= l
J

E' (

1/k2

- 2Ei(- 1!F)E * (l !F) + 4) 1

-2

r
)1

1 1k2

t)

~ -

dt

e l [ - Ei ( - t )1dt -[E i( - 1jlc2) F+ 4Ei( - l /P )h + In Cli P ) ]+ 0 1 ~ .

t
J

In these equ ations,

- E i(-x) =

1 ~ dt,
00

r oo

E* (x)= - -

-I

rlt

_C _

J-x t

where
-

(2.5)

'

roo = lim (J -'+ f oo )

J- x

- x

E-----70

with ~ > O; and 0 1 = l.099019, a constant. The functions - E i ( -x) and E* (x) are well-known
functions for which tables are available. See, for example, the book: "Tables of Sine, Cosine,
and Exponential Integrals," U.S. N fttional Bureau of Standards [1940]. The integrals
1 / k2
[

and

l /k2
1

- E i( - t )

'(

dt

~[- Ei(- t )] dt
t

are evaluated in appendix 2 to six significant digits by numerical integration .

Letting K = 20 10gIO le, we can get the following asymptotic formulas from (2.3 ), (2.4 ).
and (2 .5).
4.3429
For K > 20, R :::o: f{+- ----,cz- -2.5068
(2.6 )

T A BLE

R (0. 5)

R - R (0.5)

UR

40
35
30
25
20

0.000
. 001
. 002
.007
. 022

0. 000
. 000
.000
.000
.000

- 0.000
- . OO l
-. 002
- .007
- .022

0.061
. 109
. 194
. 346
.616

- 18
- 16

- 14

- 12
- 10

. 034
. 054
. 086
. 135
. 214

. 000
. 000
. 000
. 000
. 000

- .034
- .054
-. 086
- .136
-. 2l4

. 980
l. 238
l. 569
1. 999

-8
-5
-4
-2
0

. 335
. 524
.813
1. 249
1. 894

. 001
.Ol7
. 107
. 383
. 953

- .334
-.507
- . 705
- .866
-.941

2. 565
3.279
4.036
4.667
5. 094

2
4
6
8
10

2.808
4.006
5.448
7. 077
8.835

1. 855
3. 064
4.519
6. 155
7. 917

.953
. 942
.929
. 922
. 918

5. 340
5.465
5. 525
5. 551
5.562

12

10.679
12.580
14. 517
16. 477
18. 452

9. 763
11. 664
13. 502
15. 562
17.537

-.916
- .916
- .915
- .915
- . 915
-.915

567
559
5iO
570
5iO
5iO

[(

14
16
18
20
00

00

00

430

.776

..--------r---'--'--I -----'------------'
j

20.0

II

RI0.II - RI 0.9 / .,..-

10.0

t==t==:J==+=+, +==Yf'===1

7.0

1--+---+---+/- f-- + --I,,'It--

-1

~==t===~==~/~~R~~~==~

S.O ~

r
3.0

a:

1.0

'I

/-I-

1-----+-/ 1-1-/

R
-'-7 '1-1-/ R -+-------J

t==:I=~=t=t==t:=Jt;!;::::;;:;:~=~

1-_+-1--+1_-+_/-','_ _
II
II

0.7

us

/ /
f-------+------+/-f-H/ -H+
/) ---+-------1

RI0.5) - R _

r--; If /L rI
/ /

-30

-20

FIG U RE

-10

10

20

1. I< i n decibels.

and
(]' R ~

For

(2. 7)

4.3429 y'l .6449 - 1/ (21c'1) .

J{<
(]' R

- 20, R~ 4 .3 429F e - l /k2

(2. 8)

~ 6. 1418lc ~ 1 + lc 2/2.

(2.9)

Th e m agni tude of the error in th ese four expressions is less than 5 (10)-4 .
Values of t he m edian , the m ean, t he differ en ce between t he m ean and the median, and the
standard devia tion of R are given in table 1. These four qu an ti ties togeth er with th e interdecile r ange of R are shown as fun ctions of J{ in figure 1. The m edian and in terdecile range ar e
taken from the p aper by Norton et al. [1955].

3. Appendix 1.

Derivation of the Formulas for Rand

(JR

If we change to n atural logarithms and m ake the substitution r = ky in (2.1) and (2.2),
we get
2
(3 .1 )
R = 2(20 loglo e)e- 1 / k [(In k)A (k) + B (k)] ,
and
(3 .2 )
11/ = 2(20log 10 e)2e - 1 / k2 [(ln lc)2A(k) + 2 (ln k )B (k)+C(k )].
H ere

A(lc) =
B (lc)=

So'" ye- y2 I o(2y/k)dy= [So '" y2a- le-y2Io(2y/k)dy

1=/

r'" (1n y)ye- y2I o(2y/k )dy= [ va>.O -2~ 1'" y2a - le-y2I o(2y/k )dy]

Jo

431

,
a= l

a nd

Using the Maclaurin series expansion for I o(2y/k) and integrating term by term, we find that

wherc r (x) is the gamma function and IFI (a; c; x) is a confluent hy pel'geometric function
[Rain\'ille, 1960].
In order to complete the evaluation of A(lc) , B (k ), and C(lc), we make use of t he known
series expansion for - Ei( -x) and of the series transformation

is the nth forward difference of ao; i.e., ,:lao = ao, ,:llao= al - aO, etc .
details, we find that R is given by (2 .3), and that
R 2= (lO loglOe)2

~ 4(lnlc)2+ 4(lnk)[ln
\..

Omitting a ny furth er

(ljlc2 )-Ei (- I jlc2)]+/

+~=- 2Yh + ln (l /k 2)-Ei( - 1/F ) ]+ 2 ~

6

(t ~) } .

( _ 1)n+ I(ljlc2) n+ 1

(n+ 1)! (n + l )

71= 1

Combining (2.3) and (3 .4), we get (2.4) for (TR .

In order to pro,' e (2.5), we derive a differ ent expression for the series in (3 .4) .
that
co

71=1

1)-- i " ~ (_l)n +1t n

(_ 1) n+ l(x)n+ 1 ( 71
6 ""
(n+ 1)!(n , l ) j= l]
I

CO

,,=1

wh erc
C=

I)J

CO
(_ 1) n+lxn+1 ( 71
~ ( + 1)1( + 1) L: . .,

71= 1

.n

)= 1.7

(n+ l ).

x=1

(3 .4)

j= l ]

Observe

1)J elt + C
~

6
...,
j= l ]

= 0.1827580.

Omitti ng any furth er details, except to note that use is m ade of the series expansions for
- E i( - x) and E* (x), and that we again use the series transformation (3 .3), we find that
co

~
n= 1

1)

(_ 1) n+l xn+1 ( "

L: -c = -Ei( -x)[ E *(x)- y - ln x] + Ei ( - 1) [E * (l )-y]
(n+ l )!(n + 1) j= l ]
+ ( X~ E i( - t )elt-2 (" E i( - t) dt + y In x+ (In x)2 +C.
Jl t
JI
t
2

Usin g this expression together with (2.3) and (3 .4), we arri\'e at (2.5) .

4 . Appe ndix 2.

Evaluation of the Integrals

(1 1k2 - Ei( - t )

Jl

elt and

l/k2 et
- [- E i( - t )]clt
1

Tb e values of th e integrals
llk2 - Ei (-t)

g(1/ F ) = i
1

elt and h(l/F ) =

i1 /k2 e t
-

t I t

[- E i( -t)]dt,

giy en in table 2 for various yalues of K , were obtain ed as follows. The i th entries of the
table were computed in terms of the precedin g entries by m eans of the formula

( Xi ](t)dt = ( ("I + ( X2 + . . . + ("i ) ](t) dt

J1 JXl
) Xi_ 1

J1

432

TABLE
f(

2
!I(l/"

11k 2

- 0. 96910013

2)

h (l / k 2)

- 6
- 6.9897000

1. 2500000
1. 5848932
2.5 Jl 8864
3.9810716
5. 0000000

0.0404747
.067941 4
.09 17103
.0971504
. 0976706

0. 123376
.234919
. 402303
.5 18481
.561387

-8
- 9.0308999
- 10
- 10.791812
- 12

6.3095733
8.0000000
10.000000
12.000000
15.848932

. 097811 6
.0978393
.0978428
.0978432
. 0978432

. 597259
.62697'1
.64964 5
.665007
.683943

.0978432
.0978432
.0978432
. 0978432
.0978432

. 706139
720383
.729470
735245
742039

.0978432
.0978432
.0978432
.0978432

. 744196
.744880
745096
.745 196

- 2
- 4

- 14

25. 118864
39. 810716
63.095734
100.00000
316.22ii6

- 16

- 18
- 20
-25
- 30
- 35
- 40

1000.0000
3162.2776
10000.000

00

- 00

where eac h of t he in tegr f11s on t he l'igh t-h and side of t he eq uation was e nl,luated n umerictllly
usin g a 16-point Gau ssia n quadmture formula . Eigh t sig nifi.cant digits were used in t he
calculaLions and the r esul ts were rounded to six signifi cant digits.
A diA'erent m ethod for calculatin g t he valu es of the in tegrals co rresponding to l{= - 40
is presented h ere to check t he Yidues g i\'en in table 2. Due to t he technique u sed in prep tl,rin g
t ile table, t his will also be ,1, ch eck on Lhe accuracy of tlte rem ainin g tabuhr entries . First obsen "e th at
10 4 el

- [- Ei ( - t) ]rlt =

10 4

where

<10-

17J 1

10 .

f OO -:- [-Ei( el

- E i( - I) .

dt =

f OO I

t)]dt - O.OOOl

+f

E i( - t ) .
t
dt + 7J

I f we use th e Madau rin series for el , we ge t

( 00 ~ [-Ei( - t )]dt = ( 00 _ E i( - t ) dt+ f: ~ ( 00 t n - 1 [ - E i( - t )]dt.

t
J1
t
n= I 11 .JI

JI

From Erdelyi et aL [J 953], we find that

( 00 t n - I [ -Ei( - t ) ]dt = r (n, 1 )+ Ei( - 1)

Jl

r (n) -'Y (n, l )+E i( - 1)

where r(n, x) and 'Y(n , x) are the in complete gamma functions, and from Le Caine [1948] we get

Thus we find that

(104 -Ei( - t ) dt"",0.097843199

t

Jl

wi th an error less than 0.5 (10 )-9 in m agni t ude.

wi th an error less than 0.5 (10 )- 7 in m agnitude.

433

In the same way we may also show that

Omitting the details , we also get from the pre-

cedin g equations that

f
I

l04

.~ [- Ei ( - t )]dt "" 0.097843199+~6 +Ei( - l)[E*(l ) - 1']- L.:

00

n~ l

1)
r.;-0.0001 "" 0.74509596
n.n
(

with an error less than 0.5 (10 )-8 in magnitude. It is seen that six significant digit accuracy is
indicated for the values of the integrals given in table 2.

Thanks are due to R . E. 'Wilkerson for checking the mathematics in this paper, and to
Gail J. Stifel for helping with the computations.

~1rs.

5. References
Erdelyi, A., VV. Magnus, F. Oberhettinger, F. G. Tricomi, et a1. (1953), Hi gher t r anscendental funct ions vol. 2
(McGraw-Hill Book Co. Inc., New York, N.Y.) (see eq (16) , p . 144) .
Le Caine, J . (1948), National Research Council of Canada, Divisio n ot Atomic Energy, Document Ko. MT- 131
(N RC 1553) (see (Oq 1.6.1, p. 8).
Nakagami, Minoru (Oct . 1940), Study on the r esultant amplitude of m a ny vibrations whose phases and a mplitudes are random , N ippon Elec. Comm. Eng. No. 22, 69- 92 (see e q ll8, p . 81) .
Norton, K. A., L. E. Vogler , W. V. Mans field , and P. J. Short (Oct. 1955), The probability distribution of the
amplitude of a con stant v ector plus a Rayleigh-di stributed vector, Proc. IRE <l3 , 1\0. 10, 1354-1361.
Korton , E. A., E. L. Shultz, and H. Yarbrough (Jan. 1952), The proba bilit y distribution of the phase of the
res ultant vector s um of a constant vector plus a Rayleigh-distrib uted vecto r, J. App1. Ph ys. 23, No.1 ,
137- 141Rainville, E. D. (1960), Special fun ctions (The Macmillan Compar..y, No,,' York, N. Y. ) (sec ex. 5, p . 128).
Rice, S. O. (July 1944), Mathematical analysis of random noise , Bell Syste m Tech. J. 23 , 282- 332.
Rice, S. O. (J a n . 1945) , Mathematical analys is of ra ndom noi se, Bell Syste m. T ech. J. 2<l, 46- 156. [A lso
published as Bell Telephon e Monograph B 1589, a nd h as been included in Selec ted papers on noi se and
stochastic processes, N. Wax, Ed. (Dover Publications, New York, N.Y., 1954) (see eq 3.10- 11 in either
t h e book edited by Wax, or Bell System Tech . .T. 2<l.ll
U .S. Natio nal Bureau of Standards (1 940), T a bles of s in e, cosin e and exponential integr als, prepare d b y t he
vVork Projects Administration for the City of New York, New York Cit:l , N .Y., volunws 1 and 2.

(Paper 68D4- 358)

434