See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/224290589

Closed-Form Expression for the Bit Error Probability of Rectangular QAM

Subject to Rayleigh Fading

Conference Paper in Vehicular Technology Conference, 1988, IEEE 38th · November 2007
DOI: 10.1109/VETECF.2007.200 · Source: IEEE Xplore

CITATIONS READS
12 3,378

3 authors:

Waslon Terllizzie A Lopes Francisco Madeiro

Universidade Federal da Paraíba Universidade de Pernambuco
206 PUBLICATIONS 617 CITATIONS 180 PUBLICATIONS 873 CITATIONS

SEE PROFILE SEE PROFILE

Marcelo Alencar
Institute of Advanced Studies in Communications
647 PUBLICATIONS 1,644 CITATIONS

SEE PROFILE

All content following this page was uploaded by Waslon Terllizzie A Lopes on 09 September 2015.

The user has requested enhancement of the downloaded file.

 Closed-Form Expression for the Bit Error
Probability of Rectangular QAM Subject to
Rayleigh Fading
Waslon Terllizzie A. Lopes Francisco Madeiro
Department of Electrical Engineering Department of Electrical Engineering Federal University of Campina Grande
ÁREA1 – College of Science Polytechnic School of Pernambuco
and Technology University of Pernambuco
Salvador, BA, Brazil Recife, PE, Brazil
E-mail: waslon@area1.br E-mail: franciscomadeiro@yahoo.com.br
Abstract—In this paper a closed-form expression for the bit
error probability (BEP) of M -ary pulse amplitude modulation
(M -PAM) subject to Rayleigh fading is derived. This expression
is used to derive another general result: a closed-form expression
for the BEP of rectangular quadrature amplitude modulation
(R-QAM) subject to Rayleigh fading. In this work the fading
channel is seen as an additive noise channel whose additive noise
is modeled as the ratio between Gaussian and Rayleigh random
variables.
Index Terms—Rectangular quadrature amplitude modulation,
Rayleigh fading, bit error probability
I. I NTRODUCTION
The requirements for multimedia traffic over wireless com-
munications systems demands an ever-increasing need for
bandwidth. Spectrally efficient digital modulation schemes,
such as quadrature amplitude modulation (QAM), are attrac-
tive techniques to achieve high transmission rates without
affecting the bandwidth of those systems.
The performance evaluation of QAM for noisy channels in
terms of bit error probability (BEP) has been addressed previ-
ously (e.g. [1]–[9]). Although some approximate expressions
give accurate error rates for high channel signal-to-noise ratio
(SNR), the evaluation of the error rates with those expressions
leads to a deviation from the corresponding exact values,
when SNR is low. Therefore, the derivation of closed-form
expressions for the BEP is a relevant problem.
In a recent paper [10], a convenient method for deriving
closed-form expressions for the BEP of modulation schemes
subject to Rayleigh fading channel was presented. The method
assumes that the Rayleigh fading channel can be seen as an
additive noise channel whose additive noise, denoted by m(t),
is modeled as the ratio between a Gaussian random variable
(r.v.) and a Rayleigh r.v. The method consists on using the
cumulative density function (CDF) of this noise, given by [10]

m  
1 m
PM (m) = pM (x)dx = √ + 1 , (1)
−∞ 2 m2 + N 0
for deriving the closed-form expressions for the BEP.
1-4244-0264-6/07/$25.00 ©2007 IEEE
Marcelo S. Alencar
Institute of Advanced Studies
in Communications (IECOM)
Campina Grande, PB, Brazil
E-mail: malencar@dee.ufcg.edu.br
In the present paper, this method is firstly used to obtain a
closed-form expression for the BEP of M -ary pulse amplitude
modulation (M -PAM) for a Rayleigh fading channel. Then,
this expression is used to obtain another general result: a
closed-form expression for the BEP of rectangular quadrature
amplitude modulation (R-QAM) subject to Rayleigh fading.
The remaining of the paper is organized as follows. Sec-
tion II presents the derivation of the BEP expression of M -
PAM and shows numerical and Monte Carlo simulation results.
The BEP expression of R-QAM is derived in Section III,
which presents numerical and simulation results for some
examples of R-QAM schemes. The conclusion of the work
is presented in Section IV. In order to maintain the paper
self-contained, an appendix with the derivation of the CDF of
m(t) is also provided.
II. B IT E RROR P ROBABILITY OF M -PAM
The signal waveforms of M -ary pulse amplitude modulation
can be expressed as
s(t) = AI cos(2πfc t), 0 ≤ t < T, (2)
where AI is the signal amplitude of the in-phase components,
fc is the carrier frequency and T is the symbol interval. In
an M -PAM scheme, a serial data sequence is converted to
log2 M bits. In (2), the amplitude AI is selected from the
set {±d, ±3d, . . . , ±(M − 1)d}, where 2d is the minimum
distance between signal points, given by

3 log2 M · Eb
d= , (3)
(M 2 − 1)
where Eb is the bit energy. The received PAM signal can be
demodulated coherently.
Recently, in [11], a closed-form expression for the BEP of
M -PAM under additive white Gaussian noise (AWGN) has
been derived. In the following, results presented by Cho and
Yoon in [11] are used to obtain a closed-form expression for
the BEP of M -PAM subject to Rayleigh fading.
 Based on the consistency of the bit mapping of a Gray coded with
signal constellation [12], Cho and Yoon have derived in [11] (1−2−k )M −1

an expression for the BEP of square M -PAM for an AWGN 1 
PM -PAM,Ray (k) = w(i, k, M )·
channel. It is given by M i=0
log2 M
 
1  3(2i+1)2 log2 M
·γ
M 2 −1
Pb = Pb (k), (4) 1 −  .
log2 M 3(2i+1)2 log2 M
k=1
M 2 −1 ·γ+1
with (10)
−k
(1−2 )M −1

1  To the best of authors’ knowledge, this is a new expression
Pb (k) = w(i, k, M )· for the BEP of M -PAM subject to Rayleigh fading.
M i=0
  (5) Some numerical examples obtained from the closed-form
3 log2 M · γ expression for the BEP of M -PAM are shown in Figure 1,
erfc (2i + 1) ,
M2 − 1 which shows the BEP performance of the M -PAM scheme as
a function of the signal-to-noise ratio per bit for M = 2, 4,
where 8, 16, 32, 64 and 128 for a channel subject to to Rayleigh
 
k−1
 i·2M  k−1 i · 2k−1 1 fading. As shown in Figure 1, the Monte Carlo simulation
w(i, k, M ) = (−1) · 2 − + , (6) results strongly agree with the results obtained from (9), (10)
M 2
and (6). It can be seen in Figure 1, for instance, that 3–4 dB
γ = Eb /N0 denotes the signal-to-noise ratio (SNR) per bit, of SNR per bit has to be invested to transmit an extra bit, in
x denotes the largest integer smaller than x, and erfc(·) order to maintain the average bit error probability of 2×10−2 .
denotes the complementary error function, given by

∞ III. B IT E RROR P ROBABILITY OF R-QAM
2 2
erfc(x) = √ e−t dt. (7)
π x In an arbitrary I ×J R-QAM scheme, the signal waveforms
consists of two independently amplitude-modulated carriers in
Notice that the BEP of M -PAM subject to AWGN is
quadrature, which can be expressed as
expressed in terms of a weighted sum of complementary
error functions. The term erfc (·) in (5) corresponds to twice s(t) = AI cos(2πfc t) − AJ sin(2πfc t), 0 ≤ t < T, (11)
the probability of the additive Gaussian noise exceeding
(2i + 1) 3 log 2 M ·Eb
(M 2 −1) . For non-Gaussian additive channels, the
where AI and AJ are the signal amplitudes of in-phase
and quadrature components, respectively, fc is the carrier
weights in (6) (which incorporates the effect on the BEP of
frequency and T is the symbol interval. In an arbitrary I × J
the bit positions in a symbol with log2 M bits) can be used
R-QAM scheme, log2 (I · J) bits of serial information stream
in conjunction with the cumulative density function (CDF) of
are mapped onto a two-dimensional signal constellation using
the corresponding additive noise for determining the BEP of
Gray coding. Among the log2 (I · J) bits, log2 I bits are
an M -PAM scheme.
mapped onto the in-phase channel, the amplitude AI of which
Considering the Rayleigh fading channel, the CDF of the
is selected from the set {±dI , ±3dI , . . . , ±(I −1)dI }, where
r.v. which models the corresponding additive noise is given
2dI is the minimum distance between two signals considering
by (1). Thus, twice the probability that the one-dimensional
their projections in the in-phase axis. Similarly, log2 J bits
additive noise m(t) exceeds (2i + 1) 3 log 2 M ·Eb
(M 2 −1) is given by are mapped onto the quadrature channel, the amplitude AJ
  of which is selected from the set {±dJ , ±3dJ , . . . , ±(J −
3 log2 M · Eb 1)dJ }, where 2dJ is the minimum distance between two
2 × P m ≥ (2i + 1) =
(M 2 − 1) signals considering their projections in the quadrature axis.

∞ The demodulation of the received QAM signal is achieved by
2×  pM (m)dm = performing two parallel PAM demodulations.
3(2i+1)2 log2 M ·Eb (8)
(M 2 −1) In this section, results presented in the previous sections
 2

3(2i+1) log2 M ·Eb for the BEP of M -PAM are extended to obtain a new, exact
(M 2 −1)N0
1 − . and closed-form expression for the BEP of an arbitrary I × J
3(2i+1)2 log2 M ·Eb
(M 2 −1)N0 +1 R-QAM scheme subject to Rayleigh fading.
From the results presented in Section II, it follows that
Using (8) and the weights in (6), the expression for the the bit error rate for the k-th bit, PM -PAM,Ray (k), with k ∈
BEP of M -PAM for a channel subject to Rayleigh fading, {1, 2, · · · , log2 M } (where M is the number of symbols of
PM -PAM,Ray , is finally obtained as the PAM constellation), is given by (10), with the weights
log2 M given by (6).
1 
PM -PAM,Ray = PM -QAM,Ray (k), (9) Thus, considering an I × J rectangular QAM, the error
log2 M probability for the k-th bit of the in-phase component and the
k=1
 0
10

10−1

Bit Error Probability

2−PAM (Analytical)
10−2 2−PAM (Simulation)
4−PAM (Analytical)
4−PAM (Simulation)
8−PAM (Analytical)
8−PAM (Simulation)
16−PAM (Analytical)
−3
16−PAM (Simulation)
10 32−PAM (Analytical)
32−PAM (Simulation)
64−PAM (Analytical)
64−PAM (Simulation)
128−PAM (Analytical)
128−PAM (Simulation)
10−4
0 5 10 15 20 25
Eb/N0 (dB)

Fig. 1. Bit error probability of M -PAM as a function of the signal-to-noise ratio per bit (Eb /N0 ) for a channel subject to Rayleigh fading channel.

error probability for the l-th bit of the quadrature component PR-QAM,Ray , can be obtained by averaging (12) and (13), that
are respectively given by is
 
 log2 I log2 J
1
(1−2−k )I−1
 1  
PI (k) = w(i, k, I)· PR-QAM,Ray =  PI (k) + PJ (l) .
I log2 (I · J)
i=0 k=1 l=1
 2
 (12) (17)
3(2i+1) log2 (I·J)·γ
1 − I 2 +J 2 −2
 It important to mention that the Expression (17) derived in the
3(2i+1)2 log2 (I·J)·γ present paper is more compact than that one presented in [13],
I 2 +J 2 −2 +1
which involves hyper-geometric and gamma functions.
and Figure 2 presents some numerical results obtained from the
(1−2−l )J−1
 closed-form expression for the BEP of R-QAM subject to
1  Rayleigh fading. The figure shows the BEP as a function of
PJ (l) = w(j, l, J)·
J the signal-to-noise per bit for the schemes 8×16 R-QAM and
j=0
  (13) 16×32 R-QAM. It is observed that the numerical results are
3(2j+1)2 log2 (I·J)·γ
I 2 +J 2 −2 corroborated by Monte Carlo simulation results.
1 − 
3(2j+1)2 log2 (I·J)·γ
I 2 +J 2 −2 +1 IV. C ONCLUSION

with This paper presented new, exact and closed-form expres-

  sions for the bit error probability (BEP) of M -ary pulse
i·2k−1 i · 2k−1 1
w(i, k, I) = (−1) I 
· 2k−1 − + (14) amplitude modulation (M -PAM) and rectangular quadrature
I 2 amplitude modulation (R-QAM) subject to Rayleigh fading.
and In this work the fading channel is seen as an additive noise
  channel whose additive noise is modeled as the ratio between
 j·2J
l−1
 l−1 j · 2l−1 1
w(j, l, J) = (−1) · 2 − + . (15) Gaussian and Rayleigh random variables.
J 2 As future works, the authors will investigate the effects
It is worth mentioning that in an R-QAM I × J scheme, of channel estimation errors as well as multipath scenarios
the relation between the bit energy, Eb , and the distances dI on the bit error probability of modulation schemes subject to
and dJ is given by Rayleigh fading.
(I 2 − 1)d2I + (J 2 − 1)d2J A PPENDIX
Eb = . (16)
3 · log2 (I · J) CDF OF THE A DDITIVE N OISE
Finally, the bit error probability of an arbitrary I × J Consider the wireless system depicted in Figure 3, where
rectangular QAM scheme for a Rayleigh fading channel, the transmitter uses M -ary modulation.
 100
16x32 R−QAM (Analytical)
16x32 R−QAM (Simulation)
8x16 R−QAM (Analytical)
8x16 R−QAM (Simulation)

10−1

Bit Error Probability

−2
10

−3
10
0 5 10 15 20 25 30
Eb/N0 (dB)

Fig. 2. Bit error probability of R-QAM as a function of the signal-to-noise ratio per bit (Eb /N0 ) for a channel subject to Rayleigh fading channel.

Transmitter Channel without error, the receiver can perform the phase compensation
Input
s(t) rc (t) (multiplication of rc (t) by ejφ ). Then, the resulting received
bits
Modulator signal r(t) can be expressed as

jφ r(t) = rc (t) · ejφ = αs(t)+z(t) · ejφ = αs(t)+η(t). (20)

αe z(t)
It is important to note that the additive noise η(t) = z(t) ·
ejφ is also a two-dimensional Gaussian r.v. having zero mean
Output Phase
Detector Demodulator and variance N0 /2 per dimension. This follows from the fact
bits Compensation
that the error probability is unaffected by a rotation, since
Receiver pN (η) is spherically symmetric [14, pp. 247].
The maximum a posteriori criterion [15] establishes that the
Fig. 3. The system model.
optimum detector, on observing r(t), sets ŝ(t) = sk (t) as the
received symbol whenever the decision function
Assuming a frequency-nonselective slow fading channel, the
received signal rc (t) can be expressed as P (si (t))pr (r(t)|s(t) = si (t)), i = 0, 1, . . . , M − 1, (21)

rc (t) = αe−jφ s(t) + z(t), 0 ≤ t ≤ T, (18) is maximum for i = k.

where s(t) represents the transmitted signal, α is the fading Based on the maximum a posteriori criterion and consid-
amplitude, φ is the phase shift due to the channel, z(t) denotes ering equiprobable constellation symbols, the detector can
the additive white Gaussian noise, and T is the signaling use the following strategy for determining the most probable
interval. transmitted symbol from the noisy observation r(t): compare
The fading amplitude α is modeled as a Rayleigh r.v., whose r(t)/α with all the constellation symbols and choose as the
probability density function (pdf) is expressed as received symbol the closest one to r(t)/α, that is, choose
2 as the received symbol the one that minimizes the metric
pA (α) = 2αe−α u(α), (19)
|r(t)/α − si (t)|.
where u(·) is the unit step function. The additive noise z(t) is The decision rule for the detector is
modeled as a two-dimensional Gaussian r.v. having zero mean  
and variance N0 /2 per dimension. Without loss of generality,  r(t) 
ŝ(t) = arg min   − si (t). (22)
a normalized fading power is considered, that is, E[α2 ] = 1, si (t) α
where E[ · ] is the expected value operator.
Assuming that the fading is sufficiently slow so that the In this scheme, after providing fading compensation (division
phase shift φ can be estimated from the received signal of r(t) by α), the channel works as an additive noise channel
because R EFERENCES
 
 αs(t) + η(t) 
ŝ(t) = arg min   − si (t)
[1] J. Lu, K. B. Letaief, J. C.-I. Chuang, and M. L. Liou, “M -PSK
si (t) α and M -QAM BER computation using signal-space concepts,” IEEE
(23) Transactions on Communications, vol. 47, no. 2, pp. 181–184, February
= arg min |s(t) + m(t) − si (t)| , 1999.
si (t) [2] P.-M. Fortune, L. Hanzo, and R. Steele, “On the computation of 16-
QAM and 64-QAM performance in Rayleigh-fading channels,” IEICE
where m(t) = η(t)/α is the additive noise obtained from the Transactions on Communications, vol. E75-B, no. 6, pp. 466–475, June
ratio between a Gaussian r.v. and a Rayleigh r.v. 1992.
Now the pdf and CDF of a random variable obtained [3] L. Hanzo, R. Steele, and P.-M. Fortune, “A subband coding, BCH
coding, and 16-QAM system for mobile radio speech communications,”
from the ratio between a Gaussian r.v. and a Rayleigh r.v. IEEE Transactions on Vehicular Technology, vol. 39, no. 4, pp. 327–339,
is presented. This r.v. can be expressed as M = N/A where November 1990.
N is a Gaussian r.v. having zero mean and variance N0 /2 and [4] M. G. Shayesteh and A. Aghamohammadi, “On the error probability
of linearly modulated signals on frequency-flat Ricean, Rayleigh and
whose pdf is given by AWGN channels,” IEEE Transactions on Communications, vol. 43, no.
1 2 2/3/4, pp. 1454–1466, February/March/April 1995.
pN (η) = √ e−η /N0 , (24) [5] P. K. Vitthaladevuni and M.-S. Alouini, “BER computation of 4/M -
πN0 QAM hierarchical constellations,” IEEE Transactions on Broadcasting,
and A is a r.v. with Rayleigh distribution whose pdf is given vol. 47, no. 3, pp. 228–239, September 2001.
[6] ——, “A recursive algorithm for the exact BER computation of
by (19). generalized hierarchical QAM constellations,” IEEE Transactions on
Under these assumptions, the pdf of M is given by [16] Information Theory, vol. 49, no. 1, pp. 297–307, January 2003.

∞ [7] X. Dong, N. C. Beaulieu, and P. H. Wittke, “Error probabilities of
pM (m) = |α|pN A (mα, α)dα, (25) two-dimensional M -ary signaling in fading,” IEEE Transactions on
−∞ Communications, vol. 47, no. 3, pp. 352–355, March 1999.
[8] J. W. Craig, “A new, simple and exact result for calculating the proba-
where pN A (η, α) is the joint probability of N and A, given bility of error for two-dimensional signal constellations,” in Proceedings
by of Military Communications Conference (MILCOM’91), 1991, pp. 571–
2 2 2 575.
pN A (η, α) = √ αe−(α +η /N0 ) u(α). (26) [9] L.-L. Yang and L. Hanzo, “A recursive algorithm for the error probability
πN0 evaluation of M -QAM,” IEEE Communications Letters, vol. 4, no. 10,
Thus, the pdf of M is given by pp. 304–306, October 2000.

∞ [10] W. T. A. Lopes and M. S. Alencar, “QPSK detection schemes for
2 2 2 Rayleigh fading channels,” in Proceedings of the IEEE International
pM (m) = √ α2 e−α (1+m /N0 ) dα. (27) Telecommunications Symposium (ITS’02), Natal, RN, Brazil, September
πN0 0 2002.
Using the fact that [17, pp. 1030] [11] K. Cho and D. Yoon, “On the general BER expression of one- and

∞ √ two-dimensional amplitude modulations,” IEEE Transactions on Com-
2 −x2 π munications, vol. 50, no. 7, pp. 1074–1080, July 2002.
x e dx = , (28) [12] P. J. Lee, “Computation of the bit error rate of coherent M -ary PSK
0 4 with gray code bit mapping,” IEEE Transactions on Communications,
(27) can be expressed as vol. 34, no. 5, pp. 488–491, May 1986.
[13] D. Yoon and K. Cho, “General bit error probability of rectangular
1 N0 quadrature amplitude modulation,” Electronics Letters, vol. 38, no. 3,
pM (m) = · . (29)
2 (m2 + N0 )3/2 pp. 131–133, January 2002.
[14] J. M. Wozencraft and I. M. Jacobs, Principles of Communication
Finally, the CDF of M can be obtained by integrating the Engineering. John Wiley and Sons, 1965.
previous expression (using [18, Eq. 2.264]). Thus, the CDF of [15] J. G. Proakis, Digital Communications, 3rd ed. New York: McGraw-
Hill, 1995.
M is given by [16] A. Papoulis, Probability, Random Variables, and Stochastic Processes,

m  
1 m 3rd ed. New York: McGraw-Hill, 1991.
PM (m) = pM (x)dx = √ +1 . (30) [17] A. D. Poularikas, Ed., The Transforms and Applications Handbook.
−∞ 2 m2 + N 0 IEEE Press, 1996.
[18] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and
ACKNOWLEDGMENT Products. Academic Press, 1979.
The authors would like to express their thanks to the Brazil-
ian Council for Scientific and Technological Development
(CNPq), the Research Foundation of Bahia State (FAPESB)
and the Institute of Advanced Studies in Communications
(IECOM) for the financial support to this research.

View publication stats