IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 12, NO.
5, MAY 2013
Exact Formulations for the Throughput of
IEEE 802.11 DCF in Hoyt, Rice, and
Nakagami-m Fading Channels
Elvio J. Leonardo, Student Member, IEEE, and Michel D. Yacoub, Member, IEEE
Abstract—This paper investigates the throughput performance
of IEEE 802.11 Distributed Coordination Function (DCF) at
the Medium Access Control (MAC) layer with Hoyt, Rice and
Nakagami-m fading environments. The approach considered
includes the signal capture model with incoherent addition of
interfering signals and uniform attenuation for all terminals (or
perfect power control). For the Hoyt and Rice fading channels,
the results presented also cover the case of coherent addition
of interfering signals. Furthermore, for the Nakagami-m fading
channel, the case of unequal average power levels is also studied.
For the sake of comparison, results for the carrier sense multiple
access with collision avoidance (CSMA/CA) are also presented.
The formulations derived here are exact and unprecedented in
the literature. Some of these formulations are given in infinite
series form whereas several of them are presented in closed-
form expressions. For the series-based ones, only a few terms
are required in order to achieve high accuracy results.
Index Terms—Wireless communication, IEEE 802.11, DCF,
MAC, CSMA, Hoyt fading, Rice fading, Nakagami-m fading,
outage probability, capture.
I. I NTRODUCTION
W IRELESS local area networks (WLANs) have been
experiencing rapid development lately in part stim-
ulated by the deployment of systems compatible with the
IEEE 802.11 standards [1]. They offer data communication
capability between terminals within radio range while allowing
a certain degree of mobility. These networks are organized
either with or without a central node. In the cellular topology,
a central node is responsible for controlling the access to the
wireless medium and forwarding data to the intended users. In
ad-hoc topologies, the central node is absent and all terminals
share similar capabilities and responsibilities; the terminals
can communicate with each other either directly or by routing
their data through intermediate nodes. In either case, the aims
are to provide connectivity among terminals and to efficiently
and fairly dispense the available bandwidth while employing
little or no central coordination.
In order to serve terminals exhibiting bursty traffic be-
havior, WLANs make use of packet radio techniques with
random access to a transmission channel shared by multiple
Manuscript received June 14, 2012; revised October 24 and December 27,
2012; accepted February 8, 2013. The associate editor coordinating the review
of this paper and approving it for publication was Z. Han.
E. J. Leonardo is with the Department of Informatics, State University of
Maringa, Brazil (e-mail: ejleonardo@uem.br).
M. D. Yacoub is with the School of Electrical and Computer Engineering,
State University of Campinas, Brazil (e-mail: michel@decom.fee.unicamp.br).
Digital Object Identifier 10.1109/TWC.2013.032513.120859
1536-1276/13$31.00 
c 2013 IEEE
2261
users. Specifically, variations of carrier sense multiple access
(CSMA) protocols and, in particular, of CSMA with collision
avoidance (CSMA/CA), are generally used to access the
wireless medium [2]–[8]. The capacity of the channel is
then influenced by the probability of packet collision and
by the signal degradation due to mutual interference and
signal attenuation. In other words, it is influenced by the
medium contention resolution algorithm and by the channel
characteristics. Currently, it is fair to say that the use of
CSMA/CA is wide, and it has been growing continually, with
very recent adoption by some other IEEE standards [9], [10].
Intuitively, one might expect that original (wireline) CSMA
systems show better performance than wireless ones because
of the more hostile channel characteristics found in the latter.
However, this is not necessarily the case. For instance, in a
channel model that takes into account the effects of fading,
competing packets arriving at a common radio receiver an-
tenna will not always destroy each other because they may
show different and independent fading and attenuation levels
[11], [12]. This leads to infer that wireless CSMA systems may
actually exhibit successful reception rate higher than that of
wireline ones. In fact, Arnbak and Blitterswijk have shown this
to happen with slotted Aloha over Rayleigh fading channels
[13].
A practical and more sophisticated implementation of the
CSMA/CA protocol is the IEEE 802.11. Such a system spec-
ifies two operating modes: the Point Coordination Function
(PCF) and the Distributed Coordination Function (DCF). The
former is an access method planned to be implemented in
an infrastructured network. The latter, which is similar to
CSMA/CA, is the main focus of this work.
In this paper, we investigate the throughput performance of
IEEE 802.11 DCF with Hoyt, Rice and Nakagami-m fading
environments and capture effect. The performance of IEEE
802.11 DCF with unsaturated traffic and non-ideal channel
is presented in [14] and it is here extended to these fading
scenarios. The evolution of the performance analysis of IEEE
802.11 DCF has been conducted in the following steps.
Bianchi [15] presented analytical results and simulation for
the performance of IEEE 802.11 DCF assuming ideal channel
conditions, finite number of terminals and saturated traffic. In
[15], the MAC algorithm is modeled by a two dimensional
Markov chain. Liaw et al. [16] added an idle state to the
Markov chain of Bianchi’s model, this way extending it to
unsaturated traffic conditions, however keeping the channel
2262 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 12, NO. 5, MAY 2013
ideal. Daneshgaran et al. [14] further extended the analysis
to include non-ideal channel conditions and capture effect. In
[14], the authors assume Rayleigh fading channel and they use
simulation to validate their results.
The IEEE 802.11 DCF is a commercially available, widely
used and practical system and it is interesting to see it
evaluated with generic and potentially more realistic fading
models. In addition, for the sake of comparison, results for
CSMA/CA in similar fading conditions are also presented.
The work presented here introduces a number of exact closed-
form as well as exact series-based expressions. To the best of
the authors’ knowledge, unless otherwise cited, these results
are novel contributions. As an intermediate outcome of the
calculation, novel results for the outage probability for these
environments are also introduced. Analysis of the outage
probability for fading channels has been an active investigation
topic for the last few years [17]–[22], in part because it may be
used to estimate the performance of spectrum sharing systems.
The work presented here offers contributions in this area and
introduces many original results.
The channel models considered in this work are among
those commonly used to describe the short-term signal statis-
tics of wireless communications links subject to fading [23]–
[25]. The Rice model (also known as Nakagami-n) is used
when the random multipath signals are superimposed on a
nonfading dominant signal, for instance, when a line-of-sight
(LOS) component is present. The Hoyt model (also known as
Nakagami-q) applies to the cases when no dominant signal is
present and the in-phase and quadrature components of the
received signal have non-identical powers or, otherwise, are
correlated. Finally, the Nakagami-m model was inferred by
Nakagami [26] from experimental data. Its distribution can
approximate the Hoyt and Rice distributions, and it includes
as special cases the one-sided Gaussian and the Rayleigh
distributions.
In this paper, the channel is considered to be memoryless,
i.e., failures to capture the channel and future attempts are
uncorrelated. In addition, all packets are assumed to have fixed
length and to require p seconds to transmit, and that the 2-way
handshaking mechanism is used for the packet transmissions.
Finally, each packet is assumed to have a single destination.
This paper is organized as follows. Sections II describes
the framework used in this paper, while Section III applies it
to the particular scenario assumed here. Section IV considers
the cases of incoherent signal addition at the receiver’s antenna
with uniform attenuation for all terminals (or perfect power
control). Coherent signal addition and a model that includes
spatial coverage, such as in a cell, are also explored. Numerical
results and conclusions are given in Sections V and VI,
respectively.
II. F RAMEWORK
Consider a generic packet data communication system. If
the transmission of an arbitrary test packet is performed over
a wireline channel, it is generally assumed that a successful
reception can only occur if no other transmission attempt is
made during the test packet reception, i.e., if there is no signal
overlap at the receiver’s end. However, in wireless systems the
radio receiver may be able to be captured by a test packet,
even in the presence of n interfering packets, provided the
power ratio between the test signal and the joint interfering
signal exceeds a certain threshold during a given portion of
the transmission period tw , 0 < tw < p, to lock the receiver
[27], [28]. In such a case, the test packet is only destroyed
if ws /wn ≤ z, during tw , with n > 0, where z is the
capture ratio, and ws and wn are the test packet power and the
joint interference power at the receiver’s antenna, respectively.
Values for z and the capture window tw depend on, e.g., the
modulation and the coding employed by the network. For a
typical narrowband FM receiver, a z value of 6 dB is suggested
in [29]. The details about estimating the values of z and tw are
beyond the scope of this paper. The interested reader can find
further information about the capture effect in the literature,
including [30]–[33].
For the wireless channel, for which the signal assumes a ran-
dom behavior, the capture phenomenon should, accordingly,
be treated statistically. Let the random variable Z be defined
as the signal-to-interference ratio (SIR) Z  Ws /Wn ≥ 0,
where Ws ≥ 0 and Wn ≥ 0 are random variables representing
the desired signal power and the interference power at the
receiver’s antenna, respectively, with the latter one assumed
to be an n-signal ensemble. If Ws and Wn are statistically
independent, the resulting probability density function (PDF)
can be expressed as [34]
 ∞
fZ (z) = yfWs (zy)fWn (y)dy (1)
0
where fWs (.) and fWn (.) are the PDFs of the desired signal
power and the interference power, respectively. The cumulative
distribution function (CDF) is then expressed as
   z0
Ws
FZ (z0 ) = Prob ≤ z0 = fZ (z)dz. (2)
Wn 0
If n is known and fixed, the resulting conditional capture
probability may be expressed as Pcp (z0 |n) = 1 − FZ (z0 ).
Of course, the statistics of Ws and Wn depend on the channel
characteristics, and, in the current work, on the fading model
used.
In order to produce the unconditional capture probability,
let γ be defined as the probability that a station starts a
transmission in a randomly chosen time slot, with γ assumed
to be constant across all time slots. Considering a scenario in
which there are N stations apt to transmit, the unconditional
probability of a test packet being able to capture the receiver
in an arbitrary transmission period may be expressed by

N −1  
N
Pcp (z0 ) = γ n+1 (1−γ)N −n−1 Pcp (z0 |n). (3)
n + 1
n=1
Hence, given the conditions described above, (3) yields the
probability of successful reception.
Furthermore, let S represent the normalized channel
throughput (or channel efficiency), defined as the fraction
of time the channel is used to successfully transmit user
information. It can be expressed as S = G Pcp (z0 ), where
G is the offered traffic. In other words, S is evaluated as the
portion of the offered traffic that is successfully received.
LEONARDO and YACOUB: EXACT FORMULATIONS FOR THE THROUGHPUT OF IEEE 802.11 DCF IN HOYT, RICE, AND NAKAGAMI-M FADING . . . 2263

It is interesting to note that the CDF in (2) also represents delay is uniformly chosen in the interval [0, CWi − 1],
another important measure of performance for wireless sys- where CWi is the contention window size at the backoff
tems: the outage probability. This is defined as the probability stage i, i = 0, . . . , m. At the first transmission attempt
that the SIR at the reference receiver falls below a certain (i = 0), the contention window size is set to its minimum
specified threshold required for successful reception. There- value CW0 = CWmin . After each unsuccessful transmission,
fore Pout (z0 ) = FZ (z0 ), where z0 is the successful reception the backoff stage i is incremented up to the value of m and
threshold. the contention window is doubled up to its maximum value
Another relevant aspect of our investigation is how the CWm = CWmax = 2m CWmin .
interfering signal arises. In a wireless system, it typically Consider a setting with unsaturated traffic generated by N
results from signals arriving at the receiver’s antenna from contending stations, non-ideal transmission channel, capture
multiple transmitters. Depending on how these random signals effect, and that the number of packets generated in the network
combine during the observation interval, one of two scenarios for new messages plus retransmissions follows the Poisson
might occur [35]: coherent addition or incoherent addition. distribution, with mean generation rate of λ packets per
Coherent addition occurs if the carrier frequencies are second. For such a scenario, Daneshgaran et al. [14] use a two-
equal and if the random phase fluctuations are small during dimensional Markov process to model the protocol’s behavior
the capture time tw . For instance, coherent addition might and produces a channel throughput given by
happen when the deviation caused by the phase modulation
Pt Ps (1 − Pe ) E{P L}
is very small, and the observation interval is short compared S= , (4)
(1−Pt )σ+Pt (1−Ps )Tc +Pt Ps (1−Pe )Ts +Pt Ps Pe Te
to the modulation rate. In other words, coherent addition
should be considered if the random phases of the individual in which: (i) Pt is the probability that, in the considered time
interferers barely vary during the capture period. Incoherent slot, at least one of the contending stations is transmitting; (ii)
addition occurs if the phases of the individual signals fluctuate Ps is the probability that a packet transmission is successful;
significantly due to mutually independent modulation [13], (iii) Pe is the probability that errors due to the channel may
[36]. occur on a transmitted packet; (iv) E{P L} is the average
Let the phasor x(t) = Re{r(t) ej[wc t+θ(t)] } represent a packet payload length; (v) σ is the duration of an empty
signal reaching the receiver’s antenna, where r(t) and θ(t) time slot; and (vi) Tc , Te and Ts are the average times a
are the random envelope and phase, respectively, and wc is channel is sensed busy due to a collision, an error affected data
the carrier’s angular frequency. For the coherent addition of n frame transmission and a successful data frame transmission,
signals, the resulting phasor is [13] xn (t) = ni=1 xi (t) where respectively. In order to calculate the channel throughput in
the subscripts n and i represent the aggregate and individual (4), the analysis leads to the set of equations given below,
variables, respectively. which should be treated as a nonlinear system and should be
For the incoherent addition, the interference power wn solved numerically [14]:
experienced during the observation interval is  the sum of ⎧
n ⎪ Pt = 1 − (1 − γ)N (5)
the
n individual signals’ powers w , i.e., w = i=1 wi = ⎪
⎪
∗ ∗
i n
⎪
⎪ N γ(1 − γ) N −1
+ Pcp (z0 )
i=1 xi (t)xi (t), where xi (.) is the complex conjugate of ⎪
⎪ Ps = (6)
⎪
⎪
phasor xi (.). Considering the current work, where the signal ⎪
⎪ P t
⎪
⎪
power is a random variable, the PDF of the joint inter- ⎪
⎪ Pcol = 1 − (1 − γ)N −1 − Pcp (z0 ) (7)
⎪
⎪
ference power is therefore the convolution of the PDFs of ⎨Peq = (1 − Pe )(1 − Pcol ) = Pe + Pcol − Pe Pcol (8)
all contributing signal powers. If the individual components 2
⎪
⎪ (9)
are independent and identically distributed (i.i.d.), then the ⎪
⎪
γ= 1−(2P )m
⎪
⎪ (CW+1)+CW Peq 1−2Peqeq +2(1−Peq ) 1−q
interference power is expressed as the n-fold convolution of ⎪
⎪ q
⎪
⎪
the PDF of the individual signal power. ⎪
⎪ q = 1 − e−λE{Sts } (10)
⎪
⎪
In the next section, the concepts introduced here are tailored ⎪
⎪ }
⎪
⎩
E{S ts = (1−P t )σ + Pt (1−P )T
s c + P P
t s (1−P )T
e s
to the specific scenario of the current work.
+Pt Ps Pe Te , (11)

III. S YSTEM M ODEL
A. Medium Access Mechanism
The IEEE 802.11 DCF standard [1] is the communication
system considered in this work. It uses CSMA/CA as its
medium contention resolution algorithm and, in essence, a
terminal ready to transmit first senses the channel for a
period of at least a Distributed Interframe Space (DIFS). If
it senses the channel idle, it transmits its packet. Otherwise,
it schedules the (re)transmission of the packet to a later time
according to some randomly distributed retransmission delay.
After the retransmission delay has elapsed, the terminal repeats
the procedure described above. A binary exponential backoff
algorithm is used to determine the retransmission delay. The
in which: (i) Pcol is the probability that collisions may occur
on a transmitted packet; (ii) Peq is the probability of failed
transmission; (iii) q is the probability that there is at least one
packet to be transmitted in the buffer; and (iv) E{Sts } is the
expected time per slot.
In the current work, the Poisson model is used to describe
the traffic generation process, i.e., the packet transmission
requests from the upper layers. Accordingly, the resulting
inter-arrival time is exponentially distributed. Although the
Poisson model may not well represent bursty packet data
traffic, it offers an approach that is simple, tractable and
currently widely used [37]–[40].
For the sake of comparison, the analysis of channel through-
put for the CSMA/CA is also performed. In this case, a similar
2264 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 12, NO. 5, MAY 2013

analysis to the one presented above for the IEEE 802.11 DCF where w = a2 + 2σ 2 = 2σ 2 (κ + 1), and κ  a2 /(2σ 2 ) is
is assumed, with a difference that for the former a single- the power ratio between the dominant and scattered signals.
stage backoff algorithm is used to determine the retransmission It is easy to see that if κ is set to zero, thus eliminating the
delay, i.e., m = 0, with CWmin = CWmax dominant component, the Rice distribution simplifies to the
Rayleigh one.
B. Channel Models Similarly to the Hoyt case described above, the coherent
For the calculations presented in this subsection, as well as addition of n uncorrelated Rice phasorsproduces a Rice
n
for the remaining of this paper, let r represents the received phasor with aggregate  mean value w n = i=1 w i , dominant
2 n 2
signal envelope, w = r2 the received signal power, and w its signal power an = i=1 ai , and scattered signal power
n
average value. σn2 = i=1 σi2 . If the n phasors are i.i.d., the resulting phasor
1) Hoyt Fading Channel: The Hoyt fading model assumes has w n = nwi , a2n = na2i , σn2 = nσi2 , and κn = κi .
that the received signal is the result of the sum of a large 3) Nakagami-m Fading Channel: In a Nakagami fading
number of multipath scattered waves, without the prevalence channel, PDF of the signal envelope is given by [26]
of a single component (for instance, the LOS signal) [41].
Let x and y be two independent Gaussian processes with zero 2r2m−1  m m − mr22
fR (r) = e r̂ (16)
mean and variances σx2 and σy2 , respectively. The PDF of the Γ(m) r̂2
received signal envelope can be expressed as [41]
  2  where r̂2  E{r2 } is the mean square value, m ≥ 0 is a
2
r − r4 σ12 + σ12 r 1 1 fading parameter, and Γ(.) is the gamma function [42, Eq.
fR (r) = e x y I0 − 2 (12)
σx σy 4 σy2 σx 6.1.1]. For m = 1/2, the Nakagami distribution reduces to
the one-sided Gaussian PDF; for m = 1, it reduces to the
where Iν (.) is the modified Bessel function of the first kind
Rayleigh PDF while m → ∞ corresponds to a non-fading
and ν-th order [42, Eq. 9.6.10].
situation. The signal power PDF may be expressed as
The signal power PDF may be expressed as
√
h −h w  w  wm−1  m m − mw
fW (w) = e w I0 H (13) fW (w) = e w . (17)
w w Γ(m) w
 √ 2  
where w = σx2 + σy2 , h  14 √1η + η , H  14 η1 − η , For the Nakagami-m channel, treatment of coherent addi-
tion of signals is still under investigation by the authors. The
and η  σx2 /σy2 is the power ratio between the in-phase and
main problem is that the exact analysis leads to very intricate
quadrature signals. Knowing that η > 0, it can be seen that the
formulation and the idea to use some approximation has not
PDF in (13) is symmetrical around η = 1 [43]. Therefore, as
yet produced interesting results.
far as the signal power distribution is concerned, considering
either of the ranges η ≤ 1 or η ≥ 1 suffices. Also, it is easy
to see that if η is set to unity, thus making σx2 = σy2 , the Hoyt IV. A NALYTICAL R ESULTS
distribution simplifies to the Rayleigh one.
For the remaining of this paper and wherever applicable, the
Considering the way the Hoyt phasor is produced, with its
subscripts s, i and n are used to represent the desired signal
in-phase and quadrature Gaussian components, it is easy to
variables, the interference signal’s individual component vari-
see that the coherent addition of n uncorrelated Hoyt phasors
ables, and the joint interference signal variables, respectively.
also
n produces a Hoyt phasor with aggregate mean
n value wn =
2 2 Also, for compactness, let z̃0 be defined as z̃0  z0 /(w s /w n )
i=1 w i , in-phase signal powerσX,n = i=1 σX,i , and where the ratio w s /w n is commonly denoted as the average
2 n 2
quadrature signal power σY,n = i=1 σY,i . If the n phasors
2 2
SIR.
are i.i.d., the resulting phasor has w n = nwi , σX,n = nσX,i ,
2 2
σY,n = nσY,i , and ηn = ηi .
2) Rice Fading Channel: The Rice fading model assumes A. Perfect Power Control
that the received signal is the result of a dominant component
1) Hoyt Fading Channel: For the calculations presented in
(such as a direct LOS signal) added to a large number of
this subsection, let (13) represent the desired signal power
multipath scattered waves. Let x and y be two independent
PDF as well as, with different parameters, the signal power
Gaussian processes with zero mean and equal variances σ 2 .
PDF of an individual component of the interference signal.
The in-phase and quadrature components of the signal enve-
If coherent addition of phasors is assumed, then (13) should
lope in a Rice fading channel can be expressed as x + a and
be used to represent the joint interference signal power PDF.
y, respectively, where the constant a represents the envelope
With (1) and (2), changing the integration order, using [42, Eq.
of the dominant signal (also, the mean value for the in-phase
9.6.10], integrating over z (see [44, Eq. 3.351.1]), and using
component). The PDF of the received signal envelope can be
[45], the integral solves to
expressed as [23]
 ar  ∞  2i 
r r2 +a2
v1  (2i)! Hs
2i
fR (r) = 2 e− 2σ2 I0 . (14) FZ (z0 ) =1 − √ (1 − v1 )j
σ σ2 hs hn i=0 (i!)2 2hs
The signal power PDF may be expressed as j=0
  2  (18)
   j+1 j Hn
κ + 1 −(κ+1) w w × 2 F1 , + 1, 1, v1
fW (w) = e w I0 2 κ(κ + 1) (15) 2 2 hn
w eκ w
LEONARDO and YACOUB: EXACT FORMULATIONS FOR THE THROUGHPUT OF IEEE 802.11 DCF IN HOYT, RICE, AND NAKAGAMI-M FADING . . . 2265

TABLE I
R ELATION BETWEEN NUMBER OF TERMS AND ACCURACY IN THE individual signal power PDF which, on its turn, is expressed
INFINITE SUMMATION OF (24). by (13). This calculation gives
Parameters smallest J for accuracy of √ n+1 n2  n−1
2
 
n ηs ηn z̃0 3-decimal-place 6-decimal-place πn 2 hn wn −nhn w n wn
fWn(wn)= e w n In−1 nH
n (21)
1 0.01 0.01 0.1 27 59 Γ( n2 ) w n 2Hn w n 2 wn
1.0 117 274
5.0 167 423 where w n = nw i , ηn = ηi , hn = hi and Hn = Hi . It can be
0.1 1.0 115 265 seen that the n-fold convolution used to obtain (20) produced
5.0 169 421 the η-μ distribution for μ = n/2 [43].
0.99 1.0 104 217
5.0 179 415
With (1) and (2), changing the integration order, using [42,
0.1 0.01 1.0 14 31 Eq. 9.6.10], integrating over z (see [44, Eq. 3.351.1]), and
5.0 18 44 using [45], the integral solves to
0.99 1.0 13 26
∞
  2i 2i
5.0 19 44 vn (2i)! Hs  Γ(j +n)
0.99 0.01 1.0 1 3 FZ (z0 ) = 1− n
5.0 1 3 Γ(n) hs hnn i=0 (i!)2 2hs j=0 j!
0.99 1.0 1 3   2 
j +n j +n+1 n+1 Hn
5.0 2 3 × (1−vn ) 2 F1
j
, , , vn (22)
10 0.01 0.01 1.0 67 110 2 2 2 hn
5.0 186 337
0.99 1.0 57 86 where
5.0 181 293 nhn
0.99 0.01 1.0 1 3 vn = . (23)
nhn + hs z̃0
5.0 1 3
0.99 1.0 1 3 Changing the summation order, using [42, Eq. 6.1.18] and
5.0 1 3 [45], the signal-to-interference CDF may be expressed as
50 0.01 0.01 5.0 170 243
∞  2cj
0.99 5.0 163 217 vn (1−vn )j Hs
0.99 0.01 5.0 1 3 FZ (z̃0 ) = 1−  n
0.99 5.0 1 3 πhs hnn j=0 (j +n)B(j +1,n) hs
    2 
Γ cj + 12 j +n j +n+1 n+1 Hn
where 2 F1 (.) is the Gauss hypergeometric function [42, Eq. × 2 F1 , , , vn (24)
Γ(cj +1) 2 2 2 hn
15.1.1], and   2 
hn 1 Hs
v1 = . (19) × 2 F1 1, cj + , cj +1,
hn + hs z̃0 2 hs
Changing the summation order, using [42, Eq. 6.1.18] and
[45], the signal-to-interference CDF may be expressed as where B(.) is the beta function [42, Eq. 6.2.2], and cj as
∞
  2 already defined.
v1  j +1 j Hn Although (24) includes an infinite summation, the evalua-
FZ (z̃0 ) = 1− √ j
(1−v1) 2 F1 , +1,1, v1
πhs hn j=0 2 2 hn tion of the CDF converges rapidly for cases of interest. Let J
 2cj     2 be the number of terms in a truncated summation (as defined
Hs Γ cj+12 1 Hs earlier). Table I gives the value of J necessary to obtain
× 2 F1 1, cj + , cj +1, (20)
hs Γ (cj+1) 2 hs a three-decimal-place accuracy (error < 0.0005) and a six-
decimal-place accuracy (error < 0.0000005) for the infinite
where cj = j/2, . is the ceiling function (defined as x =
summation of (24).
k, with x ∈ R and k being the smallest integer such that x ≤ k).
Although (20) includes an infinite summation, the evalua- 2) Rice Fading Channel: For the calculations presented in
tion of the CDF converges rapidly for cases of interest. In order this subsection, let (15) represent the desired signal power
to estimate the error if the summation in (20) is truncated, PDF as well as, with different parameters, the signal power
all summands whose absolute value is larger than 10−20 are PDF of an individual component of the interference signal.
calculated, and those smaller than this value are discarded If coherent addition of phasors is assumed, then (15) should
because it was observed that they do not affect the desired be used to represent the joint interference signal power PDF.
accuracy. The error of the truncated summation may now be With (1) and (2), changing the integration order, using [42, Eq.
estimated and, of course, these results vary depending on the 9.6.10], integrating over z (see [44, Eq. 3.351.1]), and using
given set of parameters used. Let J be defined as the number [45], the integral solves to
of terms in a truncated summation, i.e., 0 ≤ j < J. The entries ∞
u1  κis
i
for n = 1 in Table I give the value of J necessary to obtain FZ (z0 ) = 1− κs +κn (1−u1 )j 1 F1 (j+1, 1, κn u1 ) (25)
e i!
a three-decimal-place accuracy (error < 0.0005) and a six- i=0 j=0
decimal-place accuracy (error < 0.0000005) for the infinite
where 1 F1 (.) is the Kummer confluent hypergeometric func-
summation of (20).
tion [42, Eq. 13.1.2], and
For the incoherent addition, assume that the interference
signal is composed of n i.i.d. variables. As a result, the joint (κn + 1)
u1 = . (26)
interference signal power PDF is the n-fold convolution of the (κn + 1) + (κs + 1) z̃0
2266 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 12, NO. 5, MAY 2013

TABLE II
R ELATION BETWEEN NUMBER OF TERMS AND ACCURACY IN THE by (15). This calculation gives
INFINITE SUMMATION OF (31).
 n+1  n−1
Parameters smallest J for accuracy of nκn +1 2 wn 2 −n(κn +1) wwn
fWn(wn ) = nκn e n
n κs κn z̃0 3-decimal-place 6-decimal-place e wn κn
1 0.01 0.01 0.1 2 2
   (28)
wn
1.0 2 3 × In−1 2n κn (κn + 1)
5.0 2 3 wn
1.0 1.0 2 3
5.0 2 3
where w n = nwi , and κn = κi . It can be seen that the n-fold
10.0 1.0 2 3 convolution used to obtain (28) produced the κ-μ distribution
5.0 2 3 for integer values of μ = n [43].
1.0 0.01 1.0 5 8 With (1) and (2), changing the integration order, using [42,
5.0 5 8
Eq. 9.6.10], integrating over z (see [44, Eq. 3.351.1]), and
10.0 1.0 5 8
5.0 5 9 using [45], the integral solves to
10.0 0.01 1.0 18 26 ∞
5.0 18 26 unn  κis  Γ(j + n)
i

10.0 1.0 19 25 FZ (z0 ) =1 − (1 − un )j

eκs +nκn i=0
i! j=0
Γ(n)j! (29)
5.0 17 26
10 0.01 0.01 1.0 2 3 × 1 F1 (j + n, n, nκn un )
5.0 2 3
10.0 1.0 2 3 where
5.0 1 3 n(κn + 1)
10.0 0.01 1.0 19 25 un = , (30)
5.0 12 25
n(κn + 1) + (κs + 1) z̃0
10.0 1.0 18 23
Changing the summation order, using [44, Eqs. 1.211.1 and
5.0 0 15
50 0.01 0.01 5.0 1 3
8.352.2], and the signal-to-interference CDF may be expressed
10.0 5.0 1 3 as
10.0 0.01 5.0 0 18 ∞
10.0 5.0 0 0 unn  (1 − un )j [1 − Q(j, κs )]
FZ (z̃0 ) =1 −
enκn j=0 (j + n)B(j + 1, n) (31)
× 1 F1 (j + n, n, nκn un ).
Changing the summation order, using [44, Eqs. 1.211.1 and
8.352.2], and the signal-to-interference CDF may be expressed If both κs and κn are set to zero in (31), the CDF simplifies to
as the Rayleigh channel model. If this result is further simplified
by assuming that w s = w i = w n /n (hence making z̃0 = nz0 ),
∞
then the same expression presented in [13, Eq. 20.a] is found.
u1  Although (31) includes an infinite summation, the evalua-
FZ (z̃0 ) =1 − (1 − u1 )j [1 − Q(j, κs )]
eκn j=0 (27) tion of the CDF converges rapidly for cases of interest. Let J
be the number of terms in a truncated summation (as defined
× 1 F1 (j + 1, 1, κn u1 )
earlier). Table II gives the value of J necessary to obtain
a three-decimal-place accuracy (error < 0.0005) and a six-
decimal-place accuracy (error < 0.0000005) for the infinite
where Q(.) is the regularized incomplete gamma function de-
summation of (31).
fined as Q(a, b) = Γ(a, b)/Γ(a), and Γ(., .) is the incomplete
gamma function [42, Eq. 6.5.3]. 3) Nakagami-m Fading Channel: For the calculations pre-
sented in this subsection, let (17) represent the desired signal
If both κs and κn are set to zero in (27), the CDF power PDF as well as, with different parameters, the signal
simplifies to the Rayleigh channel model. If this result is power PDF of an individual component of the interference
further simplified by assuming that w s = wi = wn /n (hence signal.
making z̃0 = nz0 ), then the same expression presented in [13, For the incoherent addition, assume that the interference
Eq. 20.b] is produced. signal is composed of n i.i.d. variables. As a result, the joint
Although (27) includes an infinite summation, the eval- interference signal power PDF is the n-fold convolution of the
uation of the CDF converges rapidly for cases of interest. individual signal power PDF which, on its turn, is expressed
Let J be the number of terms in a truncated summation (as by (17). This calculation yields
defined earlier). The entries for n = 1 in Table II give the  mn  
wnmn −1 mn mn wn
value of J necessary to obtain a three-decimal-place accuracy fWn (wn ) = exp − (32)
(error < 0.0005) and a six-decimal-place accuracy (error Γ(mn ) wn wn
< 0.0000005) for the infinite summation of (27). where mi and mn = nmi are the individual and the joint
For the incoherent addition, assume that the interference fading parameters, respectively, and w n = nwi is the joint
signal is composed of n i.i.d. variables. As a result, the joint mean power. It can be seen from (17) and (32) that both
interference signal power PDF is the n-fold convolution of the signal power and interference power are described by the same
individual signal power PDF which, on its turn, is expressed distribution, except that they have distinct parameters.
LEONARDO and YACOUB: EXACT FORMULATIONS FOR THE THROUGHPUT OF IEEE 802.11 DCF IN HOYT, RICE, AND NAKAGAMI-M FADING . . . 2267

Using the appropriate expressions in (1), and after some The PDF of the received packet mean power fW (w) =
manipulation, the signal-to-interference PDF is found as  dρ 
fG (ρ)  dw  is calculated using (38) and (40) and can be
1 w n ms ms −1 expressed as
fZ (z) = b (1 − b)mn +1 (33)
B(ms , mn ) w s mn 2π
fW (w) = G(w −1/α ). (41)
where α Gt w 1+2/α
w n ms z
b= . (34) The capture probability for spatial coverage is now con-
w n ms z + w s mn
sidered. Given an arbitrary spatial traffic density G(ρ), (41)
The corresponding CDF may be expressed as [46], [47] can be used to calculate the PDF of the test packet mean
FZ (z̃0 ) = Ibn (ms , mn ) = 1 − I1−bn (mn , ms ) (35) power fW s (ws ). The PDF of the mean interference power of
n packets fW n (w n ) is calculated by convolving (41) n times.
where Ix (.) is the regularized incomplete beta function [42, With these results and assuming that the signal power and the
Eq. 6.6.2], and interference power are statistically independent, the CDF of
ms z̃0 the signal-to-interference ratio can be calculated as
bn = . (36)
ms z̃0 + mn  z0  ∞  ∞
FZ (z0 ) = dz dw n fZ (z)fW n(w n )fW s(w s ) dw s (42)
0 0 0
B. Spatial Coverage
The analysis presented above for the incoherent addition where fZ (.) is given by (1). With (3) and (42), the capture
of interferers assumes that the components of the interference probability is then calculated. As an example, let us use the
signal have identical mean power wi , i = 1, . . . , n. This quasi-constant traffic density given in [13], expressed as
restriction limits the results to systems where perfect power Gt   π 
control is employed or to terminals placed at a fixed distance G(ρ) = ξ exp − ξρ4 , ξ > 0, ρ ≥ 0. (43)
π 4
from the receiving antenna (i.e., on a circular ring) and in an
environment without any shadowing effects. Let us now extend It is possible to see that the traffic density is roughly constant
the model presented above to include the case of packets inside the cell of radius ρ = 1, falling rapidly beyond the cell
arriving with different mean powers, e.g., from terminals boundary. If we select α = 4, it can be seen that
with a given spatial distribution across the cell radius and √  
ξ πξ
at different transmission distances to the receiver’s antenna. fW s (w s ) = 3/2
exp − (44)
Therefore, the statistical behavior of the packet mean power 2ws 4ws
needs to be specified and taken into account.
and √  
The mean power of a packet received from a terminal at a
n ξ n2 πξ
distance d is of the general form [23] fW n (w n ) = 3/2
exp − . (45)
2w n 4wn
w = Λ d−α (37)
Using these results in (42), and considering the Nakagami-
where α gives the channel attenuation with the distance, and m channel, where fZ (.) is given by (33), the signal-to-
Λ is a value that depends on, e.g., the transmit power and the interference CDF is given by
height and gain of the antennas. Typical values of the exponent ms
α are α = 2 in free space and α = 4 in urban land mobile c 2 F1 (ms , ms +mn , ms +1, c0 )
FZ (z0 ) = sec(ms π) 0
cellular systems. Using a similar approach to that presented ms B(ms , mn )
 
in [13], let ρ  d Λ−1/α be defined and used to rewrite (37) √ Γ mn + 2 1
− 2 c0   (46)
as Γ(ms )Γ(mn )Γ −ms + 32
w = ρ−α . (38)  
1 1 3 3
× 3 F2 , 1, mn + , , −ms + , c0
2 2 2 2
Let the offered traffic density function G(ρ) be defined as
the number of packets offered per transmission period per unit where 3 F2 (.) is a generalized hypergeometric function [44,
of area at distance
 ∞ ρ. The total offered traffic can be calculated Eq. 9.14.1], and
by Gt = 2π 0 ρ G(ρ) dρ. ms
The spatial CDF of the offered traffic as a function of the c0 = n 2 z 0 . (47)
mn
distance ρ can be expressed as
Note that the parameter ξ does not influence the results in (46).
FG (ρ) = Prob{packet generated within distance ρ} This is due to the fact that it appears as a multiplicative factor

2π ρ (39) of the mean power (Eqs. 44 and 45), and in (42) the average
= u G(u) du
Gt 0 is performed over the ratio of these powers. In physical terms,
this parameter simply indicates more or less concentration of
and the corresponding PDF is traffic within the center of the cell, which impacts equally on
2π the effect of the desired signal as well as of the interference
fG (ρ) = ρ G(ρ). (40) ones.
Gt
2268 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 12, NO. 5, MAY 2013

TABLE III 100

T YPICAL NETWORK PARAMETERS . curve 1 2 3 4 5 6 7 8
κ 0 0 1 2 5 5 10 10
MAC header 34 bytes 1z
j n 50 1 1 1 1 50 1 50
PHY header 24 bytes
z
2j
10−1 z
3j
ACK frame 14 bytes + PHY header 5j
z z
4j
packet payload 1020 bytes z
6j
channel bit rate 1 Mbps z
7j

Pout
10−2
DIFS 50 μs
SIFS 10 μs z
8j
ACK timeout 300 μs
σ 20 μs 10−3
m 5
CWmin 8
10−4
100 0 5 10 15 20 25 30
1/z̃0 [dB]

1z
j j j
z 5z Fig. 2. Outage probability for incoherent Rice channel with κ = κs = κn .
z3 j
2j 4z 6j
z
10−1 The dashed lines correspond to the Rayleigh channel (κ = 0).

100
Pout

10−2 1j
z
z2j
3j
z
4j
z
10−1 z
5j
10 −3 z
6j
curve 1 2 3 4 5 6
η 1 1 0.1 0.1 0.01 0.01 Pout
n 50 1 50 1 50 1 10−2
10−4
0 5 10 15 20 25 30
1/z̃0 [dB] 10−3
curve 1 2 3 4 5 6
Fig. 1. Outage probability for incoherent Hoyt channel with η = ηs = ηn . m 0.5 0.5 1 1 2 2
The dashed lines correspond to the Rayleigh channel (η = 1). n 50 1 50 1 1 50

V. N UMERICAL R ESULTS
This section presents numerical results assuming, wherever
applicable and unless otherwise indicated, the networks pa-
rameters listed in Table III. These parameters belong to the
IEEE 802.11b protocol. However, the mathematical models
used here hold for any wireless protocol with similar MAC
functionality.
A. Outage Probability
The outage probabilities for the Hoyt, Rice and Nakagami-
m channels, considering incoherent addition of interfering sig-
nals and perfect power control, are pictured in Figs. 1, 2 and 3,
respectively. In all cases the outage probability decreases with
higher values of average SIR, as expected. In addition, using
the results for the Rayleigh channel as a reference, the figures
allow a comparison between these channels with regards to the
outage probabilities. Considering a line of decreasing fading
intensity, the Rice channel spans from Rayleigh fading (κ = 0)
to a no-fading situation (κ → ∞); the Hoyt channel ranges
from one-sided Gaussian fading (η = 0) to Rayleigh fading
(η = 1); and the Nakagami-m channel varies from one-
sided Gaussian fading (m = 1/2) to a no-fading situation
(m → ∞), with Rayleigh fading (m = 1) in-between. The
figures show that higher fading intensity values translate into
higher outage probabilities. Also, considering the average SIR
value constant, the number of interfering signals n has only a
minor influence on the outage probability.
10−4
0 5 10 15 20 25 30
1/z̃0 [dB]
Fig. 3. Outage probability for incoherent Nakagami-m channel with m =
ms = mi . The dashed lines correspond to the Rayleigh channel (m = 1).
B. Channel Throughput: Perfect Power Control
With the results obtained in the previous sections, using
(3), (4), (31), (24) and (35), it is now possible to calculate
the IEEE 802.11 DCF channel throughput for the Hoyt, Rice
and Nakagami-m channels, assuming incoherent addition of
interfering signals and perfect power control. Let τ be the
worst case propagation delay and τ̃ = τ /p its normalized
version. For the Hoyt channel, the throughput for various
values of η = ηs = ηn , z̃0 , τ̃ and Pe is shown in Fig. 4. For
the Rice channel, Fig. 5 depicts the throughput for various
values of κ = κs = κn , z̃0 , τ̃ and Pe . For the Nakagami-m
channel, considering various values of m = ms = mi , z̃0 , τ̃
and Pe , the throughput is presented in Fig. 6.
The figures indicate that the behavior of the throughput S
as a function of the packet rate λ consists, basically, of two
regions: a linear growth region where S is a linear function of
λ, and a saturation region where S remains almost constant.
The transition between the two regions may be characterized
by a peak in the throughput, in particular if the number of
nodes N is high.
In addition, the throughput graphs clearly indicate the
important role the normalized worst case propagation delay
τ̃ plays to determine the channel throughput. Also, higher
normalized capture threshold z̃0 values mean lower channel
LEONARDO and YACOUB: EXACT FORMULATIONS FOR THE THROUGHPUT OF IEEE 802.11 DCF IN HOYT, RICE, AND NAKAGAMI-M FADING . . . 2269

1.0 1.0
j
1z
1j
z2j
z3j
zz 2j
zj
3z4jzj
4j5z j 5z j
zj
7j 6z
6z7z
j
0.8 0.8
j 8z9jz
8j
z zj
j
z
10j
9j
zz
10j
11
12z jz
j
zz
j
11z zj
j 13
0.6 12 13j z 0.6 14 15z16j
z
14j
S

S
0.4 0.4

curve 1 2 3† 4 5 6 7 8‡ 9 10 11‡ 12 13 14‡ curve 1 2 3 4† 5 6 7 8 9 10‡ 11 12 13‡ 14 15 16‡

0.2 N 4 4 4 10 10 4 4 4 10 10 10 20 20 20 0.2 N 4 4 4 4 10 10 4 4 4 4 10 10 10 20 20 20
η 1 0.1 0.1 1 0.1 0.1 1 0.1 0.1 1 0.1 0.1 1 0.1 m 5 1 0.5 0.5 1 0.5 0.5 1 5 0.5 0.5 1 0.5 0.5 1 0.5
z̃0 [dB] 6 6 6 6 6 20 20 20 20 20 20 20 20 20 z̃0 [dB] 6 6 6 6 6 6 20 20 20 20 20 20 20 20 20 20
Note: τ̃ = 0.01 and Pe = 0 except †: τ̃ = 2, and ‡: Pe = 0.1. Note: τ̃ = 0.01 and Pe = 0 except †: τ̃ = 2, and ‡: Pe = 0.1.
0.0 0.0
0 20 40 60 80 100 0 20 40 60 80 100
λ [pkt/s] λ [pkt/s]

Fig. 4. Throughput S for IEEE 802.11 DCF with 2-way handshake, Fig. 6. Throughput S for IEEE 802.11 DCF with 2-way handshake,
incoherent Hoyt channel, perfect power control and η = ηs = ηi = ηn . incoherent Nakagami-m channel, perfect power control and m = ms =
The dashed lines correspond to the Rayleigh channel (η = 1) mi = mn /n. The dashed lines correspond to the Rayleigh channel (m = 1).

1.0
b
b b
1.0
0.8
1z
j2j
z3j
zj Nakagami-m, m = 5
4z j
5z Rice, κ = 8.47
0.8
6j
zj
7z 0.6 Rayleigh
z
8j
9j
z
j
z
Nakagami-m, m = 0.51
S
10 Hoyt, η = 0.01
0.6 j
z12j
11 z
13j
z 0.4
14j
z
S

0.4
0.2
curve 1 2 3† 4 5 6 7 8‡ 9 10 11‡ 12 13 14‡
0.2 N 4 10 4 4 10 4 4 4 10 10 10 20 20 20
κ 10 10 10 0 0 0 10 10 0 10 10 0 10 10 0.0
z̃0 [dB] 6 6 6 6 6 20 20 20 20 20 20 20 20 20 0 20 40 60 80 100
Note: τ̃ = 0.01 and Pe = 0 except †: τ̃ = 2, and ‡: Pe = 0.1.
0.0 λ [pkt/s]
0 20 40 60 80 100
Fig. 7. Throughput S for IEEE 802.11 DCF with 2-way handshake,
λ [pkt/s]
incoherent Hoyt, Rice, Nakagami-m and Rayleigh channels with perfect
Fig. 5. Throughput S for IEEE 802.11 DCF with 2-way handshake, power control, η = ηs = ηi = ηn , κ = κs = κi = κn and
incoherent Rice channel, perfect power control and κ = κs = κi = κn . m = ms = mi = mn /n, τ̃ = 0.01, N = 4, z̃0 = 6 dB and Pe = 0.
The dashed lines correspond to the Rayleigh channel (κ = 0).

with [26] 
1− 1
m −1
η=  . (48)
throughput, which is expected since it indicates a diminished 1+ 1
−1
m
ability of receiver detection of the intended signal among the
interfering signals. Examining same-z̃0 data, compared to the On the other hand, the Rice channel is approximated if m > 1,
Rayleigh channel results, higher values of κ in Rice channels with [26] √
tend to produce slightly lower throughput. The reason behind m2 − m
κ= √ . (49)
it is that higher κ implies a more deterministic scenario, m − m2 − m
approaching that of the wireline case. On the other hand, lower Of course, m = 1 produces the Rayleigh channel. Fig. 7
values of η in Hoyt channels tend to produce slightly higher illustrates these situations by showing the throughput for the
throughput when compared to Rayleigh channel results. The Hoyt, Rice and Rayleigh channels and equivalent curves for
reason for this is that lower η implies a more random scenario the matching Nakagami-m channel. It can be seen that the
which may increase the chance of capture. For the Nakagami- original and the matching curves are indistinguishable. These
m channel, depending on the value of m, throughput results results show that the formulation obtained for the spatial
can be higher or lower than those seen for the Rayleigh coverage for the Nakagami-m case can be directly applied
channel. In all cases, higher fading intensity values tend to to the Hoyt (1/2 < m < 1) and Rice (m > 1) cases, for
higher throughput, although the differences observed are rather which the exact formulations are mathematically intractable.
marginal. Also, in all cases, higher probability of channel error
Pe translates into lower throughput. C. Channel Throughput: Spatial Coverage
It is noteworthy that the Nakagami-m distribution approxi- With the results obtained in Section IV-B, using (3), (4)
mates Hoyt and Rice and includes Rayleigh. The Hoyt channel and (46), the throughput can be calculated for the Nakagami-
is approximated if the parameter m is such that 1/2 < m < 1, m channel model. Fig. 8 presents the throughput for various
2270 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 12, NO. 5, MAY 2013

1.0 1.0

1z
j2j
z3j
z
4j
zj 1j
z
j
z 5z6j
z7j
z jj
z
z2 3z
0.8 0.8
8 9j
z
zj
j 4j
10 z
11 5z
j6z
j
j
z13j
12 z14j
z
0.6 0.6
zz
7j8j9z
j
S

S
0.4 0.4
curve 1 2 3 4† 5 6 7 8 9 10† 11 12 13 14
N 4 4 4 4 4 4 4 4 4 4 10 10 10 10
0.2 m 5 1 0.5 0.5 0.5 1 5 5 0.5 0.5 1 0.5 1 5 0.2 curve 1 2 3 4† 5† 6† 7‡ 8‡ 9‡
z̃0 [dB] 6 6 6 6 20 20 20 6 6 6 6 20 20 20 m 5 1 0.5 5 1 0.5 5 1 0.5
Pe 0 0 0 0 0 0 0 0.1 0.1 0.1 0.1 0.1 0.1 0.1 Note: E{P L} = 1020 bytes except †: E{P L} = 520 bytes,
Note: 802.11 except †: CSMA/CA. and ‡: E{P L} = 128 bytes.
0.0 0.0
0 20 40 60 80 100 0 20 40 60 80 100
λ [pkt/s] λ [pkt/s]

Fig. 8. Throughput S for IEEE 802.11 DCF with 2-way handshake, incoher- Fig. 9. Throughput S for IEEE 802.11 DCF with 2-way handshake, incoher-
ent Nakagami-m channel with spatial coverage, m = ms = mi = mn /n ent Nakagami-m channel with spatial coverage, m = ms = mi = mn /n,
and τ̃ = 0.01. The dashed lines correspond to the Rayleigh channel (m = 1). τ̃ = 0.01, N = 10, z̃0 = 6 dB and Pe = 0. The dashed lines correspond to
The dotted lines correspond to CSMA/CA. the Rayleigh channel (m = 1).

values of m = ms = mi , z̃0 and Pe . In addition, it
shows how the IEEE 802.11 DCF compares against a simpler
version of CSMA/CA. In this case, CSMA/CA is assumed to
implement a single-stage backoff algorithm, i.e., m = 0, with
CWmin = CWmax = 256. Similarly to the perfect power
control scenario presented above, it can be seen that lower
values of z̃0 or Pe translates into higher throughput figures.
However, it seems that the influence of the fading parameter m
is somewhat weaker when compared to the results presented
earlier. On the other hand, the results indicate that z̃0 appears
to play a more significant role in determining the channel
performance. Also, the results, presented for N = 4, show that
the IEEE 802.11 DCF outperforms the CSMA/CA although
the difference seems narrower at the both ends of the curves,
i.e., with low and high packet rates. Also, the transition
between the linear and the saturation regions for CSMA/CA
is not as sharp as the one observed for the IEEE 802.11 DCF.
Fig. 9 presents the throughput for various values of m =
ms = mi and average packet payload length E{P L}. As
expected, the throughput is lower for shorter payload packets,
although the relation between packet payload length and
throughput is not a one-to-one. For instance, a reduction of
49% in the average packet payload length (from 1020 to 520
bytes) causes the saturated throughput to drop only about 8%,
and an 87% reduction (from 1020 to 128 bytes) causes the
saturated throughput to drop about 37%.
As already mentioned, it is important to note that since
the Nakagami-m distribution approximates Hoyt and Rice, the
former may be used to estimate the spatial coverage when the
desired channel model is either one of the latter.
VI. C ONCLUSIONS
This paper investigates the throughput performance of IEEE
802.11 DCF in a packet radio network and Hoyt, Rice and
Nakagami-m fading environments. Analytical and numerical
results are presented considering the signal capture model with
coherent and incoherent addition of interfering signals. The
approach used here includes the signal capture model with
uniform attenuation for all terminals (or perfect power control)
and unequal average power levels (or spatial coverage). The
results indicate that higher fading intensity, lower capture
threshold or lower propagation delay contributes to higher
channel throughput. Also, since the fading intensity values
of the Nakagami-m channel model approximate those found
in the Hoyt and Rice channels, the results presented for the
former channel offer a good indication of what range the latter
channels may exhibit. It is certainly of interest to extend these
results to more general fading scenarios, such as those of κ-μ,
η-μ and α-μ [48].
The IEEE 802.11 DCF algorithm and its corresponding
Markov model as used here, and that has been extracted
from [14], has already been fully validated in [14] by means
of simulation. Apart from the new channel model statistics
proposed here, which have no effect on the traffic model of
[14], the formulations of [14] are kept intact. The new fading
statistics added to the model of [14] have been fully validated
by means of numerical integration. Therefore, the full model
is validated.
R EFERENCES
[1] IEEE Standard 802.11, Wireless LAN Medium Access Control (MAC)
and Physical Layer (PHY) Specifications, 1999 Edition.
[2] C. Woo Pyo, H. Harada, “Throughput analysis and improvement of
hybrid multiple access in IEEE 802.15.3c mm-wave WPAN,” IEEE J.
Sel. Areas Commun., vol. 27, no. 8, pp. 1414–1424, Oct. 2009.
[3] T. H. Kim, J. Ni, and N. H. Vaidya, “A distributed throughput-optimal
CSMA with data packet collisions,” in Proc. 2010 IEEE Workshop
Wireless Mesh Netw., pp. 1-6.
[4] J. Ghaderi and R. Srikant, “On the design of efficient CSMA algorithms
for wireless networks,” in Proc. 2010 IEEE Conf. Decision Control, pp.
954–959.
[5] Y. Chang, C. P. Lee, and J. A. Copeland, “Goodput enhancement of
VANETs in noisy CSMA/CA channels,” IEEE J. Sel. Areas Commun.,
vol. 29, no. 1, pp. 236–249, Jan. 2011.
[6] H. C. Lee, “A MAC throughput over Rayleigh fading channel in the
802.11a/g/n-based mobile LAN,” in Proc. 2011 Int. Conf. Elect. Eng.
Inf., pp. 1–7.
[7] J. Dai and Y. Yamao, “CSMA/CA performance under fading environ-
ment with two-dimensional distribution of hidden terminal,” in Proc.
2011 IEEE Veh. Technol. Conf. – Fall, pp. 1–5.
[8] H. Cheng and Y. Yamao, “Performance analysis of CSMA/CA broadcast
relay network for ITS V2V communications,” in IEEE Proc. 2012 Veh.
Technol. Conf. – Spring, pp. 1–5.
[9] IEEE Standard 802.15.4, Local and metropolitan area networks, Low-
Rate Wireless Personal Area Networks (LR-WPANs), 2011 Edition.
[10] IEEE Standard 802.15.6, Local and metropolitan area networks, Wire-
less Body Area Networks, 2012 Edition.
LEONARDO and YACOUB: EXACT FORMULATIONS FOR THE THROUGHPUT OF IEEE 802.11 DCF IN HOYT, RICE, AND NAKAGAMI-M FADING . . .
[11] S. A. Qasmi and K. T. Wong, “The probability distribution of the carrier-
to-interference ratio (CIR) of a CSMA/CA ad hoc wireless network,” in
Proc. 2005 IEEE Military Commun. Conf., vol. 2, pp. 996–1001.
[12] S. Choudhury and J. D. Gibson, “Payload length and rate adaptation for
throughput optimization in wireless LANs,” in IEEE Proc. 2006 Veh.
Technol. Conf. – Spring, vol. 5, pp. 2444–2448.
[13] J. C. Arnbak and W. V. Blitterswijk, “Capacity of slotted ALOHA in
Rayleigh-fading channels,” IEEE J. Sel. Areas Commun., vol. 5, no. 2,
pp. 261–269, Feb. 1987.
[14] F. Daneshgaran, M. Laddomada, F. Mesiti, and M. Mondin, “Un-
saturated throughput analysis of IEEE 802.11 in presence of non
ideal transmission channel and capture effects,” IEEE Trans. Wireless
Commun., vol. 7, no. 7, pp. 1276–1286, Apr. 2008.
[15] G. Bianchi, “Performance analysis of the IEEE 802.11 distributed
coordination function,” IEEE J. Sel. Areas Commun., vol. 18, no. 3,
pp. 535–547, Mar. 2000.
[16] Y. S. Liaw, A. Dadej, and A. Jayasuriya, “Performance analysis of
IEEE 802.11 DCF under limited load,” in Proc. 2005 Asia-Pacific Conf.
Commun., vol. 1, pp. 759–763.
[17] I. Trigui, A. Laourine, S. Affes, and A. Stphenne, “Performance analysis
of mobile radio systems over composite fading/shadowing channels with
co-located interference,” IEEE Trans. Wireless Commun., vol. 8, no. 7,
pp. 3448–3452, July 2009.
[18] J. F. Paris and D. Morales-Jimnez, “Outage probability analysis for
Nakagami-q (Hoyt) fading channels under Rayleigh interference,” IEEE
Trans. Wireless Commun., vol. 9, no. 4, pp. 1272–1276, Apr. 2010.
[19] D. Morales-Jimnez and J. F. Paris, “Outage probability analysis for η-
μ fading channels,” IEEE Commun. Lett., vol. 14, no. 6, pp. 521–523,
June 2010.
[20] H. Tran, T. Q. Duong, and H.-J. Zepernick, “Performance of a spectrum
sharing system over Nakagami-m fading channels,” in Proc. 2010 Int.
Conf. Signal Process. Commun. Syst., pp. 1–6.
[21] R. Giacomelli, R. K. Ganti, and M. Haenggi, “Outage probability of
general ad hoc networks in the high-reliability regime,” IEEE/ACM
Trans. Netw., vol. 19, no. 4, pp. 1151–1163, Aug. 2011.
[22] J. F. Paris, “Outage probability in η-μ/η-μ and κ-μ/η-μ interference-
limited scenarios,” IEEE Trans. Commun., accepted for publication.
[23] M. D. Yacoub, Foundations of Mobile Radio Engineering. CRC Press,
1993.
[24] M. K. Simon and M.-S. Alouini, Digital Communication over Fading
Channels: A Unified Approach to Performance Analysis. John Wiley &
Sons, 2000.
[25] 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.
[26] M. Nakagami, “The m-distribution—a general formula of intensity
distribution of rapid fading,” in Proc. 1960 Symp. Statistical Methods
Radio Wave Propag., pp. 3–36.
[27] J. H. Kim and J. K. Lee, “Capture effects of wireless CSMA/CA
protocols in Rayleigh and shadow fading channel,” IEEE Trans. Veh.
Technol., vol. 48, no. 4, pp. 1277–1286, July 1999.
[28] N. Zhang et al., “Exploiting multiuser diversity in reservation random
access,” IEEE Trans. Wireless Commun., vol. 5, no. 9, pp. 2548–2554,
Sept. 2006.
[29] J.-P. Linnartz et al., “Near-far effects in land mobile random access
networks with narrow-band Rayleigh fading channels,” IEEE Trans. Veh.
Technol., vol. 41, no. 1, pp. 77–90, Feb. 1992.
[30] A. Kochut, A. Vasan, A. U. Shankar, and A. Agrawala, “Sniffing out the
correct physical layer capture model in 802.11b,” in Proc. 2004 IEEE
Intern. Conf. Netw. Protocols, pp. 252–261.
[31] J. Lee, W. Kim, S.-J. Lee, D. Jo, J. Ryu, T. Kwon, and Y. Choi, “An
experimental study on the capture effect in 802.11a networks,” in Proc.
2007 ACM Workshop Wireless Netw. Testbeds, Experimental Evaluation
Characterization, pp. 19–26.
[32] A. Iyer, C. Rosenberg, and A. Karnik, “What is the right model for
wireless channel interference?” IEEE Trans. Wireless Commun., vol. 8
, no. 5, pp. 2662–2671, May 2009.
[33] A. Chehri, P. Fortier, and P. M. Tardif, “Throughput-delay trade-off for
slotted Aloha multiple access with capture effect,” J. Comput. Science,
vol. 5, no. 9, pp. 630–634, Sept. 2009.
[34] A. Papoulis, Probability, Random Variables, and Stochastic Processes,
3rd edition. McGraw-Hill, 1991.
2271
[35] J.-P. Linnartz, Narrowband Land-Mobile Radio Network. Artech House,
1993.
[36] R. Prasad and A. Kegel, “Improved assessment of interference limits in
cellular radio performance,” IEEE Trans. Veh. Technol., vol. 40, no. 2,
pp. 412–419, May 1991.
[37] A. Sunny, J. Kuri, and S. Aggarwal, “Delay modelling for a single-hop
wireless mesh network under light aggregate traffic,” in Proc. 2011 Int.
Conf. Commun. Signal Process., pp. 271–275.
[38] S. H. Nguyen, H. L. Vu, and L. L. H. Andrew, “Performance analysis
of IEEE 802.11 WLANs with saturated and unsaturated sources,” IEEE
Trans. Veh. Technol., vol. 61, no. 1, pp. 333–345, Jan. 2012.
[39] I. McLean and K. Christensen, “Reducing energy use: a dual-channel
link,” IEEE Commun. Lett., vol. 16, no. 3, pp. 411–413, Mar. 2012.
[40] J. Mišić, S. Rashwand, and V. B. Mišić, “Analysis of impact of TXOP
allocation on IEEE 802.11e EDCA under variable network load,” IEEE
Trans. Parallel Distrib. Syst., vol. 23, no. 5, pp. 785–799, May 2012.
[41] R. S. Hoyt, “Probability functions for the modulus and angle of the
normal complex variate,” Bell Sys. Tech. J., vol. 26, p. 318–359, Apr.
1947.
[42] M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions,
10th. printing. National Bureau of Standards, 1972.
[43] M. D. Yacoub, “The κ-μ distribution and the η-μ distribution,” IEEE
Antennas Propag. Mag., vol. 49, no. 1, pp. 68–81, Feb. 2007.
[44] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and
Products, 5th. edition. Academic Press, 1994.
[45] Wolfram Research, Mathematica, v. 5.0, Champaign, Wolfram Research,
2003.
[46] A. Wojnar, “Unknown bounds on performance in Nakagami channels,”
IEEE Trans. Commun., vol. 34, no. 1, pp. 22–24, Jan. 1986.
[47] S. A. Al-Semari and N. Grami, “A general expression for the capacity of
slotted ALOHA under Nakagami fading,” in Proc. 1999 IEEE Wireless
Commun. Networking Conf., vol. 2, pp. 849–853.
[48] M. D. Yacoub, “The α-μ distribution: a physical fading model for the
Stacy distribution,” IEEE Trans. Veh. Technol., vol. 56, no. 1, pp. 27-34,
Jan. 2007.
Elvio J. Leonardo was born in Maringá, Brazil, in
1961. He received the B.S. and M.Sc. degrees in
Electrical Engineering from the School of Electrical
and Computer Engineering of the State University
of Campinas, Brazil, in 1984 and 1992, respectively.
From 1985 to 1992, Mr. Leonardo worked at as an
Engineer at the Research and Development Center
of Telebrás, the then holding of the Brazilian state-
owned telephone operators. In 1992 he moved to
Australia, where he first worked as a Research
Assistant at the Sydney University, joining later the
Motorola Australia Software Centre as a Software Engineer. In 1998 he
moved to Libertyville, USA, still working for Motorola, where he helped
the development of cellular phones. In 2002 he moved back to Brazil, where
he started his academic career. Currently he is an Assistant Professor at the
State University of Maringá. His general research interests include embedded
systems and wireless communications.
Michel Daoud Yacoub was born in Brazil in 1955.
He received the B.S.E.E. and the M.Sc. degrees from
the School of Electrical and Computer Engineering
of the State University of Campinas, UNICAMP,
Brazil, in 1978 and 1983, respectively, and the Ph.D.
degree from the University of Essex, U.K., in 1988.
From 1978 to 1985, he worked as a Research Spe-
cialist at the Research and Development Center of
Telebrás, Brazil, in the development of the Tropico
digital exchange family. He joined the School of
Electrical and Computer Engineering, UNICAMP,
in 1989, where he is presently a Full Professor. He consults for several
operating companies and industries in the wireless communications area.
He is the author of Foundations of Mobile Radio Engineering (Boca Raton,
FL: CRC, 1993), Wireless Technology: Protocols, Standards, and Techniques
(Boca Raton, FL: CRC, 2001), and the co-author of Telecommunications:
Principles and Trends (So Paulo, Brasil: Erica, 1997, in Portuguese). He holds
two patents. His general research interests include wireless communications.