Sensing-Throughput Tradeoff for Cognitive Radio

Networks
Ying-Chang Liang, Yonghong Zeng, Edward Peh, and Anh Tuan Hoang
Institute for Infocomm Research
21 Heng Mui Keng Terrace, Singapore 119613
Email: {ycliang, yhzeng, cypeh, athoang}@i2r.a-star.edu.sg
AbstractIn cognitive radio networks, the secondary network
(users) are allowed to utilize the frequency bands of primary
network (users) when they are not currently being used. To
support this function, the secondary users are required to sense
the radio frequency environment, and once the primary user is
found to be active, the secondary users have to vacate the channel
within certain amount of time. There are two parameters related
to channel sensing: probability of detection and probability of
false alarm. The higher the detection probability, the better the
primary users can be protected. However, from the secondary
users perspective, the lower the false alarm probability, the more
chances the channel can be reused, thus the higher the achievable
throughput for the secondary users. In this paper, we study the
fundamental tradeoff between sensing capability and achievable
throughput of the secondary users. Particularly, we study the
design of sensing slot duration to maximize the achievable
throughput for the secondary users under the constraint that the
primary users are sufciently protected. Using energy detection
scheme, we prove that there indeed exists an optimal sensing
time which provides the best tradeoff. Cooperative sensing is
also studied based on the methodology of the proposed sensing-
throughput tradeoff. Computer simulations are presented to
evaluate the proposed tradeoff methodology.
I. INTRODUCTION
With the proliferation of wireless communications tech-
nology in the last couple of decades, in many countries,
most of the available radio spectrum has been allocated. This
results in the spectrum scarcity which poses a serious problem
for the future development of the wireless communications
industry. On the other hand, careful studies of the usage pattern
reveal that most of the allocated spectrum experiences low
utilization. Recent measurements by Federal Communications
Commission (FCC) show that 70% of the allocated spec-
trum in US is not utilized. Furthermore, time scale of the
spectrum occupancy varies from milliseconds to hours [1].
This motivates the concept of frequency reuse that allows
secondary networks to borrow unused radio spectrum from
primary licensed networks (users).
The core technology behind frequency reuse is cognitive
radio [3], [4], for which one of the essential components
is channel sensing [7], i.e., the wireless devices can sense
the radio spectrum environment within their operating range
to detect frequency bands that are not occupied by primary
users. In December 2003, FCC issued a Notice of Proposed
Rule Making that identies cognitive radio as the candidate
for implementing negotiated/opportunistic spectrum sharing
[2]. In response to this, the IEEE has formed the 802.22
Working Group to develop a standard for wireless regional
area networks (WRAN) based on cognitive radio technology
[6]. WRAN systems operate on the VHF/UHF TV bands,
ranging from 54 MHz to 862 MHz. While doing so, they must
ensure that no harmful interference is caused to the incumbent
primary users, which, for the VHF/UHF bands, include TV
users and the FCCs Part 74 wireless microphones [6].
Figure 1 illustrates the topology of a WRAN system where
the primary users include TV users and wireless microphones,
and the secondary users include both WRAN base station
(BS) and WRAN customer premise equipments (CPEs). The
WRAN systems are designed to provide wireless broadband
access to rural and suburban areas, with the average coverage
radius of 33 km. The operating principle of WRAN is based
on opportunistic unlicensed access to temporarily unused TV
spectrum. To achieve that, WRAN systems will be able to
sense the spectrum, identify unused TV channels, and utilize
these channels for supporting broadband connections.
The fundamental objective for WRAN is to maximize the
spectrum utilization of the TV channels when they are not used
by the primary users. On the other hand, to protect the incum-
bents, whenever a channel is declared to be used by primary
users, the WRAN system has to vacate that channel within
certain amount of time. Thus channel sensing is of signicant
importance for cognitive radio networks. In WRAN, each
frame consists of one sensing slot (subframe) and one data
transmission slot (subframe). Associated with channel sensing
are two parameters: probability of detection and probability of
false alarm. The higher the detection probability, the better the
primary users can be protected. However, from the secondary
users perspective, the lower the false alarm probability, the
more chances the channel can be reused, thus the higher
the achievable throughput for the secondary users. As both
parameters are related to the sensing time allocated, thus
there could exist a tradeoff between sensing capability and
achievable throughput for the secondary network.
In this paper, we study the fundamental sensing-throughput
tradeoff for cognitive radio networks. Particularly, we study
the design of sensing slot duration to maximize the achievable
throughput for the secondary users under the constraint that
the primary users are sufciently protected. Using energy
detection scheme, we prove that there indeed exists an optimal
sensing time which provides the best tradeoff. Cooperative
sensing is also studied based on the proposed tradeoff method-
ology.
1-4244-0353-7/07/$25.00 2007 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the ICC 2007 proceedings.
5330
While channel sensing is a conventional signal detection
problem [9], [10], it has received intensive attention recently in
the area of cognitive radio networks [12], [13], [14]. However,
the criteria considered are usually from the perspective of
protecting the primary user, i.e., maximizing the probability
of detection under the constraint of probability of false alarm.
In cognitive radio networks, in order to maximize the spectrum
usage for the secondary network, it turns out that the problem
should be minimizing the probability of false alarm, under the
constraint of probability of detection. In this paper, we study
the sensing-throughput tradeoff problem from this new angle.
This paper is organized as follows. Section II presents the
receive model for channel sensing. In Section III, we study
the sensing-throughput tradeoff problem. Distributed channel
sensing is studied in Section IV. Performance evaluation and
comparison are given in Section V, and nally, conclusions
are drawn in Section VI.
II. CHANNEL SENSING
In this section, we rst present the general model for
channel sensing, then review the energy detection scheme and
analyze the relation between the probability of detection and
probability of false alarm.
A. General Model of Channel Sensing
Suppose that we are interested in the frequency band with
carrier frequency f
c
and bandwidth W and the received signal
is sampled at sampling frequency f
s
, which is greater than
the Nyquist rate. When the primary user is active, the discrete
received signal at the secondary user can be represented as
y(n) = s(n) +u(n), (1)
which is termed hypothesis H
1
. When the primary user is
inactive, the received signal is given by
y(n) = u(n). (2)
This case is referred to as hypothesis H
0
. We make the
following assumptions.
(AS1) The noise u(n) is a Gaussian iid random process
with mean zero and variance E[|u(n)|
2
] =
2
u
;
(AS2) The primary signal s(n) is an iid random process
with mean zero and variance E[|s(n)|
2
] =
2
s
;
(AS3) The primary signal s(n) is independent of the
noise u(n).
Denote =

2
s

2
u
as the received signal-to-noise ratio (SNR)
of the primary user under hypothesis H
1
. We consider both
real Gaussian noise case and circularly symmetric complex
Gaussian (CSCG) noise cases. For the primary signal s(n), we
consider four cases: (1) BPSK modulated signal; (2) complex
PSK modulated signal; (3) real Gaussian signal and (4) CSCG
signal. Note the later two cases represent, for instance, signals
with rich inter-symbol interference, precoded signals such as
orthogonal frequency division multiplexing (OFDM) signals
or OFDM signals with linear precoding.
Two probabilities are of interest for channel sensing: proba-
bility of detection, P
d
, which denes, at the hypothesis H
1
, the
probability of the algorithm detecting the presence of primary
signal; and probability of false alarm, P
f
, which denes,
under the hypothesis H
0
, the probability of the algorithm
detecting the presence of primary signal. From the primary
users perspective, the higher the P
d
, the better protection it
receives. However, from the secondary user perspective, the
lower the P
f
, there are more chances for which the secondary
users can use the frequency bands when they are available.
Obviously, for a good detection algorithm, P
d
should be as
high as possible while P
f
should be as low as possible.
B. Energy Detector
Energy detector is the most popular channel sensing scheme.
The test statistic for energy detector is given by
T(y) =
1
N
N

n=1
|y(n)|
2
. (3)
Under hypothesis H
0
, the test static T(y) is a random variable
whose probability density function (PDF) p
0
(x) is a Chi-
square distribution with 2N degrees of freedom for complex
case, and with N degrees of freedom for real case. For a
chosen threshold , the probability of false alarm is given by
P
f
() = Pr(T(y) > |H
0
) =
_

p
0
(x)dx. (4)
Using central limit theorem (CLT), we have the following
proposition.
Proposition 1: For a large N, the PDF of T(y) under
hypothesis H
0
can be approximated by a Gaussian distribution
with mean
0
=
2
u
and variance
2
0
=
1
N
_
E|u(n)|
4

4
u

.
If u(n) is real Gaussian, then E|u(n)|
4
= 3
4
u
, thus
2
0
=
2
N

4
u
.
If u(n) is CSCG, then E|u(n)|
4
= 2
4
u
, thus
2
0
=
1
N

4
u
.
Next, we focus on the CSCG noise case for which the
probability of false alarm is given by
P
f
() = Q
__

2
u
1
_

N
_
, (5)
where Q() is the complementary distribution function of the
standard Gaussian, i.e.,
Q(x) =
1

2
_

x
exp
_

t
2
2
_
dt. (6)
Under hypothesis H
1
, denote p
1
(x) as the PDF of the test
static T(y). For a chosen threshold , the detection probability
is given by
P
d
() = Pr(T(y) > |H
1
) =
_

p
1
(x)dx. (7)
Proposition 2: For a large N, the PDF of T(y) under
hypothesis H
1
can be approximated by a Gaussian distribution
with mean
1
= ( + 1)
2
u
and variance

2
1
=
1
N
_
E|s(n)|
4
+E|u(n)|
4
(
2
s

2
u
)
2

, (8)
if s(n) and u(n) are both circularly symmetric and complex,
and

2
1
=
1
N
_
E|s(n)|
4
+E|u(n)|
4
(
2
s

2
u
)
2
+ 2
2
s

2
u

, (9)
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the ICC 2007 proceedings.
5331
if s(n) and u(n) are both real. Furthermore,
If s(n) is complex PSK modulated and u(n) is CSCG,
then
2
1
=
1
N
(2 + 1)
4
u
;
If s(n) is BPSK modulated and u(n) is real Gaussian,
then
2
1
=
2
N
(2 + 1)
4
u
;
If s(n) and u(n) are both CSCG, E|s(n)|
4
= 2
4
s
and
E|u(n)|
4
= 2
4
u
, then
2
1
=
1
N
( + 1)
2

4
u
;
If s(n) and u(n) are both real Gaussian, E|s(n)|
4
= 3
4
s
and E|u(n)|
4
= 3
4
u
, then
2
1
=
2
N
( + 1)
2

4
u
.
We focus on the complex PSK signal and CSCG noise case.
Based on the PDF of the test static, the detection probability
can be approximated by
P
d
() = Q
_
_

2
u
1
_

N
2 + 1
_
. (10)
For a target detection probability, P
d
, the detection threshold
can be determined by
_

2
u
1
_

N
2 + 1
= Q
1
(P
d
). (11)
From (5), on the other hand, this threshold is related to the
probability of false alarm as follows:
Q
1
(P
f
()) =
_

2
u
1
_

N. (12)
Therefore, for a target detection probability

P
d
, the probability
of false alarm is related to the target detection probability as
follows:
P
f
= Q
_
_
2 + 1Q
1
(

P
d
) +

N
_
. (13)
On the other hand, for a target probability of false alarm,

P
f
,
the detection probability is given by
P
d
= Q
_
1

2 + 1
(Q
1
(

P
f
)

N)
_
. (14)
Finally, for a given pair of target probabilities (

P
d
,

P
f
), the
number of required samples to achieve these targets can be
determined from (5) and (10) by cancelling out the threshold
variable . The result is given by
N
min
=
1

2
_
Q
1
(

P
f
) Q
1
(

P
d
)
_
2 + 1
_
2
. (15)
III. SENSING-THROUGHPUT TRADEOFF
In the previous section, the relation between probability of
detection and probability of false alarm has been established.
In this section, we study the fundamental tradeoff between
sensing capability and achievable throughput of the secondary
network. Using energy detection scheme, we will prove that
the there exists indeed the optimal sensing time which achieves
the best tradeoff.
A. Problem Formulation
Figure 2 shows the frame structure designed for a cognitive
radio network where each frame consists of one sensing
slot and one data transmission slot. Suppose the sensing slot
duration is and the frame duration is T. Denote C
0
as the
throughput of the secondary network when it operates in the
absence of primary users, and C
1
as the throughput when it
operates in the presence of primary users. For example, if there
is only one transmission in the secondary network and the SNR
for this link is SNR
s
. Let SNR
p
be the SNR of the primary
user received at the receiver of the secondary transmission link.
Then C
0
= log
2
(1 +SNR
s
) and C
1
= log
2
_
1 +
SNRs
1+SNRp
_
.
Obviously, we have C
0
> C
1
.
There are two scenarios for which the secondary network
can operate at the primary users frequency band.
Scenario I: The rst case is when the primary user is not
present and no false alarm is generated by the secondary
users. In this scenario, the achievable throughput is given
by
R
0
() =
T
T
(1 P
f
)C
0
. (16)
Scenario II: The second case is when the primary user
is active but it is not detected by the secondary users. In
this scenario, the achievable throughput is represented as
R
1
() =
T
T
(1 P
d
)C
1
. (17)
Let us further dene P(H
1
) as the activity probability for
which the primary user is active in the frequency band of
interest. The achievable throughput of WRAN system is then
given by
R() = P(H
1
)R
1
() +P(H
0
)R
0
(). (18)
where P(H
0
) = 1 P(H
1
).
Obviously, for a given frame duration T, the larger the
sensing time , the smaller the data transmission time T .
On the other hand, from (13), since Q(x) is a monotonically
decreasing function of x, for a given target probability of
detection, the longer the sensing time, the lower the probability
of false alarm, which corresponds to higher throughput under
the hypothesis of H
0
. The objective of sensing-throughput
tradeoff is to identify the optimal sensing duration for each
frame such that the achievable throughput of the secondary
network is maximized while there is sufcient protection to
the primary users. Mathematically, the optimization problem
can be described as
max

R() = P(H
1
)R
1
() +P(H
0
)R
0
(), (19)
s.t. P
d


P
d
, (20)
where

P
d
is the target probability of detection for which the
primary users are dened as being sufciently protected.
In practice, the target probability of detection

P
d
is chosen
to be close to but less than 1, especially for low SNR regime.
For instance, in 802.22 WRAN, we choose

P
d
= 0.9 for a
SNR of 15dB. It is pointed out that if the primary users
want 100% protection in its frequency band, there will then be
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the ICC 2007 proceedings.
5332
not allowed to exploit the secondary usage for that frequency
band. Also, we suppose the activity probability P(H
1
) of
primary users is small, say less than 0.2, thus it is economically
advisable to explore the secondary usage for that frequency
band. Furthermore, notice that C
0
> C
1
. Thus the second
term in (18) dominates the achievable throughput. Therefore
the optimization problem can be approximated by
max

R() = P(H
0
)
T
T
(1 P
f
)C
0
(21)
s.t. P
d


P
d
. (22)
B. Energy Detection
When energy detector is used, using (22) and choosing P
d
=

P
d
, we have

R() = C
0
P(H
0
)
_
1

T
__
1 Q
_
+
_
f
s

__
, (23)
where =

2 + 1Q
1
(

P
d
). Thus, from (23), we can see
that the achievable throughput of the secondary network is a
function of sensing duration .
Theorem 1: Under the assumptions (AS1) - (AS3), if the
primary signal is complex PSK modulated and the noise
process is CSCG, there exists an optimal sensing time which
yields the maximum achievable throughput for the secondary
networks.
Proof: It can be veried from (23) that

() = C
0
P(H
0
)
_

1
T
+
1
T
Q( +
_
f
s
)
+

f
s
(T )
2

2T
exp
_
( +
_
f
s
)
2
/2
_

1/2
_
.
(24)
Obviously,
lim
T

() < C
0
P(H
0
)
_

1
T
+
1
T
Q()
_
< 0, (25)
lim
0

() = +> 0. (26)
In deriving (25), we have used the fact that Q(x) is a
decreasing function and upper bounded by 1. (25) and (26)
mean that

R() increases when is small and decreases when
approaches T. Hence, there is a maximum point of

R()
within interval (0, T).
Furthermore, from (18), we have
R

() =

R

()
1
T
P(H
1
)(1

P
d
)C
1
. (27)
Thus, it is easy to show that there is a maximum point of R()
within interval (0, T).
IV. DISTRIBUTED CHANNEL SENSING
In this section, we consider cooperative channel sens-
ing using multiple distributed secondary users. The tradeoff
methodology proposed in the previous section can be applied
directly. Dene the hypotheses for the ith receiver as:
H
1
: y
i
(n) = h
i
s(n) +u
i
(n), (28)
H
0
: y
i
(n) = u
i
(n), (29)
for i = 1, , M. Denote P
d,i
and P
f,i
as the probability
of detection and probability of false alarm at the ith receiver,
respectively. By choosing the detection threshold as
0
, the
probability of detection for the ith receiver is
P
d,i
= Q
_
_

0

2
u
|h
i
|
2
1
_

N
2|h
i
|
2
+ 1
_
, (30)
and the probability of false alarm for the ith receiver is
P
f,i
= Q
__

0

2
u
1
_

N
_
. (31)
Once the decisions are made for each receiver, they can be
sent to the BS which makes the nal decision. The optimal
decision fusion rule is Chair-Varshney rule [9]. Here we only
consider three simple decision fusion rules to decide whether
the primary user is active.
A. Logical-OR Rule
LO rule is a simple decision rule described as follows: if one
of the decisions says that there is a primary user, then the nal
decision declares that there is a primary user. Assuming that all
decisions are independent, then the probability of detection and
probability of false alarm of the nal decision are, respectively,
P
d
= 1
M

i=1
(1 P
d,i
), (32)
P
f
= 1
M

i=1
(1 P
f,i
). (33)
B. Logical-AND Rule
LA rule works as follows: if all decisions says that there
is a primary user, then the nal decision declares that there
is a primary user. Again, assuming that all decisions are
independent, then the probability of detection and probability
of false alarm of the nal decision are, respectively,
P
d
=
M

i=1
P
d,i
, P
f
=
M

i=1
P
f,i
. (34)
C. Majority Rule
Another decision rule is based on majority of the individual
decisions. If half of the decisions or more says that there
is a primary user, the the nal decision declares that there
is a primary user. The probabilities for nal decision can
be formulated accordingly based on the probabilities from
individual secondary users.
V. COMPUTER SIMULATIONS
In this section, computer simulation results are presented to
evaluate the sensing-throughput tradeoff for different sensing
algorithms. The primary user is assumed to be a QPSK
modulated signal with bandwidth of 6MHz and the sampling
frequency is the same as the bandwidth of the primary user.
The additive noise is a zero-mean CSCG process. We are
interested in low SNR regime where the SNR for primary
user received at the secondary users location is SNR
p
= 15
dB. We choose the frame duration as T = 100ms, and set the
target probability of detection as

P
d
= 0.9.
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the ICC 2007 proceedings.
5333
A. One secondary user for channel sensing
First, we compare the probability of detection and proba-
bility of false alarm calculated from Monte Carlo simulation
results and the theoretical results derived in this paper. 20000
Monte Carlo simulations are performed. For Monte Carlo
simulations, for each sample size and under hypothesis H
1
,
we rst nd out the detection threshold to achieve the target
detection probability based on the 20000 test statistics, then
apply this threshold to the hypothesis H
0
and derive the
probability of false alarm. Figure 3 shows the comparison
between simulated and theoretical probabilities. It is seen
that P
d
achieves the target probability of detection, while the
probability of false alarm decreases with the increase of the
number of samples.
Next, suppose the SNR for secondary transmission is
SNR
s
= 20dB. Thus C
0
= log
2
(1 + SNR
s
) = 6.6582 and
C
1
= log
2
_
1 +
SNRs
1+SNRp
_
= 6.6137. We also assume that
P
p
= 0.2. Figure 4 shows the achievable throughput versus
the sensing time allocated to each frame for the secondary
network. Here two results are considered: R() of (18), and

R() of (23). It is seen that for both quantities, the maximum

is achieved at the sensing time of about 2.5 ms. Figure
5 illustrates the normalized achievable throughput for the
secondary network, which is dened as
_
1

T
_
(1 P
f
).
Again, it reveals a maximum point at the sensing time of
about 2.5ms. Furthermore, the simulated results match to the
theoretical results very well. Because of this consistency, in
the following evaluations, we will only consider the theoretical
results.
B. Distributed channel sensing
We consider the case when multiple secondary users are
used to sense the presence of primary user. For simplicity,
we assume that the SNRs for primary user at each secondary
user are all equal to 15dB. Figure 6 shows the optimal
sensing time versus the number of sensing users when various
decision fusion schemes are applied. It is seen that for each
fusion scheme, the optimal sensing time decreases with the
increase of sensing terminals. Figure 7 shows the maximum
normalized throughput of the secondary network with different
number of sensing users. The maximum achievable normalized
throughput increases with the increase of sensing terminals.
VI. CONCLUSIONS
In cognitive radio networks, the interests of primary user
and secondary user are contradictory. In this paper, we formu-
late the sensing-throughput tradeoff problem by considering
both users interests. Particularly, we study the design of sens-
ing slot duration to maximize the achievable throughput for the
secondary users under the constraint that the primary users
are sufciently protected. Using energy detection scheme, we
have proved that there indeed exists an optimal sensing time
which provides the best tradeoff. Cooperative sensing has also
been studied based on the proposed tradeoff methodology.
Computer simulations have shown that for a frame duration
of 100ms, and the SNR of primary user of 15dB, the
TV Station
WRAN BS
Wireless
Microphones
WRAN CPE
TV Protection
Contour
WRAN
Coverage
Microphone
Protection
Area
Fig. 1. An IEEE 802.22 WRAN deployment scenario. Each WRAN system
consists of a base station (BS) serving xed wireless subscribers called
customer premise equipments (CPE). Incumbent users are TV receivers and
FCCs Part 74 wireless microphones.
optimal sensing time achieving the highest throughput while
maintaining 90% detection probability is 2.5ms, if energy
detector is used. This optimal sensing time reduces when
distributed channel sensing is applied.
REFERENCES
[1] Federal Communications Commission, Spectrum policy task force
report, FCC 02-155. Nov. 2002.
[2] Federal Communications Commission, Facilitating opportunities for
exible, efcient, and reliable spectrum use employing cognitive radio
technologies, notice of proposed rule making and order, FCC 03-322.
Dec. 2003.
[3] J. Mitola and G. Q. Maguire, Cognitive radios: making software radios
more personal, IEEE Personal Communications, vol. 6, no. 4, pp. 1318,
Aug. 1999.
[4] J. Mitola, Cognitive radio: An integrated agent architecture for software
dened radio, Doctor of Technology, Royal Inst. Technol. (KTH),
Stockholm, Sweden, 2000.
[5] S. Haykin, Cognitive radio: Brain-empowered wireless communica-
tions, IEEE J. Selected Areas in Communications, vol.23, No.2, pp.201-
220, Feb. 2005.
[6] IEEE 802.22 Wireless RAN, Functional requirements for the 802.22
WRAN standard, IEEE 802.22- 05/0007r46, Oct. 2005.
[7] A. Sahai and D. Cabric, Spectrum sensing: fundamental limits and prac-
tical challenges, Proc. IEEE International Symposium on New Frontiers
in Dynamic Spectrum Access Networks (DySPAN) 2005, Baltimore, MD,
Nov. 2005.
[8] S. M. Mishra, A. Sahai, and R. W. Brodensen, Cooperative sensing
among cognitive radios, Proc. IEEE International Conference on Com-
munications (ICC) 2006, Istanbul, Turkey, Jun. 2006.
[9] Z. Chair and P. K. Varshney, Optimal data fusion in multiple sensor
detection systems, IEEE Trans. on Aerospace and Elect. Syst., vol.22,
pp.98-101, January 1986.
[10] W. A. Gardner, Exploitation of spectral redundancy in cyclostationary
signals, IEEE Signal Processing Magazine, vol. 8, pp. 1436, Apr. 1991.
[11] Y.-C. Liang et al., System description and operation principles
for IEEE 802.22 WRANs, Available at IEEE 802.22 WG Website
http://www.ieee802.org/22/, Nov. 2005.
[12] J. Hillenbrand, T. A. Weiss and F. Jondral, Calculation of detection and
false alarm probabilities in spectrum pooling systems, IEEE Commu-
nications Letters, vol. 9, no. 4, pp. 349351, April 2005.
[13] G. Ganesan, Y. Li, Cooperative spectrum sensing in cognitive radio
networks, Dyspan2005, pp.137-143, Nov. 2005.
[14] A. Ghasemi and E.S. Sousa, Collaborative spectrum sensing for oppor-
tunistic access in fading environments, Dyspan2005, pp.131-136, Nov.
2005.
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the ICC 2007 proceedings.
5334
Frame 1 Frame 2 Frame K
T -
Fig. 2. General operation sequence of a cognitive radio system, with
quiet period for sensing being inserted in between normal data transmission
intervals.
0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6
x 10
4
10
3
10
2
10
1
10
0
Number of samples
P
r
o
b
a
b
i
l
i
t
y

o
f

d
e
t
e
c
t
i
o
n
/
f
a
l
s
e

a
l
a
r
m
Prob. of detection simulated
Prob. of detection theory
Prob. of false alarm simulated
Prob. of false alarm theory
Fig. 3. Comparison of simulated and theoretical probability of detection and
probability of false alarm.
1 1.5 2 2.5 3 3.5 4 4.5 5
4.7
4.8
4.9
5
5.1
5.2
5.3
5.4
Sensing time (ms)
A
c
h
i
e
v
a
b
l
e

t
h
r
o
u
g
h
p
u
t

(
b
i
t
s
/
S
e
c
/
H
z
)
Fig. 4. Achievable throughput for the secondary network: o R(), +

R().
1 1.5 2 2.5 3 3.5 4 4.5 5
0.88
0.89
0.9
0.91
0.92
0.93
0.94
0.95
0.96
0.97
Sensing time (ms)
A
c
h
i
e
v
a
b
l
e

t
h
r
o
u
g
h
p
u
t

(
b
i
t
s
/
S
e
c
/
H
z
)
Simulated
Theory
Fig. 5. Normalized achievable throughput for the secondary network.
2 4 6 8 10 12 14 16 18 20
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
2.2
2.4
Number of users
O
p
t
i
m
a
l

s
e
n
s
i
n
g

t
i
m
e

(
m
s
)


LogicAND
LogicOR
LogicMajority
Fig. 6. Optimal sensing time for distributed channel sensing with various
decision rules.
2 4 6 8 10 12 14 16 18 20
0.97
0.975
0.98
0.985
0.99
0.995
Number of users
M
a
x
i
m
u
m

n
o
r
m
a
l
i
z
e
d

c
a
p
a
c
i
t
y


LogicAND
LogicOR
LogicMajority
Fig. 7. Normalized achievable throughput for distributed channel sensing
with various decision rules.
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the ICC 2007 proceedings.
5335