Sunteți pe pagina 1din 6

Effect of RF Impairments on Spectrum Sensing

Techniques
Mohammad Haroun, Hussein Kobeissi, Oussama
Bazzi, Haidar El Mokdad
Ecole Doctorale des Sciences et de Technologie, Lebanese
University, Hadath
Beirut, Lebanon
Email: hmokdad@ul.edu.lb
AbstractSpectrum sensing is the key element in the

implementation of a cognitive radio equipment allowing


the utilization of vacant bandwidth unused by the primary
users such as broadcasting white spaces. To implement
suitable spectrum sensing techniques, RF front end is
mandatory at the secondary devices. However, these frontends are usually low cost devices implying then some RF
impairments which may affect the properties and
characteristics of the received signal hence the
performance of the spectrum sensing. In this paper, we
analyze the effect of three main imperfections on the
spectrum sensing capabilities of a CR receiver: the carrier
frequency offset, the IQ mismatch and the phase noise. We
show that the energy detector is not sensitive to any of
these errors while the cyclostationnarity detector is only
sensitive to the phase noise while its performance is
independent of the carrier frequency offset and IQ
mismatch values.
Keywords- Energy Detector, Cyclostationarity detector, RF
impairments

I.
INTRODUCTION
Nowadays, Cognitive Radio (CR) technology is an
important enabler for network access, flexible resource
allocation, and context-aware services. According to S.
Haykin [1], cognitive radio is an intelligent wireless
communication system that is aware of its surrounding
environment (i.e., outside world), and uses the methodology of
understanding-by-building to learn from the environment and
adapt its internal states to statistical variations in the incoming
radio frequency (RF) incentive by making corresponding
changes in certain operating parameters (e.g., transmit-power,
carrier-frequency, and modulation strategy). Thus the first
challenge of a CR device is to sense the environment (to detect
for instance the presence or absence of a primary network) in a
low Signal to Noise Ratio (SNR) regime. In more realistic
scenario, one could imagine an LTE like signal in the vicinity
of DVB spectrum White Spaces.
In practice, a CR device is composed of two main modules:
the analog RF front-end which, among other tasks, adapts its
components to the frequency and bandwidth of interest and, the
digital processor which allows the algorithmic implementation
978-1-4799-1297-1/14/$31.00 2014 IEEE

Lise Safatly, Karim Y. Kabalan, Youssef Nasser


Electrical and Computer Engineering Department
American University of Beirut, Faculty of Engineering and
Architecture
Beirut, Lebanon
Email: youssef.nasser@aub.edu.lb
such as synchronization, estimation, and detection. The two
modules are connected by the Analog to Digital Converter
(ADC) controlled by a local oscillator working at a tunable
frequency. While the design of an RF front-end is practically
straightforward in most of the standardized systems, this task
becomes challenging in CR. Indeed, in standardized systems,
the front-end is designed to work on a very limited set of
carrier frequencies and bandwidth. Thus their specifications
could be easily deduced. However, in CR like systems, the RF
front-end should work on wide spectrum and/or several bands
in order to be able to sense the available spectral vacancies in
the vicinity. Moreover, it should be designed with a wide range
of power dynamics in order to take into account different
standards transmission specifications. For instance, the digital
TV broadcasting technologies are transmitting with higher
power values than cellular technologies. Hence, a CR device
should deal with these RF front-end constraints ([2][3]) and
propose the necessary solutions.
In literature, very few works dealt with the RF impairments
in CR. In [4], the sampling clock offset was analyzed and
corresponding compensation solution was proposed for the
feature detection receiver. In [5][6], the authors have provided
a thorough analysis of the performance of the Matched Filter
(MF) in the presence of the Carrier Frequency Offset (CFO),
the Phase Noise (PN) and time offset. Even though the work
presented in [5] has dealt with the IQ mismatch, it did not
include the other impairments in the Energy Detector (ED) and
Cyclostationarity Detector (CD) performance evaluation.
In this article, we analyze the effect of different RF
impairments created by the front-end on two potential sensing
techniques: the ED and the CD. We focus in this work on three
main impairments, known to be the most challenging ones [6]
in any RF design: the CFO, the PN and the IQ mismatch. We
show in this paper that the CFO and IQ mismatch have no
effect on the performance of ED and CD while both sensing
techniques are sensitive to the PN.
The remaining of the paper is organized as follows. Section
2 describes the system model and the sensing techniques
analyzed in this work. In section 3, we develop the analytical
models of the ED and CD detectors in the presence of the RF
impairments. Simulation results are drawn in section 4. Section
5 concludes the paper.

II.

SYSTEM MODEL AND SENSING ALGORITHMS

A. System Model
In this work, we assume the following test model:
: ( ) = ( )
: ( ) = ( ) + ( )

(1)

where k denotes the sample index, ( ) is the signal


transmitted by the primary user, ( ) denotes zero-mean
Additive White Gaussian Noise (AWGN) with variance .
( ) corresponds to the received signal by the secondary user
under two hypotheses:
and
where the null hypothesis
corresponds to the idle state and
corresponds to the
busy channel state in the tested band.
The model presented in (1) corresponds to an ideal
transmission where the non-linearities of the RF front-end are
not considered. However, in practice, these constraints should
be accounted for as they will affect the system performance.
B. Sensing Algorithms
The optimal sensing technique is realized when the channel
gain, noise power and primary user variance are known at the
secondary device. In practice, however, we might not know
any or some of these parameters. In this case, the signals are
assumed to be normally distributed with low Signal to Noise
Ratio (SNR) levels. In this work, two sensing algorithms are
considered: the energy detector and the cyclostationarity
detector.
As stated above, if no information is available about the
primary signal behavior, the transmitted signal is considered
to be a stationary zero-mean white Gaussian process,
independent of the observation noise. In this case, the
spectrum sensing problem of the ED is simply a distinction
between two mutually independent identically distributed (iid)
Gaussian sequences. Let = [ (1), . . . , ( )] denote the
vector (sequence) of N received samples, k=1,, N. Under the
assumption of iid Gaussian distribution of both transmitted
signal and additive noise, the variance of the received
sequence could be written as:

where

=
=

is the variance of transmitted sequence under

The ED is then easily seen to be equivalent to deciding


=

1
2

In this work, we consider the second order CD. A digital


signal x(n) is said to be second-order almost-cyclostationary in
the wide sense if its autocorrelation function:
( ,

| ( )| >

(2)
.
if:
(3)

where is a form of the scaled energy of the whole received


signal. It is a scaled version of a standard Chi-squared random
variable with 2N degrees of freedom whose characteristics in
terms of probability of miss detection
and probability of
false alarm
could be easily found [7].
2) Cyclostationarity Detector (CD)
The main problem of the ED resides in its sensitivity to the
SNR level as shown in literature [3]. Thus, the cyclostationarity

) { ( +

) ( )}

(4)

is an almost-periodic function of the discrete-time parameter


n. Thus, it can be expressed as:
( ,

)=

( )

(5)

where
( ) = lim

1) Energy Detector

:
:

detector is proposed to remedy to this sensitivity. Indeed, it is


able to detect a low SNR regimes and poor estimate of the
noise power. The CD detector is based on the analysis of the
higher order moments (such as the autocorrelation function) of
the received signal.
A signal is cyclostationary of order N (in the wide sense) if
and only if we can find some nth-order nonlinear
transformation of the signal that will generate finite-strength
additive sine-wave components, which result in spectral
lines [8]. For the simplest nontrivial case, which is N = 2, a
signal x(t) is cyclostationary with cycle frequency if and
only if at least some of its delay product waveforms, y(t) = x(t)x(t) or z(t) = x(t-)x*(t) (where * denotes conjugate) for some
delays , exhibit a spectral line at frequency .

1
2 +1

( ,

(6)

is the Cyclic Autocorrelation Function (CAF) at cyclic


frequency and
{
, : ( ) 0. Note that the
( ) is periodic in with
cyclic autocorrelation function
period 1.
The interest of such a detector dwells in the fact that, virtually
most of the man-made communication signals exhibit second
order cyclostationarity with cycle frequencies related to
hidden periodicities underlying the signal and other relevant
signal parameters, such as the carrier frequency, symbol
and/or chip rates, period of the spreading or scrambling codes,
modulation index, etc.
In this work, we adopt the Orthogonal Frequency Division
Multiplexing (OFDM) as the basic waveform used for CD
analysis since OFDM technology is adopted in most of the
current and future communication standards. As shown in [9],
the fundamental cyclic frequency of an OFDM signal is ,
where
=
+
is the duration of one OFDM symbol
including the guard interval, and is the sampling period.
The OFDM signal is strongly cyclostationary at time-lag =
=
. When the lag parameter of the estimated CAF
sets to
= , the CAF should reveal local peaks at
multiples of fundamental cycle frequency of the OFDM signal,
that is
, = 1,2, . . . , 1. At the cycle frequencies
other than , the magnitude of CAF should be small as
compared with local peaks. In this case, (6) becomes:

( )=

() ( + )

s(k)

(7)

(8)

( )

where is a value chosen to ensure that the cyclic frequency


is not a multiple of the fundamental cycle frequency , L is a
pre-defined value representing the number of peaks considered
in the test statistic. For = , the numerator is very large
with respect to the denominator. This means that large peaks
will appear at values which are multiple to . The reader
may refer to [9] for more information on the CAF of an
OFDM signal.
III.

RF IMPAIRMENTS MODEL IN SPECTRUM SENSING


TECHNIQUES

In communication systems, the RF front-end is a basic module


in the transmission chain. However, with the advances in
current and future systems, the RF technology is pushed to its
operational boundaries whereas the front-end becomes the
major governing component of the transmission performance.
While in most of the system level analysis, the RF front end
impairments are not considered in CR, their effect is
nevertheless affecting the received signal performance.
In cognitive radio, the sensitivity of the spectrum sensing
techniques to the RF impairments was barely investigated. In
this paper, we propose to model and analyze the effect of the
CFO, phase noise and IQ mismatch of the two detectors
presented in the previous section. In the next sub-sections, we
first describe these impairments and then include them in the
detectors expressions.
A. Carrier Frequency Offset
Let s(n) be the baseband signal at the transmitter output. After
Digital to Analog Conversion (DAC), the signal is transmitted
through the channel on a carrier frequency FTX. At the receiver
side, the receiver firstly down-converts the received signal by
utilizing a local oscillator operating at a carrier frequency
FRX=FTX+f where f is the carrier frequency offset between
the transmitter and receiver. We assume that the receiver is
aware of the carrier frequencies and bandwidth of the possible
transmitted signal. Then, as represented in Figure 1, the
baseband received signal could be written as:

: ( ) = ( )

: ( ) = ( )
= ( ).

Figure 1- Baseband model of the RF impairments


By inserting (10) in (3) we can easily deduce that the ED is
not sensitive to the CFO.

( )

r(k)

Baseband model of
the RF impairments

w(k)

In order to evaluate the CD performance, the proposed test


statistic is given by:

y(k)

( )
+ ( )

(9)

where ( ) = ( ).
has the same statistical
characteristics of ( ), ( ) is the equivalent baseband
received signal without RF impairments and ( ) is the
equivalent baseband signal affected by the CFO.

In the case of the CD, the test statistic of (8) is applied on the
received signal ( ) given in (9). In this case, the CAF of (7)
yields:
( ) =
1

( ).

+ ).
(

(10)

) for =

As expected, it is obvious that | ( )| =


( ) . This
means that the CD detector is not sensitive to the CFO.
B. Phase Noise
Phase noise is the difference between the phase of the receiver
oscillator and the phase of the carrier of the received signal
and is usually modeled as a Wiener random process or a widesense stationary (WSS) random process. The discrete model of
the Wiener phase noise is represented by [11]:
( ) = ( ).

(11)

where ( ) and ( ) are respectively the equivalent baseband


received signal without and with the phase noise inclusion.
is the phase rotation of the received signal due to the phase
noise. It is given by the Wiener process as:
=

(12)

with
is uniformly distributed on [, ) and { } is an i.i.d.
real Gaussian process with zero mean and variance =
2 , where
is called the Full-Width at Half-Maximum
(FWHM) or the diffusion factor and is the sampling period.
The insertion of (12) in (11) yields:
(

( ) = ( ).

(13)

Again, using (13) and (9) in (3), it is easily shown that the ED
is not sensitive to the phase noise.
As for the CD, the CAF of (7) becomes:

( )=

( ) ( + )

(14)

By successive computations, it becomes:

( )=

1
1

.
(15)

.
1

: {| ( )| } = {| ( )| } =
: {| ( )| } = {| ( )| } =

Equations derived in (21) are similar to those obtained in (2).


With the assumption on , this means that the ED is not
sensitive to the IQ mismatch.

where are independent identically distributed (iid) and


~ (0,2 . . ). Then the sum =
is normally
distributed with zero mean and variance = 2 . . . This
expression (15) is identical to (7) except that there is an added
noise in each components phase of the CAF. Simulation
results derived in the next section will show that this added
noise will highly affect the CAF of the OFDM signal.
C. IQ Mismatch
The IQ mismatch is due to the loss of orthogonality between
the In-Phase and Quadrature-Phase components at the receiver.

To evaluate the effect of the RF impairments on the CD


detector, the CAF should be derived. The insertion of (17) in
(7) yields:

( )=

1
1

( .
+

Figure 2- Model for Analog IQ Mismatch


In the case of IQ mismatch in the system, the baseband
received signal ( ) could be written as a linear expression of
the ideal signal ( ) and its conjugate ( ) as follows:
(

(16)

The distortion parameters and in (16) and Figure 2 are


related to the amplitude and phase imbalances between the I
and Q branches in the RF/Analog demodulation process
through a simplified model as follows
= cos

= cos

sin

sin

where and are, respectively, the phase and amplitude


imbalance between the I and Q branches. The phase imbalance
is any phase deviation from the ideal 90o between the I and Q
branches. The amplitude imbalance is defined as

=
(19)
+

IV.

According to (3), the ED evaluates the average of | ( )| ,


represented as {| ( )| }, over a large number of received
samples. Assuming that 1 , i.e. the gain difference
between the In-Phase and Qaudrature-Phase is small, we can
deduce that:

)
( )
.
( ) + | | .

( )+

( )

( )

SIMULATION RESULTS AND DISCUSSION

The target of this section is to provide simulation results on


the effects of the studied RF impairments on the spectrum
sensing techniques considered in this paper. Due to the
complexity of the CD, we focus in this part on the CD results
only. The results of the ED are straightforward.
A. Carrier Frequency Offset
1

0.8

are the gain amplitudes on the I and Q

To analyze the ED sensitivity to the IQ mismatch, we first


write the expression of | ( )| derived in (16) to be used in
(3). After some mathematical manipulations, this expression is
given by:
| ( )| = (1 + )| ( )| + 2{ ( )}
(20)

( )+

(22)

0.8

0.6
0.6

PD

where
and
branches.

+ .

Equation (22) includes three components. The first term is a


scaled version of the CAF given in (14) and highly depends on
the parameter . For small values of , this scale factor of the
first term could be neglected. The second and the third terms
contain the conjugate CAF. However, it is well known that an
OFDM signal does not have a conjugate cyclostationarity [10]
which means that the IQ mismatch will affect the CD
performance.

(17)
(18)

=(1 +

)( .

(| | .

+ .

+ . .
+| | .
= | | .
( ) +
+ .

( )= ( )+

(21)

0.4

=0.005

0.4

=0.005

=0.1

=0.1

=0.5
0.2

=0.8

=0.5

=1

0.2

0
0

0
0

0.1

0.02

0.2

0.04

0.3

0.06
PFA

0.4

0.08

0.1

=0.8

0.12

=1
0.5
P

0.6

0.7

0.8

0.9

FA

Figure 3- ROC of the CD at different values of CFO

As shown in Section 3, the ED and CD results are independent


of the CFO. Indeed, we plot on Figure 3 the Receiver
Operational Characteristic (ROC), i.e. probability of detection
PD versus probability of false alarm PFA of these detectors. This
figure shows that the CD detector has the same ROC whatever
the values of the CFO. Figure 4 represents the CAF of the
OFDM received signal r(k) in the presence of CFO. Both
figures show that the same results are obtained independently
of the CFO value = .
0.25
=0.001
=0.1
0.2

Figure 7 and Figure 8 show, as expected from the previous


analysis, that the ROC of the CD detector is independent of
the IQ mismatch whatever the values of and are.

=0.8
=1

|Ryy*()|

cyclostationarity of OFDM signal could be lost in the case of


IQ mismatch. Then, the main problem turns out to find and
plot the properties of the conjugate CAF. Figure 6 shows and
proves the results provided in [10]. Indeed, the conjugate CAF
does not show any peaks at the multiples of the fundamental
cyclic frequency. Fortunately, the values of the CAF are much
higher than those of the conjugate CAF. Moreover, Table 1
shows that the weight (1 + ) of the CAF in (22) are
equivalent or higher than those of the conjugate CAF. This
means that the latter, i.e. second and third term of (22), could
be simply neglected in the evaluation.

0.15

Table 1- coefficients of the CAF and conjugate CAF


0.1

=1dB =1dB =1dB =10dB =10dB =10dB

0.05

0
-5

-4

-3

-2

-1

=2o

=20o

=60o

=2o

=20o

=60o

| |

1.00

1.01

1.14

1.03

3.98

25.75

| |

1.58

1.56

1.43

99.96

97.01

75.25

1.25

1.26

1.28

10.14

19.66

44.01

Figure 4- CAF at time lag equal N for different values of CFO


0.2
0.18

0.18

|R *()|

0.16

|Rxx()|

xx

|Rxx *()|

0.16
0.14

Amplitude

0.14
0.12

|Rxx ()|

0.1
0.08
0.06

Amplitude

B. Phase Noise
Figure 5 shows the ROC function of the received signal
with the CD detector in the presence of phase noise with
different values of
, where is the diffusion factor defined
in (12). It is clear from this figure that the ROC degrades when
increases. For
= 5 10 , the probability of detection
becomes equivalent to the probability of false alarm. It is the
worst case scenario for RF spectrum sensing. This means that
the RF engineers should be very careful when they design their
RF front-end for CR implementation.

0.12
0.1

0.04
0.02
0

-0.2

0.2

0.4

0.6

0.8

0.08
0.06
0.04

0.02

0.9

Ts =0.00001

0.8

-3

-2

-1

Ts =0.00015

0.9

Ts =0.0002

0.8

Ts =0.0003

0.7

0.2

Ts =0.0005

0.6

0.1

0.4

PD

0.3

0.1

0.2

0.3

0.4

0.5
PFA

0.6

0.7

0.8

0.9

Ts =0.0001

0.5

Figure 6- CAF and Conjugate CAF of an OFDM signal.

Ts =0.00009

0.6

0
0

-4

Ts =0.00005

0.7

PD

0
-5

Figure 5- ROC of the CD detector at different values of


C. IQ mismatch
Equation (22) shows that there are 3 terms consisting the CAF.
The 1st term contains the CAF of the signal y(k). Meanwhile,
2nd and 3rd terms contain the conjugate CAF. The authors
of [10] have shown that the conjugate CAF of an OFDM
signal does not exist which means that the properties of

0.5
0.4
0.3

=2o

0.2

=20o

0.1

=60o

0
0

0.1

0.2

0.3

0.4

0.5
P

0.6

0.7

0.8

0.9

FA

Figure 7- ROC of the CD with different values of the angle


mismatch , with =10 dB

REFERENCES

1
0.9

[1]

S. Haykin, Cognitive Radio: Brain-Empowered Wireless


Communications, IEEE Journal on Selected Areas in Communications,
vol. 23, no. 2, pp. 201220, February 2005

[2]

R. W. Brodersen D. Cabric, S. M. Mishra, Implementation issues in


spectrum sensing for cognitive radios, in Asilomar Conference on
Signals, Systems, and Computers, 2004.

[3]

B. Natarajan H. Zamat, Use of dedicated broadband sensing receiver in


cognitive radio, in IEEE Proceedings of ICC, June 2008.

[4]

Zahedi-Ghasabeh, A., Tarighat, A., Daneshrad, B., Sampling Clock


Frequency Offset Compensation for Feature Detection in Spectrum
Sensing, International Conference on Communications (ICC), pp. 1 6,
June 2010

[5]

Jonathan Verlant-Chenet, Julien Renard, Jean-Michel Dricot, Philippe


De Doncker, Francois Horlin, Sensitivity of Spectrum Sensing
Techniques to RF impairments, 71th IEEE Vehicular Technology
Conference (VTC 2010-Spring), pp. 1-5, May 2010

[6]

Youssif Fawzi Sharkasi, Des McLernon , Mounir Ghogho, Spectrum


Sensing in the Presence of RF Impairments in Cognitive Radio,
International Journal of Interdisciplinary Telecommunications and
Networking, 4(3), 65-78, July-September 2012

[7]

A. M. Wyglinski, M. Nekovee, T. Hou, Cognitive Radio


Communications and Networks: Principles and Practice, Elsevier 2010.

[8]

W. A. Gardner, Cyclostationarity in Communications and Signal


Processing, IEEE Press.

[9]

M. Onera, F. Jondralb, Air interface identification for Software Radio


systems, ELSEVIER 2006

0.8
0.7

PD

0.6
0.5
0.4
0.3
=1dB

0.2

=10dB

0.1
0
0

=100dB
0.1

0.2

0.3

0.4

0.5
PFA

0.6

0.7

0.8

0.9

Figure 8- ROC of the CD with different values of the angle


mismatch , with =2
V. CONCLUSION
In this paper, we have presented and analyzed the
sensitivity of the energy detector and cyclostationarity detector
to the RF impairments in the case of OFDM based signals. We
have shown analytically and by simulations that the ED
detector performance is independent of the CFO, phase noise
and IQ mismatch while CD is sensitive to the phase noise. In
the future, the authors will analyze these outputs by real
measurements and then propose suitable mitigation algorithms.
ACKNOWLEDGMENTS
This paper is supported by the Lebanese National Council for
Scientific Research (LNCSR).

[10] G. Huang and J. K. Tugnait, On Cyclostationarity Based Spectrum


Sensing Under Uncertain Gaussian Noise, IEE Transactions on Signal
Processing, Vol. IEEE Trans. On Signal Processing, No. 8, April 2013
[11] T. Schenk, RF Imperfections in High-rate Wireless Systems, Impact
and Digital Compensation, Springer publications, 2008
[12] H. Ghozlan, G. Kramer, On Wiener Phase Noise Channels at High
Signal-to-Noise Ratio, .

S-ar putea să vă placă și