Sie sind auf Seite 1von 9

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 11, NO.

7, JULY 2012

2353

A Time-Domain Joint Estimation Algorithm for


CFO and I/Q Imbalance in Wideband
Direct-Conversion Receivers
Yen-Chang Pan and See-May Phoong

AbstractCarrier frequency offset (CFO) and in-phase and


quadrature-phase (I/Q) imbalance are two of the common frontend impairments in low-cost communication devices. It is known
that CFO can cause significant performance degradation in
multi-carrier modulation (MCM) systems. Also, the existence
of the I/Q imbalance usually reduces the accuracy of CFO
estimation. In this paper, we propose a new data-aided scheme
for the joint estimation of CFO and I/Q imbalance using simple
matrix formulation. The proposed algorithms utilize only the
periodicity of the generalized periodic pilot (GPP). They do
not need to know the channel impulse response and the exact
values of the training sequence. Moreover, our method has a
low complexity and its performance compares favorably with
the existing methods.

of the low-pass filters. They generate conjugate-image interference after frequency down-conversion. As a result, both
the CFO and I/Q imbalance can significantly deteriorate the
system performance [1], [10]. Furthermore, they can be quite
severe in low-cost DCRs [2].

It is usually difficult and impractical to completely eliminate


the CFO and I/Q imbalance at the front-end circuit design. Instead, discrete signal processing (DSP) techniques are widely
employed for estimation and correction. They can be generally
classified into data-aided and non-data-aided (blind) schemes
depending on whether or not a preamble or training sequence
is used. With the aid of training sequence, CFO itself is
not difficult to estimate and compensate [3], [4]. However,
it is known that the performance of CFO estimation will be
degraded by the existence of I/Q imbalance [5], and this issue
has recently been reexamined by [6]. In order to obtain an
accurate CFO estimate, the presence of I/Q imbalance must
be taken into account. On the contrary, I/Q imbalance can also
be calibrated before CFO estimation using the blind approach
[10], which has the advantage of being robust to CFO.

Manuscript received May 24, 2010; revised October 25, 2010, May 4 and
September 29, 2011, and January 4, 2012; accepted March 26, 2012. The
associate editor coordinating the review of this paper and approving it for
publication was M. Torlak.
The authors are with the Graduate Institute of Communication Engineering,
Dept. of Electrical Engineering, National Taiwan University, Taipei, Taiwan
106, R.O.C. (e-mail: d97942022@ntu.edu.tw, smp@cc.ee.ntu.edu.tw).
This work is supported by the National Science Council, Taiwan, R.O.C.,
under Grant NSC 100-2221-E-002-201-MY3.
Digital Object Identifier 10.1109/TWC.2012.051412.100882

The joint estimation of CFO and I/Q imbalance has been


studied by many researchers [11]-[22]. This paper discusses
only the data-aided schemes. In particular, a number of
different approaches utilizing the periodicity of the training
sequence were recently developed. Yan, et al. [12] proposed
a CFO estimator which is robust to the I/Q imbalance. The
I/Q imbalance is estimated using least squares method after
CFO estimation. De Rore, et al. [13] modified the above
CFO estimator using Taylor series analysis to achieve a better
performance. The same analysis was also applied to the
development of I/Q imbalance estimator. The CFO estimators
in [12] and [13] are based on the cosine function. The
performance of these estimators degrades seriously when the
CFO is small. Sun, et al. [14] solved this issue by injecting
a pseudo CFO into the received signal. In [15], Chen, et
al. proposed a new cosine-based CFO estimator. Moreover,
they also derived a complementary CFO estimator based on
the sine function. The combined cosine and sine estimator is
robust against different CFO values. In [16], they introduced
a refined sine estimator to further enhance the estimation
accuracy. It was shown that the estimators in [16] achieved
a good performance. However, the methods in [12]-[16]
considered only the frequency-independent I/Q imbalance.
The joint estimation with frequency-dependent I/Q imbalance
were stated in [20]-[22]. A compensation structure for the

Index TermsMulti-carrier modulation, carrier frequency offset (CFO), frequency-dependent I/Q imbalance, direct-conversion
receivers.

I. I NTRODUCTION
N recent years, multi-carrier modulation (MCM) has been
adopted by many broadband wireless standards due to its
high spectrum efficiency, ease of channel equalization, and
the flexible data rate. The direct-conversion receiver (DCR)
has become a major approach to the implementation of the
systems employing MCM due to its small circuit-size and
low power consumption. Although MCM systems are robust
against frequency-selective channels, they are sensitive to
synchronization errors, such as carrier frequency offset (CFO),
and to analog front-end non-idealities, such as I/Q imbalance.
CFO is caused by the frequency difference between the local
oscillators and/or by the Doppler effect in mobile systems.
It will induce the inter-carrier interference (ICI) between
the subcarriers. On the other hand, for wideband DCRs, the
I/Q imbalance consists of two parts. One is the frequencyindependent imbalance resulting from the oscillator gain and
phase mismatches between the in-phase (I-) and quadraturephase (Q-) branches. The other is the frequency-dependent
imbalance resulting from filter (frequency response) mismatch

c 2012 IEEE
1536-1276/12$31.00 

2354

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 11, NO. 7, JULY 2012

wideband I/Q imbalance has been proposed in [21] using a


modified periodic pilot (MPP) scheme. Unfortunately, their
CFO estimator needs a search algorithm and the computational
complexity is high. In [22], a generalized periodic pilot (GPP)
scheme has been proposed. Based on this scheme and the
compensation structure in [21], they developed a joint linear
least square (LLS) algorithm for estimating all the parameters
of CFO and I/Q imbalance. Its complexity is considerably low
compared with the two-step method in [21].
In this paper, we propose a new two-step estimation method
for CFO and I/Q imbalance using the GPP. A cosine-based
CFO estimator is first derived from the null space of a 22
matrix associated with the received samples and thus it has
a relatively low complexity. Then a sine estimate is derived
from an alternate matrix formulation. Combining the cosine
and sine estimates, we form a CFO estimator based on the
tangent function, which is more robust to various CFO values.
By adopting the imbalance compensation structure in [21], we
develop a new LLS algorithm for estimating the parameters
of frequency-dependent I/Q imbalance. The proposed method
are based on simple matrix formulation and they exploit
only the generalized periodicity of the GPP sequence. That
is, they do not need to know the channel impulse response
(which can be frequency-selective) and the exact values of
the training sequence. Using the proposed matrix formulation,
the proposed method does not need a search algorithm for
CFO estimation and its complexity is lower than the existing
methods. Simulation results show that the performance of
the proposed method compares favorably with the existing
methods.
The rest of this paper is organized as follows: Section II
describes the system as well as the signal models. Section III
derives the algorithms for CFO estimation. Section IV deals
with the I/Q imbalance. Section V discusses the complexity issue. Simulation results and comparisons are shown in Section
VI. Concluding remarks are given in Section VII.
Related conference papers: Parts of the material in this
paper also appear in our two conference papers [25], [26].
However in [25], we consider only the frequency-independent
I/Q imbalance. Whereas in [26], only the cosine estimator is
derived. It is well known that the cosine-based CFO estimators
degrade significantly when the cosine value is close to 1.
Moreover, in both [25] and [26], we need to explicitly estimate
the noise power before obtaining the cosine estimate. In this
paper, we introduce a new constant matrix with normalized
columns so that the noise power estimation process can be
avoided. Moreover, a new sine estimator for the CFO is
derived. By combining the cosine and sine estimators, we have
a tangent estimator for CFO, which is much more robust to
CFO variation.
Notation: Column vectors (matrices) are indicated by lowercase (uppercase) boldfaced letters. AH , A , and det(A)
denote the conjugate-transpose, pseudoinverse, and the determinant of A, respectively.
[A]ij is the (i, j)-th element of A.

The scalar j = 1 and () denotes complex conjugation.


{}, {}, and | | indicate real part, imaginary part, and
absolute value, respectively. E{} stands for statistical expectation.

II. S IGNAL AND S YSTEM M ODEL


In this paper, we assume that the GPP sequence s(n)
consists of P M samples, and it is periodic with period M
and an artificial phase rotate . That is,
s(n + M ) = ej s(n) for 0 n < (P 1)M.

(1)

At the receiver, the received baseband signal without front-end


impairments is given by
r(n) = s(n) h(n) + qch (n),

(2)

where denotes linear convolution, h(n) is the channel


impulse response, and qch (n) is the channel noise which is
assumed to be white. When the system suffers from CFO and
frequency-dependent I/Q imbalance, the received baseband
signal is no longer given by r(n) in (2). Figure 1 shows
the general structure of direct-conversion receivers. Define
y(n) = yI (n) + jyQ (n), then the received signal becomes
[21], [22]




y(n) = ejn r(n) c+ (n) + ejn r(n) c (n) (3)
where the CFO parameter = 2f Ts (f is the CFO in
Hz and Ts is the sampling period in second). Both c+ (n) and
c (n) are functions of gain mismatch , phase mismatch ,
and impulse responses of GI (f ) and GQ (f ):

1
gI (n) + (1 + )ej gQ (n) ,
c+ (n) =
2

1
c (n) =
gI (n) (1 + )ej gQ (n) ,
2

gI (n) = F 1 {GI (f )}t=nT ,
s

.
(4)
gQ (n) = F 1 {GQ (f )}
t=nTs

where F 1 {} denotes the continuous-time inverse Fourier


transform. Using (2), the received signal in (3) can be expressed as
h+ (n)


 
y(n) =ejn {s(n) h(n) c+ (n)ejn }


+ ejn {s (n) h (n) c (n)ejn } + q(n).



h (n)
(5)
with the noise




q(n) = ejn qch (n) c+ (n) + ejn qch (n) c (n). (6)
The model in (5) states that besides CFO, the desired signal
s(n) is also contaminated by its image interference s (n).
In this paper, we will show how to estimate the CFO
and suppress the image interference component from the
received signal y(n). The proposed method exploits only the
generalized periodicity property in (1) and it does not need to
know the channel impulse response h(n) and the exact values
of the GPP s(n). Thus the proposed method can be applied
to many communication systems.
III. CFO E STIMATORS
In this section, we derive two types of CFO estimators. One
is based on the cosine estimate and another is based on the
sine estimate. The estimation of I/Q imbalance parameters will
be discussed in Section 4.

PAN and PHOONG: A TIME-DOMAIN JOINT ESTIMATION ALGORITHM FOR CFO AND I/Q IMBALANCE IN WIDEBAND DIRECT-CONVERSION . . .

GI f

Re 2 r (t ) e j 2Sf c t

cos 2S ( f c  'f )t

 1  H sin 2S ( f c  'f )t  I

GQ f
Fig. 1.

Ts

Ts

yI n

yQ n

General structure of direct-conversion receivers.

A. Cosine estimator
Like [12], [13], [15], [22], we assume that P 4 and the
channel is stationary during the transmission of the training
sequence. The first period of the received signal (0 n <
M ) is dropped since it is considered as a part of the guard
interval. We assume the order of h+ (n) (as well as h (n))
in (5) is M . Because of the phase difference in (1) and the
assumption of stationary channel, it can be verified that for
M n (P 1)M 1, we have

(7)

Let y(n) be the received signal given in (5) and define the
31 vector1

y(n)
y(n) = y(n + M ) for M n < (P 2)M. (8)
y(n + 2M )
Using (5) and (7), we can write y(n) as

y(n) = A()
with

ejn {s(n) h+ (n)}


ejn {s (n) h (n)}

1
A() = ej(M+)
ej2(M+)


q(n)
+ q(n + M )
q(n + 2M )
(9)

1
ej(M+) .
ej2(M+)

Note that the 32 matrix A() is independent of I/Q imbalance parameters and the time index n. Collecting all the y(n)
vectors in (8), we form the 3(P 3)M matrix

y(M ) y(M + 1) . . . y((P 2)M 1)


Y = y(2M ) y(2M + 1) . . . y((P 1)M 1) .
y(3M ) y(3M + 1) . . . y(P M 1)
(10)
Using (9), we can write
Y = A()X + Q,

ej((P 2)M1) {s((P 2)M 1) h+ (n)}


ej((P 2)M1) {s ((P 2)M 1) h (n)}
and the noise matrix Q has the

q(M )
q(M + 1)
Q = q(2M ) q(2M + 1)
q(3M ) q(3M + 1)

(11)

1 In the absence of I/Q imbalance, it is sufficient to consider the correlations


between y(n) and y(n + M ) for CFO estimation [3], [4]. However, when
there is I/Q imbalance, the received signal y(n) in (5) is corrupted by the
image copy in addition to CFO. In order to decouple the effects of CFO and
I/Q imbalance, we collect y(n), y(n + M ), and y(n + 2M ) in a vector.
By doing so, we can generate a nullspace in the covariance matrix that is
dependent only on the CFO.

form

. . . q((P 2)M 1)
. . . q((P 1)M 1) .
. . . q(P M 1)
(12)
Computing YY H , we have a 33 matrix
H
H
H
YY H A() XX
   A () + QQ ,

(13)

where the approximation is made by assuming P M 1 and


the uncorrelatedness of the transmitted signal and the channel
noise. As the rank of the A()AH () is at most 2, there
T

exists a nonzero 31 vector v = v0 v1 v2
such that
vH A()AH ()v = 0.

s(n + M ) h+ (n) = ej {s(n) h+ (n)} ,


s (n + M ) h (n) = ej {s (n) h (n)} ,

where the 2(P 3)M matrix X is given by


jM
e
{s(M ) h+ (n)}
...
X=
ejM {s (M ) h (n)} . . .

2355

(14)

In practice, it is usually true that the 22 matrix is invertible


(because (P 3)M is usually much larger than 2). Therefore,
equation (14) implies AH ()v = 0, which is equivalent to

v0 + v1 ej(M+) + v2 ej2(M+) = 0
.
(15)
v0 + v1 ej(M+) + v2 ej2(M+) = 0
If we define a second degree polynomial V (x) = v0 + v1 x +
v2 x2 , then (15) means that both ej(M+) and ej(M+)
are the roots of V (x). In other words, we would like to
find a second degree polynomial V (x) with roots of the
form ej(M+) . That is, V (x) = v0 (1 ej(M+) x)(1
ej(M+) x), which implies
v0 = v2 .

(16)

Substituting this equality into (15), we have cos(M + ) =


v1 /(2v0 ). In order to incorporate the constraint (16) into the
CFO estimation method, we define

1


0
2
u0
. (17)
v = Bu with B = 0 1 and u =
u1
1
0
2
The columns of B are normalized so that BH B = I2 . Because
there exists a nonzero vector v such that AH ()v = 0, which
implies that we can also find a nonzero vector u such that
AH ()Bu = 0. The objective becomes finding a nonzero
21 vector u such that
uH BH A()AH ()Bu = 0.

(18)

From u,
we obtain the cosine estimator as cos(M + ) =
u1 /( 2u0 ).
In practice, YY H contains the noise term QQH . Because
qch (n) is white (by assumption), it can be shown by direct
calculation that E{q(n)q (n+M )} = E{q(n)q (n+2M )} =
0. Using this fact and (12), it can be verified that when the
length of the training sequence satisfies P M 1, QQH can
2
be approximated by a diagonal matrix QQH Q
I3 , where

2356

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 11, NO. 7, JULY 2012

2
Q
= (P 3)M N0 and N0 is the variance of q(n). Using
this approximation, we have
2
YY H A()AH () + Q
I3

(19)

yI n

B YY B = B A()A ()B +
= BH A()AH ()B +
H

2
Q
BH B
2
Q
I2 .

yQ n

Since B A()A ()B is a positive-semidefinite matrix,


the estimation of u is equivalent to solving the eigenvector
corresponding to the smallest eigenvalue of the 22 matrix
H
H
=
B
 YY B.
T If we denote the estimated eigenvector as u
u
0 u
1
, the CFO cosine estimate is obtained as
(21)

is generally complex-valued and cos(M ) is realNote that u


valued, the imaginary
part is removed by taking only the real

u0 . The CFO estimate is


part of
u1 / 2


1  1 

cos =
cos(M
+ ) ,
(22)
cos
M
As the range of the inverse cosine function is [0, ), the range
of CFO is

cos <
+
.
M
M
M
B. Tangent estimator
In the previous CFO estimation, is obtained by inverting
a cosine function. It is known that the performance of the
CFO estimator based on the cosine function degrades when
M + is close to a multiple of [16]. Moreover, since the
range of inverse cosine function is [0, ), a phase ambiguity
arises when M + is out of this range and thus the estimation
range of CFO is limited. To solve these issues, we introduce
another algorithm that generates sine estimate. By combining
the two CFO estimates (cosine and sine), we will obtain a
CFO estimator based on the tangent function, which is more
robust to the small M + . To derive the sine estimator, we
define another matrix based on Y as
1

0
2
0 1 .
BH
s Y with Bs =
1

0
2
Similar to (20), we obtain the 22 matrix
H
H
H
2
H
R s = BH
s YY Bs Bs A()A ()Bs + Q Bs Bs
H
2
= BH
s A()A ()Bs + Q I2 .

(23)

Under the assumption that the transmitted signal is circularly


symmetric2 , (23) can be further approximated as (see Appendix)



2c1 sin2 (M + )
j 2c2 sin(M + )
Rs
j 2c2 sin(M + ) c1
2
+ Q
I2 ,

1
cos I

e jn[

y'I n
tan I

G n nw

y 'Q n

(20)

u1

}.
cos(M
+ ) = {
2
u0

w1 n
w n

Then we can write the 22 matrix product

CFO
compensation

Stage 2

Stage 1

(24)

2 A signal s(n) is said to be circularly symmetric if its 0th -order complementary autocorrelation function E{s2 (n)} = 0 [10]. Circularly symmetric
signals include many commonly used constellations such as QAM and M PSK (M > 2).

Fig. 2.

Compensation structure in [21].

M
M
where c1 = (P 3)M Es ( k=0 |h+ (k)|2 + l=0 |h (l)|2 ),
M

M
c2 = (P 3)M Es ( k=0 |h+ (k)|2 l=0 |h (l)|2 ), and Es
is the average power of s(n). From above, we can compute
| sin(M )| from the diagonal elements of (24). In order to do
2
. From the
this, we need to know the estimated value of Q
derivation of (18)-(21), it can be seen that
2

Q
= the smallest eigenvalue of BH YY H B.

(25)

2
In other words,
Q
can be obtained in prior at the cosine
estimation. Also note that for most practical cases, the image
interference produced by I/Q imbalance is minor compared
to the desired signal, which leads to c2 > 0. Therefore, the
polarity of sine is the same as the polarity of {[Rs ]21 }. Using
this fact, (24), and (25), the CFO sine estimate is obtained as

 [R ]
2
Q
]
}
{[R
s
s
21 

 . (26)
  11
sin(M
+ ) =
|{[Rs ]21 }| 2 [R ]
2
Q
s 22

Combining the cosine and sine estimators, we have a tangent


estimator for CFO:





sin(M
+ )
1
1

tan
.
(27)
tan =
M

cos(M
+ )
As the range of inverse tangent is [, ), the range of the
CFO estimator becomes independent of , or equivalently

tan <
.

M
M
IV. C OMPENSATION OF I/Q I MBALANCE
A structure for the compensation of frequency-dependent
I/Q imbalance and the correction of CFO is proposed by
[21]. The compensation structure is shown in Figure 2. At
Stage 1, the filter and gain mismatches are compensated by a
length-L FIR filter w1 (n) at the I-branch and a proper delay
nw is added to the Q-branch to match the delay introduced
by w1 (n). At Stage 2, the phase mismatch can be compensated by computing yI (n) = (1/ cos()) (yI (n) w1 (n)) +
tan()yQ (n nw ). To further simplify the structure, the gain
factor 1/ cos() at the I-branch can be merged into w(n).

(n) is fed into the CFO
The resulting y  (n) = yI (n) + jyQ
compensator to complete the compensation process.
Below we will propose a new method to estimate w(n) and
tan() by adopting the compensation structure in Figure 2. As
we will demonstrate later, the proposed method has a much
better performance than [21]. In the following, we assume that
the CFO has been estimated as in (27) and thus it is known.
Assuming that there is no channel noise, it was shown in
[16] that if the I/Q imbalance is completely eliminated in

PAN and PHOONG: A TIME-DOMAIN JOINT ESTIMATION ALGORITHM FOR CFO AND I/Q IMBALANCE IN WIDEBAND DIRECT-CONVERSION . . .

Stages 1 and 2, then the signals yI (n) and yQ
(n) in Figure 2
3
satisfy

yI (n + 2M ) yI (n) = 2 sin(M + )yQ
(n + M ),



2 sin(M + )yI (n + M ) = yQ
(n + 2M ) yQ
(n).
(28)

In other words, w(n) and tan() should be designed such that



yI (n) and yQ
(n) satisfy (28). To do this, let us define

T
,
w = w(0) w(1) . . . w(L 1)
T

yIQ (n) = yI (n) . . . yI (n L + 1) yQ (n nw )
.
Then from Figure 2, we can write


w
T


yI (n) = (yIQ (n))
(n) = yQ (n nw ).
, yQ
tan()
(29)
Using the above equations, (28) can be rewritten in matrix
form:


w
C(n)
= d(n),
(30)
tan()
where


C(n) =

d(n) =

(yIQ (n + 2M ) yIQ (n))


T
2 sin(M + ) (yIQ (n + M ))
2 sin(M + )yQ (n nw + M )
yQ (n nw + 2M ) yQ (n nw )


,


2357

multiple of due to the utilization of sin(M + ) in the


equations of (28). One can avoid this by carefully designing
the phase rotate in GPP or one can do the following. If
the estimated sin(M + ) is smaller than some predefined
threshold, then other schemes (like [7]-[11]) can be applied
for I/Q estimation.
Estimation of CFO and receiver I/Q imbalance in the
presence of transmitter I/Q imbalance: Suppose that the
transmitter also suffers from the I/Q imbalance. When the GPP
sequence s(n) in (1) is sent, the transmitter I/Q imbalance
will introduce an image term and the transmitted sequence is
no longer given by s(n). Let s (n) denote the I/Q distorted
transmitted sequence. It can be verified [23], [24] that because
s(n) satisfies the generalized periodic structure in [22], the
I/Q contaminated sequence s (n) will also satisfy the same
generalized periodic structure (except for the first M samples).
That is,
s (n + M ) = ej s (n) for M n < (P 1)M.
Because the proposed estimation methods exploit only the
periodicity of the transmitted sequence, we can still use them
to estimate the CFO and the receiver I/Q imbalance even when
there is transmitter I/Q imbalance. The only modification is
that in this case, we need to drop 2M samples at the receiver.
That means, we collect y(n) for 2M n < P M .

Note that when formulating these matrices, the CFO parameter


is replaced with the estimate tan obtained in (27). In
practice, both C(n) and d(n) are noisy. By collecting all the
y(n) that satisfies (29), the I/Q compensation parameters w
and tan() can be estimated by minimizing the cost function
2


(P 2)M1 



w
 .
C(n)

d(n)
J (w, tan()) =


tan()
n=M+L1
(31)
 that minimize (31) can be
and tan()
The solutions w
obtained by solving a linear least square (LLS) equation. The
solution is given in closed-form:


d(M + L 1)
C(M + L 1)



C(M + L)
d(M + L)

.
.


..
..
tan()
d ((P 2)M 1)
C ((P 2)M 1)
(32)
where the superscript denotes the pseudoinverse.
The case of M + = 0: As noted in [19], the time-domain
estimation algorithms for I/Q imbalance suffer from a poor
performance when M + = 0. This can be seen from (5)
and (7). When M + = 0, the received samples y(n) in
(5) becomes periodic (in the absence of noise) for any I/Q
imbalance. Thus any estimation algorithm that exploits the
periodicity of training sequence for I/Q estimation will fail.
In our method, similar result is occurred when M + is a
3 Note that [16] considers only frequency-independent I/Q imbalance.
Furthermore, though the expressions in (28) were derived in [16], they were
used for CFO estimation (not I/Q imbalance). In this paper, we show how
the same expressions can be used for I/Q estimation.

V. C OMPLEXITY I SSUE
In this section, we will analyze the complexity of the
proposed algorithm and make comparisons to some of the
existing methods. For the CFO estimators derived in Section
3, the main computation is the calculation of the 33 matrix
YY H . Recall that the 3(P 3)M matrix Y has the
structure shown in (10). The second and third rows of Y
are the shifted versions of the first row of Y. We can use
a sliding window approach to compute the elements of YY H
efficiently. If we do this, by careful calculation, it can be
verified that the computation of YYH needs (3P 6)M
complex multiplications and (3P 6)M 7 complex additions.
For the estimator in Section 4, the main complexity is the LLS
solution in (32).
The complexity of the proposed algorithms is summarized
in Table I in terms of real additions (ADD) and real multiplications (MUL), along with existing methods in [21], [22].
For simplicity, we only show the most dominant term of
each operation. The CFO estimation in [21] requires a sample
covariance matrix and a one-dimensional search algorithm,
whereas the proposed method and [22] do not. The computations of the P P sample covariance matrix and each cost
function require 2P 2 M and 12P 2 ADD/MUL respectively.
Since the search algorithm depends on the search resolution
as well as the search range, we denote the number of search
points as and thus the complexity of one-dimensional search
algorithm is 12P 2 . In Table I, we assume that all the LLS
problems are solved using Householder solution [27], which
requires n2 (m n/3) ADD and MUL for an m n LLS
problem. In comparison, the complexity of [21] is much higher
than the other two if P is large. On the other hand, we can see
that when L > 2 (the length of w(n) is larger than 2), which

2358

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 11, NO. 7, JULY 2012

TABLE I
C OMPLEXITY C OMPARISON

[21]
[22]
Proposed

ADD/MUL



12P 2 + 2P M P + (L + 1)2


2P M (L + 2)2


2P M (L + 1)2 + 7

Proposed
[21]
No compensation

Example in Section 6
35

7.47 104
1.57 104
1.38 104

is true for most of the compensation schemes, the complexity


of our method is lower than [22].

Image Rejection Ratio (dB)

Method

40

25

20

15

VI. S IMULATION R ESULTS AND C OMPARISONS

10

Fig. 3.

10

15

20
25
SNR (dB)

30

35

40

IRR versus SNR with perfect knowledge of CFO (M = 0.078).

50
45
40
Image Rejection Ratio (dB)

In this section, computer simulations are carried out to


verify the performance of the proposed algorithms. Consider
the wireless LAN system of IEEE802.11a with carrier frequency 5.2GHz, N = 64 subcarriers, and bandwidth 20MHz.
The GPP is generated by adding phase rotate to the short
preamble of this system, which corresponds to P = 10 and
M = 16. The phase difference is set to /2 as in [22]. The
multipath channel is modeled as an 8-order Gaussian random
FIR channel and the taps are independent with variance (1/2)l
for 0 l < 8. At the receiver, the amplitude mismatch of the
oscillator is (1+)=1dB and the phase mismatch is = 5 . The
filters GI (f ) and GQ (f ) are modeled as 8-tap Butterworth
lowpass filters with cutoff frequencies 8MHz and 8.3MHz
respectively. The SNR is defined by the ratio of the power
of GPP and the variance of channel noise. All results are
averaged through 10000 trials. Note that the method in [21]
uses their MPP sequence with guard interval of length M .
Due to this, half of the training sequence is used as the guard
interval for [21]. The resolution and the range for the search
algorithm in [21] are set to 0.02 and [0.5, 0.5) respectively,
which corresponds to = 50. In Table I, we list the required
ADD/MUL of each method with the above settings. We see
that the search algorithm in [21] needs significantly more
computations.
At first, we evaluate the performances of the proposed I/Q
estimation in Sec. 4. We compare it with the method in [21].
For this purpose we assume that the CFO is known perfectly
at the receiver. Figure 3 shows the image rejection ratio (IRR)
[10] of these two methods for CFO M = 0.078. One can
see that both algorithms successfully increase the IRR as SNR
gets large. The IRR of our I/Q compensator is much better than
[21], especially when the SNR is high. The gain in IRR can
be as large as 5dB. We also plot in Figure 4 the IRR in the
frequency domain against the subcarrier index with SNR =
20dB. Since the cut-off frequencies of the lowpass filter are
chosen as 8MHz (dotted lines) and the bandwidth of the
system covers 10MHz, it is desirable to obtain high IRRs
for the subcarrier indices within 25. It can be seen that the
IRRs of our I/Q estimator are much higher than [21] for many
subcarriers, especially for the subcarrier indices close to the
cut-off frequencies.
The performances of joint estimation for CFO and I/Q
imbalance are examined in Figures 5-10. Figure 5 and 6
show the mean square error (MSE) of the CFO estimate
versus the SNR for CFO M = 0.078 and M = 0.37
respectively. The MSE of the CFO estimates is evaluated by

30

35
30
25
20
Proposed
[21]
No compensation

15
10
5

subcarriers within the cutoff frequencies


30

20

10

0
10
Subcarrier Index

20

30

Fig. 4. IRR versus subcarrier index with SNR= 20dB and perfect knowledge
of CFO (M = 0.078).

(N/(2)) E{( )2 }. In Figure 5, we first notice that [21]


has good performance at low SNR. But it soon floors as SNR
increases due to the limited resolution of the search algorithm.
The proposed algorithm is better than [22] for all SNR values.
Furthermore, the proposed algorithm has a nearly 5dB gain
at low SNR, which shows that the proposed algorithm is
more robust than [22] in the noisy environment. Also notice
that the proposed cosine method performs as well as the
tangent method when M = 0.078. When M = 0.37,
the proposed cosine method and [22] suffer large degradation
because in this case M + is closer to than the case
of M = 0.078. On the other hand, the proposed tangent
method has nearly the same performance for M = 0.078.
This confirms the fact that the proposed tangent method is
more robust to CFO variation. We also see from Figure 6
that both the proposed tangent method and the cosine method
significantly outperform [22]. The search method in [21] is
also robust to the CFO variation for moderate SNR values.
Figure 7 shows the MSE of the CFO estimate versus the

PAN and PHOONG: A TIME-DOMAIN JOINT ESTIMATION ALGORITHM FOR CFO AND I/Q IMBALANCE IN WIDEBAND DIRECT-CONVERSION . . .

2359

10

10
[22]
Proposed (cosine)
Proposed (tangent)
[21]

10

[22]
Proposed (cosine)
Proposed (tangent)

10

MSE of CFO estimate

MSE of CFO estimate

10

10

10

10

10

10

10

10

10

10

10

15
20
SNR (dB)

25

30

MSE of CFO estimate vs SNR with M = 0.078.

Fig. 5.

10

35

Fig. 7.

1.5

0.5

0.5

1.5

MSE of CFO estimate vs CFO with SNR=20dB.

10

10

SNR=10dB, [22]
SNR=10dB, Proposed (tangent)
SNR=25dB, [22]
SNR=25dB, Proposed (tangent)

10

10
2

MSE of CFO estimate

MSE of CFO estimate

10

10

10

10

10

10

10

10

Fig. 6.

10

[22]
Proposed (cosine)
Proposed (tangent)
[21]

10

15
20
SNR (dB)

25

30

35

MSE of CFO estimate vs SNR with M = 0.37.

actual CFO for SNR=20dB. We choose the CFO values such


that M is in [0.95, 0.95]. For clarity, the CFO estimator
using search algorithm in [21] is not shown. In the figure, the
proposed cosine estimator is better than [22] for |M | < /2.
However, both the cosine estimators ([22] and the Proposed
(cosine)) suffer from severe degradation when |M | is close
to /2. Furthermore, they do not work properly when CFO is
outside this range. The proposed tangent estimator extends the
CFO estimation range to |M | < . Moreover, it is also more
robust to the CFO variation than the two cosine estimators.
Figure 8 depicts the MSE of the CFO estimate versus
the number of periods P of the training sequence with
M = 0.078. When SNR=10dB, increasing P does not
improve the performance of [22] for P > 10. The MSE of
the proposed method lowers as P gets large because both the
2
I2 in
approximations XXH (diagonal) and QQH Q
Section 3 become more accurate for larger P . Similar results
can be seen for SNR=25dB.
Figure 9 and 10 show the bit error rate (BER) versus the
SNR for CFO M = 0.078 and M = 0.37 respectively.

10

Fig. 8.

10

15
P

20

25

MSE of CFO estimate vs P with M = 0.078.

The BER is evaluated by sending 10 OFDM data blocks


after GPP transmission. The block size of the data block
is N = 64 and the length of cyclic-prefix is 16. The
uncoded data bits are quadrature phase-shift keying (QPSK)
modulated and then transmitted through 48 subcarriers with index in {[24, 1], [1, 24]}. Zero-forcing channel equalization
with perfect channel knowledge is assumed at the receiver.
We also show the cases of no compensation and perfect
compensation for reference. The proposed method uses the
proposed tangent estimator for CFO estimation. As shown in
the figures, since both CFO and I/Q imbalance generate intercarrier interference, the performances are considerably poor
without compensation. Although the method in [21] has the
best performances at low SNR, it suffers from degradation as
SNR increases. In both figures, our method is better than the
method in [22], especially for the low and moderate SNRs.
The performance gain is larger when M = 0.37. This is
mainly due to the improvement of the CFO estimation of our
method.
In summary, we observe from the above simulations that the

2360

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 11, NO. 7, JULY 2012

our method has a better performance (especially when the


SNR is not very high) and a slightly lower complexity than
[22].

10

Bit error rate

10

A PPENDIX
D ERIVATION OF (24)

10

as

10

No compensation
[22]
Proposed
[21]
Perfect compensation

10

10

15
20
SNR (dB)

with
11 =
25

30

35

L+ L



(P 2)M1

12 =

 

22 =

L+ L



n=M

h (k1 )h (k2 )

(P 2)M1


Bit error rate

s (n k1 )s(n k2 ),

n=M

where L+ and L are the orders of h+ (n) and h (n)


respectively. Under the assumptions of (P 3)M 1,
L+ < M , L < M , and circularly symmetric s(n), we have
the following properties [10]

10

10

(P 2)M1




No compensation
[22]
Proposed
[21]
Perfect compensation

10

15
20
SNR (dB)

s(n k1 )s (n k2 )

n=M
(P 2)M1

10

Fig. 10.

k1 )s(n k2 ),

n=M

k1 =0 k2 =0

10

s(n k1 )s (n k2 ),

(P 2)M1

h+ (k1 )h (k2 )
ej2n s(n

k1 =0 k2 =0

10

h+ (k1 )h+ (k2 )

k1 =0 k2 =0
L+

BER vs SNR with M = 0.078.

Fig. 9.

After some arrangements, the matrix in can be written




11 12
= XXH =
,
(33)

12
22

(P 3)M Es ,
0,

k1 = k2
otherwise

ej2n s (n k1 )s(n k2 ) 0, k1 , k2 ,

n=M
25

30

35

BER vs SNR with M = 0.37.

proposed cosine and tangent CFO estimators perform better


than [22], especially at low SNR. The proposed tangent CFO
estimator is more robust to CFO variation than the proposed
cosine estimator and [22]. The search method in [21] has
the best performance at low SNR but it degrades at high
SNR due to the limited resolution. Moreover, its complexity
is much higher than [22] and the proposed methods (see
Table I). That is, the proposed method offers a complexity
that is significantly lower than [21] by sacrificing only a little
performance at low SNR. As a result, the proposed method
compares favorably with the existing methods in [21] and [22].
VII. C ONCLUSIONS
In this paper, we proposed two novel CFO estimators that
utilize the generalized periodic structure of the preamble
sequence. The proposed method does not need to know
the channel impulse response and the exact value of the
training samples. The parameters of the I/Q imbalance can be
estimated using a new LLS algorithm. Like [22], the proposed
CFO estimator does not need a search algorithm. Moreover,

where Es is the average power of s(n). Using the above


properties, (33) can be approximated as
 

L+
2
|h
(k)|
0
+
k=0
,
(P 3)M Es
L
2
0
l=0 |h (l)|
(34)
By substituting (34) into the first term on the right-hand-side
of (23), the equation (24) is established.
R EFERENCES
[1] T. Pollet, M. Van Bladel, and M. Moeneclaey, BER sensitivity of
OFDM systems to carrier frequency offset and Wiener phase noise,
IEEE Trans. Commun., vol. 43, no. 234, pp. 191193, Feb./Mar./Apr.
1995.
[2] B. Razavi, Design considerations for direct-conversion receivers, IEEE
Tran. Circuits and Systems II: Analog and Digital Sig. Process., vol. 44,
no. 6, pp. 428435, June 1997.
[3] P. H. Moose, A technique for orthogonal frequency division multiplexing frequency offset correction, IEEE Trans. Commun., vol. 42, no. 10,
pp. 29082914, Oct 1994.
[4] J. Li, G. Liu, and G. B. Giannakis, Carrier frequency offset estimation
for OFDM-based WLANs, IEEE Signal Process. Lett., vol. 8, no. 3,
pp. 8082, Mar. 2001.
[5] V. K.-P. Ma and T. Ylamurto, Analysis of IQ imbalance on initial frequency offset estimation in direct down-conversion receivers, in Proc.
2001 IEEE Workshop on Sig. Proc. Advances in Wireless Commun., pp.
158161.
[6] M. Inamori, A. M. Bostamam, and Y. Sanada, Performance of frequency offset estimation in the presence of IQ imbalance for OFDM
direct conversion receivers, in Proc. 2009 IEEE Pacific Rim Conf. on
Commun., Computers and Sig. Proc., pp. 826831.

PAN and PHOONG: A TIME-DOMAIN JOINT ESTIMATION ALGORITHM FOR CFO AND I/Q IMBALANCE IN WIDEBAND DIRECT-CONVERSION . . .

[7] M. Valkama, M. Renfors, and V. Koivunen, Advanced methods for


I/Q imbalance compensation in communication receivers, IEEE Trans.
Signal Process., vol. 49, pp. 23352344, Oct. 2001.
[8] D. Tandur and M. Moonen, Joint adaptive compensation of transmitter
and receiver IQ imbalance under carrier frequency offset in OFDMbased systems, IEEE Trans. Signal Process., vol. 55, pp. 52465252,
Nov. 2007.
[9] D. Tandur and M. Moonen, Joint compensation of OFDM frequency
selective transmitter and receiver IQ imbalance, in Proc. 2007 IEEE
International Conf. on Acoustics, Speech and Sig. Proc., vol. 3, pp.
III-81-III-84.
[10] L. Anttila, M. Valkama, and M. Renfors, Circularity-based I/Q imbalance compensation in wideband direct-conversion receivers, IEEE
Trans. Veh. Technol., vol. 57, no. 4, pp. 20992113, July 2008.
[11] J. Tubbax, A. Fort, L. Van der Perre, S. Donnay, M. Engels, M. Moonen,
and H. De Man, Joint compensation of IQ imbalance and frequency
offset in OFDM systems, in Proc. IEEE Global Telecommunications
Conf., pp. 23652369.
[12] F. Yan, W.-P. Zhu, and M. Omair Ahmad, Carrier frequency offset estimation and I/Q imbalance compensation for OFDM systems, EURASIP
J. Advances in Signal Proc., vol. 2007, article ID 45364.
[13] S. De Rore, E. Lopez-Estraviz, F. Horlin, and L. Van der Perre, Joint
estimation of carrier frequency offset and IQ imbalance for 4G mobile
wireless systems, in Proc. 2006 IEEE International Conf. on Commun.,
pp. 20662071.
[14] M.-F. Sun, J.-Y. Yu, and T.-Y. Hsu, Estimation of carrier frequency offset with I/Q mismatch using pseudo-offset injection in OFDM systems,
IEEE Trans. Circuits and Systems, vol. 55, no. 3, Apr. 2008.
[15] Y. Chen, A. D. S. Jayalath, J. A. Zhang, and T. Pollock, Frequency
offset estimation for OFDM systems in the presence of I/Q mismatch,
in Proc. 2007 IEEE International Symposium on Personal, Indoor and
Mobile Radio Commun., pp. 15.
[16] Y. Chen, J. A. Zhang, and A. D. S. Jayalath, Low-complexity estimation
of CFO and frequency independent I/Q mismatch for OFDM systems,
EURASIP J. Wireless Commun. and Networking, volume 2009, article
ID 542187.
[17] F. Horlin, A. Bourdoux, and L. Van der Perre, Low-complexity EMbased joint acquisition of the carrier frequency offset and IQ imbalance,
IEEE Trans. Wireless Commun., vol. 7, no. 6, pp. 22122220, June 2008.
[18] Y.-H. Chung, K.-D. Wu, and S.-M. Phoong, Joint estimation of I/Q
imbalance, CFO and channel response for OFDM systems, in Proc.
2009 IEEE International Conf. on Acoustics, Speech and Sig. Proc., pp.
25732576.
[19] M. Inamori, A. Bostaman, Y. Sanada, and H. Minami, IQ imbalance
compensation scheme in the presence of frequency offset and dynamic
DC offset for a direct conversion receiver, IEEE Trans. Wireless
Commun., vol. 8, pp. 22142220, May 2009.
[20] D. Tandur and M. Moonen, Efficient compensation of RF impairments
for OFDM systems, in Proc. 2009 IEEE Wireless Comm. and Networking Conf., pp. 16.
[21] G. Xing, M. Shen, and H. Liu, Frequency offset and I/Q imbalance
compensation for direct-conversion receivers, IEEE Trans. Wireless
Commun., vol. 4, no. 2, pp. 673680, Mar. 2005.
[22] H. Lin, X. Zhu, and K. Yamashita, Low-complexity pilot-aided com-

[23]
[24]
[25]
[26]
[27]

2361

pensation for carrier frequency offset and I/Q imbalance, IEEE Trans.
Commun., vol. 58, no. 2, Feb. 2010.
L. Anttila, M. Valkama, and M. Renfors, Frequency-selective I/Q
mismatch calibration of wideband direct-conversion transmitters, IEEE
Trans. Circuits and Systems II, vol. 55, no. 4, pp. 359363, Apr. 2008.
Y. Zou, M. Valkama, and M. Renfors, Pilot-based compensation of
frequency-selective I/Q imbalances in direct-conversion OFDM transmitters, 2008 IEEE Conf. on Vehicular Technology Fall.
Y.-C. Pan and S.-M. Phoong, A new algorithm for carrier frequency
offset estimation in the presence of I/Q imbalance, 2010 IEEE Conf.
on Vehicular Technology Spring Spring.
Y.-C. Pan and S.-M. Phoong, A new estimation algorithm for CFO
and I/Q imbalance in wideband direct-conversion receivers, 2011 IEEE
Global Telecommunications Conf.
G. H. Golub, and C. F. Van Loan, Matrix Computations, 3rd edition.
The Johns Hopkins University Press, 1996.
Yen-Chang Pan (M09) was born in Taipei, Taiwan,
in 1983. He received the B.S. and M.S. degree in
communication engineering from National ChiaoTung University (NCTU), Hsinchu, Taiwan, in 2006
and 2008 respectively. He is now a Ph.D. student at
Graduate Institute of Communication Engineering,
National Taiwan University (NTU), Taipei, Taiwan.
His current research interests include discrete-time
signal processing for the estimation of front-end
non-idealities in communication systems.

See-May Phoong (M96-SM04) was born in Johor,


Malaysia, in 1968. He received the B.S. degree
in electrical engineering from the National Taiwan
University (NTU), Taipei, Taiwan, R.O.C., in 1991
and M.S. and Ph.D. degrees in electrical engineering from the California Institute of Technology
(Caltech), Pasadena, California, in 1992 and 1996,
respectively.
He was with the Faculty of the Department
of Electronic and Electrical Engineering, Nanyang
Technological University, Singapore, from September 1996 to September 1997. In September 1997, he joined the Graduate
Institute of Communication Engineering and the Department of Electrical
Engineering, NTU, as an Assistant Professor, and since August 2006, he has
been a Professor.
Dr. Phoong has previously served as an Associate Editor for IEEE Transactions on Circuits and Systems II: Analog and Digital Signal Processing
(Jan. 2002 - Dec. 2003), IEEE Signal Processing Letters (March 2002 - Feb.
2005) and IEEE T RANSACTIONS ON C IRCUITS AND S YSTEMS I: R EGULAR
PAPERS (Jan. 2006 - Dec. 2007). His interests include multirate signal
processing, filter banks and their application to communications. He received
the Charles H. Wilts Prize (1997) for outstanding independent research in
electrical engineering at Caltech. He was also a recipient of the Chinese
Institute of Electrical Engineerings Outstanding Youth Electrical Engineer
Award (2005).

Das könnte Ihnen auch gefallen