Standard-Independent I/Q Imbalance

Compensation in OFDM Direct-Conversion

Receivers
Marcus Windisch, Gerhard Fettweis
Dresden University of Technology, Vodafone Chair Mobile Communications Systems,
D-01062 Dresden, Germany
Email: windisch@ifn.et.tu-dresden.de
AbstractThe growing number of different mobile
communications standards calls for inexpensive and
highly exible receiver architectures supporting all
standards. The direct conversion receiver is a very
attractive candidate to reach this goal. However,
unavoidable imbalances between the I- and the Q-
branch of the I/Q demodulator lead to a signicant
performance degradation at the reception of OFDM
signals.
In this paper a novel compensation algorithm
is presented to overcome this problem. The novel
approach does not depend on any standard-specic
signal components, such as pilots or a preamble.
Instead, a blind I/Q imbalance parameter estimation is
performed during the ordinary receive mode. There-
fore, the algorithm is applicable to a wide range of
present and future OFDM communications standards.
Index TermsOFDM, Direct-conversion receiver,
I/Q imbalance, Blind compensation, IEEE 802.11a
I. INTRODUCTION
Advanced receiver architectures based on I/Q
signal processing are very attractive because the
need for a bulky analog image rejection lter is
avoided. However, one of the drawbacks is I/Q
imbalance, resulting from imperfect matching of the
analog components in the I- and the Q-branch of the
receiver.
The estimation and compensation of the I/Q im-
balance in OFDM direct-conversion receivers has
been addressed in the literature before. The most
intuitive approach for an estimation of the I/Q im-
balance parameters is to feed the I/Q mixer with
dedicated calibration or training signals. In [1] this is
This work was partly supported by the German Ministry of
Education and Research (BMBF) within the project Wireless
Gigabit with advanced multimedia support (WIGWAM) under
grant 01BU370
done by feeding the RF part off-line with a receiver-
generated calibration signals. The effort of additional
hardware in the receiver can be avoided if compo-
nents of the received signal itself, such as known
pilots, are used for a calibration [2], [3]. However,
such techniques are limited to a certain class of
communications standards with the presumed pilots.
Furthermore, in a practical scenario the pilots are
likely to be affected not only by the receiver I/Q
imbalance, but also by other impairments, such as
the transmission channel and various RF impair-
ments at the transmitter and the receiver. Because
all these effects have to be considered, an accurate
pilot based I/Q imbalance compensation is likely to
be very complex.
The dependence on known pilots is avoided by
applying blind signal processing techniques. The
compensation of I/Q imbalance in the time domain
based on Blind Signal Separation (BSS) has been
proposed in [4]. In addition to the computational
complexity of BSS, the drawback of BSS is one
of its fundamental assumptions, that at most one
of the source signals to be reconstructed is Gaus-
sian distributed. However, both the I- and the Q-
component of the OFDM time domain signals are
asymptotically Gaussian distributed as the number
of subcarriers and the modulation order gets larger.
This property limits the application of BSS based
techniques for OFDM applications.
In order to cope with receiver I/Q imbalance in
a wide range of present and future OFDM com-
munications standards an advanced estimation and
compensation technique is required. In this paper we
will present a novel approach for a blind estimation
of the unknown I/Q imbalance. It does not require
any pilot signals, which makes it independent from
the targeted communications standard.
This paper is structured as follows: Section II
introduces a model for the I/Q imbalance, wich is
used in this paper. A novel approach for the blind
estimation and compensation of the I/Q imbalance
is presented in sections III and IV. In section V we
demonstrate the capabilities of this novel compen-
sation technique exemplary for the IEEE 802.11a
standard, followed by the conclusions in section VI.
II. I/Q IMBALANCE MODEL
The purpose of any receiver architecture is to
convert a frequency band of interest of an RF signal
down to base band. The real RF signal just before
down-conversion can be written as
r(t) = y(t)e
+j2f
C
t
+ y

(t)e
j2f
C
t
, (1)
where y(t) denotes the complex base band equiva-
lent of the frequency band of interest, f
C
denotes
its carrier frequency and ()

denotes (complex)
conjugation. The principle of direct-conversion is
to use a local oscillator (LO) with the frequency
f
LO
= f
C
and to apply complex (I/Q) mixing. For
reference, we will rst consider the case of no I/Q
imbalance. In this case, the down-conversion yields
the frequency band of interest
y(t) = LP{r(t)x
LO
(t)}, (2)
where x
LO
(t) = e
j2f
LO
t
denotes the time function
of the (perfectly balanced) complex LO and LP{}
denotes low pass ltering. If the direct-conversion
architecture is used for the reception of an OFDM
signal, y(t) equals the received OFDM base band
signal. This signal is further A/D-converted and
its samples y(k) are used to compute the symbols
on the subcarriers by applying the Discrete Fourier
Transform (DFT)
Y
m
(n) =
1
L
DFT
L
DFT
1

l=0
y(l)e
j2lm/L
DFT
, (3)
were L
DFT
denotes the number of subcarriers.
Y
m
(n) denotes the demodulated subcarrier symbol
at the m
th
sumcarrier of the n
th
OFDM symbol. For
a better understanding of the I/Q imbalance effects,
the analyzed interval of subcarrier indices is set to
m [L
DFT
/2; L
DFT
/2 1], where m=0 denotes
the DC subcarrier.
Unfortunately, a perfect analog I/Q mixing is not
achievable in practice. Unavoidable tolerances in the
manufacturing process lead to deviations from the
desired 90

phase shift and the desired equal gain

in the I- and the Q-branch. These imperfections can
be modelled by a complex LO with the time func-
tion x
LO
(t) = cos(2f
LO
t) jg sin(2f
LO
t + ),
Y
m
(n)
f
f
Z
m
(n)
f
LO
Y

m
(n)
f
LO
Y
m
(n)
b)
a)
K
1
e
j2fLOt
Y

m
(n)
Z
m
(n)
K
2
e
+j2fLOt
Fig. 1. Frequency domain illustration of I/Q imbalance in
OFDM direct-conversion receivers: a) Spectrum of the RF signal
r(t), b) Spectrum of the base band signal z(t)
where g denotes the amplitude imbalance and
denotes the phase imbalance. Furthermore, we dene
the complex I/Q imbalance parameters
K
1
=
1 + ge
j
2
, K
2
=
1 ge
+j
2
, (4)
in order to rewrite the time function of the complex
LO with I/Q imbalance as
x
LO
(t) = K
1
e
j2f
LO
t
+ K
2
e
+j2f
LO
t
. (5)
This means, direct-conversion with I/Q imbalance
can be interpreted as a superposition of a desired
complex down-conversion (weighted by K
1
) and an
undesired complex up-conversion (weighted by K
2
).
Consequently, the received base band signal with
I/Q imbalance z(t) is a superposition of the desired
frequency band y(t) and its image y

(t):
z(t) = LP{r(t) x
LO
(t)} = K
1
y(t) + K
2
y

(t). (6)
This superposition translates to a mutual interfer-
ence of symmetric OFDM symbols (see Fig. 1). With
(3) and (6), the received OFDM symbols in the case
of I/Q imbalance become:
Z
m
(n) = K
1
Y
m
(n) + K
2
Y

m
(n). (7)
Equivalently, we can write for the mirror subcarrier
Z
m
(n) = K
1
Y
m
(n) + K
2
Y

m
(n). (8)
Merging (7) and the conjugate of (8) leads to a
convenient matrix description of the I/Q imbalance
effects:
_
Z
m
(n)
Z

m
(n)
_
= K
_
Y
m
(n)
Y

m
(n)
_
, K =
_
K
1
K
2
K
2

K
1

_
.
(9)
Since the matrix K is always non-singular for
realistic imbalance parameters, the desired OFDM

K
1

Y
m
(n),

Y
m
(n)
D
F
T
z(t)
ADC
Collection
of statistics

K
1

K
2
Z
m
(n), Z
m
(n)
Fig. 2. Structure of the proposed I/Q imbalance compensation
algorithm
symbols Y
m
(n) and Y
m
(n) can be perfectly re-
constructed based on the interfered symbols Z
m
(n)
and Z
m
(n) using the inverse K
1
. It should be
stressed, that the reconstructed symbols Y
m
(n) and
Y
m
are not necessarily identical to the transmitted
OFDM symbols. Instead, they might be corrupted
by the channel or other RF impairments. The com-
pensation of such distortions is beyond the scope of
this paper. We focus on a cancellation of the I/Q
imbalance effects, i.e. the goal is to provide OFDM
symbols equivalent to those of a perfectly balanced
direct-conversion.
The key for the compensation is to gain knowl-
edge about the unknown mixing matrix K. An anal-
ysis of the mixing equation (9) shows its structural
similarity to the mixing equation in Low-IF receivers
[5]. In fact, considering the effects of I/Q imbalance,
the mutual interference between desired signal and
image signal in Low-IF receivers corresponds to the
mutual interference between symmetric subcarriers
in OFDM direct-conversion receivers. It has been
shown in [5], that a completely blind estimation
of the unknown mixing matrix is possible for the
Low-IF case. In this paper we will show, that the
parameter estimation approach of [5] can be adapted
to OFDM direct-conversion receivers.
III. I/Q IMBALANCE COMPENSATION
The proposed I/Q imbalance compensation ap-
proach consists of two steps: The blind determina-
tion of the product K
1
K
2
, followed by the derivation
of the inverse K
1
. For the purpose of a clear pre-
sentation we will drop the sample time index (n) in
this section. However, it should be kept in mind that
the investigated OFDM symbols are generally time
variant. In order to determine K
1
K
2
, two statistics
of the imbalanced symbols Z
m
and Z

m
will be
evaluated, namely the cross-correlation E{Z
m
Z
m
}
and the power of the sum E
_
|Z
m
+ Z

m
|
2
_
, where
E{} denotes expectation.
The only assumption we introduce is, that
E{Y
m
Y
m
} = 0 holds at the examined subcarrier
index m, i.e. Y
m
and Y

m
are uncorrelated and
have zero mean
1
. This assumption is realistic at
least for pairs of data-subcarriers, if a proper source
and channel coding is applied. Analyzing the cross-
correlation term yields
E{Z
m
Z
m
}
= K
1
K
2
E{Y
m
Y

m
} + K
1
K
2
E
_
Y
m
Y

m
_
+ K
1
2
E{Y
m
Y
m
} + K
2
2
E
_
Y

m
Y

m
_
= K
1
K
2
(P
m
+ P
m
) , (10)
where P
m
= E{Y
m
Y

m
} and P
m
= E
_
Y
m
Y

m
_
denote the power of the subcarriers. Furthermore, we
can write
Z
m
+ Z

m
= (K
1
+ K
2

) Y
m
+ (K
2
+ K
1

) Y

m
= Y
m
+ Y

m
, (11)
which holds, since denition (4) yields
K
1
+ K
2

= K
2
+ K
1

= 1. (12)
Now, the second expectation term can be written as
E
_
|Z
m
+ Z

m
|
2
_
= E
_
|Y
m
+ Y

m
|
2
_
= E{Y
m
Y

m
} + E
_
Y
m
Y

m
_
+ E{Y
m
Y
m
} + E
_
Y

m
Y

m
_
= P
m
+ P
m
. (13)
Merging (10) and (13) results in the backbone equa-
tion of the proposed imbalance parameter estimation
scheme:
E{Z
m
Z
m
}
E
_
|Z
m
+ Z

m
|
2
_ = K
1
K
2
(14)
In order to derive the compensation matrix K
1
, the
product has to be split into its composing factors.
Although K
1
and K
2
cannot be separated directly,
the parameters g and can be extracted. Denition
(4) yields K
1
K
2
=
1
4
_
1 g
2
_
j
1
2
g sin, which
can be converted into:
g =
_
1 4 Re {K
1
K
2
}
= arcsin
_

2
g
Im{K
1
K
2
}
_
.
(15)
Re {} and Im{} denote the real part and the
imaginary part, respectively. The values K
1
and K
2
are now determinable using denition (4). Hence, at
least one pair of uncorrelated subcarriers allows for a
completely blind determination of the unknown I/Q
1
A residual cross-correlation E{Y
m
Y
m
} = 0 limits the
achievable accuracy of the parameter estimation. A comprehen-
sive theoretical analysis of how the parameter estimation in a
Low-IF receiver is affected by a residual cross-correlation can
be found in [6].
imbalance parameters. The structure of the resulting
I/Q imbalance parameter estimation and compensa-
tion algorithm is depicted in Figure 2.
IV. REALIZATION ASPECTS
In a practical receiver the expectation terms of
(14) have to be replaced by sample based approxi-
mations. This can be done by an averaging operation
over multiple pairs of uncorrelated subcarriers. Fur-
thermore, the I/Q imbalance parameters change very
slowly with time. Hence, an averaging over time
is also reasonable. The estimation can be formally
written as

K
1

K
2
=

mM

nN
Z
m
(n)Z
m
(n)

mM

nN
|Z
m
(n) + Z

m
(n)|
2
.
(16)
M denotes the chosen subset of M (positive) sub-
carrier indices, N denotes the chosen subset of N
sample time indices. Obviously, the accuracy of the
estimation will be affected by the number of incor-
porated sample pairs MN. An increased subcarrier
block size M raises the computational effort at
each time instant n, whereas an increased temporal
block size N raises the duration of the parameter
estimation. Hence, the proposed parameter estima-
tion allows for a exible tradeoff between accuracy,
computational effort and measurement time.
The separation scheme of the product into the
estimates

K
1
and

K
2
is the same as sketched for the
exact values, leading to the estimated I/Q imbalance
compensation matrix

K
1
=
1
|

K
1
|
2
|

K
2
|
2
_

K

1


K
2

2

K
1
_
. (17)
Using this blindly gained compensation matrix, a
reconstruction of the unknown source signals is
possible:
_

Y
m
(n)

Y

m
(n)
_
=

K
1
_
Z
m
(n)
Z

m
(n)
_
=

K
1
K
_
Y
m
(n)
Y

m
(n)
_
.
(18)
The computational effort of the proposed I/Q imbal-
ance compensation technique is very low. In order
to implement the parameter estimation (16), only
8M real additions and 6M real multiplications are
required at the OFDM symbol rate. The more costly
calculation of the compensation matrix

K
1
can be
done at a much lower rate, depending on how fast
the I/Q imbalance changes with time.
10 15 20 25 30
10
4
10
3
10
2
10
1
10
0
SNR [dB]
S
E
R
I/Q Imbalance (no compensation)
I/Q Imbalance (with compensation)
Reference: no I/Q Imbalance
N=10
N=100
N=1000
Fig. 3. Mean symbol error rate for an AWGN channel and I/Q
imbalance parameter estimation based on all data-subcarriers
(64-QAM, g=1.05, =5

)
V. SIMULATION RESULTS
We will demonstrate the capabilities of the pro-
posed I/Q imbalance compensation algorithm con-
sidering the IEEE 802.11a WLAN standard [7],
which is a widely used OFDM-based wireless com-
munications standard. The highest modulation order
(64-QAM), which is also most sensitive to I/Q
imbalances, is used in the simulations.
For each simulation setting, we rst perform a pa-
rameter estimation based on the proposed approach,
followed by a measurement of the resulting symbol
error rate (SER), if the estimated compensation
matrix

K
1
is used for correction. The estimation
is done based on all data-subcarriers, which results
in M=24 for the IEEE 802.11a signal [7]. Because
the parameter estimation is done based on the ac-
tual quasi-random realization of the incorporated
symbols Z
m
(n), the accuracy of the estimation and
consequently the resulting SER will vary. Therefore,
instead of a single realization SER, it is more rea-
sonable to plot the mean SER. The following results
are calculated based on 1000 independent trails of
the parameter estimation for each simulation setting.
We start our analysis with the simple case of an
AWGN channel. Figure 3 shows the performance
degradation due to the I/Q imbalance for g=1.05,
=5

. The parameter estimation was done using all

data-subcarriers (M=24) for a varying block size
N. The simulations show a moderate sensitivity to
I/Q imbalance for an AWGN channel (3 dB SNR-
degradation at SER=10
3
). Even a small block size
(N=100, equivalent to 0.4 milliseconds measure-
ment time) is sufcient for the parameter estimation
to push this SNR-degradation below 0.1 dB.
10 20 30 40 50 60
10
4
10
3
10
2
10
1
10
0
SNR [dB]
S
E
R
I/Q Imbalance (no compensation)
I/Q Imbalance (with compensation)
Reference: no I/Q Imbalance
N=10000
N=1000
N=100
N=10
Fig. 4. Mean symbol error rate for an ETSI-A channel and
I/Q imbalance parameter estimation based on all data-subcarriers
(64-QAM, g=1.05, =5

)
10 20 30 40 50 60
10
4
10
3
10
2
10
1
10
0
SNR [dB]
S
E
R
I/Q Imbalance (no compensation)
I/Q Imbalance (with compensation)
Reference: no I/Q Imbalance
g=1.05, =5

g=1.02, =2

g=1.1, =10

Fig. 5. Mean symbol error rate for an ETSI-A channel and
I/Q imbalance parameter estimation based on all data-subcarriers
(64-QAM, N=1000)
The impact of I/Q imbalance is substantially more
critical for a frequency selective channel. In our
simulations we use the ETSI-A channel model [8].
Figure 4 shows, that even for a high SNR the SER
never falls below a specic limit, which is prede-
termined by the parameters of the I/Q imbalance
(here: 5.3 % SER for g=1.05, =5

). Such a lower
limit still exists if the proposed compensation is
applied. However, depending on the requirements of
the chosen communications standard, the error oor
can be arbitrarily decreased by a proper choice of
the block size N.
Furthermore, the actual I/Q imbalance parameters
of a practical receiver are hardly predictable. Instead,
they will vary with the manufacturing process. As a
consequence, the limiting SER error oor due to I/Q
imbalance will be arbitrary (Fig. 5). In contrast, for
a xed choice of M and N, the mean SER with
compensation is independent from the actual values
of the I/Q imbalance parameters. The error oor
due to I/Q imbalance is kept at an predetermined
arbitrary low level. Therefore, by using the proposed
digital compensation scheme, the demands to the
analog part of the receiver can be drastically reduced.
VI. CONCLUSION
A novel approach for a blind estimation and
compensation of I/Q imbalances in OFDM direct-
conversion receivers has been proposed. Because
the parameter estimation does not depend on any
structures of the received signals (such as pilots), this
technique is applicable to a wide range of present
and future OFDM based communications standards.
The accuracy of the I/Q imbalance parameter estima-
tion and consequently the performance of the com-
pensation is arbitrarily scalable, allowing for a ex-
ible tradeoff between performance, computational
effort and measurement time. In order to fulll the
requirements of a chosen communications standard,
the effects of the I/Q imbalance can be compensated
digitally, while scaling down the demands for the
analog part of the receiver.
REFERENCES
[1] S ebastien Simoens, Marc de Courville, Francois Bourzeix,
and Paul de Champs, New I/Q imbalance modeling and
compensation in OFDM systems with frequency offset, in
Proc. IEEE PIMRC 2002, Sept. 2002, vol. 2, pp. 561566.
[2] Andreas Schuchert, Ralph Hasholzner, and Patrick Antoine,
A novel IQ imbalance compensation scheme for the recep-
tion of OFDM signals, IEEE Trans. Consumer Electron.,
vol. 47, no. 3, pp. 313318, Aug. 2001.
[3] Jan Tubbax, Boris C ome, Liesbet Van der Perre, Luc
Deneire, St ephane Donnay, and Marc Engels, Compensa-
tion of IQ imbalance in OFDM systems, in Proc. IEEE Intl.
Conf. on Communications (ICC 2003), May 2003, vol. 5,
pp. 34033407.
[4] Piotr Rykaczewski, Volker Blaschke, and Friedrich K. Jon-
dral, I/Q Imbalance Compensation for Software Dened
Radio OFDM Based Direct Conversion Receivers, in Proc.
7th Intl. OFDM-Workshop, Hamburg, Germany, 2003.
[5] Marcus Windisch and Gerhard Fettweis, Blind I/Q Im-
balance Parameter Estimation and Compensation in Low-
IF Receivers, in Proc. 1st Intl. Symposium on Control,
Communications and Signal Processing (ISCCSP 2004),
Hammamet, Tunisia, 21-24 Mar. 2004.
[6] Marcus Windisch and Gerhard Fettweis, Performance
Analysis for Blind I/Q Imbalance Compensation in Low-
IF Receivers, in Proc. 1st Intl. Symposium on Control,
Communications and Signal Processing (ISCCSP 2004),
Hammamet, Tunisia, 21-24 Mar. 2004.
[7] IEEE, Part11: Wireless LAN Medium Access Control
(MAC) and Physical Layer (PHY) specications, IEEE
Std 802.11a-1999, 1999.
[8] ETSI EP BRAN, Channel models for HIPERLAN/2 in
different indoor scenarios, Mar. 1998.