IQ Imbalance Correction for OFDMA Uplink
Systems
Hisham A. Mahmoud, H seyin Arslan
u Mehmet Kemal Ozdemir, Francis E. Retnasothie
Electrical Engineering Dept., Univ. of South Florida Logus Broadband Wireless Solutions
4202 E. Fowler Ave., ENB-118, Tampa, FL, 33620 1408 Birchmount Road Toronto, ON, M1P 2E3 Canada
E-mail: {kemal, francis}@loguswireless.com
E-mail: abp4ig@r.postjobfree.com, abp4ig@r.postjobfree.com
OFDMA-uplink (UL) systems, IQ imbalance problem be-
Abstract Direct conversion receivers are attractive for low
cost systems as they avoid intermediate frequency (IF) lters. comes more complicated. First, the non-idealities in multiple
However, the direct conversion from radio frequencies (RF) to user front-ends are expected to be different. Second, the total
complex inphase (I) and quadrature (Q) baseband signals in one
IQ imbalance effects on OFDMA systems results in multiuser
mixing step introduces additional front-end distortions. These
interference (MUI) as opposed to inter-carrier interference
IQ distortions lead to a degradation in the system performance.
(ICI) in OFDM systems. In addition, in OFDMA-UL systems,
The problem becomes more signi cant in orthogonal frequency
division multiple access (OFDMA) systems where multiple user preambles are not used for channel estimation. Instead, the
signals with different IQ impairments are combined in the uplink channel is divided into subchannels -that can be assigned
(UL) signal. In this paper, detection methods for OFDMA-UL
to users- with dedicated pilots for each subchannel. Without
signals corrupted by IQ distortions are investigated. The received
preambles, many of proposed algorithms above are inapplica-
signal as a function of transmitted signals, IQ parameters,
ble.
and communication channels is mathematically formulated. We
designed a novel pilot pattern that is used by two proposed In this paper, we consider the effect of IQ imbalance on
estimation and compensation methods of IQ impairments to the the OFDMA-UL system. The received signal as a function
signal. Proposed methods were shown to signi cantly improve
of multiuser transmitted signals, IQ parameters, and commu-
the system performance.
nication channels is mathematically formulated. This signal
model is then used to investigated methods to estimate and
I. I NTRODUCTION
equalized both the multiuser channels and IQ distortions. A
Physical layers (PHY) based on orthogonal frequency divi- novel pilot pattern is designed which is then used by two
sion multiplexing (OFDM) have been proposed for various proposed methods to ef ciently mitigate signal distortions
wireless standards such as IEEE 802.11a, IEEE 802.11g, caused by the combined effect of multipath dispersive channels
digital audio broadcasting (DAB), digital video broadcasting and IQ imbalances of multiple users. The proposed methods
(DVB), and IEEE 802.16e [1] [5]. The direct conversion are shown to signi cantly reduce the impact of IQ imbalance
receiver is an attractive architecture since it avoids costly on OFDMA signals.
intermediate frequency (IF) lters, reduces power consump-
II. S YSTEM M ODEL
tion, and allows for easier integration than super-heterodyne
structure [6]. However, direct conversion receivers cause more We consider an OFDMA-UL system employing N subcar-
distortions to the baseband signal due to the imbalance be- riers per symbol. The DC subcarrier and guard subcarriers are
tween the inphase (I) and quadrature (Q) branches. The IQ disabled. The remaining subcarriers Nused are used for data
imbalance results in a degradation in the system performance. transmission with equal number of subcarriers on each side
This drawback of direct conversion receivers becomes more of the DC subcarrier. The above assumptions are common
signi cant with OFDM systems as they are known to be for most practical implementations of OFDMA systems. Used
sensitive to receiver front-end non-idealities [7]. Thus, there subcarriers are divided into M subchannels, the indexes of
is a need for methods to nullify or reduce the IQ distortions. each subchannel m is in the set Mm . Subchannels can then be
Several techniques have been proposed to reduce the degra- assigned to multiple users. Every user m generates a sequence
dation caused by IQ imbalance in OFDM-based systems [8]. of modulated symbols xm which are mapped to subcarriers of
In [9] [11], the IQ imbalance caused by the receiver front- assigned subchannels. The signal in then fed to an inverse fast
end is investigated. The combined effects of IQ imbalance at Fourier transform (IFFT) block and the cyclic pre x (CP) is
the transmitter and receiver front-ends is investigated in [12] added. The nth sample of the transmitted symbol of user m
[15]. An interesting approach that uses the in uence of IQ is,
imbalance to obtain diversity gain is proposed in [16].
xm (k )ej 2 nk/N,
sm (n) = Ng n N 1
In all of the above, the effects of IQ imbalance on multiuser
k Mm
OFDM systems -also known as orthogonal frequency division
(1)
multiple access (OFDMA) systems- are not considered. In
where Ng + N is the total number of samples per symbol and represents the ICI for v = u or the MUI for v = u, and
Ng is the size of the CP. the last term represents the noise. The interference term can
be canceled by jointly detecting xu (k ) and x (k ). In such a
The IQ imbalance distorts the ideal Tx signal as fol- v
case, the receiver needs to estimate Hm for m = 1, . . ., K .
lows [10],
The pilot subcarriers Np are divided among users/subchannels.
sm = m sm + m sm, (2)
In an IQ distortion free OFDMA system, the system es-
timates the channel response at Nused subcarriers using Np
where is the complex conjugate, and m and m are
pilots. However, for a system suffering from IQ distortion,
related to the IQ imbalance as follows,
every user m transmitting over subcarriers k Mm causes
m = cos m + j sin m, (3a)
m
interference at k as well, as shown in (7). This means that in
/
m = m cos m j sin m, (3b) worst case scenario, where k Mm k, the user effectively
transmits over twice the number of assigned subcarriers. Re-
where m and m are the gain imbalance and the phase
ceiver then needs to estimate two user channels per subcarrier
mismatch, respectively. For the remainder of the paper, m
or a total of 2Nused channel responses using the same number
and m are referred to as the IQ parameters. At the receiver,
of pilots. In addition, estimation of K sets of IQ parameters
the sum of all transmitted user signals is received,
{ m, m } is needed to fully cancel the interference introduced
K
by image carriers as shown in (8).
y (n) = rm (n) + w(n), (4)
III. C HANNEL /IQ E QUALIZATION
m=1
where K is the total number of users in the system at the In this section, equalization of the channel effects and IQ
current symbol, w(n) is thermal noise, and rm is the received distortions in the received signal is discussed. First, the UL
signal of the mth user. frame structure needs to be introduced. In OFDMA systems,
(5) used subcarriers are divided into smaller units, also known
rm (n) = sm (l)hm (n l),
as slots or tiles, which represent the minimum allocation
l
unit (see Fig. 1). The tiles are then assigned to subchannels
where hm (n) is the sampled channel impulse response (CIR) either randomly or depending on user channels. The overall
of the mth user at time t = nTs and Ts is the sampling time of received UL frame is a sum of all current user signals,
received signal. In order to eliminate any interference between which are transmitted over different communication channels.
adjacent OFDMA symbols, i.e. inter-symbol interference (ISI), The use of sophisticated channel estimation algorithms at the
the CP size, Ng is chosen such that Ng l 1 for all user receiver can signi cantly increase the system complexity and
channels. At the receiver, the CP is removed and the signal processing delay. Therefore, pilots are assigned within each
is fed to an N -point fast Fourier transform (FFT) block. The tile. The receiver can then perform channel estimation on
output of the FFT can be expressed as, each tile individually using low complexity channel estimation
N 1 techniques (e.g. linear interpolation). The drawback of this
y (n)e j 2 nk/N,
Y (k ) = 0 k N 1, (6) method is the low spectral ef ciency since more pilots are
n=0 needed to perform suf ciently accurate channel estimation.
where k is the subcarrier index. If received signal is ISI free, An example of how the UL frame is divided into tiles is
shown in Fig. 1. Used subcarriers are divided into Q tiles,
then received subcarriers of user u are,
with equal number of tiles on each side of the DC subcarrier.
Y (k ) = u Hu (k )xu (k ) + v Hv (k )x (k ) + (k ), k Mu
v This means that for every tile q, there is an image tile q =
(7)
Q q + 1. Three different tile structures are shown in the
where (k ) is thermal noise modeled as additive white Gaus- gure, where one tile dimensions are three symbols by four
sian noise (AWGN) with power spectral density (PSD) of N0, subcarriers. The tile bandwidth, Btile = 3 f, where f is the
k is the mirrored subcarrier of k over the DC subcarrier, subcarrier spacing, and the tile duration, Ttile = 3Tsym, where
v {1, . . ., K } is the index of user which subcarrier k Tsym = (N + Ng )Ts is the symbol duration. For the remainder
is assigned to, and Hm is the mth user channel frequency of this paper, Tile C structure is used. Note that a similar
response (CFR) (i.e. Hm is the N -point Fourier transform frame structure is employed by the mobile WiMAX standard.
of hm ). If Hu is perfectly known to the receiver, then the In a system with no IQ distortion, the receiver would
equalized signal of user u is, consider each tile independently. The four pilots within a
tile would be used to obtain least squares (LS) estimates.
Hv (k ) (k )
Y (k ) = u xu (k ) + v xv (k ) +, k Mu . The LS estimates are then used to estimate the channel at
Hu (k ) Hu (k )
the remaining data subcarriers using 2-D linear interpolation.
desired signal
interference If the received signal is distorted by the IQ imbalance, the
(8) receiver processes each tile q along with its image tile q .
It can be seen from the above equation that the rst term is Simpli ed block diagrams of both a conventional OFDMA
the desired signal with an attenuation u, the second term receiver and an IQ distorted OFDMA receiver are depicted in
k=1 q=1
1-tile channel
Subcarriers-to-tiles Mapper
estimation &
Parallel-to-Serial &
equalization
Serial-to-Parallel
Permutation
FFT
RF front-end
1-tile channel
estimation &
N Q
equalization
(a)
k=1 q=1 2-tile
channel+distortion
Subcarriers-to-tiles Mapper
estimation &
Q
Parallel-to-Serial &
equalization
Serial-to-Parallel
Permutation
FFT
RF front-end
Q/2
2-tile
channel+distortion
Q/2+1
N estimation &
equalization
(b)
Fig. 2. Receiver block diagram of (a) conventional OFDMA-UL system, and (b) An OFDMA-UL system for signals with IQ distortion.
Fig. 2. Without the loss of generality, we assume that tile q is in (10) is not suf cient to estimate the channel/IQ parameters
assigned to user u and tile q is assigned to user v = u. The
matrix Zi,j . Two methods are proposed to solve this problem.
received subcarriers of tile q and q are,
The rst proposed method, Method A, assumes that the tile
bandwidth, Btile is less than the channel coherence bandwidth,
yi,j = u Hu,i,j xi,j + v Hv,i,j x, yi,j q, (9a)
i,j
BC, for all users. In this case, the CFR over the tile band is
u Hu,i, x,
yi, = v Hv,i, xi, + yi, q,
(9b) almost constant (i.e. Hm,i,j Hm,i,j +3 for m {1, . . ., K }).
j j j j i,j j
Based on the above assumption, the second pair of pilots at
where the subscript {i, j } represents the ith OFDMA symbol
symbol i and subcarriers {j + 3, 3} can be represented as,
j
and j th subcarrier as shown in Fig. 3. Using (9a) and (9b),
the received signals as function of transmitted signals, users yi,j +3 u Hu,i,j +3 v Hv,i,j +3 xi,j +3
=
CFR, and users IQ parameters is expressed as,
x 3
yi, 3 u Hu,i, 3 v Hv,i, 3
j j j i,j
yi,j u Hu,i,j v Hv,i,j xi,j
xi,j +3
= (10)
x
yi, u Hu,i, v Hv,i, Zi,j (12)
x 3
j j j i,j
i,j
Zi,j
Using (10) and (12), the system can solve for Zi,j . The same
To jointly detect transmitted subcarrier xi,j and its image xi,, procedure is then repeated for the two pilot pairs at symbol
j
the receiver needs to estimate 8 unknowns (4 CFR samples and i + 2 to obtain Zi+2,j . To insure that this set of equations
4 IQ parameters), if the estimation of the channel responses has a unique solution, pilot vector pairs are chosen to be
and IQ imbalances are done separately. Alternatively, the orthogonal [17], [18] such that,
receiver can estimate only the 4 unknowns of Zi,j, if the
xi,j +3
effects of the channels and IQ imbalances are estimated jointly. x
xi,j = 0, (13a)
x 3
i,j
However, in this case, the receiver does not have an estimate of i,j
the IQ distortion introduced by each user. The LS estimation xi+2,j +3
x +2,
xi+2,j = 0. (13b)
of transmitted signal is given by, x +2, 3
i j
i j
xi,j yi,j
= Z 1 (11) Next, Zi+1,j can be estimated using 1-D interpolation between
x i,j y
i,
i,j j
Zi,j and Zi+2,j . Finally, using (11), {Zi,j, Zi+1,j, Zi+2,j } are
IV. C HANNEL /IQ E STIMATION used to estimate the remaining data subcarriers at symbols
Let s assume that {xi,j, xi,, xi,j +3, xi, 3 } are pilots {i, i + 1, i + 2}, in this order. The receiver repeats this process
j j
for each pair of tiles (Q/2 times).
as shown in Fig. 3. Knowing the value of the two pilots
TABLE I
S IMULATION PARAMETERS .
$
Parameter Value
2.5 GHz
Carrier frequency
FFT size, N 1024
CP size, Ng 128
I
Number of used subcarriers, Nused 840
Sampling frequency, 1/Ts 11.2 MHz
Number of tiles, Q 210
Number of subchannels, M 35
G
each pair of tiles. The pilots orthogonality is achieved as
' F & #
follows,
xi+2,j
G
xi,j xi, = 0, (15a)
x +2,
j
i j
xi+2,j +3
xi,j +3 xi, 3 = 0. (15b)
x +2, 3
j
i j
H
To enable the receiver to use either methods, we propose to
use the following pilot pattern, which satis es the conditions
in (13a), (13b), (15a), and (15b).
d
"
xi,j xi,j +3 d1
! #
2
=,
@9
(16a)
4
d
xi+2,j xi+2,j +3 d1
7
2
8
ED 7
d
xi, 3 xi, d2
BC
A6
j j 1
=,
(16b)
d
xi+2, 3 xi+2, d2
j j
1
5
4
where d1 and d2 are arbitrary symbols that belong to the
( $
modulation symbol set used for pilots. The above pilot pat-
)
3210
tern achieves orthogonality between pilot pairs in both the
subcarrier dimension and the symbol dimension. Thus, the
Fig. 1. UL frame structure and subcarrier mapping to tiles.
receiver can improve the system performance by adaptively
using either Method A or Method B depending on the user
symbols
channel conditions (i.e. delay spread and Doppler spread) or
Tile q
Tile q
by using both methods and choosing the estimates with the
i+2
least error vector magnitude (EVM) values.
i+1
V. S IMULATION R ESULTS
i
subcarriers
A system based on the WiMAX standard [5] is considered.
DC j-3 j-2 j-1 j
j+1 j+2 j+3
j
Typical simulation parameters are chosen based on [19] and
are summarized in Table I. We assume three users are transmit-
Fig. 3. Subcarrier indexing of a tile pair.
ting in all UL frames. The available 35 subchannels are divided
among users with 12, 12, and 11 subchannels assigned to
user 1, user 2, and user 3, respectively. Subchannels assigned
The second proposed method, Method B, assumes that the
tile duration, Ttile is less than the coherence time of the to UL users are randomized for every frame. The modulation
channel, TC, for all users. In this case, Hm,i,j Hm,i+2,j for used for data subcarriers is QPSK, and for pilots, BPSK is
used (d1 = d2 = 1). The system is tested over three channels
m {1, . . ., K }. Based on the above assumption, the second
pair of pilots (to be coupled with (10)) at symbols {i, i + 2} based on the ITU channel models [20]: Indoor A, Pedestrian B,
and subcarriers {j, } can be represented as, and Vehicular A, with Doppler spread of 0 Hz (0 km/h), 6.9 Hz
j
(3 km/h), and 138.9 Hz (60 km/h), respectively. It is assumed
yi+2,j xi+2,j
Zi,j that the channel is independent between different frames and
(14)
yi+2, xi+2,
j j between different users. The receiver processes each frame
Using (10) and (14), the system solves for Zi,j . Similarly, individually. At the receiver, the uncoded bit error rate (BER)
the pilot pairs at subcarriers {j + 3, 3} can be used to
j is measured and averaged over all users.
solve for Zi,j +3 . Next, the receiver uses 1-D interpolation For all generated simulations, it is rst assumed that there
to calculate {Zi,j +1, Zi,j +2 } from {Zi,j, Zi,j +3 }. Finally, the are no IQ imbalances introduced to the signal. The receiver
above values of Z are used to estimate the remaining data uses all pilots per tile to estimate and equalize only the channel
subcarriers in both tiles. The procedure is then repeated for effects using 2-D linear interpolation technique. This case is
labeled as Ideal IQ . Next, the UL user signals are distorted 0
10
by IQ imbalances. The IQ gain imbalance and phase mismatch IQ Imbalance / No compensation
IQ Imbalance / Method A
of user 1, user 2, and user 3 are {0.2, 2 }, {0.35, 5 }, and IQ Imbalance / Method B
{0.5, 8 }, respectively. The uncoded BER is measured while 1
Ideal IQ
10
the receiver operates assuming no IQ imbalance is present.
The results are labeled as IQ Imbalance/ No compensation . 2
10
Finally, Method A and Method B are used to estimate and
BER
equalize the combined channel and IQ distortions. The results
are shown in Fig. 4, Fig. 5, and Fig. 6. In all gures, the 3
10
system performance degrades signi cantly when the receiver
ignores the IQ distortions to the signal. At a signal-to-noise 4
10
ratio (SNR) of 35 dB, the IQ distortions increase the BER
by two orders of magnitude. On the other hand, proposed
methods manage to reduce the BER considerably. For the
**-**-**-**-**-** 40 45 50
results over the Pedestrian B channel in Fig. 4, the delay spread SNR
is high while the Doppler spread is low. As a result, Method
B outperforms Method A, since it assumes low Doppler Fig. 4. Average uncoded BER of QPSK signals received over Pedestrian B
spreading. Method B signi cantly reduces the IQ distortion channel and with IQ impairments.
effect on the signal. The loss is around 2 dB from the ideal
IQ case at BER = 10 3 . For the results over the Vehicular A 0
10
channel in Fig. 5, both delay spread and Doppler spread are IQ Imbalance / No compensation
high. In this case, Method A slightly outperforms Method B, IQ Imbalance / Method B
IQ Imbalance / Method A
since its approximation is more accurate. While both methods 1
Ideal IQ
10
A and B improve the BER signi cantly, the loss from the ideal
IQ case is considerable. Under such highly selective channels
2
10
and with IQ distortions, the system can choose to reduce signal
BER
bandwidth which leads to an improved performance of Method
A. Finally, for the results over the Indoor A channel in Fig. 6, 3
10
both delay spread and Doppler spread are low. As expected
in this case, both methods A and B perform very well. The
4
BER degradation due to the IQ distortions is recovered with 10
around 2.5 dB loss from the ideal IQ case at BER = 10 3 .
Note that there is a loss from the ideal IQ case even in
channels with low Doppler spread and delay spread. This **-**-**-**-**-** 40 45 50
SNR
loss is a result of the noise averaging effect. In ideal case,
the estimation of channel response, which is almost constant
Fig. 5. Average uncoded BER of QPSK signals received over Vehicular A
over slot bandwidth and/or slot duration, is averaged by the channel and with IQ impairments.
2D interpolation over twice the number of pilots compared
to proposed methods. A receiver capable of detecting the
presence of IQ distortions can avoid this loss by switching to R EFERENCES
conventional channel estimator if received signal is free from
[1] Supplement to IEEE standard for information technology telecom-
IQ distortions.
munications and information exchange between systems - local and
metropolitan area networks - speci c requirements. Part 11: wireless
LAN Medium Access Control (MAC) and Physical Layer (PHY) spec-
i cations: high-speed physical layer in the 5 GHz band, The Institute
VI. C ONCLUSION
of Electrical and Electronics Engineering, Inc. Std. IEEE 802.11a, Sep.
1999.
In this paper, a model for OFDMA-UL systems with IQ [2] IEEE standard for information technology- telecommunications and
information exchange between systems- local and metropolitan area
impairments is introduced. The received signal as a function of
networks- speci c requirements Part 11: wireless LAN medium access
transmitted signals, IQ parameters, and communication chan- control (MAC) and physical layer (PHY) speci cations, IEEE Std
nels is mathematically formulated. The signal model is used to 802.11g-2003 (Amendment to IEEE Std 802.11, 1999 Edn. (Reaff
2003) as amended by IEEE Stds 802.11a-1999, 802.11b-1999, 802.11b-
design a novel two-dimensional orthogonal pilot pattern. Two
1999/Cor 1-2001, and 802.11d-2001) Std., 2003.
low complexity methods are proposed to equalize and detect [3] U. H. Reimers, DVB-the family of international standards for digital
IQ distorted signals using designed pilot pattern. Proposed video broadcasting, Proceedings of the IEEE, vol. 94, no. 1, pp. 173
182, 2006.
methods were shown to improve the system performance
[4] P. Shelswell, The COFDM modulation system: the heart of digital au-
signi cantly over conventional detection methods for signals dio broadcasting, Electronics & Communication Engineering Journal,
suffering from IQ distortions. vol. 7, no. 3, pp. 127 136, 1995.
Phase I/Q Mismatch Compensation Using Orthogonal Pilot Sequences,
0
10
EUSIPCO 2006, Sep. 2006.
IQ Imbalance / No compensation
[18] R. Chrabieh and S. Soliman, IQ Imbalance Mitigation via Unbiased
IQ Imbalance / Method A
Training Sequences, Global Telecommunications Conference, 2007.
IQ Imbalance / Method B
1
Ideal IQ GLOBECOM 07. IEEE, pp. 4280 4285, Nov. 2007.
10
[19] W. Forum, WiMAX system evaluation methodology, Sep. 2007.
[20] Recommendation ITU-R M.1225, Guidelines for Evaluation of Radio
Transmission Technologies for IMT-2000, International Telecommuni-
2
10
cation Union (ITU), 1997.
BER
3
10
4
10
**-**-**-**-**-** 40 45 50
SNR
Fig. 6. Average uncoded BER of QPSK signals received over Indoor A
channel and with IQ impairments.
[5] IEEE Standard for Local and metropolitan area networks Part 16: Air
Interface for Fixed and Mobile Broadband Wireless Access Systems
Amendment 2: Physical and Medium Access Control Layers for Com-
bined Fixed and Mobile Operation in Licensed Bands and Corrigendum
1, IEEE Std 802.16e-2005 and IEEE Std 802.16-2004/Cor 1-2005
(Amendment and Corrigendum to IEEE Std 802.16-2004) Std., 2006.
[6] A. A. Abidi, Direct-conversion radio transceivers for digital communi-
cations, IEEE J. Solid-State Circuits, vol. 30, no. 12, pp. 1399 1410,
1995.
[7] B. Come, R. Ness, S. Donnay, L. Van der Perre, W. Eberle, P. Wambacq,
M. Engels, I. Bolsens, and H. IMEC, Impact of front-end non-idealities
on bit error rate performance of WLAN-OFDM transceivers, Radio
and Wireless Conference, 2000. RAWCON 2000. 2000 IEEE, pp. 91 94,
2000.
[8] A. R. Wright and P. A. Naylor, I/Q mismatch compensation in zero-
IF OFDM receivers with application to DAB, Acoustics, Speech,
and Signal Processing, 2003. Proceedings.(ICASSP 03). 2003 IEEE
International Conference on, vol. 2, 2003.
[9] A. Tarighat and A. H. Sayed, On the baseband compensation of IQ
imbalances in OFDM systems, Acoustics, Speech, and Signal Process-
ing, 2004. Proceedings.(ICASSP 04). IEEE International Conference on,
vol. 4, 2004.
[10] J. Tubbax, B. Come, L. Van der Perre, L. Deneire, S. Donnay, and
M. Engels, Compensation of IQ imbalance in OFDM systems, Com-
munications, 2003. ICC 03. IEEE International Conference on, vol. 5,
2003.
[11] H. Sha ee and S. Fouladifard, Calibration of IQ imbalance in OFDM
transceivers, Communications, 2003. ICC 03. IEEE International Con-
ference on, vol. 3, 2003.
[12] J. Lin and E. Tsui, Joint adaptive transmitter/receiver IQ imbalance
correction for OFDM systems, Personal, Indoor and Mobile Radio
Communications, 2004. PIMRC 2004. 15th IEEE International Sym-
posium on, vol. 2, 2004.
[13] A. Tarighat and A. H. Sayed, OFDM systems with both transmitter
and receiver IQ imbalances, Signal Processing Advances in Wireless
Communications, 2005 IEEE 6th Workshop on, pp. 735 739, 2005.
[14] T. C. W. Schenk, P. F. M. Smulders, and E. R. Fledderus, Estimation
and compensation of Tx and Rx IQ imbalance in OFDM-based MIMO
systems, Radio and Wireless Symposium, 2006 IEEE, pp. 215 218,
2006.
[15] M. Valkama, Y. Zou, and M. Renfors, On I/Q imbalance effects in
MIMO space-time coded transmission systems, Radio and Wireless
Symposium, 2006 IEEE, pp. 223 226, 2006.
[16] Y. Jin, J. Kwon, Y. Lee, D. Lee, and J. Ahn, Obtained Diversity
Gain in OFDM Systems under the In uence of IQ Imbalance, IEICE
Transactions on Communications, vol. 91, no. 3, p. 814, 2008.
[17] L. Giugno, V. Lottici, and M. Luise, Low-Complexity Gain and
.dvi