Вы находитесь на странице: 1из 4

M2B.

41

Joint Compensation of IQ imbalance and Frequency Offset in OFDM systems.


Jan Tubbax1† , Boris Côme1 , Liesbet Van der Perre1 , Stéphane Donnay1, Marc Engels1∗ , Marc Moonen2, Hugo De Man1∗
Email: tubbax@imec.be - Phone: +32-16/28.82.50 - Fax: +32-16/28.15.15
1
IMEC, Leuven, Belgium - 2 KUL, Leuven, Belgium

Abstract— Zero-IF receivers are gaining interest because of their poten- given by [11] as
tial to enable low-cost OFDM terminals. However, Zero-IF receivers in-
troduce IQ imbalance which may have a huge impact on the performance. riq = (1 + ) cos ∆φ<{r} − (1 + ) sin ∆φ={r}
Rather than increasing design time or component cost to decrease the IQ
imbalance, an alternative is to tolerate the IQ imbalance and compensate + [(1 − ) cos ∆φ={r} − (1 − ) sin ∆φ<{r}] (1)
it digitally. Current solutions require extra analog hardware or converge = (cos ∆φ +  sin ∆φ) · r + ( cos ∆φ −  sin ∆φ) · r∗
too slowly for bursty communication. Moreover, the tremendous impact of
a frequency offset on the IQ estimation/compensation problem is not con- = α · r + β · r∗ (2)
sidered. In this paper, we analyze the joint IQ imbalance-frequency offset
estimation and propose a low-cost, highly effective, all-digital compensation with riq the time domain signal with IQ imbalance, <() denotes
scheme. For large IQ imbalance ( = 10%, ∆φ = 10o ) and large frequency the real part, =() the imaginary part and ()∗ the complex conju-
offsets, our solution results in an average implementation loss below 0.5 dB.
It therefore enables the design of low-cost, low-complexity OFDM receivers. gate and
α = cos ∆φ +  sin ∆φ (3)
β =  cos ∆φ −  sin ∆φ (4)
I. I NTRODUCTION
Frequency domain signals are underscored, while time domain
OFDM is a widely recognized and standardized modulation
signals are not. Signals are indicated in bold and scalar parame-
technique [1], [2]. Unfortunately, OFDM is also sensitive to
ters in normal font.
non-idealities in the receiver front-end [3]. This sensitivity leads
Throughout the rest of the paper, the term IQ parameters
either to heavy front-end specifications and thus an expensive
refers to α and β for calculations and estimations ; to indicate
front-end or to large performance degradations. IQ imbalance
physical parameters, however, we use the more direct  and ∆φ.
has been identified as a key front-end effect for OFDM systems.
We also analyze the effect of the IQ imbalance in the fre-
However, OFDM is also very sensitive to frequency off-
quency domain. If d is the incoming OFDM symbol (thus
sets [4]. Therefore, we investigate the joint IQ imbalance and
d = dt · c + n with dt the transmitted OFDM symbol, c the
frequency offset estimation and compensation and introduce a
channel and n the noise and · an element-wise multiplication),
low-complexity compensation scheme to combat the IQ imbal-
then IF F T (d) is the corresponding time domain signal. Apply-
ance and Carrier Frequency Offset (CFO). To the best of our
ing the IQ imbalance (2) and taking the F F T to convert back to
knowledge, no such solution has been proposed so far. [5], [6]
the frequency domain leads to
propose a calibration solution which requires extra, carefully de-
signed, analog hardware. [7] investigates only the impact of IQ diq = F F T {α · IF F T (d) + β · [IF F T (d)]∗ }
imbalance on the CFO estimation. [8], [9] offer a IQ compen- = α · d + β · d∗m (5)
sation with slow convergence and ignore the impact of the fre-
quency offset. In [10] we proposed an IQ compensation with a where diq is the OFDM symbol with IQ imbalance and dm the
fast convergence, but it only handles small frequency offsets. incoming OFDM symbol, mirrored over the carriers: (d)m (i) =
In this paper, we propose an all-digital, low-complexity IQ (d)(mod(Nsc − i + 2, Nsc )), with Nsc the number of sub-
compensation which converges within 1 OFDM training symbol carriers in the OFDM symbol, 1 ≤ i ≤ N and mod the modulo
and can handle large frequency offsets. operation. Carrier 1 is the DC carrier.
In this paper we focus on IQ compensation for bursty com-
II. IQ IMBALANCE munication, for which channel estimation is performed on the
basis of a known training symbol. Both IEEE802.11a [1] and
This section is a brief summary of [10] to introduce the IQ
HIPERLAN-II [2] provide such a Long Training Symbol (LTS)
model and the frame work to compare other algorithms. More
(BPSK symbol t). The effect of IQ imbalance on channel esti-
details can be found in that paper.
mation can be calculated based on (5) (and ignoring the noise for
the moment ; its effect is taken into account for the performance
A. Effect/Model
analysis)
IQ imbalance can be characterized by 2 parameters: the am-
plitude imbalance  between the I and Q branch, and the phase
orthogonality mismatch ∆φ. The complex baseband equation h = t · [α(t · c) + β(t · c)∗m ]
for the IQ imbalance effect on the ideal time domain signal r is = α · c + β · t0 · c∗m (6)
† Jan Tubbax is also a Ph.D Student at the KULeuven. where h is the channel estimate calculated from the LTS, c is
∗ is also a professor at the KULeuven. the exact channel vector and t0 = t · tm .
1.5 analog digital
^

h c
TS IQ
IQ FFT LTS
*
est
()m α,β
1
time
|H|

data IQ -IQ FFT EQ

0.5
Fig. 2. Overview of the IQ estimation/compensation.

channel without IQ
channel with IQ
channel with IQ corrected the data. As we already have an estimate of the IQ parameters
0
0 10 20 30
carrier
40 50 60 70
before the data symbols arrive we can compensate the effect in
the time domain, i.e. before the FFT: this allows us to correct
Fig. 1. The effect of IQ imbalance and correction on channel estimation. the IQ imbalance as early as possible. The setup is shown in
figure 2.
B. Estimation/Compensation To obtain the corrected signal r̂ from the observed time do-
main signal riq , we need to solve (2) for r, i.e.
From (6) the corrected channel response is
α∗ · riq − β · r∗iq
α∗ h − βt0 · h∗m r̂ = (11)
ĉ = (7) |α|2 − |β|2
|α|2 − |β|2
Since this algorithm is based on a frequency-domain estima-
The estimation of α and β is based on the information that the
tion of the IQ parameters, throughout the rest of the paper we
corrected channel response should have a smooth channel char-
refer to it as the ’IQ-Frequency Domain’ or IQ-FD algorithm.
acteristic: since the coherence bandwidth of the channel is (a
lot) larger than the inter-carrier-spacing in a WLAN system, the C. Performance
channel response does not change substantially between succes-
sive frequency taps (the x-line in figure 1). With IQ imbalance, To test the performance of the estimation/compensation
sharp transitions occur in the measured channel response h due scheme, we ran simulations for coded 64QAM in a multi-path
to the β degradation term (the o-line in figure 1). Thus, cor- environment. The multi-path channel are obtained through ray-
recting the IQ imbalance means making the channel response tracing simulations [12]. The impact of severe IQ imbalance
’smooth’ again. Therefore, we select the set of IQ parameters ( = 10%, ∆φ = 10o ) on coded (R = 3/4 from the IEEE or
(α, β) which renders the corrected channel response ĉ as smooth HIPERLAN standard) 64QAM transmission is quite dramatic:
as possible ; in other words we minimize the Mean Square Error it renders the system useless, causing a error floor of 5.10−1 .
(MSE) between consecutive channel coefficients Our combined channel and IQ estimation/compensation scheme
X gets the degradation at a BER of 10−5 down to 0.25 dB (fig-
M SE = |ĉl+1 − ĉl |2 (8) ure 4). The compensation scheme with the exact knowledge of
l the IQ parameters fully eliminates the IQ imbalance.

The derivation can be found in [10]. The results are III. IQ C OMPENSATION IN THE PRESENCE OF CFO
0 0 A. Effect/Model
P
l (hP
l+1 − hl )(t −l−1 h−l−1 − t −l h−l )
β̂ = 0 0 2
(9)
l |t −l−1 h−l−1 − t −l h−l | In the time domain a Carrier Frequency Offset (CFO) causes
q
<{β̂}={β̂} an additional phase rotation proportional with time t on an in-
α̂ = 1 − =2 {β̂} −  q (10) coming signal r
1 − =2 {β̂} rcfo = r · e2πCF O·t (12)

Figure 1 shows that we can correct the influence of the IQ In the frequency domain, a frequency offset causes leakage [13].
imbalance on the channel estimate extremely well: the corrected This means an OFDM symbol d is received under a frequency
channel response (the ∆-line) coincides (almost) perfectly with offset as
NX
sc −1
the exact channel response (the x-line). Note that carriers 28 to
(dcfo )i = dj · γi−j (13)
38 are zero carriers, which means no channel estimate is needed
j=0
on those carriers.
This algorithm provides us with a corrected channel response where γ is the inter-carrier-interference caused by the frequency
and an estimate of the IQ parameters α and β (or equivalently offset CFO
 and ∆φ). Since  and ∆φ and thus α and β are typically
CF O Nsc −1 sin π( CF O
∆f − n)
static over many symbols, we can use their estimates from the γn = eπ( ∆f −n) Nsc (14)
channel correction also for the correction of the IQ imbalance on sin Nπsc ( CF O
∆f − n)
analog digital
with ∆f is the inter-carrier-spacing and Nsc the total number of r1 h1 ^
-CFO FFT LTS c
carriers in the system. TS
CFO IQ CFO IQ
est est
Mathematical analysis of the RF signals shows that the effect +CFO
r2
()* FFT LTS
h2
α,β
time
of the IQ imbalance and CFO can be modeled at baseband by ^
riq r
first applying the frequency offset CF O, followed by all IQ im- data CFO IQ -IQ -CFO FFT EQ

balance contributions (from RF, LO and baseband) combined in


(α,β). The resulting signal rcfo−iq is thus Fig. 3. Overview of the joint IQ-CFO estimation/compensation.

rcfo−iq = α · r · e2πCF O·t + β · r∗ · e−2πCF O·t (15)


based on h1 = f f t(r1 ) · t and h2 = f f t(r∗2 ) · t (as indicated
where r is the input signal. in figure 3) as follows
B. Estimation/Compensation α∗ · h1 − β · h2
ĉ = (19)
(13) shows that if no CFO is present, the symbol transmitted |α|2 − |β|2
on carrier i will be received at carrier i. However, in the presence
(19) provides the corrected channel response if we know (α,β).
of CFO, each carrier is spread out over all other carriers accord-
ing to (13)-(14), the so-called inter-carrier-interference or leak- As in the previous section, we put the constraint that the cor-
rected channel response is as smooth as possible. In other words,
age. Because of this leakage, the assumptions used to derive
we minimize the MSE from (8).
existing IQ estimation/compensation algorithms are no longer
valid and thus their performance will degrade. For example, the This yields
degradation for IQ-FD algorithm at a BER of 10−5 for coded P ∗ ∗
l (h1,l+1 − h1,l )(h2,l+1 − h2,l )
64QAM already exceeds 1 dB at a CFO of 20 kHz and exceeds β̂ = P 2
(20)
5 dB at 40 kHz. In practice, the CFO can easily exceed 40 kHz. l |h2,l+1 − h2,l |
This clearly illustrates the need for an algorithm to take both IQ <{β̂}={β̂}
q
imbalance and frequency offset into account. α̂ = 1 − =2 {β̂} −  q (21)
We derive an IQ compensation scheme according to the same 1 − =2 {β̂}
principles as for the IQ-FD algorithm (as in [10]), but now tak-
ing the frequency offset into account. Once (α,β) are known, (11) can be used to correct the IQ
We start from the time domain signal (15) and assume we can imbalance on the received data signal, followed by a frequency
accurately estimate the frequency offset ; later we comment on offset correction.
this assumption. If we compensate for the frequency offset we The resulting block diagram is depicted in figure 3.
obtain If no frequency offset is present, then the joint IQ-CFO com-
r1 = α · r + β · r∗ · e−22πCF O·t (16) pensation (figure 3) reduces to the IQ compensation (figure 2)
after rearranging some blocks and (20-21-19) reduce to (9-10-
We observe two contributions in the resulting signal: the first 7).
one contains a scaled version of the desired signal ; the second Since the estimation is done in the frequency domain, it can
contribution contains a scaled version of the complex conjugate be easily redesigned to incorporate frequency-dependent IQ. In
of the signal, rotated over −2 · CF O. In the frequency domain, that case also the compensation is best handled in the frequency
the former gives rise to the desired signal, while the latter yields domain through a correction similar to (19).
a mirror image (because of the complex conjugate), subject to
leakage (caused by the −2 · CF O). C. Performance
On the other hand, if we compensate (15) by +CF O, we
We illustrate the performance of this joint IQ-CFO compen-
obtain
sation scheme for a WLAN case study with coded (R = 3/4
r2 = α · r · e22πCF O·t + β · r∗ from the IEEE or HIPERLAN standard) 64QAM transmission.
In the absence of CFO, this scheme is equivalent to the IQ-
⇒ r∗2 = β ∗ · r + α∗ · r∗ · e−22πCF O·t (17)
FD algorithm and thus has the same performance: for large IQ
As in (16), we see the same two contributions: the desired signal imbalance ( = 10%, ∆φ = 10o ) it reduces the degradation at a
and the complex conjugate of the signal rotated over −2 · CF O, BER of 10−5 down to 0.25 dB, as shown in figure 4 and 5.
but with different weighting factors. Considering (16) and (17) Figure 5 illustrates the IQ compensation works for large fre-
a set of equations in the variables r and r∗ · e−22πCF O·t and quency offsets as well: the degradation doesn’t exceed 0.5 dB
solving for r leads to at a BER of 10−5 coded 64QAM even for frequency offsets ex-
ceeding 100 kHz.
α∗ · r1 − β · r∗2 The CFO is estimated by means of the average phase rotation
r̂ = (18)
|α|2 − |β|2 between the 2 Long Training Symbols

(18) yields the corrected time domain signal based on r1 and r2 . angle(E[x2 · x∗1 ])
Equivalently, we write the equation for the channel correction CFˆ O = (22)
2πNsc Ts
0
10 IV. C ONCLUSIONS

−1
In this paper, we introduced a low-complexity estima-
10
tion/compensation scheme that tackles the IQ imbalance caused
by Zero-IF receivers. The scheme converges on one OFDM
10
−2
training symbol. It removes large IQ imbalance ( =
10%, ∆φ = 10o ) and reduces the average remaining degra-
BER

10
−3
dation down to 0.5 dB, even in the presence of large fre-
quency offsets. This allows to relax the IQ mismatch specifi-
10
−4 cations significantly. In turn, this allows to use cheaper com-
Reference ponents and to shorten the design time. Therefore, our esti-
No IQ compensation
−5 IQ compensation mation/compensation scheme enables low-cost, low-complexity
10
14 16 18 20 22 24
SNR (dB)
26 28 30 32 34 OFDM receivers.

Fig. 4. The performance of the IQ compensation scheme for coded (R=3/4) R EFERENCES
64QAM and ( = 10%,∆φ = 10o ) in the absence of a frequency offset. [1] IEEE standard 802.11a-1999 - part 11: wireless LAN medium access con-
trol (MAC) and physical layer (PHY) specifications: high-speed physical
layer in the 5 GHz band. 1999.
[2] HIPERLAN type 2 standard - functional specification data link control
1
(DLC) layer. October 1999.
0.9 [3] J. Tubbax, B. Côme, L. Van der Perre, L. Deneire, and M. Engels. OFDM
vs. single-carrier with cyclic prefix: a system-based comparison. In IEEE
0.8
VTC Fall, volume 2, pages 1115–1119, 2001.
[4] T. Pollet, M. Van Bladel, and M. Moeneclaey. BER sensitivity of OFDM
degradation @ 10−5 [dB]

0.7
systems to carrier frequency offset and wiener phase noise. IEEE Trans-
0.6
actions on Communications, 43(2/3/4):191–193, Feb./Mar./Apr. 1995.
0.5 [5] S. Simoens, M. de Courville, F. Bourzeix, and P. de Champs. New I/Q
imbalance modeling and compensation in OFDM systems with frequency
0.4
offset. PIMRC.
0.3 [6] I. Bouras, S. Bouras, T. Georgantas, N. Haralabidis, G. Kamoulakos,
C. Kapnistis, S. Kavadias, Y. Kokolakis, P. Merakos, J. Rudell,
0.2
S. Plevridis, I. Vassiliou, K. Vavelidis, and A. Yamanaka. A digitally cali-
0.1 brated 5.15-5.825 GHz transceiver for 802.11a wireless LANs in 0.18µm
CMOS. IEEE Internat. Solid-State Circuits Conf., pages 352–353, 2003.
0
0 20 40 60 80 100 120 140 160 [7] Vincent K.-P. Ma and Tommi Ylamurto. Analysis of IQ imbalance on ini-
CFO [kHz]
tial frequency offset estimation in direct down-conversion receivers. 2001.
[8] Andreas Schuchert, Ralph Hasholzner, and Patrick Antoine. A novel IQ
Fig. 5. Remaining degradation of the IQ compensation @ 10−5 coded 64QAM imbalance compensation scheme for the reception of OFDM signals. IEEE
as a function of the frequency offset. Transactions on Consumer Electronics, 47(3):313–318, August 2001.
[9] M. Valkama, M. Renfors, and V. Koivunen. Advanced methods for I/Q
imbalance compensation in communication receivers. IEEE Transactions
on Signal Processing, 49(10):2335–2344, October 2001.
[10] J. Tubbax, B. Côme, L. Van der Perre, S. Donnay, and M. Engels. IQ im-
with x1 and x2 the received training sequences, Nsc Ts the time balance compensation for OFDM systems. accepted at IEEE International
between 2 corresponding samples of the training sequences. Conference on Communications ICC, 2003.
[11] Behzad Razavi. RF Microelectronics. Prentice Hall, 1998.
This CFO estimation performs well in the presence of IQ im- [12] Patrick Vandenameele. Space Division Multiple Access for Wireless Lo-
cal Area Networks. PhD thesis, KULeuven in co-operation with IMEC,
balance: the simulation results from figure 5 are based on the October 2000.
CFO estimates and show the estimates are accurate enough to [13] Claus Muschallik. Influence of RF oscillators on an OFDM signal. IEEE
produce a small residual degradation in the IQ imbalance esti- Trans. on Consumer Electronics, 41(3):592–603, Aug. 1995.
[14] I. Koffman and V. Roman. Broadband wireless access solutions based
mation/compensation. on OFDM access in IEEE 802.16. IEEE Communications Magazine,
40(4):96–103, April 2002.
The IQ estimation/compensation algorithm requires no extra
analog hardware and a very small additional digital complexity.
The IQ imbalance may occur anywhere in the receiver, because
the RF, LO and baseband contributions are jointly estimated and
compensated.
Since the IQ estimation only requires a known Training Sym-
bol, this scheme is applicable to any system which uses a Train-
ing Symbol to estimate the channel (e.g. Wireless LAN [2],
[1] and Broadband Fixed Wireless Access [14]). Moreover, the
scheme is also independent of the data that follows the Train-
ing Symbol. This means it is applicable to any constellation size
and to OFDM as well as Single-Carrier with Frequency-Domain
Processing.

Вам также может понравиться