You are on page 1of 5

Partial Transmit Sequence-Active Interference Cancellation for UWB OFDM Transmission

Poramate

Tarasak, Zhiwei Lin, Xiaoming Peng and Francois Chin

Institute for Infocomm Research, Agency for Science Technology and Research (A*STAR) 1 Fusionopolis Way, #21-01 Connexis, Singapore, 138632 Email: ptarasak@i2r.a-star.edu.sg

AbstractA novel algorithm called partial transmit sequenceactive interference cancellation (PTS-AIC) is proposed to implement interference avoidance functionality in an ultra-wideband (UWB) OFDM device. It can be considered as another application of the PTS algorithm that has been applied to an OFDM transmission to reduce its peak-to-average power ratio. Specically, PTS-AIC with adjacent subblock partitioning and interleaved subblock partitioning are proposed and compared. It is shown that the PTS-AIC can provide deep notch on the interference band. The x%-excess interference power is dened as a performance measure for PTS-AIC. Compared to existing approaches, PTS-AIC provides more exibility in parametrizing the algorithm for trade-off between performance and complexity.

I. I NTRODUCTION UWB, a Gigabit transmission enabler, is a spectrum underlay system in which its spectrum is very wide and lies below other narrowband spectrums. It is expected that a UWB device must endure interference caused by narrowband devices but it cannot generate the interference back to them. In this respect, the UWB device is usually considered as a secondary device as opposed to the narrowband device which is a primary device. If the spectrum usage of both devices is not handled properly, one or both devices may experience severe performance degradation due to interference from the other. Evidently, several papers have shown the detrimental effect of interference crossing between UWB and other wireless systems [1][4]. To prevent such performance degradation, a UWB device must be capable of detecting narrowband spectrums and modifying its transmission in a smart way. The term detectand-avoid (DAA) is introduced to refer to this functionality [5]. The importance of DAA is not only limited in the UWB context but DAA is generally a function in cognitive radio devices operating under the environment of coexistence between two or more wireless systems [6][10]. Current important narrowband systems that share parts of the UWB spectrum are WiMAX at the 3.5GHz frequency range and IEEE802.11a at the 5GHz frequency range. In the future, it is expected that more emerging wireless technologies might share the same part of available spectrum with each other, e.g., body area networks, wireless regional area networks, vehicular adhoc network, etc. Therefore, DAA functionality will be an increasingly important component of a wireless device. This paper focuses on the avoidance part of DAA for

the UWB system that employs OFDM transmission. In the WiMedia standard [11], tone nulling at the overlapped narrowband spectrum is suggested as an avoidance technique. The problem of tone nulling is that sidelobe interference from the remaining tones is still present in the interference band [12]. One efcient technique to solve this is active interference cancellation (AIC) proposed in [12]. In addition to tone nulling the subcarriers that lie inside the interference band, AIC sacrices two more subcarriers beside them and replaces them with the computed AIC tones. The purpose of placing AIC tones is to generate negative interference in order to cancel the sidelobe interference from the remaining subcarriers. Many extensions and improvements of the AIC have been proposed recently. AIC has been applied to an antenna selection system in [13]. The problem considered is when antenna selection is done per subcarrier. In this case, AIC algorithm has to be applied to all antennas not just one antenna transmitting on the interference band since sidelobe interference comes from all antennas. The formulation of AIC in [13] is to minimize the remaining interference power with the constraints on total transmission power of the edge AIC tones from all antennas not exceeding the spectral mask. Few low complexity AIC algorithms have been proposed in [14]. The algorithm is based on simplifying the matrix computation in AIC. The complexity saving in [14] comes at a price of degraded interference cancellation performance. However, it is claimed that the algorithm is still better than the original AIC in the worst case, i.e., the power of the maximum peak in the interference band is still lower that that of the original AIC. The so-called extended AIC has been proposed in [15] to increase the notch depth of AIC. The idea is to insert the extended AIC tones between the usual AIC tones and they together generate better negative interference However, since the extended AIC tones are placed in between the usual subcarrier positions, they will interfere with the data subcarriers and cause error oor to the bit error rate performance. Further enhancement of AIC was also proposed in [16]. The enhancement is achieved by cyclic shifting, phase shifting, or joint cyclic and phase shifting the data subcarriers in order to minimize the remaining interference after performing AIC. The algorithms in [16] increase the cancellation performance of AIC signicantly with the drawbacks of high complexity and the requirement of side information.

AIC tone Information

Interference tones AIC tone Information

84 85 86 87 88 Interference Band

Fig. 1. Interference band, interference tones and AIC tones, assuming the primary-user band overlaps with the 85th-87th OFDM subcarriers [12]

This paper proposes a novel algorithm called partial transmit sequence-active interference cancellation (PTS-AIC). The proposed algorithm is motivated by the efcient PTS algorithm to reduce peak-to-average power (PAPR) of an OFDM system [17], [18]. The idea is to partition the data block into several subblocks. Each subblock is multiplied with a weighting factor chosen from a constellation set. Then, a data block is reconstructed and AIC is performed. The algorithm chooses the optimum set of weighting factors such that the remaining interference power is minimized. Similar to the approach in [16], there is a requirement on side information. Two types of subblock partitioning are considered. The rst one is adjacent subblock partitioning and the second one is interleaved subblock partitioning. It will be shown that PTSAIC is very effective and provides exible trade-off between interference cancellation performance and complexity. II. AIC We describe the AIC algorithm [12] below as it is a required block for PTS-AIC. The denitions of interference band, interference tones and AIC tones are shown in Fig. 1. It is assumed that the transmitter and the receiver know the locations of the interference band. This assumption is generally not critical since the overlapped narrowband spectrum is usually xed or changes only slowly. Suppose X(k), k = 1, . . . , N, represents original frequency-domain data symbols where N is the number of subcarriers (FFT size). The time-domain signal is represented by N 1 nk X(k)exp j2 . (1) x(n) = N
k=0

Let Y = [Y (0), . . . , Y (M N 1)]t , X = [X(0), . . . , X(N 1)]t , []t is a transpose. We can write 1 (4) Y = P X, N where P is an upsampling matrix of size M N N whose N 1 n l element P (l, k) = n=0 exp j2 N k M . Suppose the interference tones start from the pth to (p + Ni 1)th subcarriers where Ni is the number of interference tones before upsampling. Dene a nulling matrix T which is constructed from an N N identity matrix by putting zeros at the diagonal elements corresponding to the interference and AIC tones. Now the interference generated from the sidelobes of other subcarriers can be computed as dI = P s T X, (5)

where P s is a submatrix of P by taking its row corresponding to upsampled spectrum in the interference band, i.e., from the (M p + 1)th to (M (p + Ni 1) + 1)th rows of P . Let a new set of subcarriers to be placed at the positions of AIC and interference tones be a column vector h of length Ni + 2. It has to generate negative interference to cancel dI , i.e., P n h = dI , (6) where P n is a submatrix of P s by taking its column corresponding to the positions of AIC tones and interference tones, i.e., from the pth to (p + Ni + 1)th columns of P s . The size of P n is then (M (Ni 1) + 1) (Ni + 2). Since P n is not a square matrix, to solve (6), one constructs a least-squared problem which nds h that minimizes squared error [12], SE = P n h + dI
2

(7)

A well-known least-squared solution to (7) is the MoorePenrose generalized inverse (pseudoinverse) [19], in the form, h = (P t P n )1 P t dI = V T X, n n (8) where V = (P t P n )1 P t P s is of size (Ni + 2) N and n n can be pre-computed. Then, h is inserted at the interference and AIC tones before performing IFFT to construct an OFDM symbol for transmission. III. PTS-AIC Fig. 2 illustrates the transmission model for PTS-AIC. First, a symbol block of length N , X, is partitioned into Nb subblocks of equal size N/Nb . Each subblock X nb is multiplied by a weighting factor Wnb . The weighting factor in this paper is chosen from a usual unit-energy complex PSK constellation of size Nw , e.g., = {1, 1} when Nw = 2, = {1, j, 1, j} when Nw = 4, and so on. Then, a symbol block is reconstructed from each subblock as X before it is processed with the AIC algorithm. The algorithm determines an optimum Nb -tuple weighting factor W opt = {W1 , W2 , . . . , WNb } such that the remaining interference power inside the interference band after performing AIC is minimum, i.e., W opt =
W Nb

The AIC process starts by upsampling the corresponding spectrum by a factor of M . The upsampled frequency spectrum is given by Y (l) = 1 N
N 1

x(n)exp j2
n=0

n l NM

(2)

where l = 0, . . . , N M 1. Substituting (1) into (2), we have Y (l) = = 1 N 1 N


N 1 N 1

X(k)exp j2
n=0 k=0 N 1

n N

l M (3)

X(k)P (l, k).


k=0

min

P n V T X(W ) + P s T X(W )

(9)

X1

Block Reconstruction
X

N/Nb

Subblock partition

. . XN . b

X 2 W1

1 1 1 1
AIC IFFT

...
(a)

2 2 2 2

...

W2

index of optimal weights W


Fig. 2.

WN b

...
opt

1 2 3 4 1 2 3 4
(b)

...

as side information

PTS-AIC transmission model

where X(W ) is a reconstructed symbol block computed by an Nb -tuple weighting factor W , Nb is a set of Nb -tuple weighting factor where its elements are chosen from . The PTS-AIC algorithm is characterized by the parameters Nb and Nw . We can adjust both parameters such that the interference cancellation performance meets the target while the complexity is affordable. As either Nb or Nw increases, the performance will improve while the complexity will increase. The complexity of PTS-AIC is determined by the number of all possible weighting factors to nd W opt which is (Nw )Nb . Since the complexity grows exponentially with Nb , one cannot increase the number of subblocks to a very large value to improve the performance as the number of possible W will be prohibitively large. There are two types of subblock partitioning considered in this paper. The rst type is called adjacent subblock partitioning in which each subblock is constructed from adjecent subcarriers of the original symbol block. The second type is called interleaved subblock partitioning in which each subblock is constructed from the subcarriers of distance Nb in the original symbol block. Both types of partitioning are depicted in Fig. 3. It will be shown that these two types of partitioning have difference performance. Since PTS-AIC modies the transmission block by the weighting factors, the receiver must be aware which set of weighting factors is applied so that it can recover the original symbol block. This can be done by sending the index of the optimum weighting factor as side information. The number of bits required for this side information is Nb log2 (Nw ). Once the receiver has the index of the optimal weighting factors, multiplication of their complex conjugates to the received signal block after FFT yields the original data block. Note one main difference between PTS-AIC and conventional PTS for reduction of PAPR. While PTS-AIC measures the interference power of the upsampled frequency spectrum before IFFT, PTS for reduction of PAPR measures the PAPR of the (upsampled) time-domain signal after IFFT. A. Performance Measure In interference avoidance mechanism, we are interested in how much interference power remaining inside the interference band. For AIC and PTS-AIC, this is equivalent to the squared error terms after performing interference cancellation

Fig. 3. (a) Adjacent subblock partitioning (b) Interleaved subblock partitioning

as dened in (7). More generally, the instantaneous remaining signal power present inside the interference band can be found from
M (p+Ni 1) l=M p

I=

|Y (l)|2 .

(10)

This is the remaining power of the upsampled spectrum at the interference band after AIC or PTS-AIC. Although the mean of the remaining interference power is normally used to compare between algorithms as in [16], a more complete picture is captured by computing the complementary cumulative distribution function (CCDF), P (I > Is ), where Is is a target remaining interference power. This is in analogy with the performance measure of a PAPR reduction algorithm of OFDM for which the CCDF is widely used. With a given Is , the CCDF determines the probability of remaining interference power above the target Is . We refer to Is of which P (I > Is ) = x% as x%-excess interference power. IV. N UMERICAL R ESULTS AND D ISCUSSION Data symbols are randomly generated from a QPSK constellation with equally likely probability. The number of subcarriers, N , is 128. Frequency domain oversampling factor, M , is four. Among 128 subcarriers, the subcarrier indices from the 85th to 87th are assumed to lie in the interference band. This corresponds to 2 4.125MHz = 8.25MHz according to the WiMedia standard. After processing the interference cancellation, the remaining interference power is measured as in (10). To compute CCDF, at least 10,000 symbol block realizations are simulated for all cases. Fig. 4 shows an example of power spectrum of PTS-AIC with Nb = 16, Nw = 2 and adjacent subblock partitioning of one symbol realization. It is shown that deep notch is achieved in the interference band. Fig. 5 shows the effect of different number of subblocks on the PTS-AIC performance with adjacent subblock partitioning when Nw = 2. As the number of subblocks increases, the remaining interference power is reduced. For Nb = 4, the 1%excess interference power is slightly less than -30dB while for Nb = 8, 16 the 1%-excess interference powers are about -35dB and slightly higher than -50dB, respectively. For Nb = 16,

10 10
0

(3)

10 Power spectrum

10

(5)

(2)

P(I>I )

(1) 10
2

10

(4) (1) Nb = 4, Nw = 2

10

10

(2) Nb = 4, Nw = 4 (3) N = 4, N = 8
b b b w w w

(4) N = 8, N = 2 (5) N = 8, N = 4 10
8

20

40

60 subcarrier

80

100

120

10 80

70

60

50 I (dB)
s

40

30

20

Fig. 4. Power spectrum of PTS-AIC with adjacent subblock partitioning. Nb = 16, Nw = 2.

Fig. 6. CCDF of remaining interference power of PTS-AIC, adjacent subblock partitioning. Effect of different size of weighting factor set .

10

10 (1)

adjacent subblock partitioning 10


1

10 (3) (2)

P(I>I )

10

10

10

I (1) Nb = 4 (2) N = 8
b

10

10 80

(3) Nb = 16 70 60 50 I (dB)
s

interleaved subblock partitioning 40 30 20 10


8

20

40 60 Realization index

80

100

Fig. 5. CCDF of remaining interference power of PTS-AIC, adjacent subblock partitioning, Nw = 2. Effect of different number of subblocks, Nb .

Fig. 7. Remaining interference powers of adjacent and interleaved subblock partitionings measured for 100 realizations with parameters Nb = 16, Nw = 2.

there is a large performance improvement but the price to pay is much higher complexity in searching for the optimum weighting factors. The number of possible 16-tuple weighting factors with Nb = 16, Nw = 2 is 216 = 65536. Fig. 6 shows the effect of different sets of weighting factors compared at xed Nb for the adjacent PTS-AIC. For Nb = 4, there is not much gain arising from using a larger constellation of weighting factors whereas for Nb = 8, a larger performance gain is clearly seen. For Nb = 8, Nw = 4, the 1%excess interference power is about -40dB. This should be compared with the case of Nb = 16, Nw = 2 in Fig. 5 since both cases have the same number of possible weighting factors in the computation. The 1%-excess interference power of Nb = 8, Nw = 4 case is about 10dB higher than that of the Nb = 16, Nw = 2 case. From this observation, it is suggested that the incurred complexity of PTS-AIC should be invested in

increasing the number of subblocks Nb rather than increasing the size of the constellation of weighting factors. Fig. 7 compares the remaining interference power of PTSAIC between adjacent and interleaved subblock partitionings with Nb = 16, Nw = 2. Both types of partitioning are applied on the same symbol block for 100 realizations. It is shown that in most cases interleaved subblock partitioning outperforms adjacent subblock partitioning but not always. Fig. 8 compares the CCDF performance among our proposed PTS-AIC schemes and joint cyclic and phase shifting AIC. The joint cyclic and phase shifting AIC is the best algorithm presented in [16] with the same complexity as PTS-AIC with Nb = 16, Nw = 2 in terms of the number of comparisons required in optimization. It can be seen that interleaved subblock partitioning outperforms adjacent subblock partitioning signicantly at any CCDF values. The improvement is about

10

trade-off between interference cancellation performance and complexity.


(5) (3)

ACKNOWLEDGMENT The rst author would like to thank Weiqiang Zhang for some discussion on implementation of the algorithm.

10

(2) P(I>I ) (4) 10


2

R EFERENCES
[1] S. Niranjayan, A. Nallanathan and B. Kannan, Modeling of multiple access interference and BER derivation for TH and DS UWB multiple access systems, IEEE Trans. on Wireless Commun., vol. 5, pp. 27942804, Oct. 2006. [2] C. Snow, L. Lampe and R. Schober, Analysis of the Impact of WiMAXOFDM Interference on Multiband OFDM, in Proc. IEEE International Conference on Ultra-Wideband (ICUWB), pp. 761-766, Sept. 2007. [3] A. Nasri, R. Schober and L. Lampe, Analysis of narrowband communication systems impaired by MB-OFDM UWB interference, IEEE Trans. on Wireless Commun., vol. 6, pp. 4090-4100, Nov. 2007. [4] K. Shi, Y. Zhou, B. Kelleci, T.W. Fischer, E. Serpedin and A.I. Karsilayan, Impacts of narrowband interference on OFDM-UWB receivers: analysis and mitigation, IEEE Trans. on Sig. Proc., pp. 1118-1128, Mar. 2007. [5] S.M. Mishra, R.W. Brodersen, S. ten Brink and R. Mahadevappa, Detect and avoid: An ultra-wideband/WiMAX coexistence mechanism, IEEE Commun. Mag., pp. 68-75, June 2007. [6] K. Sodeyama, K. Watanabe, K. Ishibashi, R. Kohno, Performance analysis of spectrum management technique by using cognitive radio, in Proc. European Conf. on Wireless Technology, pp. 119-122, Oct. 2008. [7] C.-K. Yu and K.-C. Cheng Chen, Multiple systems sensing for cognitive radio networks over Rayleigh fading Channel, in Proc. IEEE Spring Vehicular Technology Conference, pp. 1574-1578, May. 2008. [8] A. Swami, B. Sadler and J. Turner, On the coexistence of ultra-wideband and narrowband radio systems, in Proc. IEEE Military Conference, pp. 16-19, Oct. 2001. [9] R. Giuliano and F. Mazzenga, On the coexistence of power-controlled ultrawide-band system with UMTS, GPS, DCS180, and xed wireless systems, IEEE Trans. on Veh. Tech., vol. 54, pp. 62-81, Jan. 2005. [10] A. Rahim, S. Zeisberg and A. Finger, Coexistence study between UWB and WiMax at 3.5 GHz band, in Proc. IEEE International Conference on Ultra-Wideband (ICUWB), pp. 915-920, Sept. 2007. [11] ECMA International, High rate ultra wideband PHY and MAC standard, ECMA-368, 1st Edition, Dec. 2005. [12] H. Yamaguchi, Active interference cancellation technique for MBOFDM cognitive radio, in Proc. 34th European Microwave Conference., pp. 1105-1108, Oct. 2004. [13] Y. Wang and J. Coon, Active interference cancellation for systems with antenna selection, in Proc. IEEE International Conference on Communications (ICC), pp. 3785-3789, May 2008. [14] S.-G. Huang and C.-H. Hwang, Low complexity active interference cancellation for OFDM cognitive radios, in Proc. IEEE Wireless Communications and Networking Conference (WCNC), pp. 1279-1283, 31 Mar.-3 Apr. 2008. [15] Z. Wang, D. Qu, T. Jiang and Y. He, Spectral Sculpting for OFDM based opportunistic spectrum access by extended active interference cancellation, in Proc. IEEE Global Telecommunications Conference (Globecom), pp. 1-5, 31 Nov.-4 Dec. 2008. [16] P. Tarasak, F. Chin, Z. Lin and X. Peng, Further enhancement for active interference cancellation on MB-OFDM UWB transmission, in Proc. IEEE Fall Vehicular Technology Conference, pp. 1-5, Sept. 2008. [17] S. H. Muller and J. B. Huber, OFDM with reduced peak-to-average power ratio by optimum combination of partial transmit sequences, IEE Electronics Letters, vol. 33, no. 5, pp. 368369, Feb. 1997. [18] L. J. Cimini Jr. and N. R. Sollenberger, Peak-to-average power ratio reduction of an OFDM signal using partial transmit sequences, IEEE Communication Letters, vol. 4, pp. 8688, Mar. 2000. [19] G.H.Golub and C.F.Van Loan, Matrix Computations, The John Hopkins University Press, 1983.

(1) N = 16, N = 2, Adjacent 10


3

(2) Nb = 16, Nw = 2, Interleaved (3) N = 8, N = 2, Adjacent


b w

(1)

(4) Nb = 8, Nw = 2, Interleaved 10 90
4

(5) Joint CycPhase AIC 80 70 60 I (dB)


s

50

40

30

Fig. 8. CCDF of remaining interference power, comparison among PTSAIC with adjacent and interleaved subblock partitionings and joint cyclic and phase shifting AIC TABLE I N UMBER OF SIDE INFORMATION BITS REQUIRED FOR PTS-AIC

Nb 4 4 4 8 8 16

Nw 2 4 8 2 4 2

Number of side information bits 4 8 12 8 16 16

10dB at 1%-excess interference power for Nb = 8 and more than 10dB for Nb = 16. Compared with joint cyclic and phase shifting AIC, PTS-AIC with interleaved subblock partitioning performs slightly better. Moreover, in practice, the PTS-AIC might be more attractive. This is because for PTS-AIC with Nw = 2, the required signal resolution entering the AIC block does not change from what is required to implement original data symbols (in this case QPSK), whereas for joint cyclic and phase shifting AIC, the required resolution of the signal could be as ne as 2/N radian [16]. The latter would be more sensitive to quantization noise and possibly more complicated architecture is needed in the implementation. Table I summarizes the number of side information bits required for PTS-AIC which also reects the complexity of PTS-AIC for each set of parameter Nb , Nw . V. C ONCLUSION Partial transmit sequence-active interference cancellation (PTS-AIC) is proposed for UWB OFDM transmission. It is shown that adjacent subblock partitioning outperforms interleaved subblock partitioning signicantly. Two parameters of PTS-AIC which are the number of subblocks and the size of constellation set for weighting factors can be adjusted to meet the interference power target with affordable complexity. Overall, PTS-AIC is very effective and provides exible

You might also like