WC33-1: A New UWB Dual Pulse Transmission ... - Semantic Scholar

A New UWB Dual Pulse Transmission and Detection Technique Xiaodai Dong Member, IEEE

Alfred C.Y. Lee

Lei Xiao Student Member, IEEE

Dept. of Electrical & Computer Engr. University of Victoria, Canada Email: [email protected]

Dept. of Electrical & Computer Engr. University of Alberta, Canada Email: [email protected]

Department of Electrical Engineering University of Notre Dame, USA Email: [email protected]

Abstract— A new UWB signaling technique based on dual pulse transmission is presented in this paper. The proposed technique is simple, robust, and based on an autocorrelation type of receiver structure. Several detection schemes with different implementation complexity are presented and their performances are studied by Monte-Carlo simulations. Comparison of the proposed new designs to the transmitted reference system is also carried out in the paper.

I. I NTRODUCTION The ultra-wideband (UWB) impulse radio has attracted enormous interests from both academia and industry recently. Typically, a UWB impulse radio communication system employs very narrow pulses for transmission and the extremely short duration of these pulses leads to high multipath resolution and fading immunity. Although the resolvable multipaths provide diversity that can be employed to enhance performance, the challenge for the receiver is how to efficiently capture the energy from all these multipaths. If a rake structure is used, a large number of rake fingers must be implemented, which is prohibited in practice due to high complexity and high cost. Moreover, a UWB channel may distort the shape of the transmitted pulse [1]. Due to the ultra wide bandwidth, distinct frequency components in a signal may react differently to propagation environments. A receiver filter matched to the transmitted pulse in coherent detection may not work well if the pulse shape is distorted by the channel. All these considerations motivate the design of the new transmission and detection scheme in this paper. The idea is rather simple: using a reference sub-pulse together with a modulated sub-pulse to constitute a dual pulse (DP) structure as the basic pulse transmission unit, as shown in Fig. 1. Since the two narrow sub-pulses are of the same shape and one after another, it is expected that the channel affects them in a similar manner. For each resolvable multipath, the first half portion of the received pulse can then be used as a noisy reference for detection and energy collection of the second half, i.e., the modulated sub-pulse. This autocorrelation receiver performs essentially non-coherent detection. Fig. 2 depicts the receiver block diagram. An improved dual-pulse transmission scheme and receiver structure to remove the “self interference” in the received dual pulses are illustrated in Figs. 3-4. Although the initial idea of the new scheme is focused on the pulse design perspective, it can be considered as a special form of the transmitted reference (TR) system. The UWB transmitted reference system was first proposed in [2] and [3], where a reference pulse and a modulated pulse seperated by

delay D seconds constitute a pulse pair to represent one bit information. The delay D is larger than the maximum delay spread of the channel plus one pulse duration to avoid the interference between the received reference pulse and the data pulse [2]. It was demonstrated that UWB-TR systems have simple implementation and robust performance. Performance analysis of UWB transmitted reference was first presented in [4], while optimal and suboptimal receivers were derived and analyzed in [5]. The authors in [6] presented a generalized optimal receiver structure that takes into account variable number of reference and data pulses. A different generalization of the TR technique was proposed in [7], where a signaling set is composed of sequences of pulses with different delays and weights. In [8], the authors studied pilot waveform assisted modulation scheme which can be considered as another generalization of the TR method. In [9], the performance of multiple pulse multiple delay modulation for UWB multiple access was investigated. Also, a differential UWB scheme was proposed in [10]. The TR method in general has several advantages over a coherent receiver. It does not require explicit channel estimation, or a large number of fingers in a rake receiver. It is robust to possible channel distortion on pulse shape. Easy and simple synchronization makes it a good candidate for bursty traffic. The drawbacks of the TR system are the inferior performance to ideal coherent detection and lower data rates because of the transmission of reference signals. In this paper, we propose a dual pulse scheme that not only maintains the advantages of TR systems, but also achieves higher data rate and provides more benefits than the conventional TR method. The remainder of this paper is organized as follows. Section II describes the dual pulse transmission scheme and several detection schemes. In Section III, an improved version of transmission and detection scheme is proposed to avoid the potential self interference between the received dual pulses caused by closely spaced multipaths. Section IV summarizes the merit of the proposed schemes and discusses simulations results. The performance of the proposed technique is then compared to the transmitted reference system. Finally, conclusions are drawn in Section V. II. T RANSMISSION AND D ETECTION T ECHNIQUE In this newly proposed system, a basic ultra-wideband pulse p(t) of duration Tw is composed of two sub-pulses: the subpulse gtr (t) which has non-zero values in the first half interval [0, Tw /2] and the sub-pulse s2 (t) in the second half interval [Tw /2, Tw ]. The energy of gtr (t) is denoted as Eb /2. The subpulse s2 (t) has a fixed relationship with the first sub-pulse

0-7803-8939-5/05/$20.00 (C) 2005 IEEE

gtr (t) depending on the particular modulation format used. For example, Tw Tw ), ≤ t ≤ Tw 2 2 and hence the UWB pulse p(t) is given by s2 (t) = gtr (t −

Tw ), 0 ≤ t ≤ Tw 2 for on-off keying (OOK) and pulse position modulation (PPM). In these cases, the basic unit of a UWB pulse is composed of two identical sub-pulses. For binary phase shift keying (BPSK), however, p(t) = gtr (t) + gtr (t −

s2 (t) = b · gtr (t −

Tw Tw ), ≤ t ≤ Tw 2 2

Dl =

and Tw ), 0 ≤ t ≤ Tw 2 where b = +1/ − 1 is the information bit. The energy of the UWB pulse p(t) is Eb . Fig. 1 illustrates the dual pulse structure for OOK and BPSK. The modulated UWB pulses are transmitted with intervals of Tf , denoted as the frame length. The same dual pulse can be sent Ns times to increase the transmission reliability. Although this paper deals with the single user scenario only, multiple access techniques such as time-hopping or spreading sequence can be used along with this pulse scheme. A BPSK modulated signal for transmission can then be expressed as p(t) = gtr (t) + b · gtr (t −

str (t) =

∞

gtr (t − iTf ) + bi/Ns gtr (t − iTf −

i=−∞

Tw ) (1) 2

where · denotes the floor function. The DP scheme can also be viewed as having the first half pulse shape either the same as or inverse to the second half shape. The design of the pulse shape gtr (t) should take into account the FCC spectrum mask, the potential narrowband interference, implementation issues, etc., but is out of the scope of this paper. We consider the UWB channel model proposed by the IEEE 802.15.3a Working Group [11]. Without loss of generality, the UWB multipath channel is assumed to have unit average energy. The frame interval Tf is assumed to be larger than the length of the channel impulse response plus the dual pulse duration Tw for intersymbol interference (ISI) free transmission. The received signal over one symbol duration (one-shot) for BPSK signaling is given by r(t) =

N K s −1

[αk grx (t − iTf − τk )

i=0 k=1

+

accounts for the general case where there is mismatch between grx (t) and gtr (t). The receiver first passes the received signal through a lowpass filter that has one-sided bandwidth W and unit magnitude. The bandwidth of the lowpass filter is large enough that the information bearing signal is passed without distortion and the noise is limited to within the filter bandwidth. The noise process after the lowpass filter is denoted by n(t). The resolvable paths are assumed to arrive at integer multiples of Tw , but the exact delays {τk } are not known or not estimated. The receiver first multiplies the received signal by its T2w delayed copy and then integrates the product every Tw seconds for the latter T2w duration, i.e.,

Tw − τk ) + n αk bgrx (t − iTf − ˜ (t) 2

(2)

where αk and τk are the amplitude and delay of the k-th multipath, n ˜ (t) is the zero mean additive white Gaussian noise with variance N20 and grx (t) is the received pulse shape corresponding to the transmitted gtr (t). This representation

N s −1 jTf +lTw +Tw j=0

r(t)r(t−

jTf +lTw + T2w

Tw )dt, 2

l = 0, . . . , Lt −1

(3) where Lt = TTmds is the total number of possible paths w and Tmds is the maximum delay spread of the channel. If consider the realistic IEEE channels CM1 to CM4 [11] that have multipaths spaced closer than a pulse duration Tw , we calcuate Dl ’s by T T N s −1 jTf +l w + w 2 2 Tw r(t)r(t− )dt, l = 0, . . . , Lt −1 Dl = Tw 2 jT +l f 2 j=0 (4) where Lt = 2Tmds /Tw . This is identical to that shown in Fig. 2, but different from (3). Although not shown here, simulations find (4) leads to better performance than (3) in CM1 to CM4. The decision variable D can be generated from Lt independent Dl ’s as D=

L t −1

T (Dl )

(5)

l=0

where T (Dl ) is a testing function that can take three different forms. The purpose of different combining schemes is to try to collect Dl ’s that contain signal components as a result of multipath arrivals, not the Dl ’s corresponding to the noise inbetween paths. We use 1) Generalized selection combining. In this method, we select and sum the Dl ’s with L largest absolute values. The test function for GSC is given by Dl , if |Dl | ≥ |D(L) | (6) T (Dl ) = 0, if |Dl | < |D(L) | where D(L) is the one that has the L-th largest absolute value among all Dl ’s. 2) Absolute threshold generalized selection combining. This method compares each autocorrelation output Dl to a fixed threshold Dth (> 0). All the Dl ’s with absolute values larger than Dth are then selected and combined. The absolute threshold test function is given by Dl , if |Dl | ≥ Dth T (Dl ) = . (7) 0, if |Dl | < Dth

0-7803-8939-5/05/$20.00 (C) 2005 IEEE

3) Normalized threshold generalized selection combining. Normalized threshold GSC differs from AT-GSC in how the threshold is determined. Instead of using a preset absolute threshold value, NT-GSC forms its threshold as a fixed fraction ηth of Dmax = max |Dl |, i.e., Dth = ηth Dmax . Define Dl , if |Dl | ≥ ηth Dmax . (8) T (Dl ) = 0, if |Dl | < ηth Dmax Due to space limitations, we have to leave out a unified analysis developed for the performance of noncoherent GSC, AT-GSC and NT-GSC which will be presented in [12]. Note that unlike the diversity combining in rake receivers, the three noncoherent combining schemes proposed here are easy to implement, since the single integrate-and-dump device has already output all Dl ’s and it is only a matter of how to compute (likely in a DSP) the final decision statistics D. An even simpler detection method referred to as “DP over Ttr ” that requires the same samping rate as the TR scheme is given by N s −1 jTf +Ttr Tw )dt (9) D= r(t)r(t − 2 jT f j=0 where Ttr ≤ Tf is the autocorrelator integration length for the DP and TR scheme. Since the received first sub-pulse grx (t) of each symbol is identical for all symbols, we can reduce the noise variance on the received first sub-pulse (reference) by averaging the received signals over several frames. The number of frames Np to average over is dependent on the fading rate of the channel and the implementation complexity constraint. III. I MPROVED DP T RANSMISSION AND D ETECTION In practice, there may be multiple paths that are spaced much closer than a dual pulse width Tw . These unresolvable multi-paths will probably distort the transmitted dual pulse and cause self interference between the reference sub-pulse and the data sub-pulse. Here we present an improved version of the dual pulse transmission and detection scheme to overcome this problem. Often the same UWB dual pulse is transmitted more than once to increase the received SNR for detection. Suppose Ns = 2. We consider two dual pulses, p1 (t) and p2 (t), as one processing unit and invert the reference sub-pulse in the second dual pulse p2 (t) before transmission, as shown in Fig. 3. As shown in the receiver block diagram Fig. 4, we construct the reference template and the received data signal as rref (t) = r1 (t − Tf ) − r2 (t)

(10)

rdat (t) = r1 (t − Tf ) + r2 (t)

(11)

and where r1 (t) and r2 (t) are the received pulses corresponding to p1 (t) and p2 (t), respectively. The decision variable is given by D = rref (t − T2w )rdat (t)dt where the integration can be performed over the whole frame as in the TR system, or only every Tw /2 duration and then follow the combining

methods described in Section II. There is another advantage of the proposed dual pulse structure for Ns = 2. Since the transmitted reference sub-pulses have possitive and negative alternating amplitude levels, the power spectrum of this type of dual pulse signals has no discrete spectral lines regardless of the sub-pulse shape gtr (t), which is a very desirable characteristic. Increased implementation complexity of this improved scheme is the price paid for better performance. IV. D ISCUSSION AND N UMERICAL R ESULTS The attractive properties of the proposed dual pulse scheme can be summarized as follows: 1) The data rate of the proposed dual pulse system doubles the conventional TR system, since the reference subpulse and the modulated sub-pulse are side by side, eliminating the spaced frame length delay between a reference pulse and a data pulse in a TR system. 2) The delay unit in the receiver is only half the duration of the transmitted dual pulse, and therefore is much shorter than the frame delay required in the conventional TR system. Note that implementing accurate short delays is easier than long delays [2]. 3) The dual pulse system is less sensitive to time variation of the channel compared to the TR method and the pilot waveform assisted modulation, because there is no time gap between the reference sub-pulse and the modulated data sub-pulse. This also means within a fixed channel coherence time, there are more received reference subpulses to be averaged over to reduce the noise effect. 4) The dual pulse system retains the same merits of TR systems, such as robust performance, simple implementation (especially for DP over Ttr ) and easy timing acquisition. Frame timing is required for the DP system with noise averaging, but there is no need to estimate either the multipath delays or the multipath gains. The sub-pulse used in the simulation is the second derivative of the Gaussian pulse gtr (t) = {1 − 4π[(t − Tw /2)/Td ]2 } exp{−2π[(t−Tw /2)/Td ]2 }, where Td = 0.2877 ns and the sub-pulse width Tw is set at 0.7 ns. The receiver lowpass filter with Hamming window has 50 taps and a bandwidth of 14.4 GHz. The simulation sampling rate is 30 GHz. For each channel model, the bit error probability is obtained from averaging the BER over 100 channel realizations. Figs. 5-8 plot the BER performance of the DP system with different detection schemes, and the conventional TR technique in UWB channels CM1-CM4, respectively. Averaging the received reference sub-pulse signal over Np = 50 frames to reduce the noise in the reference template is also performed in these figures. As expected, the performance is significantly improved with noise averaging in all plots. For the TR and one type of DP systems, the integration of the product r(t)r(t− T2w is over an interval Ttr less than the whole frame duration so that the more noisy trail close to the end of the frame is not included in the decision variable. This results in improved performance over integrating over the whole frame.

0-7803-8939-5/05/$20.00 (C) 2005 IEEE

For the DP schemes in the four channel models studied, the GSC, AT-GSC, NT-GSC and the DP over Ttr receivers have similar performance using the parameters shown in Figs. 58, with or without reference sub-pulse noise averaging. The number of branches selected in GSC and the normalized threshold in NT-GSC are parameters whose values will influence the error probability of the respective system. The TR system has close performance to the DP systems in CM2 to CM4. In the CM1 channel, the TR scheme slightly outperforms the DP systems at the high signal-to-noise ratio (SNR). This is because the channel model CM1 has a lineof-sight (LOS) component and the majority of the channel energy is concentrated in the first few closely spaced paths, which may result in strong self interference for the DP system and the interference effect becomes more dominant at high SNR’s. For CM2 to CM4 channels, the multipaths are more widely spreaded into longer time durations, resulting in less inter-path interference. Moreover, the system performances (both DP and TR) in different channel models degrade from CM1 to CM4, as expected. The degradations from CM1 to CM4, however, are not very significant for either the DP or TR system studied, which demonstrates the robustness of the autocorrelation receivers in different channel environments. The performance of the improved DP scheme is plotted for CM1 to CM4 in Figs. 9-12, respectively. Since Ns = 2, these figures have at least 3 dB performance gain over the corresponding Figs. 5-8. In Fig. 9 for CM1, the improved scheme has less interference between the received reference sub-pulse and the data sub-pulse, and hence closing the performance gap at high SNR’s between the DP and TR systems. In Figs. 10-12, the improved DP scheme slightly outperforms the TR system with noise averaging.

[4] J. D. Choi and W. E. Stark, “Performance of ultra-wideband communications with suboptimal receivers in multipath channels,” IEEE J. Selected Areas Commun., vol. 20, pp. 1754-1766, Dec. 2002. [5] Y.-L. Chao and R. A. Scholtz, “Optimal and suboptimal receivers for ultra-wideband transmitted reference systems,” Globecom’2003, San Fransisco, USA, pp. 759-763, Dec. 2003. [6] S. Franz and U. Mitra, “On optimal data detection for UWB transmitted reference systems,” Globecom’03, pp. 744-748, Dec. 2003. [7] H. Zhang and D. L. Goeckel, “Generalized transmitted-reference UWB systems,” IEEE Conf. Ultra Wideband Systems and Techno., pp. 16-19, 2003. [8] L. Yang and G. Giannakis, “Optimal pilot waveform assisted modulation for ultra-wideband communications,” IEEE Trans. Wireless Commun., pp. 1236-1249, July 2004. [9] F. Nekoogar and F. Dowla, “On the performance of multiple pulse multiple delay UWB modulation,” Wireless 2003, Calgary, Alberta, Canada, pp. 219-225, July 2003. [10] M. Ho, V. Somayazulu, J. Foerster and S. Roy, “A differential detector for an ultra-wideband communications systems,” Vehicular Technology Conf., pp. 1896-1900, May 2002. [11] J. Foerster, Channel modeling subcommittee report (Final), IEEE802.1503/490r1, Feb. 2003. [12] X. Dong, A.C.Y. Lee and L. Xiao, “Design and analysis of a novel UWB dual pulse transmission and detection technique,” to be submitted to the IEEE journal. OOK Tf "1"

"0"

"0"

BPSK

T

f

"1"

"0"

Fig. 1. Illustration of the transmitted dual pulse structure for on-off keying and binary phase shift keying.

V. C ONCLUSION A novel dual pulse transmission and auto-correlation detection scheme for UWB communications has been presented. It has several advantanges over the conventional transmitted reference scheme, such as higher data rate and implementation edge, while retaining the many benefits a TR system possesses. A new improved structure has further enhanced the system performance. The performance of several different detection and combining schemes has been studied by simulations. The proposed dual pulse scheme permits a simple, low cost, and robust UWB transceiver. To further increase the transmission data rate, mechanisms for combating intersymbol interference caused by reducing frame duration must be designed and included in the receiver.

Fig. 2.

Receiver block diagram. BPSK Ns=2 b=0 p2(t)

p1(t) T

f

BPSK Ns=2 b=1 p (t) 2

p1(t) Tf

Fig. 3. Illustration of the improved dual pulse structure for binary phase shift keying with the pulse repetition number Ns = 2.

R EFERENCES [1] M. Z. Win and R. A. Scholtz, “Characterization of ultra-wide bandwidth wireless indoor channels: a communication-theoretic view,” IEEE J. select. Areas Commun., vol. 20, pp. 1613-1627, Dec. 2002. [2] R. Hoctor and H. Tomlinson, “Delay-hopped transmitted-reference RF communications,” IEEE Conf. Ultra Wideband Systems and Techno., pp. 265-269, May 2002. [3] N. Van Stralen, A. Dentinger, K. Welles, II., R. Gaus, R. Hoctor and H. Tomlinson, “Delay hopped transmitted-reference experimental results,” IEEE Conf. Ultra Wideband Systems and Techno., pp. 93-98, May 2002.

Fig. 4.

Receiver block diagram of the improved system.

0-7803-8939-5/05/$20.00 (C) 2005 IEEE

0

0

10

−1

10

10

−1

Bit Error Probability


10

−2

10

TR (T =28ns, N =1) tr

p

DP, Int (Ttr=28ns, Np=1)

−3

10

DP, GSC (Lt=40, L=20, Np=1)

tr

tr

t

DP, NT−GSC (ηth=0.1, Np=1)

TR (T =28ns, N =50)

TR (Ttr=28ns, Np=50)

th

p

p

−4

DP, Int (T =28ns, N =50)

10

tr

p

DP, GSC (Lt=40, L=20, Np=50)

DP, AT−GSC (D =0.1, N =50)

DP, AT−GSC (Dth=0.1, Np=50)

p

th

p


DP, NT−GSC (η =0.1, N =50) 2

4

th

p

6

8

10 12 Average Eb/N0 (dB)

14

16

18

20

0

22

2

4

6

8 10 Average E /N (dB) b

Fig. 5. Performance of the DP system employing noncoherent GSC, AT-GSC and NT-GSC for CM1 channels.

12

14

16

18

0

Fig. 9. Performance of the improved DP system employing noncoherent GSC, AT-GSC and NT-GSC for CM1 channels.

0

0

10

−1

10

10

−1


10


p

DP, GSC (L =40, L=20, N =50) t

−2

10

TR (Ttr=42ns, Np=1) DP, Int (T =42ns, N =1) tr

−3

10

p

DP, GSC (Lt=60, L=30, Np=1)

−2

10

TR (Ttr=42ns, Np=1) DP, Int (Ttr=42ns, Np=1)

−3

10

DP, GSC (L =60, L=30, N =1) t

DP, AT−GSC (Dth=1.5, Np=1)

DP, NT−GSC (η =0.1, N =1)


p

th

p


TR (T =42ns, N =50) tr

−4

p

−4

DP, GSC (L =60, L=30, N =50)

DP, GSC (L =60, L=30, N =50)

DP, AT−GSC (D =0.1, N =50)

DP, AT−GSC (D =0.1, N =50)

DP, NT−GSC (η =0.1, N =50)

DP, NT−GSC (η =0.1, N =50)

t

2

4

t

p

th

0


10


10

p

DP, AT−GSC (D =1.5, N =1) th

th

p

6

8

p

th

p

p

th


14

16

18

20

0

22

2

p

4

6


0


12

14

16

18

0


0

0

10

−1

10

10

−1


10


p

DP, NT−GSC (η =0.1, N =1) tr

−2

10

TR (T =77ns, N =1) tr

p

Dp, Int (T =77ns, N =1) tr

−3

10

p

DP, GSC (L =110, L=60, N =1) t

p

−2

10

TR (Ttr=77ns, Np=1) Dp, Int (T =77ns, N =1) tr

−3

10

t

p

DP, AT−GSC (D =1.5, N =1)


DP, NT−GSC (η =0.1, N =1)

th

tr

p

th

p

TR (T =77ns, N =50) tr

p

−4

p


10

DP, Int (Ttr=77ns, Np=50) DP, GSC (L =110, L=60, N =50)

DP, GSC (L =110, L=60, N =50)

DP, AT−GSC (D =0.1, N =50)

DP, AT−GSC (D =0.1, N =50)

DP, NT−GSC (η =0.1, N =50)

DP, NT−GSC (η =0.1, N =50)

t

2

4

t

p

th

0

p

DP, GSC (L =110, L=60, N =1)

DP, AT−GSC (Dth=1.5, Np=1) TR (T =77ns, N =50) −4

10

p

6

8

p

th

p

th

p

th


14

16

18

20

0

22

2

p

4

6


0


12

14

16

18

0


0

0

10

−1

10

10

−1


10


p

DP, GSC (L =40, L=20, N =1)

p


0

p

DP, Int (T =28ns, N =1)

DP, AT−GSC (D =1.5, N =1) th

−4

TR (T =28ns, N =1) −3

10

DP, AT−GSC (D =1.5, N =1)

th

10

−2

10

−2

10

TR (Ttr=126ns, Np=1) DP, Int (T =126ns, N =1) tr

−3

10

p

DP, GSC (Lt=180, L=100, Np=1)

p

t

p



TR (T =126ns, N =50)


p

−4

DP, Int (T =126ns, N =50)

10

tr

p

DP, GSC (L =180, L=100, N =50)

DP, GSC (L =180, L=100, N =50)

DP, AT−GSC (D =0.1, N =50)

DP, AT−GSC (D =0.1, N =50)


DP, NT−GSC (η =0.1, N =50)

t

4

t

p

th

2

tr

DP, GSC (L =180, L=100, N =1)

p


0

DP, Int (T =126ns, N =1) DP, AT−GSC (Dth=1.5, Np=1)

tr

−4

TR (Ttr=126ns, Np=1) −3

10

DP, AT−GSC (D =1.5, N =1) th

10

−2

10

6

8

p

th

p

p

th


14

16

18

20

22

0

2

4

p

6


0


12

14

16

18

0

Fig. 12. Performance of the improved dual pulse system employing noncoherent GSC, AT-GSC and NT-GSC for CM4 channels.

0-7803-8939-5/05/$20.00 (C) 2005 IEEE