Performance Comparison of Pulse Shapes in STDL and IEEE 802.15 ...

T2G.1

Performance Comparison of Pulse Shapes in STDL and IEEE 802.1 5.3a models of the UWB Channel Mohammad Upal Mahfuz, Kazi M. Ahmed, Rabindra Ghimire and R.M.A.P. Rajatheva Telecommunications Program Asian Institute of Technology (AIT) Pathumthani, Thailand. {upalm, kahmed, rajath}(ait.ac.th Abstract-This paper demonstrates performance comparison of Ultra Wideband (UWB) systems in the Stochastic Tapped Delay Line (STDL) and IEEE 802.15.3a models of the UWB channel with RAKE receiver in presence of Multiple User Interference (MUI). System models are based on Direct Sequence Code Division Multiple Access (DS-CDMA) principle with BiPhase Modulation (BPM). The system has been studied in presence of multiple user interference (MUI) and Gaussian noise. During the analysis, various types of narrow, sub-nanosecond pulses with same power were used and extensive simulations were run. The simulation results show that pulse shape has noticeable impact on the BER performance with the STDL channel model. The performance of the pulse waveform based on third derivative of Gaussian pulse has proved to be better than other pulses that were used with STDL channel model. On the other hand, maintaining the same parameters but by using IEEE 802.15.3a channel model shows that pulse shapes have no effect on BER performance. Spectral behavior of higher order Gaussian pulses has been explained. Finally, it is concluded that for STDL model, the third derivative of Gaussian pulse is the most suitable pulse shape for fulfilling the FCC regulated power spectral density (PSD) requirement and BER performance in multi-user interference environment, whereas using IEEE 802.15.3a channel, at least 4th or 5th derivative of Gaussian pulse should be chosen since increasing derivative order increases the peak emission frequency but decreases signal bandwidth.

FBW 2

INTRODUCTION

Ultra wideband (UWB) technology is a promising technique for future data communication systems, high accuracy (indoor) geo-location devices and sensor applications. UWB systems utilize carrierless transmission with very low power spectral density. UWB techniques are based on the transmission of nanosecond level short pulses that generate extremely wide spectrum. The narrow pulse waveform spreads the signal energy over the large frequency band. According to the regulation of Federal Commission of Communications (FCC), a signal is defined as UWB signal if it has a -10 dB fractional bandwidth, FBW greater than (or equal to) 0.20 or it occupies at least 500 MHz of the spectrum [1], FBW being expressed as (1)

0-7803-9282-5/05/$20.00 ©2005 IEEE

L>

0.20

(1)

where fH and fL correspond to the -10 dB bandwidths. Although UWB impulse radio can be termed as 'Spread Spectrum' technique because of its extremely large 7.5 GHz operational bandwidth, UWB radio transmits information without carrier, rather by using sub-nanosecond pulses. This paper discusses and compares the UWB system performances for Stochastic Tapped Delay Line (STDL) and IEEE 802.15.3a channel models for indoor UWB propagation. The paper uses Direct Sequence (DS) spread spectrum concept with a chip waveform of a very narrow pulse [2]. Also each user has an individual Pseudo Random code for user separation and spectrum smoothing. In order to analyze the potential of UWB systems, models describing the UWB propagation channel must be properly known. The channel models help analyze system performance and make design trades. The channel model presented in [3] was formulated as a Stochastic Tapped Delay Line (STDL) model of the UWB indoor channel. The IEEE 802.15.3a UWB channel model is the most recent channel model that IEEE 802.15.3a Standardization Group has developed for the evaluation of ultra wideband communication systems [4]. In this paper, to develop the UWB system model, DSCDMA spreading approach, both Stochastic Tapped Delay Line (STDL) and IEEE 802.15.3a channel models of the channel [3, 4] and RAKE receiver in presence of multiple user interference are considered. The BER performances of using STDL and IEEE 802.15.3a models and RAKE receiver with multiple user interference are shown. The effect of pulse shapes on channel characteristics has also been indicated. Finally, it has been concluded that third derivative of Gaussian pulse is the most suitable pulse shape for simultaneously fulfilling the FCC regulated power spectral density (PSD) and reasonable BER performance using STDL channel model and that 4th or 5th derivative of Gaussian pulse is the most suitable pulse shape for IEEE 802.15.3a model. The paper is organized as follows: Section II provides with a brief description of the concept of DS-CDMA based transmitting scheme. Ultra wideband pulse characteristics have been explained in Section III. The discussion is followed by Section IV and V describing UWB propagation channels under investigation. The receiver

Keywords- Ultra wideband, STDL, IEEE 802.15.3a, channel model, pulse shape. I.

f

fH+fL

811

ICICS 2005

Frequency domain analysis of RAYLEIGH pulse derivatives

section is described in Section VI. Simulation results have been explained in Section VII. Finally, Section VIII concludes the paper.

10 -

Right shift of center frequency cD 10

DIRECT SEQUENCE BI-PHASE MODULATION (DS-BPM) TRANSMISSION SCHEME A DS-UWB system is basically identical to an ordinary DS-SS system except that the bandwidth spreading effect is achieved by pulse shaping. A DS-UWB signal can be written as [6] II.

SDS (t)aDS

Z Wtr(t JTm) nj(12D

N

)

0

1 0th derivative

lo

~1 0 COL1 First derivative

10 2

(2)

where n is a pseudorandom code that takes values {+1} and aDS is the maximum amplitude of the DS-UWB pulse train. Each information bit consists of NDS pulses and has a duration of Tb;= NDSTp, as shown in Figure 1.

4

6

8

Frequency [Hz]

10 x

12 10

(a) Frequency spectrum: Rayleigh pulse Frequency domain analysis of CUBIC pulse derivatives 10o

ULTRA WIDEBAND PULSE CHARACTERISTICS Firstly, the basic pulses used in simulation of this study are the nth order derivatives of Gaussian pulse expressed as shown III.

N

0

a) 10

in(3)

0

Q

n

(t)

C/cnLU

(3)

ed n

In addition to above, the channel effects using Rayleigh pulse, cubic pulse, truncated Sinc pulse and dual Gaussian pulse were also investigated. For the same pulse width, the frequency spectrums of Rayleigh and Cubic pulse have been shown in Fig. 2(a) and Fig. 2(b) respectively. To ensure that the frequency spectrum of the pulse fulfills the FCC PSD requirement, the appropriate derivative order and the decaying factor have to be chosen carefully. As shown in Fig. 2(c) the different order pulse derivatives based on Gaussian pulse have been used for selecting an optimum pulse-decaying factor of 0.168 ns. This has been calculated following the bi-section method approach as shown in [7]. Finally, Gaussian pulses are given more importance in this paper to compare the two channel models because Gaussian pulses are easier to generate at the pulse generator.

10

10

Frequency [Hz]

x 10

(b) Frequency spectrum: Cubic pulse

E 0 co

a N

a/) cn

Tb

-A~~~~~~l 4~~~~~~~~ l

A/

IH-I Tf

N \A >1

T=

0

NAWIIW

(c) Power spectral densities of higher order derivatives of Gaussian pulse

t

Tf =Tp

Figure 2: UWB pulse characteristics

Figure 1: UWB system based on Direct Sequence (DS) technique

IV.

THE STDL CHANNEL MODEL

The Stochastic Tapped Delay Line (STDL) channel model is a derivative of Saleh-Valenzuela (S-V) model. The

812

users

Transmitter Section

Channel Model

Figure 3: Block diagram of system model

transmitted Bi-Phase modulated signals, xj (t) pass through the STDL propagation channel model, which is based on measurements made in a typical office environment for UWB radio channel [3]. The model characterizes the shape of the power-delay profile (PDP) in terms of path gains, GK and delays, -K of multipath components. The path resolution of the considered system was A-= 2ns. The STDL model represents the statistics of path gains and its dependence on the delays. The large-scale and the small-scale fading statistics have also been distinguished. The global and local parameters as described in [3] have been shown in Table I.

TABLE I. EXPRESSIONS OF GLOBAL AND LOCAL PARAMETERS [3]

Global parameters Path Loss

L

Power Ratio r

d < lm

LN (-4;3) Local parameters

EnergyGains Gk F(Gk; mk ) -

mk

m Values

-

TN (/m. ('Zk ); ¢m

'k)

I'm (Z-k ) = 3.5 Zk 73 a2(ZJk) = 1.84

16k

TABLE II. MODEL PARAMETERS IN IEEE 802. 15.3A UWB CHANNEL [4]

K

1=0 k-=O

do)

-

V. IEEE 802.15.3A UWB CHANNEL MODEL The IEEE 802.15.3a standard model has been described in detail in [4]. Since channel measurements showed multipaths arriving in clusters, the model used a Saleh-Valenzuela (S-V) approach for the time-of-arrival statistics calculation. Lognormal distribution rather than a Rayleigh distribution for the multipath gain magnitude has been recommended. In addition, independent fading is assumed for each cluster as well as each ray within the cluster. The discrete impulse response of the multipath model can be described as [4] hi(t) = X IC0 I)

20.4.log10(d

-56+74.log10(d/do) d>lm DecayConstant £ LN(16.1;1.27) Shadowing Gtt LN (-PL;4.3)

Model Parameters

(4)

Cluster arrival rate A (1/nsec) Ray arrival rate, X (1/nsec) Cluster decay factor, F Ray decay factor, y Std. dev. of cluster log-normal fading a1 (dB) Std. dev. of ray log-normal fading o2 (dB) Std. dev. of total multipath log-normal fading a, (dB)

where { ak / } are the multipath gain coefficients, { Tr } is the delay of the Ith cluster, { - } is the delay of the kth multipath component relative to the Ith cluster arrival time (T'), { X } represents the lognormal shadowing, and i refers to the ith realization. The IEEE 802.15.3a channel model has been explained for 4 special cases depending on transmitter to receiver distance and the availability of LOS path between them. In this paper, Case C (4-1Om, NLOS) condition of the channel has been used and the corresponding channel parameters are shown in Table II.

813

Case C (NLOS at 410 meters) 0.0667 2.1 14.00 7.9 3.3941 3.3941 3

various types of sub-nanosecond pulses as well as multiple user interference (MUI) have been presented. The arrival of multiple users is assumed to be Gaussian in nature and exist over the entire period. Figures 4 and 5 show the simulation results for different pulse shapes with 8 selective RAKE arms at a distance of 5m with STDL and IEEE 802.15.3a models respectively with AWGN only without multiple-user interference. For each case, synchronization and channel estimation are assumed to be perfect. The results show that the pulse shape has noticeable impact on the performance of the system if the channel model is assumed as STDL channel model. Also, the performance of third derivative of Gaussian pulse is better than any other pulses in STDL channel model. For example, for a BER of 10-4, using 8 RAKE arms at a distance of 5m as shown in Fig. 4, the third derivative of Gaussian pulse has 3.5 dB Eb/No gain over that of Gaussian monopulse. On the other hand, Fig. 5 shows the performance of different Gaussian pulse in the same conditions as for Fig. 4 case but using the recently accepted IEEE 802.15.3a UWB channel model. As shown in Fig. 5, higher order derivative pulses have no effect on BER performance. As shown in Fig. 2, pulse derivative order only shows on which portion of the frequency spectrum the pulse energy is located. As a

VI. RECEIVER SECTION The received signal is the sum of replicas of the transmitted signals. The received signal is, therefore, expressed as Np,th

r(t

i=l

(5)

(c, x(t - -C, ) + n(t) + nf (t))

where n(t) is zero mean AWGN and nf (t) is the interference signal. For the simulation of this study RAKE receiver with a maximum of 8 arms is used. The signals in the correlators are despread by Walsh Hadamard Code of length 16. All decision statistics are weighted by a weighting factor, a to form overall decision statistics. The signals are then integrated over the entire period. The integrated signal is then compared with the appropriate threshold value to receive a better estimate of the transmitted signal.

VII. RESULTS In this section, results based on the simulation of an ultra wideband system using DS-CDMA multiple access technique, both Stochastic Tapped Delay Line (STDL) and IEEE 802.15.3a channel models and RAKE receiver with 8 airms for 10° 10

10

Rayleigh Pulse Third Derivative Gaussian Second Derivative Rayleigh Truncated Sinc Gaussian Monopulse Third Derivative Rayleigh Fifth Derivative Gaussian Dual Gaussian monopulse

=i

102

*AWGN No Interference; RAKE, 8 arms No Interference; RAKE, 12 arms 1 Interference; RAKE, 8 arms 1 Interference; RAKE, 12 arms 5 Interference; RAKE, 8 arms 5 Interference; RAKE, 12 arms

10° 10 1-

m

Lmco

0o-3 1 o-4 -

10

10

Eb/No in dB

15

20

1n _0

25

15

2o

Figure 6. Comparison of the performance at a distance of 5m with third derivative Gaussian pulse using STDL channel model

Figure 4. BER performance of different pulses in STDL model of the UWB channel at d=5m in the presence of AWGN with 8 RAKE arms

DS-BPM UWB system: d=5m, 8 SRAKE arms

10°

10 Eb/No in dB

5

MUI Performance of DS-BPM UWB System

-

10

-_- 1st deri. Gaussian -A- 2nd deri. Gaussian -V- 3rd deri. Gaussian -e- 4th deri. Gaussian --E 5th deri. Gaussian

-6t -A-

No MUI 1 interferer

PL L3 interferers lll

x

1 1

-

-

X

-

-6- 5 interferers --*- 7 interferers

10

w LL]

w LL]

co

co

10o2 -

10

10-3

0

2

4

6

8

10

12

Eb/No in dB

14

16

18

20

0

Figure 5. Performance of different Gaussian pulses in IEEE 802.15.3a model with DS-BPM, d=5m im the presence of AWGN with 8 RAKE arms

2

8

10

12

Eb/No in dB

14

16

18

20

Figure 7: Multiple user performance of DS-BPM UWB system using IEEE 802.15.3a model at d=5m

814

consequence, it refers to the appropriate fulfillment of FCC regulated PSD mask. Moreover, it is also to be mentioned here that the system has also been simulated with some other pulses than Gaussian pulse but finally those pulses are not considered because of increased transmitter complexity involved. Gaussian based pulses are easier to generate in the pulse generator. Those pulses also do not fulfill the FCC PSD mask properly for the same pulse width. Finally, it has been concluded that using STDL channel model, third derivative of Gaussian pulse performs the best in single user case. On the other hand, using IEEE 802.15.3a channel model, higher order pulse shapes have no effect on BER performance. As derivative order increases, the peak emission frequency increases and signal bandwidth decreases [7], so choosing the most appropriate derivative order is a trade off with pulse decaying factor for a desired BER level. Bandwidth maximization is also an important factor in choosing the derivative order of the UWB pulse transmission. Keeping this point in consideration, the 4th order derivative of the Gaussian pulse has been chosen in this study. Fig. 6 shows system performance in presence of multiple user interference at a distance of 5m using STDL channel model using third derivative of Gaussian pulse. As shown in Fig. 6, using the STDL channel model results in less degradation in BER performance of third derivative of Gaussian pulse than that of Rayleigh pulse and second derivative of Rayleigh pulse in the presence of MUI. The third derivative of Gaussian pulse shows better performance than all other pulses. On the other hand, Fig. 7 shows the MUI performance for a DS-UWB system using IEEE 802.15.3a channel model with 4th derivative of Gaussian pulse. Fig. 7 clearly suggests that DSBPM system is quite robust in MUI environment.

REFERENCES [1] H. Yomo, P. Popovski, C. Wijting, I. Z. Kovacs, N. Deblauwe, A. F. Baena and R. Prasad, "Medium Access Techniques in Ultra-Wideband Ad Hoc Networks", IST-2001-34157 Power Aware Communications for Wireless OptiMised personal Area Network (PACWOMAN), partially funded by the EC, 2001. [2] D. Barras, F. Ellinger, H. Jackel, "A comparison between ultrawideband and narrowband transceivers", TRLabs, IEEE Wireless 2002, Calgary, Canada, July 2002. [3] D. Cassioli, M. Z. Win and A. F. Molisch, "The Ultra-Wide Bandwidth Indoor Channel: From Statistical Model to Simulations," IEEE Journal on Selected Areas in Communications, August, 2002, Vol. 20, No. 6, pp. 1247-1257. [4] J. R. Foerster, M. Pendergrass and A. F. Molisch, "A Channel Model for Ultrawideband Indoor Communication", Wireless Personal Multimedia Communications (WPMC-03), Kanagawa, Japan, October 19-22, 2003, Vol. 2, pp. 116-120. [5] J. Foerster, "Channel Modeling Sub-committee Report (Final)", IEEE P802.15 Working Group for Wireless Personal Area Networks (WPAN), December 3, 2003. [6] B. Mielczarek, M.-O. Wessman and A. Svensson, "Performance of Coherent UWB RAKE Receivers with Channel Estimators", IEEE Vehicular Technology Conference (Fall), Orlando, Florida, USA, October, 2003. [7] H. Sheng, P. Orlik, A. M. Haimovich, L. J. Cimini, Jr. and J. Zhang,

"On the spectral and power requirements for Ultra-wideband transmission", IEEE 2003 International Conference on Communications (ICC), May 2003, AK, USA, pp. 738-742.

VIII. CONCLUSIONS At present, UWB radio is an emerging technology especially in high data rate short distance indoor communications. In this paper the two channel models for

UWB communications, named Stochastic Tapped Delay Line (STDL) and IEEE 802.15.3a, have been compared in view of the effects of pulse shape on BER performance. DS-CDMA has been used in the transmitter section because it is

comparatively less complicated and is a cheaper multiple access technique currently available. Results show that third derivative of Gaussian pulse is the most suitable pulse shape

using STDL channel model whereas higher order derivatives have no effect on BER using IEEE 802.15.3a model, even though it is suggested that 4th or 5th derivative of Gaussian pulse with appropriate pulse decaying factor should be chosen with IEEE 802.15.3a channel model. Consideration of the antenna effects is referred to as future work of the current research. Gaussian pulse and its derivatives have been given more importance because they can be generated in the pulse generator at the easiest way. Finally, this paper concludes that pulse shape has an important impact on the BER performance of a UWB system depending on the channel model in use and at the same time appropriate fulfillment of the FCC regulated PSD mask.

815