ICT 2004 Page 1 of 1

On behalf of the ICT 2004 Technical Program Committee - TPC, I certify that the paper entitled "PERFORMANCE
COMPARISON BETWEEN BPSK AND BPPM TH-UWB RADIO SYSTEMS IN THE PRESENCE OF NARROW-BAND
INTERFERENCE USING PROPER WAVEFORM DESIGN" authored by HASSAN KHANEE and PAEIZ AZMI has been
accepted for publication and presentation at ICT 2004 Poster Session. Final versions of ICT 2004 accepted Poster papers
will be published in CD-ROM which ISBN is 85-904524-2-5.

file://G:\ICT2004\html\certificados\cor\Cert_color_0000000121.html 11/9/2004
 Performance Comparison between BPSK and BPPM Time-Hopping UWB Impulse
Radio Multiple-Access Systems in the Presence of Narrow-Band Interference Using
Proper Waveform Design
Hassan Khanee, Paeiz Azmi

Electrical and Computer Engineering Department Signal Processing and Multimedia Research Lab.
Tarbiat Modarres University Iran Telecommunication Research Center
Tehran-Iran Tehran-Iran

Abstract- Ultra Wide Band (UWB) radio systems, because of Following this introduction, in section II, the mathematical
their huge bandwidth, must co-exist with many narrow-band model of TH-UWB-IR signals is considered. In section III, the
systems in their frequency band. This coexistence may cause proper waveform design that leads to the best NBI rejection is
significant degradation in the performance of both systems. applied to BPPM TH-UWB-IR. Section IV shows the simula-
Several methods for narrow-band interference (NBI) tion results, and section V concludes this paper.
suppression in UWB radio systems have been proposed. One of
them is based on mitigating the NBI effects through proper II. MODELING OF TRANSMITTED AND RECEIVED SIGNALS
waveform design. In [1] it has been shown that using doublet IN A TH-UWB-IR SYSTEM
pulses can significantly reduce the NBI effects on Binary Phase Sending one pulse per frame time Tf , and Ns pulses per
Shift Keying Time-Hopping UWB Impulse Radio (BPSK TH-
UWB-IR) systems. In this paper, we extend the technique data symbol the transmitted signal of the kth user of a TH-
proposed in [1] to Binary Pulse Position Modulation (BPPM) UWB-IR system is [4, 5, 6]:
∞

∑a
TH-UWB-IR systems. We show that the proper waveform
design can properly work for NBI suppression in BPPM TH- strk (t ) = k
[ j / Ns ] wtr (t − jT f − C kjTc − δ b[kj / N s ] ) (1)
j = −∞
UWB-IR systems. In this paper, we also compare BPSK and
BPPM based TH-UWB-IR systems in different conditions. where [x] denotes integer part of x , Tf is the frame time,
wtr (t ) is the transmitted waveform, usually referred to as
Keywords: Ultra Wide Band Impulse Radio (UWB-IR),
Narrow-Band Interference (NBI) Suppression, Time-Hopping monocycle, δ is the parameter of BPPM modulation, and
(TH), Binary Phase Shift Keying (BPSK), Binary Pulse 0 ≤ C kj < Nh is the pseudo random time-hopping sequence of
Position Modulation (BPPM), and Proper Waveform Design.
the kth user in the jth frame with period of N p used for
I. INTRODUCTION
multiple access capability while Tc represents a chip time. Both
In recent years, UWB radio communication systems have
drawn a lot of attention from both the research community and BPSK and BPPM TH-UWB-IR schemes are foreseen in (1). In
industry. This is mainly because of the properties that make BPSK TH-UWB-IR systems, the information is conveyed by
them viable candidates for short-range high-data rate communi- series a i , which takes the values ±1 while bi = 0 for all i . In
cations in dense multipath environments [1-6]. BPPM based systems, the information is conveyed by series
UWB Impulse radio system is a kind of UWB systems
transmits information data by modulating impulse-like ultra
b i , which takes the values 0 and 1 while a i = 1 for all i . Each
short pulses called monocycles. Using ultra short pulses results data symbol is transmitted over Ns monocycles for achieving
a very big bandwidth from near DC to several GHz [4].
Occupying an ultra wide bandwidth causes new problems for processing gain. Therefore the symbol rate, represented by Rs ,
this kind of radio. This is because there may be many other can be simply given by Rs = 1 Tf Ns [1,4,6].
communication systems, which are simultaneously working at
The effect of the receiving antenna is modeled as
different parts of UWB systems bandwidth. This indicates that
derivative operation [6], so the received signal at the output of
external narrow-band sources may degrade the performance of
the receiver antenna is modeled as:
UWB systems.
Nu
r (t ) = ∑ g k S rec
In [1], it has been shown that the proper waveform design
can suppress the NBI effects on the performance of BPSK TH-
k
(t − τ k ) + I (t ) + n(t ) (2)
k =1
UWB-IR systems. In this paper, we extend this method to ∞
BPPM TH-UWB-IR systems and give a comparison between
BPSK and BPPM TH-UWB-IR systems with respect to the
where Srec (t ) =
k
∑a
j =−∞
k
[ j / Ns ] wrec (t − jTf − C kjTc − δb[kj / Ns ] ) is the
SNR and BER performance degradation due to NBI. Our
results show that in the presence or absence of NBI, the BPSK received signal of the kth user, wrec(t) denotes the derivative of
based system has better performance than the BPPM one.
 Pulse Train Correlator
(i + 1) NsTf + τ k
( i + 1) N s T f + τ k
∫
r(t)
(.) dt
iN s T f +τ k

∑ v(t − jT
j
f − C kjTc − τ k ) Hold
Circuit

d ik
Demodulated data of user k Compare to
zero
Figure 2. Block diagram of the correlator receiver of kth user for both
BPSK and BPPM TH-UWB-IR.

For the user of interest (3) can be reduced to:

Figure 1: Transmitted and Received monocycle waveforms that are first
and second-order derivatives of Gaussian pulse shape. d i = s i + c i + ei + n i (4)
where si corresponds to the signal of the user of interest and is
wtr (t ) [6], g k and τ k denote the channel gain and delay of
equal to Ns g1a i Rw (0) and Ns g1[ Rw (bi δ ) − Rw (δ [1 − bi ])]
1 1 1
the kth user, I (t ) represents external NBI, Nu is the number
for BPSK and BPPM based systems, respectively, and
of active users, and n(t ) represents receiver thermal noise with ∞
Rw(τ ) = ∫ wrec(t)wrec(t +τ )dt .
two-sided power spectral density of N0 / 2 . Fig. 1 shows the −∞
The second term on the right hand side of (4) accounts for
normalized waveforms wtr (t ) and wrec (t ) extracted from [6]
the cross-talk due to the multiple access interference (MAI) and
and are used in this paper. can be found as follows:
The block diagram of the correlator receiver for both BPSK Nu ( i +1) Ns −1
and BPPM TH-UWB-IR schemes is shown in Fig. 2. The ci = ∑ ∑ga k
k
j −α k Rw ((C 1j − C kj−α k )Tc − β k ) (5)
[ ]
receiver uses a template waveform denoted by v(t ) derived k =2 j =iNs Ns
from received monocycle waveform signal. The template Nu ( i +1) Ns −1

waveform v(t ) is equal to wrec (t ) and wrec (t ) − wrec (t − δ ) ci = ∑ ∑ g {R k w ((C 1j − C kj−α k )Tc − b kj −α k δ − β k ) −
[ ]
k =2 j =iNs Ns
for BPSK and BPPM based TH-UWB-IR schemes, respec- (6)
tively. We use this notation for template waveform in the rest of Rw ((C 1j − C kj−α k )Tc − δ (b kj −α k − 1) − β k )}
k [ ]
this paper. The receiver computes the following statistic d i for Ns

demodulating ith data symbol of the kth user: for BPSK and BPPM based systems, respectively.
( i +1) NsTf +τ k The third term on the right hand side of (5) represents the
d ik = ∫ r (t )∑ v(t − jTf − C kjTc − τ k )dt effect of the external NBI and can be found as follows:
iNsTf +τ k
j
( i +1) Ns −1 ([ i +1] Ns − j )Tf −C jTc
1
(3)
∑ ∫ I (t + jT
( i +1) Ns −1
( i +1) NsTf +τ k
ei = + C 1jTc )v(t )dt
=∫
iNsTf +τ k
r (t ) ∑ v(t − jT
j =iNs
f − C kjTc − τ k )dt
j =iNs ( iNs − j )Tf −C1jTc
f (7)

The last term on the right hand side of (5), i.e., ni

If d i > 0 , then for BPSK and BPPM based schemes the
k
corresponds to the random noise process with variance
received data symbol is declared as "1" and "0”, respectively, δ n2i = Ns Rv (0)N0 / 2 at the input of the threshold detector. It
and if d < 0 , then for BPSK and BPPM based schemes the
k
i can be found as follows:
received data symbol is declared as "-1" and "1", respectively. ( i +1) Ns −1 ([ i +1] Ns − j )Tf −C jTc
1

Without loss of generality, it is assumed that the delay of the

user of interest (herein the first user) τ 1 = 0 and the channel
ni = ∑
j =iNs
∫ n(t + jT f + C1jTc )v(t )dt (8)
( iNs − j )Tf −C1jTc
gain associated with the user of interest, g1 = 1 . With these Since v(t ) has an ultra short duration and the integration
assumptions each user's delay relative to the delay of first user
interval [(iNs − j)Tf − C1jTc , ([i + 1]Ns − j)Tf − C1jTc ] contains
can be written as τ k = αkTf + βk , where αk is an integer
this ultra short duration for any integer j in [iNs , (i + 1) Ns −1] ,
number and β k is uniformly distributed in [0, T f ] [1,4].
(7) can be written as:
 ( i +1) N s −1 ∞
ei = ∑ ∫ I (t + jT
j =iN s
f + C 1j Tc )v(t )dt (9)
30
BPSK

−∞
Assuming I (t ) is zero mean, the variance of ei can be 25 f0=3 GHz
computed as:

SNR Degradation due to NBI (dB)

f0=2 GHz
( i +1) Ns −1 ( i +1) Ns −1 ∞

∑ ∑ ∫ R (τ )
20 f0=4 GHz
δ = E[e ] =
2
ei
2
i I (10) f0=4.5 GHz
j =iNs j =iNs
'
−∞
15
Rv (τ − ( j − j ' )Tf − (C 1j − C 1j ' )Tc )dτ
f0=1 GHz f0=5 GHz
where Rv (τ ) and RI (τ ) are the autocorrelation functions of 10

v(t ) and I (t ) , respectively.

Since, Rv (τ − ( j − j )Tf − (C j − C j ' )Tc ) is nonzero only
' 1 1 5
f0=6 GHz

for j = j ' , (10) can be written as: 0

-40 -35 -30 -25 -20 -15 -10 -5 0
SIR (dB)
∞
δ e2i = E[ei2 ] = N s ∫ RI (τ )Rv (τ )dτ (11)
30
BPPM

−∞
Using Parseval's theorem [1], (11) can be written as:
25
∞ f0=3 GHz

δ = E[e ] = Ns ∫ SI ( f )Sv ( f )df

SNR Degradation due to NBI (dB)

2 2 (12)
ei i
−∞ 20

This implies that by properly designing the pulse shape wtr (t ) , f0=2 GHz

15 f0=5 GHz
NBI can be mitigated by generating zeros in the power spectral
density of the pulse shape around frequencies that the NBI is
strongly present [1]. It is worth noting that using this approach 10
f0=4 GHz
reduces the UWB interference on narrow-band systems that f0=1 GHz
appears in noise level increment in such systems. Thus we can 5
achieve mutual interference suppression between narrow-band f0=6 GHz
and UWB based systems using proper waveform design. 0
-40 -35 -30 -25 -20 -15 -10 -5 0
If MAI is modeled as a Gaussian noise process, the SNR of SIR (dB)
the user of interest at the input of the threshold detector is
Figure 3. SNR degradation in BPSK and BPPM TH-UWB-IR systems
calculated for BPSK and BPPM cases as follows: caused by tone interference with parameter f 0 represents NBI center
Ns2 g12 Rw2 (0) frequency.
SNR( Nu ) = (13)
δ n2i + δ e2i + δ MAI
2
when RI (τ ) = P cos( 2πf0τ ) which corresponds to the tone
N g [ Rw (0) − Rw (δ )]
2 2 2
interference I (t ) = 2 P cos(2πf0t ) . Fig. 3 shows the effect
SNR( Nu ) = s 1
(14)
δ n2i + δ e2i + δ MAI
2
of this tone interference on SNR(20) for both schemes as SNR
Nu degradation versus the signal to interference ratio (SIR) with the
where δ MAI = E[ci ] = Nsδ
2 2 2
∑g
k=2
2
k is MAI variance using parameter f0 represents NBI center frequency. For both
schemes, the received pulse shape is assumed to be the second
whole frame time for time-hopping range to increase multiple order derivative of Gaussian pulse shape.
∞
Since the communication channel is symmetric, the
1
access performance. While δ is equal to ∫ Rw2 (τ ) d τ
2
probability of error can be written as:
2T f −∞

1 ∞ Pe = p(error | a i1 = 1) = p (d i < 0 | a i1 = 1) (15)

and
Tf ∫ −∞
{Rw2 (τ ) − Rw (τ ) Rw (τ + δ )}dτ for BPSK and
Pe = p(error | bi1 = 1) = p(d i > 0 | bi1 = 1) (16)
BPPM based schemes, respectively.
In order to evaluate the NBI effects, we consider a case with for BPSK and BPPM based schemes, respectively. Assuming
20 active users communicating at 25Mbps. In the absence of Gaussian distribution for MAI, and perfect power control, i.e.,
external NBI, we set SNR(20) to 10dB for BPSK case while g k / g1 = 1 , we get:
with the same conditions (fixed signal and noise power) the
SNR for BPPM case is computed as 9.155dB. Consider the
  
  (17)
 Rw (0 ) 
Pe = Q  
 ( N u − 1 ) δ MAI
2
δ 2
N 0 Rw (0 ) 
− BPSK
+ +
ei
 2 2 
 Ns g N1 s 2Ns 
 
 
  (18)
R w ( 0 ) − R w (δ )
Pe = Q  
 (Nu − 1 ) δ MAI
2
δ ei
2
N 0 Rv (0 ) 

− BPPM
+ 2 2 + 
 Ns g1 N s 2Ns 
for BPSK and BPPM based schemes, respectively.
Fig. 4 shows the probability of error performance of BPSK
and BPPM TH-UWB-IR in the presence of an external tone
interference with SIR=-20dB when 20 users are communicating
at 25Mbps. The curves show probability of error versus the tone
interference frequency. Solid curves correspond to using a
single second order derivative of Gaussian pulse shape (Mono
pulse). Fig. 5 demonstrates the PSD of mono and doublet
waveforms shown in Fig. 6. As can be observed from Figs 4-5,
the probability of error performance in the presence of NBI
decreases more rapidly at frequencies where the power spectral
density (PSD) of the mono pulse is stronger. The dotted curves
in Figs 4-5 correspond to using a doublet pulse shape with
Tgap =1ns, which will be described in the next section. Fig. 4
also shows that in the same conditions, BPSK based scheme
outperforms BPPM based scheme.

III. INTERFERENCE SUPPRESSION

As mentioned before, the degradation due to NBI is
maximized whenever its frequency is placed at the point that the
Figure 4. Probability of error of BPSK and BPPM TH-UWB-IR systems
PSD of the received signal is maximum. To perfectly mitigate a versus tone interference frequency with and without using doublet pulse.
tone NBI with frequency f0, a doublet pulse must be used whose
gap time equals 1/f0 [1]. The waveform of a doublet pulse is as
follows:
1
wdoublet (t ) = ( wrec (t ) − wrec (t − Tgap )) (19)
2
where Tgap is the gap time between two monocycles. Fig. 6
shows doublet waveform with Tgap =1ns. Power spectral
density of doublet template waveform, vdoublet (t ) , denoted by
Svdoublet ( f ) is computed for BPSK and BPPM based schemes,
as follows:
2 2
W (f) 2Wrec ( f )
Svdoublet ( f ) = doublet = sin 2 (πfTgap ) (20)
Tf Tf
2 Figure 5. PSD of monocycle and Doublet pulses. The nulls in doublet pulse
8Wrec ( f ) PSD are placed in the frequencies that the NBI is present.
Svdoublet ( f ) = sin 2 (πfTgap ) sin 2 (πfδ ) (21)
Tf have two degrees of freedom to generate nulls in the BPPM
By controlling Tgap we can generate nulls in frequencies where based spectrum those are Tgap and δ , i.e. the spectrum of
the strong NBIs are present and mitigate their effects. For received signal has nulls at f = k Tgap in the BPSK based
example, if Tgap is set to 1ns, the NBIs around multiple
system and at f = k Tgap and f = k δ in the BPPM based
integers of 1GHz are completely removed. As can be seen, we
system, where k is an integer number. Fig. 4 shows that no
 0
BPSK
10

BW= 50, 100 and 200 MHz

-1 with SIR= -20dB
10

Probability of Error
BW= 50 MHz
with SIR= -17 dB

-2
10

-3
10
0 1 2 3 4 5 6 7
NBI Center Frequency (GHz)
Figure 6. Doublet and Monocycle waveforms.
0
BPPM
10
BER performance degradation exists in such frequencies, i.e.,
f0 = 1, 2, 3, …,10 GHz, regardless of how strong the tone NBI
at these frequencies is. BW= 50, 100 and 200 MHz
with SIR= -20dB
Fig. 7 shows the SNR degradation in the presence of NBIs -1
10
Probability of Error

that have flat power spectral densities over 50, 100 and 200
MHz bandwidths with SIR=–20dB versus NBI center frequ-
BW= 50 MHz
ency, while 20 users are on the air and communicating at with SIR= -17 dB
25Mbps. Using doublet pulses reduces SNR for BPSK and
-2
BPPM based schemes to 9.3dB and 8.47dB, respectively, in the 10

absence of NBI. As Fig. 7 shows assuming flat power spectral

density for NBI over a fixed bandwidth degrades the
performance smoothly and with a fixed SIR the wider bandwi-
dths cause more performance degradation. 10
-3

0 1 2 3 4 5 6 7
For better comparison between BPSK and BPPM TH-UWB- NBI Center Frequency (GHz)
IR systems, Fig. 8 shows the probability of error versus signal
Figure 7. Probability of error of BPSK and BPPM TH-UWB-IR systems
to Gaussian noise power ratio in the absence of NBI for both versus NBI center frequency with different bandwidths.
mono-pulse and doublet pulse waveforms with different
modulation schemes while 20 users are communicating at 0
10
25Mbps. From Fig. 8, it can be seen that in the same conditions, BPSK
BPPM
BPSK based scheme outperforms BPPM based scheme. It can 10
-1
Double t BPSK
also be seen that using doublet pulses degrades the performance -2
Double t BPPM
10
in the absence of NBI. This result conforms to the results that
Probability of Error

we found from Fig. 4 by considering frequency region 10

-3

above7GHz. Since the support of doublet pulse is wider than

-4
mono pulse, using doublet pulses increases the probability of 10

catastrophic collisions and leads to increment in multiple access 10

-5

interference (MAI) and decreases achievable data rate.

-6
10

IV. SIMULATION RESULTS -7

10 20 users are communicating at 25 Mbps
In the following section, simulation results for two schemes
(i.e. BPSK and BPPM TH-UWB-IR) are presented. We assume 10
-8

0 5 10 15 20 25
perfect power control, fully random TH sequence, coherent Eb /N0 (dB)
detection and asynchronous multiple access. The BER Figure 8: Probability of error of BPSK and BPPM TH-UWB-IR systems
performances of BPSK and BPPM schemes in the presence of a versus signal to noise ratio with and without using doublet pulse.
tone NBI with SIR=-20dB are given in Figures 9 and 10,
respectively. It is apparent that NBIs with center frequencies Figures 9 and 10 with Fig. 4 shows that Gaussian approxima-
near DC or out of correlator band does not degrade the BER tion for multi-user interference is over optimistic for TH-UWB-
performance of TH-UWB-IR. As can be seen, BPSK scheme IR, which utilizes very large bandwidth. Table 1 lists the differ-
has better BER performance than the BPPM one. Comparing ent parameters used in this simulation.
 10
0
BP S K V. CONCLUSION
S ingle P ulse One of the techniques for narrow-band interference
Double t P u lse w ith T g ap = 1 ns
suppression in TH-UWB-IR systems is based on using properly
designed UWB pulse waveform. This goal can be achieved by
10
-1 designing waveforms that have nulls in frequencies where NBI
exists. Using doublet pulses for NBI mitigation has been
Bit Error Rate

proposed for BPSK scheme in [1,2]. In this paper we have

extended doublet pulse technique to BPPM scheme and have
10
-2 compared these two schemes in AWGN channels in the
presence of multiple-access and narrow-band interferences. It
has been shown that BPSK scheme slightly outperforms BPPM
scheme while BPPM scheme has two degrees of freedom for
10
-3
NBI suppression by controlling Tgap and δ compared to BPSK
0 1 2 3 4 5 6 7 8
Fre que ncy (GHz )
scheme that has one degree of freedom, i.e. Tgap . Also, it is
Figure 9. BER performance simulation of BPSK TH-UWB impulse radio
versus NBI center frequency. seen that using doublet pulse increases multiple access interfer-
ence and decreases achievable data rate because doublet pulse
BP P M
10
0 has larger support than single pulse.
S in g le P u lse
Do u ble t P u lse w ith T gap = 1 n s

10
-1
REFERENCES
[1] A. Taha, Keith M. Chugg, “ A Theoretical Study on the
Bit Error Rate

Effects of Interference on UWB Multiple Access Impulse

Radio” , in Proc. of IEEE conference on Signal and System and
10
-2
Communications, pp. 728-732, 2002
[2] M. Hamalainen, V. Hovinen, J. Inatti, M. Latvaho, “In-
Band Interference Power Caused by Different Kinds of UWB
Signals at UMTS/WCDMA Frequency Bands, ” IEEE Radio and
-3
10
0 1 2 3 4 5 6 7 8
Wireless Conference, pp. 97-100, 2001.
F re qu e ncy (G Hz ) [3] R. A. Scholtz, R. Weaver, E. Homier, J. Lee, P. Hilms, A.
Figure 10. BER performance simulation of BPPM TH-UWB impulse Taha, R. Wilson, “UWB Radio Deployment Challenges,” IEEE
radio versus NBI center frequency. Personal Indoor Mobile Radio Communications, London, U.
K., pp. 620-625, September 2000.
Table 1. Simulation Parameters.
[4] M. Z. Win, R. A. Scholts, “Ultra-Wide Bandwidth Time-
Parameter Value
Hopping Spread Spectrum Impulse Radio for Wireless
Tc 2 ns
Multiple-Access Communications,” IEEE Transactions on
Tf 40 ns Communications, Vol. 48, No. 4, pp. 679-691, April 2000.
Tgap 1ns [5] Ali Taha, Keith M. Chugg, “Multipath Diversity Reception
Nh 20 of Wireless Multiple Access Time-Hopping Digital Impulse
Nu 20 Radio,” IEEE Conference on Ultra Wideband Systems and
Ns 1 Technologies, Baltimore, M.D., May 2002.
Np ∞ [6] Itsik Bergel, Eran Fishler, Hagit Messer, “Narrow
Rs 25Mbps Band Interference suppression in Time-Hopping Impulse-Radio
SIR -20dB systems,” IEEE Conference on Ultra Wideband Systems and
Technologies, Baltimore, M.D., pp. 303-307, May 2002.