Performance Analysis of Impulse Radio

Ultra-Wideband Differential Detection Schemes for

Body Area Networks
Igor Dotli c

and Ryuji Kohno

Medical ICT Institute, National Institute for Information and Communications Technology
3-4, Hikarino-oka, Yokosuka, 239-0847, Japan
Email: {dotlic, kohno}@nict.go.jp
AbstractIn this work differentially coherent communication
schemes for Impulse Radio Ultra-Wideband (IR-UWB) that
transmit single uninterrupted package of energy per symbol,
expected to be included in the IR-UWB IEEE 802.15.6 wireless
Body Area Networks (BAN) standard are considered. Models
for two receiver architectures, suitable for such communication
schemes and feasible to implement in current CMOS technology
- duty-cycled sampling receiver and chirp receiver are derived.
After that, two differential detection techniques for employed
receiver architectures - ordinary DPSK and Sample-wise DPSK
(S-DPSK) are described. Theoretical analysis of probability of
error for S-DPSK is also provided. Performances of the sampling
and the chirp receiver architectures with DPSK and S-DPSK
detection methods employed are analyzed under narrowband and
Frequency-Modulated UWB (FM-UWB) types of interference.
I. INTRODUCTION
Since its Effective Isotropic Radiated Power (EIRP) levels
are limited by regulations to be very low, application of un-
licensed UWB radio is limited to short-range communication
systems, such as Personal Area Networks (PAN) or BAN. IR-
UWB is a class of UWB devices suitable for low to medium
data rate communications in which transmitted symbols con-
sist of pulses of energy. Thanks to the fact that IR-UWB data
rates are typically low compared to the bandwidths used, duty
cycle of IR-UWB communication schemes can be relatively
low. This fact gives an opportunity to a IR-UWB devices to
employ duty-cycling to its Radio Frequency (RF) front-end.
Since transceivers RF front-end is typically responsible for
a large portion of its overall power consumption, employing
RF front-end duty-cycling, i.e. turning off RF front-end for
signicant percentage of communication time gives a potential
for signicant power consumption savings. Thus, IR-UWB
is attractive for BAN, where power consumption and battery
life are critical in many applications, and it is currently being
standardized as one of wireless BAN radio interfaces [1].
Classic and widespread IR-UWB communication scheme
of using isolated short pulses as chips [2], although inherently
having a low duty cycle, is not particularly convenient for
duty-cycling of transceivers RF front-end; it requires turning
RF front-end on and off for each chip transfered and, depend-
ing on a data rate, there can be many chips per symbol. In
contrast, communication scheme employed by IEEE 802.15.4a
IR-UWB standard [3] of transmitting single uninterrupted
package of energy (short-pulse burst) as symbol is much more
convenient for the duty-cycling of transceivers RF front end,
since it requires only a single on-off cycle per symbol.
When coherence of a IR-UWB communication scheme is
considered, we can state that it is true that non-coherent,
energy-detection IR-UWB systems are easier to design and
have lower power consumption than (differentially) coherent.
However, their sensitivity to different kinds of interference
is high. This, together with the fact that number of UWB
channels of 500 MHz bandwidth available worldwide is only
3 [1] limits their usage in signicant number of applications
such as medical BAN in which many networks can be co-
located. Thus, our attention is directed towards coherent and
particularly differentially coherent systems.
Differential phase modulation enables both fully coherent
detection and differential detection. Some modalities of differ-
ential detection allow gathering of a signicant portion, if not
all, of multipath energy available in the channel without chan-
nel estimation. For this reason, differential detection schemes
have been studied in IR-UWB, where channel estimation is
not trivial and can signicantly increase receivers complexity.
However, all of IR-UWB differential communication schemes
studied in available literature are based on isolated short-pulse
chips
1
.
As mentioned above, it is very likely for a IR-UWB system
to operate in the presence of interference and for many appli-
cations, such as medical BAN, performance under interference
can be critical. Some differential detection methods, especially
if no channel estimation is employed, are more sensitive to
interference compared to the fully coherent detection methods.
Therefore, it is of interest to evaluate how well IR-UWB
differential detection schemes perform under interference.
In this work we develop a model of differential detection
for differential communication schemes, included in the draft
of IR-UWB IEEE 802.15.6 wireless BAN standard, that for a
sake of efcient duty cycling of transceivers RF front-end use
single uninterrupted package of energy per symbol. Receiver
architectures that are considered are the sampling receiver
[5], however, with RF front-end duty-cycling employed and
1
Fairly complete list of references can be found in [4].
the chirp receiver
2
[7]. The performance of two receiver
architectures considered are examined in different multipath
channels and under two types of interference: from a narrow-
band system and from another UWB system suitable for low
data-rate communication - FM-UWB [8]. Furthermore, two
differential detection schemes are examined: ordinary DPSK,
which includes the channel estimation and S-DPSK, which
does not require a channel estimation.
II. SYSTEM MODEL
A. Transmission waveforms
There are two transmission waveforms that are considered.
First is linear chirp pulse with carrier frequency f
0
, duration
T
c
and frequency sweep of f
c
, complex envelope of which
can be expressed as
c(t) = exp
_
j2
_
f
0
t +
K
c
2
t
2
__
w(t, T
c
/2, T
c
/2). (1)
Function w(t, t
1
, t
2
) is the rectangular window function with
limits t
1
and t
2
and K
c
is a chirping slope dened as
K
c
=
f
c
T
c
. (2)
In this work we will use f
c
= 550 MHz since it ts the chirp
pulse well under IEEE 802.15.4a narrower spectrum mask of
600 MHz width.
Time-bandwidth (TB) product of any waveform represents
the number of independent complex samples needed to char-
acterize the waveform. For chirp pulse it is expressed as
TB
c
= f
c
T
c
. (3)
The second transmission waveform considered is a short-
pulse burst, used in IEEE 802.15.4a standard [3], complex
envelope of which can be expressed as
b(t) =
B1

i=0
(1)
z
i
p(t iT
p
) exp(j2f
0
t), (4)
where p(t) is the IEEE 802.15.4a mandatory root raised-cosine
pulse waveform [3], with duration T
p
= 2 ns; z
i
{0, 1}, i =
0, . . . , B1 is a binary scrambling sequence of the burst. From
(4) it is clear that duration of the burst T
b
is T
b
= BT
p
.
B. Transmitter
The transmitter uses phase modulated waveforms of the
form (1) or (4) as symbols. The transmitter uses time hop-
ping reduce Multi-User Interference (MUI) in the multi-user
environments and distribute it evenly among users in the
system. Transmitted phase-modulated chirp pulse symbols can
be expressed as
a
k
(t) = exp (j
k
) c (t l(k)T
h
kT
sym
) , < k < .
(5)
2
It was concluded that receiver architectures based on Quadrature Analog
Correlation (QAC) [6] would require too high timing precision to work well
at low data rates with the communication scheme used.
l(k) is a time-hopping sequence with values
l(k) {0, . . . , (T
sym
T
g
) /T
h
1} , (6)
T
g
< T
sym
is the guard interval employed to reduce Inter-
Symbol Interference (ISI) in the system. Transmitted short
pulse burst can be dened analogous to (5) with symbol c
being substituted with b. Transmitted symbols are differen-
tially encoded

k
= (
k
+
k1
) mod 2 for < k < , (7)
where
k
is the information-carrying phase. In this work we
will refer to DBPSK where
k
{0, } and Gray-encoded
DQPSK where
k
{0, /2, , 3/2}.
Regardless of which of the two described waveforms is
transmitted, its duration is related to the data rate used through
the constant duty cycle (DC) as
T
c
= T
b
= DC T
sym
. (8)
The constant duty cycle gives a constant ratio of instantaneous,
i.e. pulse radiated power to average radiated power (EIRP).
Thus, constant pulse power, which is convenient for the
transmitter implementation, results in a constant EIRP, which
is through power spectrum bounded by the regulatory and the
standard spectral masks.
C. Propagation through channel
If a pulse of the form (5) goes through a multipath channel,
the result will be r
k
(t) = a
k
(t) h(t), where denotes
convolution, h(t) is the impulse response of a channel. For
[1], [3] channel models, h(t) is of the form
h(t) =
N

n=1
h
n
(t
n
), (9)
where h
1
, . . . , h
N
are complex multipath coefcients and

1
, . . . ,
N
are corresponding arrival times. Thus, received
signal is of the form
r
k
(t) =
N

n=1
h

n
a
k
(t
n
). (10)
D. Duty-cycled sampling receiver
In the duty-cycled sampling receiver, as shown in Fig. 1(a),
received signal r
k
(t) is mixed with the pulse g
k
(t,
R
),
R
being timing of the receiver, to produce complex baseband
signal v
k
(t,
R
)
v
k
(t,
R
) = r
k
(t)g

k
(t,
R
). (11)
Assuming time hopping sequence synchronization, g
k
(t,
R
)
is given by
g
k
(t,
R
) = d(t kT
sym
l(k)T
h

R
), (12)
with d(t) being carrier pulse of duration T
d
dened as
d(t) = exp(j2f
0
t)w(t, T
d
/2, T
d
/2). (13)
Since duration of the carrier pulse in the receiver T
d
is,
generally, shorter than a duration of the received symbol r
k
(t),
Carrier
generator
X
X
90
ADC
ADC
Data
( )
k
r t
( , )
k R
g t
( , )
Q
k R
v t
( , )
I
k R
t s
( )
Q
k R
s
Digital detection
and timing
generation
( )
I
k R
s
( , )
I
k R
v t
( , )
Q
k R
t s
(a)
Chirp pulse
generator
X
X
90
ADC
ADC
Data
( )
k
r t
( , )
k R
g t
( , )
Q
k R
v t
( , )
I
k R
t s
( )
Q
k R
s
Digital detection
and timing
generation
( )
I
k R
s
( , )
I
k R
v t
( , )
Q
k R
t s
(b)
Fig. 1. Receiver architectures considered. (a) Sampling receiver. (b) Chirp receiver.
the windowing will result in a loss of energy quantied through

m
parameter:

m
=
(E
b
/N
0
)
r
(E
b
/N
0
)
v
=
+
_

|r
k
(t)|
2
dt
+
_

|v
k
(t,
R
)|
2
dt
. (14)
If ISI is neglected and, for the sake of simplicity and without
loss of generality, 0-th symbol is considered, signal after low-
pass ltering can be expressed as
s
0
(t,
R
) = v
0
(t,
R
) h
LP
(t) = exp(j
0
)u(t,
R
), (15)
where h
LP
(t) is the impulse response of the anti-aliasing low-
pass lter. Sampling of s
0
(t,
R
) with sampling frequency
f
s
= 1/T
s
, which is larger or equal to the Nyquist frequency,
will produce a vector of samples
s
0
(
R
) = exp(j
0
)u(
R
), (16)
with u(
R
) being the column vector of samples of u(t,
R
)
u(
R
) = [u(
R
T
d
/2 + mT
s
,
R
)]
M
m=1
, (17)
where M = T
d
/T
s
is the number of samples. If ISI is
neglected, the vector of samples of the k-th received symbol
can be expressed as
s
k
(
R
) = exp(j
k
)u(
R
). (18)
When ISI is taken into account (18) is modied and becomes
s
k
(
R
) =

i
exp(j
i
)u(
R
+ (l(k) l(i)) T
h
+ (k i)T
sym
).(19)
Receivers timing acquisition is done by nding
R
that
maximizes u(
R
):

Ropt
= arg
_
max

R
u(
R
)
_
arg
_
max

R
E {s
k
(
R
)}
_
.
(20)
Low-pass ltering and digitalization will introduce further
SNR loss that can be quantied with a coefcient which is
expressed as a ratio of SNR after mixing (E
b
/N
0
)
v
and SNR
in the digital part of the receiver (E
b
/N
0
)
s

LP
=
(E
b
/N
0
)
v
(E
b
/N
0
)
s
=
+
_

|h
LP
(t)|
2
dt
+
_

|v
k
(t,
R
)|
2
dt
u(
R
)
2
.
(21)
Total loss of the SNR in the receiver will be equal to
=
(E
b
/N
0
)
r
(E
b
/N
0
)
s
=
m

LP
=
+
_

|h
LP
(t)|
2
dt
+
_

|r
k
(t)|
2
dt
u(
R
)
2
.
(22)
E. Chirp receiver
While the duty-cycled sampling receiver can be used for
any type of transmitted waveform, the chirp receiver, shown in
Fig. 1(b), can be used only if the chirp pulse (1) is transmitted.
In contrast with the sampling receiver, received signal is not
mixed with an unmodulated carrier pulse, but with the chirp
pulse generated locally in the receiver. Under assumption T
c
>
2

, where

is a RMS delay spread of the channel, this has

an effect of compressing the signal after mixing v
k
(t,
R
) in
frequency to RMS bandwidth denoted
f
equal to

f
K
c

. (23)
Since duration of the signal is equal to T
c
, TB product of
v
k
(t,
R
) will be
TB
v
f
c

. (24)
Thus, the number of complex samples needed to sample
the signal does not depend on a duration of the transmitted
waveform, if condition T
c
> 2

is satised. This technique

can signicantly reduce the number of the complex samples
needed in the receiver, especially when duration of the trans-
mitted waveform is high.
Apart from compressed spectrum of the signal after mix-
ing v
k
(t,
R
) and thus lower sample rate needed, analytical
description of the chirp receiver is very similar to the one of
the sampling receiver. Indeed, expressions for the signal in the
digital domain of the receiver are of the same form (15)-(19)
and the timing acquisition is done by using the same criteria
(20). Furthermore, loss parameters of the chirp receiver can be
dened in the same way as (14), (21) and (22). More about
the chirp receiver can be found in [7].
F. Digital detection methods
There are two symbol detection methods that are considered.
First is ordinary DPSK detection [9]; symbols samples s
k
are,
rst, approximately match-ltered

k
= u
H
(
R
)s
k
(
R
), (25)
then, information-carrying phase shift is estimated in the
following manner

k
=
_

k1

k
_

d
. (26)

d
is estimation of phase drift coupled with the timing drift in
the receiver. Practically,
d
does not change from symbol to
symbol and can be easily estimated through time averaging,
i.e. via all-digital PLL. This technique requires estimation of
u(
R
). Timing drift complicates nding u(
R
), since it causes
continuous change of u(
R
) vector itself.
Second digital domain detection scheme considered can be
viewed as a simplication of the aforementioned DPSK. In
it, samples of two consecutive symbols are correlated in the
digital domain to produce a phase difference estimate

k
=
_
s
H
k1
(
R
)s
k
(
R
)
_

d
. (27)
We will refer to this detection scheme as Sample-wise Dif-
ferential Phase Shift Keying (S-DPSK), since no other appro-
priate term was found in available literature. S-DPSK does
not require any implementation of u(
R
) estimation, since it
indirectly uses s
k1
(
R
) samples for estimating u(
R
). In-
deed, this makes S-DPSK more practical than DPSK for low-
complexity system implementation. However, the complexity
reduction does not come without a price, since S-DPSK is
suboptimal compared to the DPSK.
We could not nd a theoretical analysis of the probability of
error in the presence of noise for the S-DPSK in the available
literature. However, clear analogy may be observed between
probability density functions of decision variables of the S-
DPSK and multichannel DPSK with Equal Gain Combining
(EGC) [9, Ch. 9]. Indeed, if S-DPSK samples are made with
innite resolution and noise in the digital domain of the
receiver is assumed to be white, the probability of error for
the S-DPSK can be found from the analytical expression for
the multichannel DPSK with EGC [9, Eq. 9.103] by replacing
the number of channels in the multichannel DPSK with the
number of samples in the receiver M and total SNR of
multichannel DPSK with SNR in the digital domain of the
receiver (E
b
/N
0
)
s
. Thus, if ISI is regarded as negligible, the
probability of bit-error for the S-DPSK with ideal sampling in
white noise can be expressed as
P
e
=
1
2
+
1
2
2M1
M

m=1
_
2M m
M m
_

[Q
m
(a

t
, b

t
) Q
m
(b

t
, a

t
)], (28)
with
t
= (E
b
/N
0
)
s
, Q
m
(, ) being generalized Marcum
Q-function of m-th order, while a

and b

are standard
coefcients for DPSK error calculation; for binary S-DPSK (S-
DBPSK) a

= 0, b

2 and for Gray-encoded quaternary

0 2 4 6 8 10 12 14 16 18
10
8
10
7
10
6
10
5
10
4
10
3
10
2
10
1

t
(dB)
P
e


DBPSK
SDBPSK, M=8
SDBPSK, M=16
SDBPSK, M=32
SDBPSK, M=64
SDBPSK, M=128
DQPSK
SDQPSK, M=8
SDQPSK, M=16
SDQPSK, M=32
SDQPSK, M=64
SDQPSK, M=128
Fig. 2. Bit-error probabilities of S-DBPSK and S-DQPSK with different
number of samples (M).
S-DPSK (S-DQPSK) a

=
_
2

2, b

=
_
2 +

2. For
M = 1 (28) gives standard bit-error probability expressions
for DBPSK and DQPSK [9].
It is interesting to note from (28) that S-DPSK bit-error
probability does not depend on a symbol signature shape, but
only on (E
b
/N
0
)
s
and M. Hence, if above assumptions hold,
one can determine DPSK and S-DPSK bit-error performance
of any sampling or chirp receiver by evaluating (22) of that
receiver on a given multipath channel.
The bit-error probabilities curves of S-DBPSK and S-
DQPSK for different number of samples are shown in Fig.
2. It can be observed that, unlike in the ordinary DPSK
detection [9], performance of both S-DPSK detection schemes
considered deteriorates with an increase of the number of
samples. Thus, a receiver that is able to gather a given portion
of multipath energy with a smaller number of samples, apart
from a lower computational complexity, has an advantage in
S-DPSK bit-error rate performance compared to a receiver that
gathers the same portion of multipath energy, i.e. achieves the
same , by using a larger number of samples.
Although DBPSK outperforms DQPSK by almost 2 dB, S-
DQPSK deteriorates slower in performance with the increase
of the number of samples compared to S-DBPSK. Thus, per-
formance of S-DQPSK and S-DBPSK for M = 16 complex
samples is approximately the same for high (E
b
/N
0
)
s
, while
for the number of samples larger than M = 16 S-DQPSK
outperforms S-DBPSK.
III. NUMERICAL EVALUATION
As 490 kbps is currently under consideration for mandatory
raw data rate for BAN [1], it was of our interest to examine
the performance of the two receiver architectures considered
at this data rate. For modulation DBPSK is chosen as less
28 26 24 22 20 18 16 14 12 10 8
10
6
10
5
10
4
10
3
10
2
10
1
SIR (dB)
P
e


1 bit sampling SDBPSK
Ideal sampling SDBPSK
Analytical LS approx. SDBPSK
1 bit sampling DBPSK
Ideal sampling DBPSK
Analytical LS approx. DBPSK
(a)
28 26 24 22 20 18 16 14 12 10 8
10
6
10
5
10
4
10
3
10
2
10
1
SIR (dB)
P
e


1 bit sampling SDBPSK
Ideal sampling SDBPSK
Analytical LS approx. SDBPSK
1 bit sampling DBPSK
Ideal sampling DBPSK
Analytical LS approx. DBPSK
(b)
Fig. 3. Performance of the sampling receiver under interference. (a) Narrowband interference. (b) FM-UWB interference.
28 26 24 22 20 18 16 14 12 10 8
10
6
10
5
10
4
10
3
10
2
10
1
SIR (dB)
P
e


1 bit sampling SDBPSK
Ideal sampling SDBPSK
Analytical LS approx. SDBPSK
1 bit sampling DBPSK
Ideal sampling DBPSK
Analytical LS approx. DBPSK
(a)
28 26 24 22 20 18 16 14 12 10 8
10
6
10
5
10
4
10
3
10
2
10
1
SIR (dB)
P
e


1 bit sampling SDBPSK
Ideal sampling SDBPSK
Analytical LS approx. SDBPSK
1 bit sampling DBPSK
Ideal sampling DBPSK
Analytical LS approx. DBPSK
(b)
Fig. 4. Performance of the chirp receiver under interference. (a) Narrowband interference. (b) FM-UWB interference.
complex to implement on the transmitter side than DQPSK,
thus T
sym
= 2051 ns. The guard interval is chosen to be
T
g
= T
sym
/2. Duration of the transmitted waveform, regard-
less of its type (chirp pulse or short-pulse burst) is dictated by
the duty cycle supported by the system (8). Duty cycle of 3.1
%, corresponding to the mandatory mode of IEEE 802.15.4a
standard [3], is adopted here, since with peak-to-peak voltage
of 600 mV, which is feasible to implement at power supply
voltages under 1 V, and antenna with the standard impedance
of 50 it produces EIRP of approximately -15 dBm, which,
for both types of the transmitted waveform used, ts well un-
der the spectral mask of IEEE 802.15.4a [3]. Hence, durations
of the transmitted waveforms are T
c
= T
b
= 64 ns. While the
sampling receiver detects the short-pulse burst (4), the chirp
receiver detects the long chirp pulse (1).
The chirp receiver and the sampling receiver that are con-
sidered both have the same integration time: T
d
= T
c
= 64 ns.
However, while sampling receiver uses Nyquist sampling
TABLE I
ACHIEVED FOR TWO RECEIVERS ON DIFFERENT CHANNEL MODELS
TOGETHER WITH RMS DELAY SPREAD (

) OF THOSE CHANNEL MODELS.

Channel model
(dB)

Chirp receiver Sampling receiver (ns)

IEEE 802.15.4a CM1 1.8 1.7 16.0
IEEE 802.15.4a CM2 2.1 1.9 18.6
IEEE 802.15.4a CM3 1.1 1.5 10.4
IEEE 802.15.4a CM4 1.7 1.7 12.8
IEEE 802.15.6 CM3 2.3 2.0 19.4
IEEE 802.15.6 CM4-1 3.1 3.0 37.8
frequency f
s
= 500 MHz (M = 32 samples), the bandwidth
of the signal being sampled in the chirp receiver is reduced
compared to the sampling receiver (23) and thus its sampling
frequency is lower f
s
= 250 MHz (M = 16 samples). For
its low implementation complexity, integrate and dump lter
is chosen as anti-aliasing low-pass lter in both receivers
(h
LP
(t) = 1 for 0 t T
s
and 0 otherwise).
When T
g
used and

of [1], [3] channel models, shown

in Table I, is considered, it can be concluded that ISI can be
regarded as negligible. Thus, the performance of the two sys-
tems in question in a multipath channel and white noise can be
deduced from their performance on a given channel. Mean
performances, for different channel models, are displayed in
Table I. Observe that performance of the chirp receiver and
the sampling receiver is very similar. Indeed, as shown in Fig.
2, for the same , if S-DPSK is employed in both receivers,
M = 16 of the chirp receiver gives it 1 dB advantage over
M = 32 of the sampling receiver.
Performances of the two receiver architectures considered
are examined under two types of interference, both having the
same carrier frequency as the victim IR-UWB system. First
interferer considered is a narrowband system with 10 Mbps
BPSK modulation. Second interfering system, chosen as a rep-
resentative of UWB systems, is FM-UWB [8] with 250 kbps
data rate and 550 MHz bandwidth. Performances of the sam-
pling and the chirp receiver under two aforementioned types of
interference with DPSK and S-DPSK detection, examined on
IEEE 802.15.4a CM1 channel model with (E
b
/N
0
)
r
= 30 dB
are shown in Figs. 3 and 4 respectively. In both Figs. ana-
lytical least squares approximation of a interference impact,
customary in the spread-spectrum literature, is also included.
In it, impact of interference is approximated by the impact of a
white noise with approximately the same power in the digital
portion of the receiver, i.e. in (28)
t
= G
p
SIR/, where G
p
is the processing gain of the system, e.g. for the chirp system
G
p
= f
c
T
sym
and SIR is ratio of mean powers of useful
signal and interference at the input of the receiver.
The chirp receiver somewhat outperforms the sampling
receiver under the FM-UWB interference regardless of a
detection technique employed and a sampling resolution used.
The analytical least squares approximation gives a pessimistic
estimation of the interference impact to the receivers perfor-
mances for both chirp and sampling receiver and both DBPSK
or S-DBPSK detection. Usage of 1-bit sampling instead of
sampling with a ideal resolution results in a performance loss
of roughly 6 dB for all cases studied under the FM-UWB
interference.
Situation is signicantly different when the impact of the
narrowband interference is considered. While performance of
the chirp receiver is quite similar to its performance under
the FM-UWB interference, the sampling receiver performs
signicantly poorer under the narrowband interference. For a
ideal sampling and DBPSK detection, the sampling receiver
gives approximately 5 dB inferior performance under the nar-
rowband interference compared to the FM-UWB interference,
while for S-DBPSK detection performance is roughly 3 dB
poorer. While 1-bit sampling deteriorates performance of the
sampling receiver with DBPSK detection approximately 4 dB,
S-DBPSK detection seems to be much more sensitive to the
same reduction in the sampling resolution.
IV. CONCLUSIONS
Performances of the sampling receiver and the chirp receiver
were examined under two types of interference with two dif-
ferential detection techniques employed. The analytical least
squares approximation does not give good predictions of the
interference impact for all cases studied. While performance
of the sampling receiver is very sensitive to the bandwidth of a
interference, performance of the chirp receiver varies little with
the interference bandwidth. This property of the chirp receiver
can be related to the fact that, in contrast with the sampling
receiver, analog portion of the receiver spreads the spectrum
of the interference before low-pass ltering and sampling,
while the spectrum of the useful signal is compressed. Thus,
regardless of which of the two types of the interference
considered is present, properties of the signals being sampled
in the chirp receiver are relatively similar.
REFERENCES
[1] IEEE 802.15 WPAN task group 6 body area networks (BAN).
[Online]. Available: http://www.ieee802.org/15/pub/TG6.html
[2] M. Win and R. Scholtz, Impulse radio: how it works, IEEE Communi-
cations Letters, vol. 2, no. 2, pp. 3638, Feb 1998.
[3] IEEE 802.15 WPAN low rate alternative PHY task group 4a (TG4a).
[Online]. Available: http://www.ieee802.org/15/pub/TG4a.html
[4] M. Pausini, G. Janssen, and K. Witrisal, Performance enhancement of
differential UWB autocorrelation receivers under ISI, IEEE Journal on
Selected Areas in Communications, vol. 24, no. 4, pp. 815 821, april
2006.
[5] F. Lee and A. Chandrakasan, A BiCMOS ultra-wideband 3.1-10.6 GHz
front-end, IEEE Journal of Solid-State Circuits, vol. 41, no. 8, pp. 1784
1791, Aug. 2006.
[6] M. Verhelst, J. Ryckaert, Y. Vanderperren, and W. Dehaene, A low
power, recongurable IR-UWB system, IEEE International Conference
on Communications, 2008. ICC 08., pp. 37703774, May 2008.
[7] I. Dotli c and R. Kohno, Low complexity chirp pulsed ultra-wideband
system with near-optimum multipath performance, IEEE Transactions
on Wireless Communications, accepted for publication.
[8] J. F. M. Gerrits, M. H. L. Kouwenhoven, P. R. van der Meer, J. R.
Farserotu, and J. R. Long, Principles and limitations of ultra-wideband
FM communications systems, EURASIP J. Appl. Signal Process., vol.
2005, no. 1, pp. 382396, 2005.
[9] M. K. Simon and M.-S. Alouini, Digital Communication over Fading
Channels: A Unied Approach to Performance Analysis. John Wiley,
2000.