IEEE International Conference on Ultra-Wideband, Zurich, September 2005

Location accuracy of an UWB localization

system in a multi-path environment
Michael Tchler, Volker Schwarz, Alexander Huber

AbstractIn this paper, we analyze the achievable location

accuracy of a wireless localization system using pulse arrival time
measurements of ultra-wideband (UWB) pulses. We briefly
review the limits on the ranging accuracy given the regulatory
constraints of UWB emissions and the parameters of the UWB
transceiver hardware used for the localization system. We show
that the accuracy of the arrival time measurements does not
depend on the pulse spectrum in a very small measurement
interval equal to half the inverse of the pulse center frequency.
For measuring the arrival time outside this (usually tiny)
measurement interval, we show that the measurement accuracy
declines considerably and that it furthermore depends on the
pulse bandwidth. We also show why short transmit pulses are
necessary in a multi-path radio frequency channel to maintain an
acceptable accuracy of the pulse arrival time.

I. INTRODUCTION
An application utilizing the radio frequency (RF) channel is
commonly called ultra wide-band (UWB), if the covered
bandwidth is >0.5GHz or >20% of the center frequency. Until
recently, only niche applications like radar or military
communications deserved to be called UWB. This is different
today, where numerous companies and universities investigate
applications utilizing the radio frequency band from 3.1GHz
to 10.6GHz, whose usage has passed standardization through
the Federal Communications Commission (FCC) [1] in the
U.S.A. and is soon expected to be standardized by the
European Telecommunications Standards Institute (ETSI) [2]
in Europe.
Besides the established companies with know how in UWB
technology such as TimeDomain [3], the majority of
commercial market players in this area currently focuses on
short-range medium- to high-rate communication applications
for so-called wireless personal area networks (WPAN). A
number of companies already offer chip sets for the high datarate WPAN standard IEEE802.15.3a [4] implementing the
multi-band orthogonal frequency division multiplex (MOFDM) approach such as [5-8]. Impulse radio (IR) utilizing
pulse position modulation (PPM) or chirp spread spectrum
(CSS) modulation are two methods for UWB communication,
which have been approved as candidates for the low data-rate
WPAN standard IEEE802.15.4a [9]. The potential of location
and tracking applications has been discussed by this task
The authors are with the Center for Microelectronics Aargau, University of
Applied Sciences NorthWest Switzerland, Steinackerstr. 1, 5210 Windisch,
Switzerland, e-mail: {m.tuechler, v.schwarz, a.huber}@zma.ch.

group as well, since in particular the IR approach allows to

measure the arrival time of the transmitted pulses precisely.
This information is the basis for calculating the time-of-flight
of the transmitted pulse by utilizing one-way or two-way
ranging algorithms. Finally, the time-of-flight information
allows to locate the devices sending and receiving the UWB
pulses [10].
We currently study integrated circuit (IC) designs in
complementary metal oxide semiconductor (CMOS)
technology at our institute, which implement an UWB
transceiver for transmitting UWB pulses and measuring the
arrival time of UWB pulses with high precision. This IC is
part of a tag, which is subject to precise localization in indoor
and/or outdoor environments. We show analytically and via
simulations that the IC can measure the arrival time of UWB
pulses with an accuracy that is comparable to or even better
than that of current state-of-the-art technology. The system
can be adapted to comply with the FCC or ETSI regulations
for UWB emissions. It can cope with signal interference of
narrow-band communication services in the 3.1-10.6GHz
band such as the wireless local area network (WLAN)
standard IEEE802.11a. The fully digital architecture also
promises low power consumption and flexibility.
In preliminary work [11], we analyzed the localization system
in a multi-path-free environment to study how the imperfect
hardware (time jitter in the samplers, sample quantization,
unsynchronized clocks) affects the accuracy of the arrivaltime measurements. It turned out that neither of these
hardware issues significantly deteriorates this accuracy. In this
paper, we draw our attention on the multi-path nature of the
radio channel indoors (residential and office environment) and
outdoors. Unfortunately, the measurement errors increase
significantly in such an environment, even when there is a
clear open path between the UWB pulse transmitter and
receiver. This scenario is referred to as the line-of-sight (LOS)
condition. We show that the reason for these measurement
errors is that the pulse arrival time estimator in the receiver
cannot discriminate the received UWB pulse from its replica
added in the multi-path RF channel. We also show that if the
so-called LOS component of the received pulse can be
identified, the achieved measurement accuracy is close to that
in a multi-path-free RF channel. Reasons for the obtained
results and possible remedies will be discussed.
The next section continues with the description of the
application for which we develop the UWB transceiver IC.
Section III derives theoretical limits of the time measurement

IEEE International Conference on Ultra-Wideband, Zurich, September 2005

accuracy using the proposed IC design given a multi-path RF
channel. These limits are compared to simulation results in
Section IV. The paper concludes with Section V.

occurring independently in each base station, where Ti is the

time-of-flight of the UWB pulse transmitted by the tag and
received at base station i. To calculate the ranging accuracy FR
from FT, we apply for simplicity the relationship

R 2 .5 c T

II. LOCALIZATION SYSTEM DESCRIPTION

Consider the wireless localization system depicted in Figure 1.
A (large) number of small tags transmit a sequence of UWB
pulses to a set of base stations, which report the measured
arrival time of the pulse train to a processing hub. The hub
calculates the tag coordinates using these arrival times. There
exist numerous approaches to convert pulse arrival times into
the tag coordinates. See for example [12] and [10].

taken from [12] (p. 183), where c is the speed of light. This
formula holds for small FT by directly calculating the tag
coordinate (x,y,z) from the time-of-flight measurements.
Similar studies on FR have been conducted in [15-17], but
without incorporating the UWB regulatory constraints and the
effects of a multi-path RF channel.
CMOS integrated circuit

Base
station

cable

Sampler

Base
station
Antenna

LNA

Sampler

Signal
Correlator processing

Sampler

Tag

TX/RX
switch

Tag
Base
station

Delay lock
loop

Clock

Processing
hub
Base
Tag

station

Figure 1: A wireless localization system.

In this paper, we do not elaborate on the reservation scheme

governing the pulse emissions of the tags and the base stationto-tag and tag-to-tag communication. We focus on the
achievable accuracy of the pulse arrival time measurements in
the base stations.
However, to be able to compare our transceiver proposal with
existing localization systems, we need to relate the achieved
time measurement accuracy to an achievable ranging
accuracy. We define the ranging accuracy of a localization
system as the standard deviation FR of the error

eR = ( x x ) 2 + ( y y ) 2 + ( z z ) 2
between the true coordinates (x,y,z) of a tag and its estimate.
Commercially available wireless localization systems achieve
an accuracy FR down to 10-15cm under LOS conditions in a
50m working range [13] or FR=30cm under LOS conditions in
a working range of 100m [14]. For the applications we have in
mind, a significantly improved accuracy in the cm range under
LOS conditions indoors is required. The desired working
range is at least 20m. Our goal is to determine the ranging
accuracy FR for our system depending on the RF channel
characteristics (multi-path distribution, path loss, noise). Note
that there are numerous other parameters affecting FR such as
the antenna (geometry, aperture, gain, reference point) and the
transceiver circuitry (noise figure, amplifier bandwidth,
sampling rate, sampling time jitter, clock stability). These
circuitry-level parameters as well as some system-level issues
(base station clock synchronization, algorithm for coordinate
calculation, pulse repetition rate, ) are considered in [11].
In the sequel, we obtain lower bounds and simulation results
on the standard deviation FT of the time measurement errors

eT , i = Ti Ti

Ampli.

Pulse
shaper

Control

Figure 2: Building blocks of the UWB transceiver IC.

III. ANALYSIS OF THE TIME MEASUREMENT ACCURACY

Consider the architecture of the transceiver IC depicted in
Figure 2, which is part of the tags as well as the base stations
in the wireless localization system depicted in Figure 1. The
tags transmit a series of the pulse sT,raw(t), which at the i-th
base station becomes a series of received pulses given by

s R (t) = G path sT (t Ti ) + a k sT (t Ti k ) ,
k =1

sT (t) = sT ,raw(t)*hTX (t)*hRX (t),

where Gpath is the path loss, sT(t) is the effective transmit pulse
including the impulse responses of the TX and RX antenna,
respectively, and ak is the coefficient of the k-th multi-path
component of the transmitted pulse arriving at the base station
with delay k relative to the LOS component. We apply a
modified Saleh-Valenzuela channel model [18] to describe the
RF channel, i.e., the distribution of the parameters ak and k.
Figure 3 depicts some channel realizations (channel impulse
responses, CIRs) for this model, which were taken from [9].
For zero gain TX and RX antennas, the path loss Gpath is
approximately given by

Gpath = (c / 4f cg ) 2 d 2 ,
where fcg is the geometric center frequency of the pulse and d
is the distance between the antennas [11].
The received pulse sR(t) is amplified with a CMOS wide-band
low noise amplifier (LNA). State-of-the-art CMOS LNAs for
the frequency band from 3.1GHz to 10.6GHz such as that
proposed in [19] achieve an average power gain Glna of 9dB
and an average noise figure knf of 5.2dB in that band. The total
noise power is knf N0 BR, where BR is the receiver bandwidth

IEEE International Conference on Ultra-Wideband, Zurich, September 2005

and N0=-174dBm/Hz.
CM1 model: LOS, residential environment

where IF is the Fisher information [20]. The analysis in [11]

shows that this bound evaluates as follows for transmission
over a multi-path-free RF channel, i.e., all ak are zero:
no quantization, no sample time jitter:
T >

CM3 model: LOS, office environment

d
P

k nf N 0
2

,
2
c G lna E T

no quantization, with sample time jitter:

k nf N 0
2
2
d

+ J ,
T >
2
Z
c G lna E T
P
1Bit quantization, with sample time jitter:
2
2 k nf N 0
d

2
+ J,
T >
2 c Glna ET
Z
P

CM5 model: LOS, outdoors

where ET is the energy of sT(t), i.e.,

+

ET =

(t ) 2 dt

Figure 3: Channel impulse responses of UWB RF channels (CM1:

LOS, residential; CM3: LOS, office; CM5: LOS, outdoors) taken
from the IEEE802.15.4a task group database at [9] (documents
04/688r0, 04/662r0, 04/714r0, and 04/671r0).

A bank of S parallel samplers in the receiver is driven via a

delay lock loop (DLL) from a clock with period Tclk, i.e., the
effective sampling time is Ts=Tclk/S. The samplers yield 1Bitquantized samples given by

sn = q Glna Gpath sR (nTs + jn ) + wn ,

where q(x)=+1 if x0 and q(x)=-1 if x<0, jn is a Gaussiandistributed sample time jitter independent over the index n
with zero mean and standard deviation FJ, and wn is a
Gaussian-distributed noise sample with zero mean and
variance FW2. We set the jitter parameter FJ to 10ps for the
chosen IC technology [11] and the noise variance FW2 to
knfN0/(2Ts), i.e., the receiver bandwidth BR is 1/(2Ts).
Assume now that the transmitter generates a series of P pulses
sT(t), which are spaced apart by the pulse repetition period Tp.
The samples sn are accumulated in the R=Tp/Ts registers rk of
the correlator, whose contents are given by

rk = sk +nR .
n =0

The registers are red out for further processing if |rk| for some
k exceeds a predefined threshold. In this case, the base station
assumes that a pulse has been received.
The smallest possible standard deviation FT of the time
measurement error given the data set {rk} with any type of
signal processing is determined by the Cramer-Rao bound

T I F 1 ,

and Z is the number of samples in which the transmit pulse

sT(t) is non-zero. The time measurement accuracy does not
depend on the actual pulse shape nor the pulse center
frequency or the pulse bandwidth. It is a function of the
available signal-to-noise ratio (SNR) ET/N0, the receiver
parameters knf and Glna, and the distance d between the TX and
the RX antenna, only. This result seems to contradict the
rule-of-the-thumb in high resolution RADAR applications
that FT declines proportionally to the pulse bandwidth [20].
This rule holds of course as well for the wireless localization
system considered here, but the decline FT due to an
increasing pulse bandwidth (exact: quadratic center frequency
of the pulse spectrum [11]) is compensated for by a rise due to
the path loss Gpath, which declines proportionally to the square
of the pulse bandwidth (exact: geometric center frequency of
the pulse spectrum [11]). Thus, the pulse arrival time
measurement accuracy of the wireless localization system is
determined by the regulatory limit on the allowed transmit
pulse energy ET and some receiver circuit parameters, only.
Figure 4 depicts lower bounds on the standard deviation FT of
the time measurement error for the three receiver circuit
models defined above (with or without sampling time jitter
and 1Bit quantization). The bounds are shown as function of
the distance d between the TX and RX antenna. The FCC
regulation on UWB emissions [1] restricts the effective
isotropically radiated power (EIRP) via the power spectral
densities -34dBm/MHz (peak) and -41.3dBm/MHz (average),
which correspond to a transmit power of 3mW (peak) and
0.56mW (average) when the entire spectrum from 3.1 to
10.6GHz is covered up to the EIRP limit. The chosen peak
power of 0.8mW in Figure 4 attains 92% of the peak power
limit at -34dBm/MHz.
From Figure 4 follows that a localization system using our
transceiver IC can achieve a ranging error standard deviation

IEEE International Conference on Ultra-Wideband, Zurich, September 2005

FR of at most 2.5cFT=2.5c0.5ps=0.4mm at d=50m distance

between the RX and the TX antenna. This accuracy is by far
sufficient for most application scenarios we have in mind (we
require a cm accuracy) and it is most likely that other system
aspects such as the synchronization between the base stations
and the relative antenna position cause much larger systematic
ranging errors. Obviously, the time jitter and the 1-Bit
quantization affect FT very little for the chosen IC parameters.

against the geometric center fcg frequency of the pulse

spectrum (linear rise of FT) as it was done in [11]. Instead, the
quadratic center frequency fcq,e of the spectrum of the
envelope of the transmit pulse sT(t) must be considered, which
yields
f
T ' = cq T .
f cq ,e

TX
RX

Ti

Ti

Figure 5: Time-of-flight Ti of UWB pulses (solid line) between the

TX and the RX antenna. The transmit pulses are cosine waves
modulated with a Gaussian envelope. The bandwidth of the envelope
function on the left is three times the bandwidth of the envelope
function on the right side. It follows that the arrival time of the pulse
on the right side is harder to detect than that on the left.
Figure 4: Lower bounds on the standard deviation FT of the time
measurement error while measuring the arrival time of UWB pulses
transmitted over a multi-path-free RF channel. The analysis takes
parameters of the receiver circuit (sample time jitter, 1-Bit quantizat.)
into account. The parameters of the transmit pulse sT(t) are:
pulse shape: Gaussian, modulated with a cosine wave at frequency
fc=4GHz, pulse bandwidth: B=2GHz, peak transmit power:
0.8mW, pulses per burst: P=500, pulse repetition time: Tp=200ns,
transmit burst length: PTp=100s, circuit parameters: knf=3.3,
Tclk=1GHz, S=10, Glna=8, and FJ =10ps.

The results of Figure 4 should be interpreted carefully.

Consider Figure 5, which depicts how the example transmit
pulse sT(t) from Figure 4 arrives at the RX antenna at time Ti.
The Cramer-Rao analysis on the standard deviation FT of the
time measurement error derived in [11] does not recognize the
periodic nature of the transmit pulse sT(t), i.e., the bounds in
Figure 4 do not vary significantly with increasing length of the
pulse (or decreasing pulse bandwidth B, respectively) as long
as the pulse energy ET and the center frequency fc are constant.
Consider the extremal case that the transmit pulse sT(t) has
zero bandwidth, i.e., sT(t) is a cosine wave at frequency fc. In
this case, the estimation algorithm for the pulse arrival time in
the receiver is able to measure Ti with standard deviation FT
within the time interval [0,1/(2fc)], only, since the received
signal is periodic with period 1/fc and the sign of the received
pulse need not coincide with that of the transmitted pulse.
This period in time is way too small for this range estimation
application. For example, given the pulse defined in Figure 4
with center frequency fc=4GHz, we can measure the pulse
arrival time in the interval [0,125ps], which corresponds to a
maximum TX-RX distance of only c125ps=3.75cm. In order
that the arrival time Ti can be measured in the interval [0,), a
transmit pulse sT(t) with non-zero bandwidth B has to be used.
Clearly, a second Cramer-Rao analysis on the standard
deviation FT of the error of measuring Ti in the interval [0,)
is required. This time, we cannot trade the quadratic center
frequency fcq of the pulse spectrum (linear decline of FT)

Thus, as shown in Figure 5, the smaller fcq,e, the wider the

transmit pulse becomes and the harder it is to measure the
arrival time Ti outside the interval [0,1/(2fc)]. For the example
pulse in Figure 4, we find that fcq=4.02GHz, fcg=3.86GHz, and
fcq,e=503MHz holds [11]. It follows that the standard deviation
FT of measuring Ti outside the interval [0,1/(2fc)] is fcq/fcq,e=8
times the standard deviation FT of measuring Ti inside the
interval [0,1/(2fc)]. As depicted in Figure 4, this results in a
ranging error standard deviation FR of at most
2.5cFT=2.5c80.5ps=3mm at d=50m distance between the
RX and TX antenna (considering time jitter and 1-Bit
quantization). Thus, the wireless localization system should
use a transmit pulse sT(t) with large bandwidth to improve the
accuracy of measuring the pulse arrival time in the receiver.
Finally, we also have to take the multi-path effects of the RF
channel into account. The effects of multi-path propagation on
the accuracy on time of arrival (TOA) measurements were
analyzed in [21;22]. They also derive algorithms for
estimating the channel impulse response (CIR) to improve the
measurement accuracy, which is in our setup a function of the
coefficients ak and the delays k,. We attempt to estimate the
arrival time of the transmitted pulse sT(t) by considering the
LOS component sT(t-Ti), only. As shown in Figure 3, the
received signal contains many multi-path copies of sT(t-Ti)
delayed by k and attenuated by ak, which are possibly
stronger than the LOS path sT(t-Ti). Using an appropriate
threshold on the register contents |rk| of the correlator in the
receiver IC, we aim to discriminate the LOS path sT(t-Ti) from
the multi-path components. Clearly, a small threshold yields a
large probability of detecting the LOS path, but this increases
as well the probability of false alarm, i.e., some noisy samples
rk of the correlator are mistaken as received pulse.
At the end of this section, we state a final argument for
choosing a short UWB pulse as transmit pulse sT(t) in the
wireless localization system. The longer sT(t) becomes in time
(with or without having a large bandwidth), the more likely it

IEEE International Conference on Ultra-Wideband, Zurich, September 2005

is that sT(t-Ti) mixes with its multi-path versions aksT(t-Ti-k).
It follows that with the same data set {rk}, many more
variables than just Ti have to be estimated (coefficients ak and
delays k). Clearly, a third Cramer-Rao analysis on the
standard deviation FT of the error of measuring Ti is required
given that the delays k and the coefficients ak must be
estimated as well. From the Cramer-Rao lower bound on the
estimation error standard deviation for a vector of parameters
[20] follows that FT increases sharply compared to FT if the
data set {rk} cannot be split into subsets of samples depending
on Ti OR k, respectively. This split is possible only if the LOS
path sT(t-Ti) is separated in time from the multi-path versions
aksT(t-Ti-k).

only those measurements from Figure 7 for which the LOS

component in the received pulse could be found correctly. In
this case, the ranging error standard deviation
FR=2.5c8.5ps=6.4mm at the distance d=20m is again close to
the Cramer-Rao bound of 2.5cFT=2.5c80.4ps=2.4mm.

IV. RESULTS
From Figure 4 and the analysis in the previous section
followed that a wireless localization system achieves a
ranging error standard deviation FR of at most
2.5cFT=2.5c80.5ps=3mm at the distance d=50m between
the TX and RX antenna given the system and pulse
parameters defined in Figure 4 and a multi-path free RF
channel. This bound can be compared to the simulation results
in Figure 6. For determining the arrival time Ti, maximum
likelihood estimation was applied [20]. To synchronize the tag
and the base station clock, the tags transmit an initial
synchronization sequence, which is used by the base station to
reduce the clock skew (chosen uniformly distributed in the
range 20ppm) down to <1ppm. The simulation results show
that the time measurement error standard deviation is very
close to the bound FT in a working range of up to d=20m. For
example, from Figure 4 and the ratio fcq/fcq,e=8 follows that
FT=80.4ps=3.2ps holds, which coincides with the simulation
result in Figure 6. For larger distances d, our synchronization
procedure starts to become unreliable causing large estimation
errors. Ongoing work includes improving the synchronization
algorithm applied so far in order that the working range of the
system (which strongly depends on the number P of
transmitted pulses) increases. The slight increase of the error
for small distances d in Figure 6 is due to the applied 1Bit
quantization, since the arrival time estimation algorithm
requires a very low SNR to work properly. This does not hold
for small distances d.
The simulation results in Figure 7 were obtained in a
residential indoor environment. Even though the
corresponding channel model CM1 taken from [9] is defined
for the distance range 0<d<4m, we applied it for 0<d<20m in
order that the simulation results focus on the dependency of
the decaying SNR with increasing d. Given this multi-path RF
channel model, the increase in the measurement error is
tremendous. The ranging error standard deviation at d=20m is
now in the meter range, since FR=2.5c12ns=9m holds.
However, a closer look at the histograms of the estimation
error eT,I reveals that the loss in measurement accuracy is
because the LOS component has been missed by the estimator
for some particular channel realizations. In Figure 8 are shown

Figure 6: Time measurement error standard deviation for determining

the arrival time of UWB pulses transmitted over a multi-path-free RF
channel, including sampling time jitter and 1-Bit quantization in the
receiver circuit. Also shown is the actual histogram of the
measurement error eT,i over 100 measurements for some distances d.

V. CONCLUSIONS
The achievable time measurement accuracy of our UWB
transceiver design is much lower than that of commercially
available localization systems [13;14] or that obtained in
related work on UWB localization systems [21;22]. However,
we focused on the accuracy of a single pulse transmission
experiment and did not include other system aspects such as
the synchronization of the base stations time bases into the
error analysis. We showed that the achievable arrival time
measurement accuracy does not depend on the pulse center
frequency fc or bandwidth B in a measurement interval equal
to [0,1/(2fc)], which corresponds to a few cm if the considered
UWB pulses have their fc in the UWB frequency band from
3.1 to 10.6GHz. In the measurement interval [0,), the time
measurement accuracy declines approximately by the ratio of
the pulse bandwidth B and the center frequency fc. In order to
keep this decline short, a pulse with large bandwidth B must
be used. We also showed why short transmit pulses are
necessary in a multi-path RF channel to maintain the arrival
time measurement accuracy. Our simulation results revealed
the problem of determining the LOS component of the
received UWB pulse in a multi-path RF channel. Ongoing
work should therefore focus on how to improve the
probability that the LOS components of the UWB pulses
transmitted by the tags can be detected. This is hard to achieve
with a minimal configuration of tags and base stations, i.e., the
localization system must offer redundant base stations to
obtain further measurement information, e.g., through a sensor
network of tags, which are able to transmit and receive UWB
pulses.

IEEE International Conference on Ultra-Wideband, Zurich, September 2005

Figure 7: Time measurement error standard deviation for determining

the arrival time of UWB pulses transmitted over a multi-path RF
channel. The chosen channel model is the CM1 model (LOS,
residential environment) from [9] (document 04/688r0). Also shown
is the actual histogram of the measurement error eT,i over 100
measurements for some distances d.

Figure 8: Time measurement error standard deviation for determining

the arrival time of UWB pulses transmitted over a multi-path RF
channel. The chosen channel model is the CM1 model (LOS,
residential environment) from [9] (document 04/688r0). Compared to
Figure 7, the standard deviation FT was calculated by considering the
measurements in the main peak of the histograms in Figure 7, only.
The width of this peak is 100ps.

[5]

General Atomics, "General Atomics UWB Solutions",

http://photonics.ga.com/uwb/index.html, Last update: 2004

[6]

PulseLINK, Inc., "PulseLINK", www.pulse-link.net, Last update:

2004

[7]

Staccato Communications, "Staccato Communications Wireless

USB", www.staccatocommunications.com, Last update: 2004

[8]

Wisair, "Wisair UWB chip sets", www.wisair.com, Last update: 2004

[9]

Institute for Electronics and Electrical Engineers, "IEEE 802.15

WPAN Low Rate Alternative PHY Task Group 4a (TG4a)",
www.ieee802.org/15/pub/TG4a.html, Last update: 31-3-2005

[10]

Roberts, R., "Ranging Subcommittee Final Report: 04/581r7",

www.ieee802.org/15/pub/TG4a.html, Last update: 30-6-2005

[11]

Tchler, M., Schwarz, V., and Huber, A. "Accuracy of an UWB

localization system based on a CMOS chip". pp. 211-220. 17-3-2005.
Aachen, Shaker Verlag. Proceedings of the 2nd Workshop on
Positioning, Navigation, and Communication.

[12]

Oppermann, I., Hmlinen, M., and Iinatti, J., UWB Theory and
Applications Wiley, 2004.

[13]

Ubisense Ltd., "Ubisense - The Smart Space Company",

http://www.ubisense.net, Last update: 13-12-2004

[14]

Multispectral Solutions Inc, "Multispectral Solutions Inc.- A

Tradition of Excellence in Innovation",
http://www.multispectral.com, Last update: 10-12-2004

[15]

Chung, W. C. and Ha, D. S. "An Accurate Ultra Wideband (UWB)

Ranging for Precision Asset Location". pp. 389-393. 2003. Ultra
Wideband Systems and Technologies, 2003 IEEE Conference on. 1611-2003.

[16]

Catovic, A. and Sahinoglu, Z. "The Cramer-Rao Bounds of Hybrid

TOA/RSS and TDOA/RSS Location Estimation Schemes". Vol. TR2003-143. 1-1-2004. Mitsubishi Electric Research Laboratory.

[17]

Lovelace, W. M. and Townsend, J. K. "The effects of timing jitter on

the performance of impulse radio". pp. 251-254. 2002. Ultra
Wideband Systems and Technologies, 2002. Digest of Papers. 2002
IEEE Conference on.

[18]

Siwiak, K. and McKeown, D., Ultra-Wideband radio technology

Wiley, 2004.

[19]

Bevilacqua, A. and Niknejad, A. "An Ultra-Wideband CMAS LNA

for 3.1 to 10.6 GHz Wireless Receivers". pp. 382-383. 2004. IEEE
International Solid-State circuits conference. 2004.

[20]

Kay, S., Fundamentals of Statistical Signal Processing, Estimation

Theory Prentice Hall, 1993.

[21]

Lee, J.-Y. and Scholtz, R. A., "Ranging in a dense multipath

environment using an UWB radio link," Selected Areas in
Communications, IEEE Journal on, vol. 20, no. 9, pp. 1677-1683,
2002.

[22]

Li, X. and Pahlavan, K., "Super-resolution TOA estimation with

diversity for indoor geolocation," Wireless Communications, IEEE
Transactions on, vol. 3, no. 1, pp. 224-234, Jan.2004.

VI. REFERENCES
[1]

FCC, F. C. C. "Ultra-Wideband Operations FCC Report and Order".

Vol. US 47 CFR Part15. 22-4-2002. Federal Communications
Commission.

[2]

Directorate General of the European Commission. "ETSI,

Harmonized Standards Covering Ultrawide Band (UWB)
Applications". Vol. DG ENTR/G/3M/329. 25-2-2003.

[3]

Time domain Corporation, "Time domain - The Pulse of the Future",

http://www.timedomain.com, Last update: 16-12-2004

[4]

Institute for Electronics and Electrical Engineers (IEEE), "IEEE

802.15 WPAN High Rate Alternative PHY Task Group 3a (TG3a)",
www.ieee802.org/15/pub/TG3a.html, Last update: 2004