A NEW SIGNAL DETECTION METHOD FOR TR-UWB

By Time Delayed Sampling & Correlation (TDSC)

Muriel Muller, Yang Ni, Roger Lamberti, Charbel Saber

GET-Institut National des Télécommunications

9, rue Charles Fourier, 91011 Evry France,

Keywords: Ultra-Wideband (UWB) communications, Transmitted Reference (TR) UWB, non-coherent detection.

Abstract: This paper introduces a new signal detection method for Low cost, Low power and Low complexity (L3)

TR-UWB systems for medium to low data rate applications such as sensors networks. This new detection

method is based on a time_to_space conversion realizable by an analog waveform sampler. This method

overcomes the major difficulties in a traditional TR-UWB detection methods based on wide band delay

lines. Finally the relaxed timing precision needed in symbol synchronization contributes further to lower the

system power consumption. This concept has been validated by simulation with real data from experimental

setup. The results will be presented and compared also with other solutions.

1 INTRODUCTION

Many applications (Metha, 2004), (Oppermann,

2004), (Duo, 2004) require low cost, low power and

low complexity (L3) short range wireless

communications means. Most of the current wireless

communications systems are designed for the data

transmission in a computer related context and do

not satisfy completely these criteria. Our motivation

in this research activity is to design wireless

communications devices with L3 characteristics

based on Impulse Radio (IR) UWB concept. The

apparent hardware simplicity of an UWB system,

hides a lot of design challenges (Bettayed, 2003).

The high switching speed for the short duration

UWB signal generation and the high timing

resolution for UWB signal detection and

synchronization between transmission and reception

generate a considerable extra hardware complexity

associated with very important power consumption

overhead.

In an impulse radio based UWB system, the pulse

generation can be implemented by using fast

switching devices such step recovery diode (SRD)

with a reasonable power consumption and

complexity. This is not the case for UWB pulse

detection.

Two main UWB detection methods can be

distinguished.

The coherent detection method, based on a signal

correlation between the incoming UWB signal and a

local generated template (Time Domain, 2001),

(Mielczarek, 2003), is requiring channel estimation.

The other one is based on non-coherent energy

detection method (Doré,2005), (Stoica, 2005).

Theoretically speaking the coherent method gives

better detection result than that of the non-coherent

one.

But in reality, the high precision synchronization

and local template generation in a multi-path

environment are extremely difficult to be

implemented with simple hardware.

Many UWB receivers use some number of analog

correlators to collect the signal energy in front-end

rake receiver architecture. The need to capture a

large amount of the transmitted energy involves to

use a great number of paths, besides the propagation

which deforms the pulses shape, path to path, leads

to a very high complexity both on hardware and

software (estimation channel) (Durisi, 2004).

Some non-coherent detection based simple systems

use signal energy detection which consists of

measuring the energy in the incoming signal. The

detection efficiency of this method is conditioned by

the degree of the temporal energy concentration of

Muller M., Ni Y., Lamberti R. and Saber C. (2006).

A NEW SIGNAL DETECTION METHOD FOR TR-UWB - By Time Delayed Sampling & Correlation (TDSC).

In Proceedings of the International Conference on Wireless Information Networks and Systems, pages 19-25

 SciTePress

an UWB signal. This detection method is usually

implemented by using fast switching devices such as

Schottky diodes or tunnel-effect diodes.

In spite of obvious hardware simplicity, this type of

methods presents several limitations. The necessary

low-pass post-filtering decreases considerably the

amplitude of the signal at the output of the detector,

spreads out this same signal over the axis of time.

This leads to a much lower temporal discrimination

power than that of initial UWB impulses and results

in an unacceptably low sensitivity. Fig. 1 gives the

comparison between the instant signal power and the

low pass filtered signal. We can observe an

important loss of signal amplitude due to the energy

dilution in time.

Figure 1: Energy detection systems.

From these observations, we think that the

Transmitted-Reference (TR) UWB architecture

(Rushforth, 1964), (Hoctor, 2002) could be a

solution to our L3 UWB systems for several reasons.

Firstly in TR-UWB system, each transmitted signal

pulse is preceded or followed by a reference pulse.

Due to the short time duration between these pulses,

the propagation channel can be considered as

constant and the two pulses will keep a strong

correlation at the reception point even in a strong

multi-path environment. So the TR-UWB signal can

be detected by using a pseudo coherent scheme

where the TR-UWB signal is correlated with a time

delayed copy of this signal as shown in Fig. 2.

Secondly the time delay between the signal pulse

and the reference pulse can be seen as a parameter of

signal diversification for channel coding and

multiplexing. Besides TR-UWB can be as efficient

as Rake-receiver for energy collection (Zasowski,

2004), but with a much simpler hardware.

Despite the advantages of TR-UWB, the major

difficulty in the realization is the broadband delay

line with wide bandwidth, highly linear phase,

perfect impedance adaptation and highly stable

delay (Goeckel, 2005). An implementation using

transmission line presents several problems,

especially the delay precision and stability. The lack

of real-time programmability is another limitation in

a real exploitation of TR-UWB. Moreover, a real

transmission line presenting an exploitable delay

value remains impossible to integrate in a miniature

circuit simply because of its physical dimension.

Figure 2 : Simplified TR-UWB communication system.

In this article we proposed a very different approach

based on a time-space conversion using an analog

sampler. The TR-UWB signal is sampled by two

analog waveform samplers with pre-defined delay

which matches the time delay between the pulses in

a TR-UWB frame. The signal detection will be done

by applying a waveform correlation between the two

TR-UWB signal samples. We called this new

concept for UWB data transmission, the Time

Delayed Sampling and Correlation (TDSC) which

will be detailed in section II. In section III, we will

report the first experimental validation results made

in different propagation conditions. These results

demonstrate the potential advantages of this TDSC

concept in L3 UWB communications systems.

Finally in section IV, we will give some conclusions

and perspectives in this research works.

2 TDSC - A NEW APPROACH

FOR TR-UWB SIGNAL

DETECTION

Recently, other fields such as the physics of high

energy particles use high speed waveform samplers

in order to capture and record highly impulsive

signals from different detectors. These signals have

WINSYS 2006 - INTERNATIONAL CONFERENCE ON WIRELESS INFORMATION NETWORKS AND SYSTEMS

the similar temporal and electrical properties as

those of UWB. The most significant work is done by

(Kleinfelder, 2003) in CMOS technology. The

realized circuit can capture several hundred points of

a signal at 10 GHz sampling rate.

Our proposed TDSC detection scheme is highly

inspired from this work. The analog waveform

sampler as shown in Fig. 4 uses an asynchronous

delay line composed of simple inverters to generate

the sampling commands at the different moments.

This asynchronous implementation permits a

sampling rate much higher with much lower power

consumption than that of a synchronous design by

using a global and explicit clock.

By using two waveform samplers as shown in Fig. 3,

one is activated at the instant T and the other at

T+D, two time delayed and sampled waveforms of a

TR-UWB signal can be obtained. If a TR-UWB

signal pulses fall in these sampling windows, we

will have two similar waveforms because the double

pulses in a TR-UWB frame have the same waveform

distortion after the propagation in a time invariant

channel as explained in section I. The absence of

TR-UWB signal will result in two totally

independent waveforms from the samplers. By

consequent, the TR-UWB signal detection here can

be easily done by using a simple correlation

operation between the time delayed sampled

waveforms. This principle is called Time Delayed

Sampling and Correlation. Fig. 5 gives a graphical

representation of this detection concept.

By using this method, we can see that the TR-UWB

will be sampled twice with a constant time delay.

The role of the broadband delay line is replaced by a

time delayed double sampling which removes

completely the need of an explicit analog signal

delay line and all the design difficulties associated

with. In this case, the delay is only applied to the

sampler’s command digital signals, which can be

generated easily and programmably in real-time

from the system clock with an extremely high

precision and high stability. So the parameter D in

TR-UWB can be used for channel coding and

multiplexing. Fig. 3 gives the overall structure of

TDSC (Ni, 2005).

Figure 3 : General scheme of Time Delayed Sample &

Correlation system, based on two samplers temporally

delayed.

Figure 4 : The principle of an asynchronous delay line

based analog waveform sampler.

The TDSC detection cannot be done continuously in

time. So a synchronisation between the TDSC

operation and the incoming TR-UWB signal is

needed. But in contrast to other detection methods

(Time Domain, 2001), TDSC method imposes much

lower constraint on the precision of this

synchronisation. As shown in Fig. 5, when the UWB

pulses fall inside the sampling windows, it can be

detected reliably.

This means that the timing precision for this

synchronisation is indexed at the TR-UWB frame

duration which is much larger than that for pulse

detection.

Input

signal

τ τ

Acquisition

SamplerA

SamplerB

ArrayCorrelator

TR‐UWB

Signal

Output

signal

Acquisition

signal

….

A NEW SIGNAL DETECTION METHOD FOR TR-UWB - By Time Delayed Sampling & Correlation (TDSC)

Figure 5: Principle of TDSC method. a) Waveforms of the

time delayed sampled TR-UWB signals captured by the

samplers A and B. b) The detection of impulse doublet is

obtained with a simple correlation operation of the two

delayed sampled waveforms.

This low timing precision necessary in detection

synchronisation reduces not only considerably the

signal acquisition time but also the complexity of

synchronisation tracking. This characteristic is

particularly interesting for sensor networks like

applications where the data transmission is sporadic

with very low duty cycle. The fast acquisition and

tracking can reduce significantly the system power

consumption by reducing the activation time which

is often conditioned by acquisition time for small

data packages.

3 EXPERIMENTAL VALIDATION

TDSC detection method has been validated by using

simulation on the data captured from an

experimental setup. This experimental setup gives

realistic input signals for the TDSC simulation

program in order to be as close as possible to the real

TR-UWB working conditions. The results of the

validation give some design guide to further VLSI

implementation.

3.1 Experimentation Setup

The experimentation setup is shown in Fig. 6. We

used a off the shelf pulse generator in order to

generate a rectangular pulses train as shown in Fig.

7. The symmetric rising and falling edge transition

time is less than 5ns (Pulse Generator E-H research

laboratories model 137A), and we set the rectangular

pulse width to 100 ns.

A simple thick monopole antenna has been used for

this experimentation. The inevitable ringing in this

kind of antenna permits to test the tolerance of

TDSC detection method vis-à-vis to the impedance

adaptation problems in real implementation.

Monopole antennas will reshape the digital pulses

into wide band limited pulses. So the rising and

falling edges of each digital pulse will generate an

impulse doublet with opposite polarities and a

temporal delay D corresponding to the digital pulse

width. Then, the impulse doublets are sent,

propagated in an indoor environment and received

by a digital oscilloscope with memory of 300MHz

bandwidth. An example of the captured TR-UWB

signal is shown in Fig. 8. Finally, we use the Matlab

software to simulate the TDSC detection method on

these sampled TR-UWB signals.

Figure 6: Experimentation setup: A off the shelf pulse

generator, two monopole antennas and a digital

oscilloscope as sampler, were used.

Figure 7: Rectangular input signal. The temporal delay D

between the two pulses of the doublet is around 100ns.

Two doublets will be separated by 300ns. A silence space

between two couple of doublets is equal to 500ns.

0,E+00 5,E-07 1,E-06 2,E-06 2,E-06 3,E-06

Time (second)

Sampler A

Sampler B

Delay D between A & B

Correlation between A & B

Waveforms of the Time Delayed Sampled TR-UWB signal

Signal detection by the Correlation between the Time Delayed

Sam

les

Detection of

impulse doublet

TTL/ CMOS

pulse

Mono

ole

Digital

Oscilloscope

Mono

ole

WINSYS 2006 - INTERNATIONAL CONFERENCE ON WIRELESS INFORMATION NETWORKS AND SYSTEMS

Figure 8: Real output doublets sampled and recorded

thanks to a digital oscilloscope. The reference pulse and

the data pulse are separated by a temporal delay D.

3.2 Experimentation Results

In the TDSC principle, the TR-UWB signal

detection is done by correlating two temporally

delayed sampled TR-UWB signal. This is simulated

by applying cross-correlation function to two

temporal windows on the recorded TR-UWB signal.

These temporal windows correspond to the double

time delayed waveform samplers. In order to skip at

first the problem of synchronisation, these windows

will slide on the recorded TR-UWB signal in a

contiguous way, as shown in Fig. 9-a. In the same

figure (Fig. 9-b), the result of the simulated TDSC

detection method with a TR-UWB signal sampled at

5Ghz rate (200ps). It is demonstrated that the basic

cross correlation function gives a satisfied results

and the virtually null output during TR-UWB signal

absence makes the decision threshold setting very

easy.

In order to evaluate the performance of TDSC

method, different configurations and parameters

have been tested.

Figure 9: a) Principle of the algorithm used to validate the

TDSC concept. One sampling window is sliding on

another window about a fixed temporal delay equal to D.

b) Correlation result between the two sampled signals.

Here the sampling rate was equal to 0,2ns.

In the above example, the receiving antenna was

placed 250 cm away from the transmitting antenna

with the same height.

Many others measurements were carried out where

the distance varied from 80 cm to 5 meters, with

constant emission power. The sampling rate also

varied from 200ps to 2ns. For all these cases, the

TDSC method gave very good detection

performance. Note that above a 2ns sampling rate,

the TDSC detection failed. This is consistent with

Shannon theorem.

These results are very promising for the TDSC

concept and the future realization circuits. Based on

these promising results we decided to investigate

others properties of TDSC method.

3.2.1 Synchronisation Tolerance and

Temporal Discrimination

The high speed impulses used in IR-UWB need a

high precision temporal synchronisation between the

transmitter and the receiver. In a classic IR-UWB

system, this temporal synchronisation precision

should be higher than the minimum pulse width.

This requirement represents one of the major

challenges in a real hardware implementation.

In the TDSC method, the detection can be effective

when the impulse doublet falls inside the temporal

sampling windows. This means that the temporal

synchronisation precision is indexed to the symbol

rate but not that of individual impulses. The

maximum width of the sampling window is to the

delay D and this gives a large synchronisation

facility and tolerance.

But another interesting point here is that this large

synchronisation tolerance has no impact on the

temporal discrimination power of the TDSC

detection because the temporal discrimination power

of the TDSC method is conditioned by the cross-

correlation function applied onto the sampled

signals. This high temporal discrimination capability

is illustrated by the following aspect: - the separation

between two pulses inside a TR-UWB doublet.

In order to code one data bit, two pulses are

transmitted, separated by a known delay D. This

delay can be seen as selectivity parameter (similar to

frequency in a frequency division system) at the

reception. A high selectivity based on this parameter

can not only reduce noise but also give higher

channel capacity for coding and multi-access.

A NEW SIGNAL DETECTION METHOD FOR TR-UWB - By Time Delayed Sampling & Correlation (TDSC)

In order to validate this selectivity, we carried out

simulations by varying the delay D at TDSC

detection stage for a recorded TR-UWB signal with

D=127ns at transmission here. We varied the value

of D +/-10 sampling points around its nominal value.

The simulation result is shown in Fig. 10. Here the

sampling period is 200ps, the temporal

discrimination can be evaluated at +/-2ns.

This characteristic can not be found in other

detection methods. For example, an energy detector

integrates all incoming energies and results in a low

temporal discrimination. Besides some information

on the incoming signal can be lost such as pulse

polarity. The coherent detection method gives a

theoretically better performance but the difficulties

in the local template generation and ultra-high

precision synchronisation make it impractical in

simple low cost and low power systems.

(a)

(b)

Figure 10: a) TDSC results for small variations around the

value of delay D. b) The variation of the maximum

detection versus D.

3.2.2 Detection and MP Energy Collection

under LOS & NLOS Contexts

We conducted experiments with and without a line

of sight propagation path in order to evaluate the

performance of TDSC detection in the two cases.

For the first case (Fig. 11), the two monopole

antennas are in line-of-sight, at a distance of about

2,50 meters. The second case (Fig. 12), is performed

with a metallic obstacle between the two monopole

antennas, in order to obtain a non line-of-sight

scheme.

In a multi-path (NLOS) context, the received TR-

UWB impulse doublets can be strongly confused

together as shown by the upper waveform in Fig. 12.

When the TDSC method is applied to this signal, the

successive TR-UWB doublet can be clearly

separated and detected, despite the energy spreading

as shown in the lower waveform in Fig. 12.

Figure 11: Simulation result for LOS experiment.

Figure 12: Simulation result for NLOS experiment.

4 CONCLUSION &

PERSPECTIVES

In this paper, we have introduced a new TR-UWB

signal detection concept -The Time Delayed

Sampled and Correlation (TDSC). This method has

been presented in details and validated

experimentally. This detection method gives

numerous advantages such as large synchronisation

tolerance, high multipath signal energy collection

efficiency in strong multi-path environment, high

signal selectivity and modular CMOS friendly

design with low power consumption, etc. Actually

this validation has been done in 300MHz band, a

WINSYS 2006 - INTERNATIONAL CONFERENCE ON WIRELESS INFORMATION NETWORKS AND SYSTEMS

CMOS waveform sampler circuit capable to operate

at 10GHz is on the way and this circuit will permit a

full validation at the lower end of FCC UWB band.

In parallel, the theoretical investigations on this

method by using different channel models proposed

by IEEE P802.15-WPANs (Foerster, 2003),

(Molisch, 2003) are on the way.

REFERENCES

Bettayeb, M.; Shah, S.F.A.; 2003 State of the art ultra-

wideband technology for communication systems: a

review. In Proceedings of the 2003 10th IEEE

International Conference on Electronics, Circuits and

Systems, Vol. 3, pp. 1276 – 1279

Doré J.B., Uguen B., Paquelet S. and Mallégol S., 2005,

UWB Non-Coherent High Data Rates Transceiver

Architecture and Implementation. In international

workshop on convergent technologies (IWCT).

Durisi, G. and Benedetto, S. 2005. Comparison between

coherent and noncoherent receivers for UWB

communications. In EURASIP journal on applied

signal processing UWB state of the art, vol.3,pp.1-9.

Foerster J. 2003. Channel Modeling Sub-committee

Report Final, IEEE P802.15-02/490r2-SG3a, 2003

IEEE P802.15 Working Group for Wireless Personal

Area Networks (WPANs)

Goeckel Dennis L. and Qu Zhang, 2005, Slightly

Frequency-Shifted Transmitted-Reference

Ultrawideband Radio: TR-UWB without the Delay

Element, In the 2005 IEEE Military Communications

Conference (MILCOM 2005).

Hoctor R. and Tomlinson H. 2002. Delayed-hopped

transmitted-reference RF communications, IEEE

Conference on UWB systems and technologies, 2002,

pp 265-270.

Kleinfelder S. 2003. GHz Waveform Sampling and

Digitization Circuit Design and implementation. In

IEEE Transactions on Nuclear Science, Vol 50, NO 4.

Metha V. and Magda El Zarki, 2004.An Ultra Wide Band

(UWB) based Sensor Network for Civil Infrastructure

Health Monitoring. EWSN.

Mielczarek, B. and Wessman, M.O. and Svensson, A.

2003. Performance of coherent UWB Rake receivers

with channel estimators. . In Vehicular Technology

Conference, IEEE 58th. Volume 3, pp:1880 - 1884 .

Molisch, A. and F, Foerster, J.R. and Pendergrass, M. ,

2003, Channel models for ultrawideband personal area

networks. In Wireless Communications, Vol. 10 ,

Issue: 6, pp: 14 - 21

Oppermann . I and Stoica. L. and al.. 2004. UWB

Wireless Sensor Networks: UWEN — A Practical

Example. In Communications Magazine, IEEE Vol.

42, Issue 12, pp:S27 - S32

NI Y. , 2005, Récepteur UWB et procédé et système de

transmission de données. Brevet Français déposé par

GET/INT, No. 0500886 ;

C.K. Rushforth, 1964, Transmitted-reference techniques

for random or unknown channels. In IEEE Trans. Inf.

Theory, vol 10, n°1,pp39-42,

Stoica L., and S. Tiuraniemi, and I. Oppermann, 2005. An

ultra-wideband low complexity circuit transceiver

architecture for sensor networks. In Circuits and

Systems, ISCAS. IEEE International Symposium on, pp

364-367.

Time Domain, 2001. PulsON® Technology Overview,

Time Domain , http://www.timedomain.com.

Duo X., Torikka, T. Li-Rong Zheng, Ismail,

M.;Tenhunen, H. Tjukanoff, E. 2004. A DC-13GHz

LNA for UWB RFID applications. In Norchip

Conference, 2004. Proceedings ,8-9 pp:241 - 244

Zasowski T., Althaus F., and Wittneben A., 2004. An

Energy Efficient Transmitted-Reference Scheme for

Ultra Wideband Communications. In International

Workshop on Ultra Wideband Systems (IWUWBS)

Joint with Conference on Ultra Wideband Systems and

Technologies (UWBST), Kyoto, Japan.

A NEW SIGNAL DETECTION METHOD FOR TR-UWB - By Time Delayed Sampling & Correlation (TDSC)