&52667$/. $1$/<6,6 21 35,17(' &,5&8,7 %2$5'

%

< 0($16 2) :$9(/(76

A

. Kacha , F. Grenez , P. De Doncker

Service Electricité Générale, Faculté des Sciences Appliquées, Université Libre de Bruxelles, Av. F. D. Roosevelt 50,

CP 165/51, B-1050 Bruxelles, Belgium

Phone: +32-2-650-3087, Fax: +32-2-650-1347, E-mail: akacha@ulb.ac.be

K. Benmahammed

Département d'Electronique, Faculté des Sciences de l'Ingénieur, Université de Setif, Setif 19000, Algeria

Phone/Fax : +213-36-97-09, E-mail : khierben@ieee.org

.(<:25'6

C

One of the main problems that the designer may be faced

to is determining when prototype failures are due to logic

rosstalk

analysis,

electromagnetic

interference,

errors or to crosstalk. Reducing crosstalk effects allows to

wavelets.

focus only on logic errors. Therefore, an efficient method

$%675$&7

for identifying the source line of crosstalk is required.

In

geometric

identification

methods,

crosstalk

in the identification problems of the source lines of

requirements are specified in voltage which is converted

crosstalk on printed circuit boards, the requirements on

to a maximum distance. Only tracks up to the specified

the electromagnetic interference are specified in voltage

maximum distance are considered for the purpose of

which is converted to a maximum distance. Only tracks

disturbance line identification. Unfortunately, tracks

up to the specified maximum distance are considered for

beyond the specified maximum distance can result in

the  purpose  of  disturbance  line  identification.

large crosstalk. Determining rules in geometric methods

Unfortunately, tracks beyond the specified maximum

is an important issue because there are no general rules

distance can result in large crosstalk. This paper proposes

that allow to identify efficiently all tracks which may

a wavelet-based approach which takes into account all

cause significant crosstalk . Recently, wavelets have

tracks that are susceptible to produce interferences

proven to be useful in many problems related to crosstalk

regardless the separating distance between the coupled

prediction and diagnostic of faults in transmission lines.

tracks. A wavelet-decomposition of the disturbed signal

In (Buccella and Orlandi 2000), wavelets has been applied

allows to characterize the source of disturbance by means

to generate crosstalk signature in order  to identify the

of the energy of the detail coefficients of the

faulty line and in (Antonini and Orlandi 2001), a wavelet

decomposition which makes easier the identification of

packet decomposition has been used for the purpose of

the source of disturbance.

,

feature

extraction

and

classification

of

the

T ,1752'8&7,21

electromagnetic source of the disturbance. In this paper,

we propose a wavelet-based approach for the purpose of

he increase in speed of digital electronic circuits has led

disturbance source line identification. In the simulation,

to several problems which may be taken into account in

we consider two parallel tracks on a PCB separated by

the earlier stage of design for the success of the system

some distance. To the first track, we apply a high-

performance (Khan and Costache 1989). Fundamentally,

frequency trapezoidal voltage characterized by small rise

three problems may be considered: Signal integrity,

time and fall time and we predict the crosstalk generated

conducted and radiated emissions and susceptibility of the

in the second track. In fact, the propagating signal in the

circuit (Gravelle and Wison 1992). Signal integrity is

disturbed line is composed of the superposition of the

strongly affected by crosstalk which is defined as the

induced crosstalk and the original signal without

unwanted interference signal generated by nearby tracks

distortion. By performing a decomposition of the overall

in the printed circuit board (PCB). As the switching speed

signal into coarse approximation and fine detail at

increases and the PCBs miniaturized, the crosstalk

different scales, it is possible to detect and quantify the

between transmission lines imposes a significant signal

crosstalk. As the distance between the two lines increases,

integrity problem, mainly between the long parallel lines

the generated crosstalk decreases and this behaviour will

(Sohn et al. 2001). If the crosstalk is sufficiently large in

be seen by inspecting the detail signal. Thus, by

magnitude and duration, it can result in a switching fault

considering the detail in the wavelet decomposition, one

(Yeargan et al. 1988, Parker et al. 1994). Thus, it is of

can get an estimation of the distance between the source

great importance to detect crosstalk using simulations

line of disturbance and the disturbed line. Examination of

during the design in order to eliminate or at least reduce it

the crosstalk feature makes easier the identification of the

to acceptable levels before constructing the prototype.

different

lines

generating

the

electromagnetic