Proc.

of the 2013 International Symposium on Electromagnetic Compatibility (EMC Europe 2013), Brugge, Belgium, September 2-6, 2013

Crosstalk analysis in complex aeronautical bundle

Charles Jullien, Philippe Besnier Michel Dunand Isabelle Junqua
Université Européenne de Bretagne, Dept. R&T ONERA/DEMR/CEM
INSA, IETR, UMR CNRS 6164, LABINAL/Safran Engineering ONERA – The French Aerospace Lab,
Rennes, FRANCE Services F-31055,
charles.jullien@safran- Villemur sur Tarn, FRANCE Toulouse, France
engineering.com

Abstract— An analysis of crosstalk inside a complex aeronautical of cross sections of bundles can be randomly generated as in
bundle from an experimental and from a numerical point of view the TWISTCAB tool presented later on.
is proposed. The numerical model is built as a succession of
uniform transmission lines generated randomly. Modeling of the A study of a simple case with a non-shielded twisted pair
geometrical pattern of twisted pairs is included. Crosstalks cable (TPC) and a victim wire has shown that the behavior of
between these twisted pairs and single wires inside the bundle are the current resulting from the coupling between these two
measured and computed. In order to save computation time, a cables is very specific and can be simply explained [5,6]. This
simplified model of the bundle is developed and validated. study led to the development of a simplified model to reduce
the size of the topological network by averaging each term of
Keywords-component; Cable Bundle; Crosstalk; p.u.l. matrices along the non-uniform bundle. Here, in this
Multiconductor Transmission Line; Statistical; Random; Simplified paper, we investigate crosstalk in a more complex
model configuration of bundles. We first designed and set up a
complete test bench to handle a very large number of cables.
I. INTRODUCTION This test bench enabled to collect various crosstalk
measurement results. Second, we had to find a way to build up
With the increasing use of composite materials in aircrafts,
a reference model, looking for the best possible agreement with
EMC requirements and especially internal coupling in bundles
measurements. Then we investigated a possible validation of
must be re-examined. Indeed, these harnesses include a large
the previously mentioned simplified model at the scale of such
number of functions performed by various cable types and in
a large bundle.
particular twisted cables and non-shielded twisted pair cables.
These cables are very well known to be a potential solution for After a presentation of the complex bundle (Section II), we
reducing electromagnetic interference, but their effect depends describe the test bench. Measurements are gathered in Section
on various parameters such as their relative position within a III. These measurements refer to current induced on a victim
bundle and arrangement of cable extremities. single cable considering an incident interference as a
differential voltage generator on unshielded twisted pair cables.
Since the late 70’s, important theoretical work has been
Then we model the whole bundle and simulate the test bench to
done to model twisted pair cables. Since the end of the 80’s,
have reference results. The reference model is introduced in
theoretical and numerical models are focused on multi-
section IV. Finally we discuss in Section V the application and
conductor harnesses. We may first mention the work done to
validation of the simplified model to reduce the time
describe distributions of multiconductor interconnections based
computing and data storage as in [6].
on the electromagnetic topology principles [1-3]. This theory
splits the complex problem into sub-elementary problems and
uses the equations of transmission line theory to handle II. DESCRIPTION OF CASE STUDY
calculation of electromagnetic interactions in bundles. Since The complex bundle must fulfill different requirements:
the late 90’s, modeling the effects of non-uniformity in bundles
to consider interlacing and twisting has become essential to It has to be built by an aeronautical harness
correctly evaluate electromagnetic interactions on realistic manufacturer like LABINAL to be representative of
geometries of cables bundles. avionic technology.
ONERA has developed a software based on the formalism Aeronautical connectors are used to ensure good contact
of Electromagnetic Topology [4], named CRIPTE. It enables to between the base and the back shell.
build up the representation of a cable network as a topological
It has to be composed of different types of twisted
network composed of tubes (short uniform transmission line)
cables and single wires to take into account the diversity
and junctions (that connect tubes between them and terminal
of conductors as in a real bundle.
loads). Each tube is characterized by matrices of per unit length
(p.u.l.) electrical parameters, namely the p.u.l. resistance (R), Consequently, the selected bundle manufactured by
inductance (L), conductance (G) and capacitance (C) matrices. LABINAL (Fig. 1) is made of 41 composed cables:
These (RLGC) parameters are derived from the geometrical
description of each conductor inside the bundle and their
electrical features by solving the Laplace equation. Geometry

978-1-4673-4980-2/13/$31.00 © 2013 IEEE 253

 Proc. of the 2013 International Symposium on Electromagnetic Compatibility (EMC Europe 2013), Brugge, Belgium, September 2-6, 2013

From AWG 24 to 18 for copper cables and from AWG loads, the twisted pair and victim single wire to SMA
24 to 20 for aluminum cables as found in sensitive connections as shown in Fig.2.
harnesses.
32 cables are single wires, 5 non-shielded twisted pair
cables (3 different pitches: two with 67twists/m, two
with 50twists/m and one with 50twists/m) and 4 non-
shielded twisted quadrifilar cables with a pitch of
25twists/m, that is to say 58 elementary conductors (or
41 composed cables).
Length is arbitrary fixed to 3m to limit the extension of
the test bench.
Each conductor is identified by labels at its extremities.
In addition; tyraps are placed along the bundle to keep
the compactedness.
Aeronautical connectors have been included at each
extremity.

Figure 2. Schematic of the bench for the complex bundle

Instrumentation boxes are connected to the equipment

boxes. They include a current clamp which is used to inject a
pure differential source with no common mode current noise
on the twisted pair, terminal impedances on cables under study
(a differential 100Ohm for the twisted pair and 50Ohm for the
victim single cable) as shown in Fig. 3.

Figure 1. Picture of the complex bundle composed of 58 cables

We have chosen to analyze 9 crosstalk configurations

between 3 twisted pair cables and 3 single wires, examining
the current induced on each single wire for a differential
excitation on one of the twisted pairs.
The twisted cables are ended by differential 100Ω loads at
each extremity and the single cables are ended by 50Ω loads
connected to the ground plane. Other cables are loaded by
resistances which have been evaluated by taking into account
aircraft manufacturer rules. Under these considerations, 3 Figure 3. Description of the experimental bench
groups of loads have been defined:
R < 5Ω, low impedance circuits, R is set to a short-
circuit, III. MEASUREMENTS OF CROSSTALKS BETWEEN TWISTED
5Ω < R < 29Ω, R is set to 10Ω PAIRS AND SINGLE WIRES
R > 29Ω R is set to 50Ω Considering a differential injection via a current clamp on a
The bundle and its connectors are finally installed on the twisted pair, 2 types of measurements have been performed:
test bench above a metallic ground plane (Dimensions: 3.500
x 0.750 x 0.003m) and between 2 metallic equipment boxes in
which neighboring cables are connected to the pre-defined

254
 Proc. of the 2013 International Symposium on Electromagnetic Compatibility (EMC Europe 2013), Brugge, Belgium, September 2-6, 2013

- Measurement of the common mode current on the different levels of current between configurations can be
twisted pair. This measure enables to fully control the explained by geometrical considerations and relative positions
injection. between the twisted pair and the single cable inside the
- Measurement of the current induced on a single wire bundle.
which characterizes crosstalk between the twisted pair IV. NUMERICAL MODELLING AND RESULTS
and the single wire.
These measurements are carried out with a network Bundle Model
analyzer as B/A where A and B are respectively connected to The model of the bundle as a succession of uniform
the injection current clamp and to the receiving current clamp. transmission lines represented as tubes in the CRIPTE network
The frequency range is 1MHz to 100MHz due to measurement is built in 2 steps. Firstly, a cross section of the bundle is
constraints. Fig. 4 illustrates the variation of the common mode randomly generated by TWISTCAB: Each cable is randomly
current on the twisted pair cable and the crosstalk current for placed in free space respecting a constraint of non-
all coupling configurations measured between each of the 3 superimposition of cables and a constraint of compactedness of
TPC and the 3 victim cables, therefore 9 induced current the bundle [7]. In this first geometry, TWISTCAB takes into
curves (versus frequency) in total. Maximal, minimal and mean account the fact that a pair is twisted by considering a bulk
values of these currents have been extracted from this set of volume around it for future twisting as in Fig. 5.
data.

Figure 5. First geometry of the complex bundle

The second step consists in generating the following

geometries. In these geometries, we take into account only the
twist of twisted cables. We made the hypothesis that other
cables could be modeled as straight cables and are therefore
located at the same positions in all cross sections. It had been
shown in [6] by a convergence study that a twist may be
discretized as a succession of 8 uniform transmission line
sections (or tubes), each one being associated with its uniform
cross-section geometry. Here, there are 4 different pitches of
twisting, defined for 1m of cable. Considering that the
minimum number of discretization has to be applied for the
smallest pitch (67 twists/m), 536 geometries are necessary to
Figure 4. Measurement of common mode currents on twisted pairs and model 1m of the bundle. The last 2m can be modeled with the
crosstalk current on single cables for the 9 coupling configurations from TPC same regular pattern.
to victim cables Finally let us note that resistive and conductive losses have
been introduced as R and G per unit length parameters.
The common mode current on the twisted pair which
carries the electromagnetic source is quite constant for all Bench Model
twisted pairs inside the bundle in the whole frequency range. The CRIPTE software has been used to model the test
Induced currents on victim cables follow the same pattern as a bench. The minimal topological network would be made of
function of the frequency. 1608 tubes (536x3). But in order to reduce the computation
The distribution of the currents induced on victim cables is time, the notion of compaction available in the software has
typical of a crosstalk between a TPC and a victim wire. The been exploited. Indeed, the software can compact a network as

255
 Proc. of the 2013 International Symposium on Electromagnetic Compatibility (EMC Europe 2013), Brugge, Belgium, September 2-6, 2013

a “black box” to give an equivalent junction characterized by The differential injection via the current clamp in the
its S-parameters and which can be re-used in any network. experimentation is modeled in the topological network as a
Here, two levels of compaction were considered: localized voltage generator applied on one wire of the twisted
The first level consists in computing 54 equivalent pair. This excitation is labeled on the red-colored tube in Fig.
junctions. 6.
At a second level we split again the remaining Simulation and results
network to compute 9 equivalent junctions.
These 9 equivalent junctions represent 1m of the bundle. As in the experimentation, common mode current on the
The final network is composed of 27 (3x9) equivalent twisted pair carrying the source and induced current on the
junctions to model the bundle on the ground plane and 16 single victim wire are computed between 1MHz and 100MHz
tubes to model extremity boxes (all loads and source) as at one extremity of the bundle. 9 crosstalk situations are also
illustrated in Fig. 6. computed for each coupling between the 3 twisted pair cables
and the 3 victim cables. Maximal, minimal and mean values
over this set of data are plotted on Fig. 7.

Figure 7. Computation of common mode current on twisted pair cables and

of current on victim single cables

The behavior and levels of the current are similar in

simulations and measurements. The level of crosstalk current
is the same for maximum and minimum value. The computed
and measured mean induced current on the victim single cable
are slightly different in terms of magnitude. Finally, one can
note some discrepancies in the resonance domain. But these
similarities between experimental and simulation results show
that it is not necessary to represent the random position of
conductors in this case of coupling.
Figure 6. Principle of compacting and final network

256
 Proc. of the 2013 International Symposium on Electromagnetic Compatibility (EMC Europe 2013), Brugge, Belgium, September 2-6, 2013

Difficulties fictive geometry to describe the TPC for any length. The
Even if the modeling of this test case can be considered as consequence is a reduction from 536 tubes for 1m of the
very satisfactory in terms of results and comparisons to bundle to 1 tube for 3m of the bundle as illustrated in Fig. 8.
measurements, two main difficulties appeared:
Computing all tubes has taken 5.6 days to get resistive,
conductive, capacitive and inductive matrices. The two
levels of compacting lasted 2.4 days. Finally each
simulation required 1h for the computation of the last
network to obtain the crosstalk results. All these
calculus are made on a computer using CPU Pentium 4
3.4GHz and 1Go RAM.
The storage of all tubes represents 8Go on HDD and all
results of compacting networks and final networks
need 500Mo.
Let us remember that the bundle under study is only 3m
long and is made of 58 elementary conductors. In aeronautical,
harnesses can be composed of more than 300 cables and can
reach 30m long which would lead to huge networks and
unrealistic computation time. These difficulties can be revised
by reducing the number of cables or decreasing the number of
tubes. Here, the second solution has been considered as a Figure 8. Application of the simplified model to the case study
possible trail of simplification.
Fig. 9 illustrates the variation of the crosstalk current for all
V. SIMPLIFIED MODEL APPROACH distribution of coupling between TPC and victim cables
A non-negligible part of computational time is spent in computed with the reference model previously detailed in
operations of compaction. If we reduce the total number of section IV and the averaging simplified model.
tubes, we may avoid to use compacted sub-networks. The
model that we propose consists in arithmetically averaging per
unit length parameters of the (L,C) matrices along the 536
tubes modeling 1m. This simplification strategy is named
AVE and the objective is, here, to reduce the number of tubes
[5-6-8-9] to a single tube containing the averaged parameters
and to apply this model to crosstalk problems. In fact, this
model consists of computing the average value of each
element of the capacitance and inductance matrices. Eq (1)
highlights this procedure for the p.u.l. inductance matrix. In
this equation N stands for the number of tubes (N=536 in the
case depicted in this paper) and i,j are indices related to
individual wires.

L1ij ... LijN

. .
. LijAVE . N (1)
AVE
L1
jj ... LNjj
. L jj . . .
N
. . . . . .

This averaging procedure suggests that the mutual terms as

seen from the two wires of a same TPC could be identical
since the neighboring cables are straight ones. It takes
however into account the other twisted pairs which do not
have the same pitch. In fact these mutual terms are within the
same range of magnitude. If one would try therefore to
identify the fictive position of the two conductors they would
Figure 9. Comparison of crosstalk current on single cables for the 9 coupling
appear as, at least partly overlapped. We must be aware that from TPC to victim cables between reference model and AVE model
we lose details inside each twist; as a consequence of this
model. With this type of equivalent model, there is a single

257
 Proc. of the 2013 International Symposium on Electromagnetic Compatibility (EMC Europe 2013), Brugge, Belgium, September 2-6, 2013

In this figure, we observe that currents on victim cables storage and computation time. Results extracted from this
from the simplified AVE model are superimposed with bundle under study, look very promising.
currents from the reference detailed model. They have the same
behavior and same resonances. The benefits are: Additional work would be also necessary to extend the
comparison between AVE model and experimental results to a
In time computing, for 536 tubes 5.6 days with statistical study in order to evaluate the confidence interval of
reference model Æ for 1tube 15min with AVE computed and measured crosstalks inside such aeronautical
model, for all network 2.44 days with reference bundles.
model Æfor all network 4h with AVE model
In data storage, for 536tubes 8Go with reference REFERENCES
model Æ for 1tube 15Mo with AVE model [1]
L.Paletta, ”Démarche Topologique pour l'étude des couplages
électromagnétiques sur des systèmes de câblages industriels de grande
In fact, these simulations results confirm the good dimension”, Thesis, September 1998
[2]
agreement with measurement with reasonable time computing C.E Baum, T.K. Liu, F.M. Tesche, ”On the Analysis of General
and data storage. Multiconductor Transmission Line Networks Interactions Notes”, Notes
350, January 1988
[3]
VI. CONCLUSION C.E Baum, “Electromagnetic Topology for the Analysis and Design of
Complex Electromagnetic System”, Fast Electrical and Optical
In this paper, we have presented an experimental and Measurements, Vol. 1, pp. 467-547, Martinus Nijhoff, Dordrecht, 1986
numerical study of a complex aeronautical bundle. As a first [4]
J.P. Parmantier, S. Bertuol, “CRIPTE: User guide”, s.l. :
step, the bundle under study and the test bench have been ONERA/DEMR -009/06 – S,2006. Vol. CRIPTE 4.2
designed by taking into account realistic constraints in terms of [5]
C. Jullien, P. Besnier, M. Dunand and I. Junqua, “Advanced modeling of
terminal impedances, selection of cables and connectors. The crosstalk between an unshielded twisted pair cable and an unshielded
numerical reference model has been built in order to consider wire above a ground plane”, IEEE Transactions on EMC, July 2012
[6]
the full discretization of twisting cables based on previous C. Jullien, P. Besnier, M. Dunand and I. Junqua, “Analysis of the
Current Distribution Induced on a Victim Wire by a Differential Voltage
articles [5-6] in order to match as closely as possible the fine Source applied onto a Twisted Pair Cable”, Asia-Pacific EMC
details of the geometry of industrial bundles. This reference Symposium, Singapore, May 2012
model has been validated by comparison of crosstalks between [7]
C. Jullien, P. Besnier, M. Dunand and I. Junqua, “Towards the
twisted pairs and single victim cable to measurements. This generation of industrial bundles through a random process under
reference model being time and memory consuming, a realistic constraints”, EMC Europe, Wroclaw, September 2010
[8]
simplification that consists in averaging RLCG matrices, terms Bellan, D., Spadacini, G. et Pignari, S. A., “Prediction of twist
by terms, has been proposed. This model is based on a nonuniformity and twist-residual effects on crosstalk in twisted-wire
pairs.”, Zurich, Proceedings EMC Zurich International Symposium
reduction of number of tubes representing the global model of Electromagntic Compatibility, 2003. pp. 181-186.
the bundle. It has been also validated by comparison to the [9]
Pignari, Sergio A. et Spadacini, Giordano., “Plane-Wave Coupling to a
reference model and the gain is substantial in terms of memory Twisted-Wire Pair Above Ground.”, s.l. : IEEE Transactions on
Electromagnetic Compatibility, Mai 2011. Vol. 53, 2.

258