Simulating Crosstalk and EMI in Cables - CST

Simulating Crosstalk and EMI in Cables - CST
Simulating Crosstalk
and EMI in Cables
C
ables not only transfer the power
needed to run electrical equipment,
but also the data signals needed to
operate them. To prevent errors and device
failures, the same attention must be paid to the
choice and installation of the cabling as is paid
to the rest of the system.
Dealing with data means considering signal
integrity (SI). Since data signals are modulated electrical pulses, anything that introduces
noise into the cable can corrupt the information, causing equipment to lose performance,
malfunction or simply fail altogether. Crosstalk
and impedance mismatch are common sources of SI problems; cables generally consist of
multiple wires travelling together. The fields
generated in one wire can, without proper
shielding, couple to others and induce currents
in them, while signals can be reflected at the
interfaces between cables if the impedances do
not match. Analyzing these issues uses RF and
microwave techniques and expertise.
It is not just interference from other data
signals that one needs to worry about, however.
Electromagnetic interference (EMI) can come
from a range of sources within the system
and the wider environment. Switched-mode
power supplies generate noise, while lightning
strikes and electrostatic discharge (ESD) introduce transients that often cause damaging
current surges in devices. Even the interaction
between the equipment and its casing can be
enough to interfere with data signals. As well as
being immune to external radiation, the cables
themselves should not radiate either. Electromagnetic compatibility (EMC) is a legal requirement and this means that they must pose
little interference risk to other devices.
Every advance in technology pushes cable
design requirements further. High-speed devices demand cables capable of handling everhigher bit-rates. Automotive and aeronautical systems, increasingly reliant on electronic
control and communications systems, need
cables that are lightweight yet also measure
up to stringent safety regulations. Consumer
electronics meanwhile are pushing toward
standardized multipurpose cables, where one
lead might be used for anything from charging a mobile phone to controlling a printer or
transferring data to a hard drive.
In light of these developments, designers
have turned to cable harnesses, where multiple
cables – sometimes a hundred or more – are
tied together and travel along the same conduit
as well as hybrid cables, which contain both signal and power wires together. The most familiar example of a hybrid cable is probably USB,
but custom cables of many configurations are
used in industrial applications. For such complex cable designs, traditional design rules to
calculate the cable’s properties become unwieldy. Simulation offers a way to develop and
check an arbitrary cable design and optimize
its layout and shielding for better performance.
OvErvIEw Of CablE SIMulatIOn
Applying full 3D EM simulation methods to
simulate detailed cabling within a large structure such as an automobile is impractical, as
David Johns and Patrick DeRoy
CST of America, Framingham, MA
Reprinted with permission of MICROWAVE JOURNAL® from the March 2013 Supplement issue.
©2013 Horizon House Publications, Inc.
Cables
& Connectors
Supplement
Cables
& Connectors
Supplement
TABLE I
VARIOUS APPROACHES TO
CABLE SIMULATION
Electrostatic Simulation
Electro- and magneto-static simulations
can be used to calculate the L and C
contributions to the line impedance. This
approach is fast but only accurate at DC.
Full-Wave Simulation
Full-wave 3D simulation lets the frequency
dependent effects be taken into account as
well, modeling the dispersion of signals as
they propagate down the line. This approach
is very accurate for broadband signaling, but
slower than a specialized cable simulation.
Specialized Cable Simulation
 Fig. 1
The cabling system inside a car.
Aluminium Foil
ALUMINUM FOIL
RS485
Twisted
RS485Pair
PAIR
(3TWISTED
IN. Twist
Rate)
(3
IN.
TWIST
36 AWG RATE)
36
AWG
Conductors
CONDUCTORS
FeP
FEP Insulation
INSULATION
 Fig. 2
Copper Braid
COPPER BRAID
PowerPOWER
Wires
WIRES
14
AWG
14 AWG
PVC
PVC
Insulation
INSULATION
A cross section of a hybrid cable carrying five wires.
shown in Figure 1. Each individual
wire might be less than a millimeter
in diameter, yet tens of meters long,
bending as it carries high frequency
signals through an electromagnetically complex environment. A conventional simulation of such a system
would need an incredibly fine resolution to capture the fields in the cable,
extended over a very large volume, resulting in slow, computationally-intensive calculations. The arbitrary cross
sections of cables and changes in the
surrounding environment along its
path make analytic solutions similarly
hard to find.
Specialized cable simulators can
model complex cables and cable harnesses far more efficiently. The cable
can be divided into segments, each
having a constant cross-section. A
2D electromagnetic field solver can
then be applied to extract the elec-
trical properties for
each segment. The
properties of the
entire cable may be
found by cascading
the electrical models into an equivalent network. The
specialized
cable
solver may also cosimulate with a full
3D field solver to
simulate coupling to
and radiation from
the cable.
The cable shown in Figure 2 carries five wires – two power wires (red
and white), two signal wires (gray
twisted pair) and a ground (green).
The signal wires are shielded from the
power wires by foil and the whole cable is shielded by a copper braid. It is
an example of a hybrid cable for which
an SI and EMI analysis would be important. The signal wires are not only
potentially at risk from external interference, they could also pick up noise
from the power wires. The shields
and the use of twisted pair wires are
both meant to reduce noise in the
signal; simulation allows the engineer
to investigate their performance. The
cable is only a few millimeters wide,
but could be meters long.
lInE IMPEDanCE SIMulatIOn
One important property of any
The cable cross-section is analyzed using
a 2D field solver and transmission-line
network analysis applied to simulate
propagation. This approach is both accurate
and efficient.
data cable is its line impedance. Ideally, the impedance of the data line
should match the impedance of the
load. If there is a mismatch, signals at
the interface between the two will be
partially reflected and interfere with
the signal, which can lead to SI problems in the connections between the
line and components. Reflection is especially problematic for bidirectional
cables, where the equipment at both
ends acts as both source and receiver.
Calculating the line impedance of
a cable is an obvious application of
simulation. Multiple approaches exist
for impedance calculations: the static
approach, a full-wave simulation and
a specialized cable simulation. Each
has its advantages and disadvantages,
as shown in Table 1.
A full-wave simulation also permits
one to model the connectors at the
end of the cable. Impedance matching and EMI analysis are very important when designing or choosing
a connector. This is especially true if
the connector differs from the cable
in some way – for example, if a twisted
wire terminates in a straight pin, or if
the connector is shielded in a different manner to the cable. The connectors are critical in impedance matching, since they provide the transition
between the cable and the load. Often
one wants to know which cable di-
Cables
& Connectors
Supplement
Cables
& Connectors
Supplement
Simulation can
also calculate the
scattering parame0.8
DiffR
ters (S-parameters),
0.6
which describe the
0.4
cable’s characteris0.2
tics concisely and
0
0
20
40
60
80
100
120
140 160
180
200
show how the losses
TIME (ns)
vary with respect
 Fig. 3 Eye pattern of a 20 m long cable carrying a 10 Mb/s signal. to signal frequency.
Standard measurements used in the
1.5 in.
3.0 in.
0.0002
lab such as the eye
diagram, shown in
0.0001
Figure 3, can also
0
be replicated with
–0.0001
simulation. To produce an eye diagram,
–0.0002
a series of random
0
100 200 300 400 500 600 700 800 900 1000
bits are fed into the
TIME (µs)
and the output
 Fig. 4 Induced differential mode currents for twist length of 1.5 in. cable,
at
the
other
end cap(red) and 3 in. (green) of the hybrid cable, assuming no shielding.
tured and graphed.
mensions will give suitable cable charLayering the output signals gives a useacteristics to minimize reflection and
ful illustration of the rounding of the
improve data transmission. To do this,
pulses caused by attenuation.
an optimizer is used. The optimizer
simulates multiple possible values for
CrOSStalK SIMulatIOn
model parameters – for instance, the
The potential for crosstalk arises
thickness of an insulator or the posiwhenever two or more wires are
tion of a wire – using sophisticated
coupled. In the example, the power
algorithms to reduce the amount of
conductors can couple with the sigguesswork involved in finding the best
nal lines. This sort of broad spectrum
configuration.
noise may be difficult to filter out of
TD VOLTAGES [REAL PART]
CURRENT (A)
VOLTAGE (V)
1
CalCulatInG lOSSES
Losses in the cable cause signals to
be attenuated, with the degree of attenuation depending on the frequency. High frequencies typically suffer
from greater loss, which has the effect
of rounding the sharp edges of data
pulses, limiting the data transmission
rate.
The AC resistance (skin effect) of
the wires is one obvious contribution
to losses. Another is the loss tangent
of the dielectric insulation materials.
Signals may also leak through nonperfect shields, introducing additional
losses into the circuit. Full-wave and
specialized cable simulation both allow designers to calculate the losses
for a cable. Full-wave simulations of
short cable sections can be cascaded,
allowing losses over the whole length
to be found without having to solve a
full model of the structure.
signals. Instead, the best option for reducing crosstalk is to shield the wires
well and make sure the coupling is
minimized.
Since the data is carried by differential signaling, it is the difference between the two wires (the differential
mode) that matters. If the cable system
is perfectly balanced, the noise coupled
to each conductor in the differential
pair will be equal and the receiver will
be able to reject the noise when the
voltage signals are subtracted.
Simulating Twisted-Pair Wire
Twisted-pair cabling, commonplace in communication systems for
over a century, winds wires around
each other in sets of two. This minimizes the loop area between the wires
to reduce mutual inductive coupling.
It also helps to maintain balance in the
line, equalizing the exposure of each
conductor to external fields. Coupling
is reduced as long as the twist length is
small compared to the wavelength of
the interference. The precise choice
of twist length can make a big difference to how immune the cable is to
crosstalk; for cables carrying several
pairs of signal wires, such as the ubiquitous Cat-5, giving each set of wires
a slightly different number of twists
per meter can reduce the crosstalk
between them significantly.
With simulation, the cable designer can examine the interference
experienced by the wires and test
how different wire configurations affect the signal characteristics. Cable
simulation lets one easily adjust the
distance between twists (twist rate) in
the computer model, and simulate the
induced currents caused by switching
noise in each case (see Figure 4). For
this cable, making the twists shorter
makes a big improvement to the differential mode noise rejection characteristics of the line. These simulations are performed without a shield
around the twisted pair to study the
noise rejection due to twisting alone.
Wire twisting is most effective if
the two wires are well-balanced. In
a real system, with varying loads and
imperfectly manufactured components, this will not always be the case.
The cable model can be attached to a
circuit made up of lumped elements
to replicate the equipment at either
end and one can see how well the
cable performs when the manufacturing tolerances are taken into account.
In this example, shown in Figure 5,
a 10 percent imbalance is introduced
in the shunt inductance (a drop of 10
µH) and a 4 percent imbalance in the
series capacitance (a 20 nF decrease).
This proves to have a huge impact
on the coupling. The peak induced
current in the imbalanced system is
ten times greater than in a perfectly
balanced system, leaping from 0.2 to 2
mA – a 20 dB increase in the intensity
of the crosstalk (see Figure 6).
Shielding
For an extra layer of protection,
which should guard against crosstalk
and interference even if the cable terminations are not balanced, the wires
can be shielded. Different shield types
Cables
& Connectors
Supplement
Cables
& Connectors
Supplement
A length of the cable is attached to a lumped element circuit in CST DESIGN STUDIO™. The imbalanced section is highlighted.
ZT =
1 dV
I0 dx
where I0 is the current flowing on one
side of the shield and dV/dx is the
voltage per unit length along the opposite side. The transfer impedance
effectively provides a measure of to
what extent the shield’s construction
prevents fields passing through. The
environment does not affect the transfer impedance – it is solely an intrinsic
characteristic of the shield itself. The
lower the transfer impedance, the
better the shielding.
In a foil shield, the skin effect improves shielding performance at high
frequencies as shown in Figure 7.
Current diffuses through the shield
at low frequencies, but is confined to
the surface at high
1.5 in.
frequencies. This
3.0 in.
0.002
3.0 in. imbalanced
means that external
0.001
fields cannot pen0
etrate the shield at
–0.001
high frequencies,
–0.002
provided that the
0
100 200 300 400 500 600 700 800 900 1000
shield is perfectly
TIME (µs)
closed and has no
apertures. Howev-  Fig. 6 Differential mode crosstalk in the imbalanced system
er, in a real shield, (blue) massively outweigh the crosstalk in the balanced system.
fields can also pass
through the seam
formed where the
layers of foil overlap, and for a more
accurate
simulation, this seam can
also be taken into
account.
Braided shield,
by contrast, behaves
the opposite way.
The transfer impedance of a braided
shield, as described
Frequency (MHz)
by Kley,2 depends
on a number of ef®,
Fig. 7 Creation tool box for a foil shield in CST CABLE STUDIOTM
fects. The skin ef- 
showing the transfer impedance curve over a range of frequencies.
fect still plays a role,
but more significant
ideal candidates for designing and
at high frequencies are the small aptesting with cable simulation software.
ertures between the strands and the
Arbitrary shields, whose properties
mutual inductive coupling, as well as an
are found experimentally, can also be
inductive interaction, known as “porused in simulations, simply by importpoising,” when strands cross each other.
ing their transfer impedance profile.
These drive up the transfer impedance,
If there are multiple shields, their
so that a braided shield is most effective
combined transfer impedance can be
at low frequencies. Figure 8 shows the
found analytically by taking into actransfer impedance curve for a braided
count the internal impedance of each
shield across the frequency spectrum.
shield and the inductance between
Because mathematical models exthem.3 Cable simulation software inist to describe the behavior of shields
corporating such calculations is capabased on their properties, they are
ble of simulating multi-layer shielding
CURRENT (A)
exist for different applications: the
main two are foil shields, consisting of
a thin sheet of metal such as aluminum,
and braided shields, which are made up
of many thin wires woven into a tube.
Shields can either go around the entire
cable to keep out external interference,
or they can be placed within the cable
to reduce crosstalk.
However, shielding not only adds
bulk and weight to the cable, it can also
decrease its flexibility and drive up the
manufacturing costs. With many types
of shielding available, each with different properties and suited toward
different types of interference, it is
not always easy to balance a cable’s
noise characteristics against the design requirements.
The performance of a shield is
measured by its transfer impedance,
as derived by Schelkunoff.1 The transfer impedance ZT is given by:
Magnitude
 Fig. 5
Cables
& Connectors
Supplement
Cables
& Connectors
Supplement
As a demonstration of the effects of grounding and shielding on EMC, and to show the accuracy of cable simulation,
multiple cables – standard RG58 coax with a braid, RG6
coax with a combined foil/braid shield, twisted pair with a
foil and drain wire, shielded twisted pair (STP) with a braid,
and the unshielded versions of the cables – had their EMI
properties studied, both experimentally and using computer
modeling. These cables were terminated in TNC connectors, which contain a metal shell or shield, enabling an ideal
360° connection between the cable and connector shield. To
provide a further comparison, an additional length of twisted
pair wire was terminated in a non-shielded connector with a
“pigtail” connection from the cable shield to the inside of the
enclosure housing (see Figure 12).
The interference source was a wire loop radiating across
a range of different frequencies. This loop can couple to the
signal wires in the cable – the extent to which the cables reject
this interference enables the shielding effectiveness to be assessed. The system would be difficult to solve accurately in
 Fig. 10
Magnitude
EXPErIMEntal vErIfICatIOn
full 3D due to the small details in geometry and the specialized cable simulator makes it possible to solve the problem in
a fraction of the time (see Figure 13).
Figure 14 shows the results of the study, where increasingly negative coupling corresponds to higher shielding
effectiveness. The unshielded results serve as a reference.
The spiral/drain cable with a pigtail termination provides
an additional 20 dB or so of shielding at low frequencies,
but its effectiveness degrades at higher frequencies. The
same cable was simulated with a TNC termination (dashed
blue line) for comparison, and provides further improvement in shielding, as expected. The STP, RG58 and RG6
Frequency (MHz)
 Fig. 8
The same dialog for a braided shield. The transfer impedance rises dramatically at high frequencies.
CURRENT (A)
arrangements. Because foil and braided shields work best
at different frequencies, combining the two in the cable
gives it a good broadband noise rejection profile. The
crosstalk analysis can now be repeated, but with shields in
place – in this case, either a foil shield or a combined foil
and braid shield around the signal wires – to see how this
affects the noise characteristics of the cable.
The combined two-shield cable has far better coupling rejection than the cable with just a foil shield. In
fact, the common mode noise is so low on this cable that
it is too small to see in Figure 9 – the peak current is
less than 4 µA.
The cable and the connector can also be combined together in one simulation. Circuit co-simulation, combining results from multiple full-wave and cable simulations, connects
components so they can be simulated as a system. Figure 10
shows a circuit for one such simulation: the power lines are
driven by a periodic switching voltage and the load is modeled by lumped elements. The output from the cable – the
blue box on the middle-left – is simply linked to the input of
the connector from Figure 11 – the gray tube on the middleright. Because the S-parameters of the connector have already been calculated, the circuit solver can run very quickly;
there is no need to rerun the entire 3D full-wave calculation.
Combined foil braid
Foil only
0.02
0.01
0
–0.01
–0.02
0
 Fig. 9
100
200
300
400
500 600
TIME (µs)
700
800
900
1000
Common mode noise for different shield arrangements. The
induced current when both shields are used is too small to see.
A circuit for testing a cable linked to the load by a shielded connector.
Cables
& Connectors
Supplement
Cables
& Connectors
Supplement
the simulation even
identified a significant resonance at
approximately
65
MHz that was subsequently seen in
the experiment.
2(1, +)
2(4, +)
2(2, +)
tranSIEnt COSIMulatIOn
2(5, +) 2(3, +)
COUPLING (dB)
Calculating the
 Fig. 12 The experimental setup showing the pigtail connector
properties of a (left) and the signal cable (right).
cable gives a good
 Fig. 11 A shielded connector compatible
with the hybrid cable showing the ports for
idea of its intrinsic
simulation.
characteristics, but
cables show good shielding effectivein the real world,
ness at all frequencies. Unlike a 360°
cables are rarely if
shield termination, which provides a
ever completely isolow impedance connection from the
lated from the enshield to the ground, the pigtail convironment. Nearby
nection adds inductance and the corstructures, external
responding frequency-dependent imfields and the route
pedance. Furthermore, the aperture
of the cable can all
in the enclosure is no longer closed
have an effect on
and may leak electromagnetic fields to
signals within the
the interior of the load box, coupling
cable. The transient
noise to the exposed signal wires.
®
nature of many in-  Fig. 13 Model of the same system in CST CABLE STUDIO™
The measurements all agree closeterference effects, showing the wire cross section for each segment.
ly with the simulation. With highly
however, means that
a complete picture of how a cable beshielded cables, the coupling is very
a full-wave simulation is necessary to
haves in a complex system. Combinsmall and the induced signals may
model the precise behavior of many
ing them into a hybrid co-simulation
be below the noise floor of the expotential sources of interference, such
task means that the advantages of
perimental equipment. This explains
as ESD and electromagnetic pulses
both – the fast, accurate cable models
some differences for coupling levels
(EMP).
and the versatility of full-wave – can
below –90 dB. Although it only used
Simulation is therefore a useful
be brought to the fore.
a very simple representation of the
tool for full system design, but neither
Two types of cable co-simulation
system, cable simulation was able to
specialized cable simulation nor full
exist: unidirectional and bidirectional.
model the results very accurately –
3D simulation alone is ideal for giving
In unidirectional simulation, the cable
is assumed to be either a transmitter
0
or a receiver, while in bidirectional
simulation, it is both. Bidirectional
–20
simulation is therefore most useful for
examining how the coupling between
UNSHIELDED RESULTS
–40
a cable and its surroundings affects
SPIRAL/DRAIN PIGTAIL
signals on the cable itself.
Figure 15 shows a typical appli–60
SPIRAL/DRAIN TNC
cation of bidirectional co-simulation.
The helicopter contains a number of
STD TNC
–80
RG58 TNC
–100
–120
0.1
RG6 TNC
0.2
0.3 0.4
0.6 0.8 1
2
3
4 5 6 7 8 10
20
30 40 50 70
100
200
FREQUENCY (MHz)
 Fig. 14
Coupling between the shield and the transmitter for different types of cables, simulated (solid lines) and measured (dotted lines).
 Fig. 15
A simple wiring system inside
a helicopter, showing cables for the control
system, antenna (center) and rotor (right).
Cables
& Connectors
Supplement
Cables
& Connectors
Supplement
cables, carrying both power and signals, following complex curved routes.
The cables may directly couple if routed too closely together, and radiated
fields may excite resonances inside the
vehicle, which can lead to increased
coupling. Transient co-simulation allows a number of different scenarios
to be modeled. One example would
be the potential susceptibility due to
an electromagnetic pulse (EMP). A
pulse striking the aircraft may couple
to the inside through imperfectly conducting panels and seams inducing
current in the cabling.
COnCluSIOn
Simulating cables requires careful consideration. Specialized cable
simulation software can be incorporated into the cable design workflow
to speed up the process and give the
designer an idea of how the cable will
behave once connected to the system.
These calculations can give not only
the electrical properties of the cable
itself, but also permit to observe how
fields propagate along the cable and
interact with the environment. Properly set-up, cable simulations can replicate real-world situations very accurately, with the results of simulations
agreeing very closely with measurements in the laboratory or field. ■
aCKnOwlEDGMEntS
The authors want to thank Jeffrey
Viel at NTS for providing EMC test
facilities.
References
1.
S.A. Schelkunoff, “The Electromagnetic
Theory of Coaxial Transmission Lines and
Cylindrical Shields,” Bell System Technical
Journal, Vol. 13, No. 10, October 1934, pp.
532-579.
2.
3.
T. Kley, “Optimized Single-braided Cable
Shields,” IEEE Transactions on Electromagnetic Compatibility, Vol. 35, No. 1,
February 1993, pp. 1-9.
E.F. Vance, “Shielding Effectiveness of
Braided Wire Shields,” IEEE Transactions
on Electromagnetic Compatibility, Vol. 17,
No. 2, May 1975, pp. 71-77.
David Johns is the VP of engineering and
support at CST of America and is based in the
Framingham, MA office. He holds a Ph.D.
in Electrical Engineering from Nottingham
University UK, specializing in the development
and application of the TLM method. He has
been actively involved in the modeling of EMC,
EMI and E3 problems for over 20 years.
Patrick DeRoy is an application engineer at CST
of America and is based in the Framingham, MA
office. He recently completed his M.S. degree
in Electrical Engineering at the University
of Massachusetts, Amherst. His Master’s
project focused on cable shielding and transfer
impedance modeling using CST STUDIO SUITE
and validating simulation results by comparison
with measurements.
For more information or to request a trial license,
please contact your local area representative.
CST AG – European Headquarters
Bad Nauheimer Str. 19
64289 Darmstadt
Germany
CST of America®, Inc. – US Headquarters
492 Old Connecticut Path, Suite 505
Framingham, MA 01701
United States
[email protected]
http://www.cst.com
Make the Connection
Find the simple way through complex EM
systems with CST STUDIO SUITE
Components don’t exist in electromagnetic isolation. They influence their neighbors’ performance. They are affected by
the enclosure or structure around them.
They are susceptible to outside influences.
With System Assembly and Modeling,
CST STUDIO SUITE helps optimize component and system performance.
Emissions from a server housing
Involved in emc/emi analysis? You can
read about how CST technology is used for
EMC/EMI analysis at www.cst.com/emc.
If you’re more interested in microwave
components or signal integrity analysis,
we’ve a wide range of worked application
examples live on our website at
www.cst.com/apps.
Get the big picture of what’s really going
on. Ensure your product and components
perform in the toughest of environments.
Choose CST STUDIO SUITE –
Complete Technology for 3D EM.
CST of America®, Inc. | To request literature (508) 665 4400 | www.cst.com
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement