prediction of pcb radiated emissions (emc symposium zurich 1998)
DESCRIPTION
The paper shows the experimental validation of predictive results of radiated emissions of a multilayer pcb. The radiated field is calculated from simulated results of pcb signals obtained from DWN analysis of interconnects.TRANSCRIPT
1
THE USE OF AN ACTIVE TEST BOARD TO VALIDATE A METHOD
THAT PREDICTS DIFFERENTIAL MODE RADIATED EMISSIONS
FROM PRINTED CIRCUIT BOARDS
Emmanuel Leroux,
Andrea Giuliano,
Carla Giachino High Design Technology
Corso Trapani 16
10139 Torino Italy
Ronald De Smedt,
Jan De Moerloose,
Willem Temmerman Alcatel Bell Telephone
Francis Wellesplein 1
B-2018 Antwerpen Belgium
Paolo Fogliati,
Piero Belforte,
CSELT
Via Reiss Romoli 274
10148 Torino Italy
Bernard Demoulin
University of Lille
59655 Villeneuve d’ascq
France
Abstract - The design of new Printed Circuit Boards
(PCBs) could be conveniently supported by the use of a
fast and reliable estimations tool of the Radiated
Emissions (RE). In this paper an active test board is
used to validate a method aimed at predicting RE due to
differential mode currents on PCB traces. The method
takes into account effects of dielectric layer in the field
calculation and does not need to discretise traces in
short segments. It can simulate all traces of a PCB in a
reasonable computation time. The simulation results are
compared with measurements in a semi-anechoic
chamber in far field. Radiation from the routed power
net, on board batteries and batteries holders is put in
evidence though measurement in near field. The paper
shows the importance of modeling of components and of
taking into account all couplings in the measurement set-
up when comparing RE results. The method validated in
this paper is included in the Telecom Hardware
Robustness Inspection System (THRIS) for evaluation
of telecom hardware.
I. INTRODUCTION
The introduction of severe international
standards in the field of Electromagnetic Compatibility
(EMC) in the last years had as result an increasing
interest for computing the Radiated Emissions (RE)
from electronic products. In fact the manufacturers of
electronic systems are interested in optimising from
EMC point of view their products before arriving to the
compliance tests in order to reduce time to market and
contain the costs. In most of cases electronic products
are composed of Printed Circuit Boards (PCBs). The
possibility of having a fast and reliable estimation of the
radiated field of PCB at design level is so highly
desirable.
In this paper an active test board is used to validate a
method aimed at predicting RE due to differential mode
currents on PCB traces.
II. MODELING OF PHYSICAL PHENOMENA
As it will be explained in the part III, the active test
board used to validate the method is made of microstrip
structures. The proposed method is now explained as
applied to an embedded microstrip made of only one
rectilinear trace. If more rectilinear traces are present,
the result is made of the vectorial summation of the
contributions of each trace.
As dielectric layers can severely modify radiation
patterns [1], it is important to take into account for each
microstrip structure the actual medium existing between
the metal plane and the air. The presented method is
able to calculate the RE of any microstrip structure, as
shown in figure 1, which shows two dielectric layers but
the presented method can take into account an arbitrary
number of layers.
P
P'
y
x
z
R
observation point
L
w
Figure 1: Representation of an embedded microstrip
structure, L: trace length, R: ”antenna” position where
the electric field is computed
A way to take into account dielectric layers in the RE
calculation is to use the dyadic Green's function of the
actual layered structure [2]. Then Sommerfeld integrals
have to be solved in a numerical way and this leads to a
long computation time. The presented method uses some
technics to give an analytical solution to the problem.
The starting point is the determination of the actual
current distribution along each trace. The method just
needs the knowledge of voltage and current on one of the
two extremities of each rectilinear trace. This
information is given in Time Domain by PRESTO [3]
2
(Post-layout Rapid Exhaustive Simulation and Test of
Operation) environment. It is a post-layout quality check
software that performs electrical simulations of PCBs to
evaluate Signal Integrity (SI). A Fast Fourier Transform
(FFT) is then performed to obtain this information in
Frequency Domain. Then the current waveform at any
abscissa x on the trace is determined by means of the
Transmission Line Theory (TLT) assuming that only the
quasi-Transverse Electric Magnetic (TEM) mode is
present along the trace. Then, the radiated
ElectroMagnetic (EM) field can be calculated using
dyadic Green’s function.
The electric field radiated from a surface current
distribution is obtained by means of the Green Dyadic G r r( ) of the layered structure, which can be
interpreted as a transfer function between the surface
current distribution J and the electric field as shown
in (1):
E r j G r r J r d ro( ) ( ). ( ) ( ) (1)
where:
- r is the coordinate of the point where the electric
field is computed;
- r is the coordinate of a point placed on the
rectilinear trace.
In general,
G r r( ) does not admit a close form
expression. However, with the assumption of being in
far field conditions, the Green Dyadic can be
substantially simplified.
For Information Technology Equipment (ITE) the
values of limits for RE in the frequency band 30 MHz to
1 GHz are given by EN 55022 [4] standard. The far
field condition applies approximately in the whole
frequency range described in the EN 55022 standard for
RE, which justifies the use of far field Green’s function.
Being difficult to directly calculate the electric field due
to the current density of a segment buried in dielectric
layers, the far field method applies the same current
source on the observation point where the EM field has
to be calculated and exploits the theory of reciprocity
[5] to transform the radiated emission calculation in a
problem of the induction of a plane wave on the
stratified structure.
Then the method assumes that the field arriving at the
air/dielectric interface is also locally a plane wave which
can be divided into two components, the transverse
electric (TE) and transverse magnetic (TM) modes. By
the use of modal analogy [2], Transmission Line Theory
(TLT) can be then applied to the propagation of these
two modes in the embedded microstrip structure to
produce two transfer functions for the actual medium
between the metal plane and the air. The following
expression of the Electric field in far field conditions is
obtained for any rectilinear radiating trace as shown in
Figure 1.
E R j
Z
Re e P x P y P z
o jk R jk h
ex ey ezo o
2 0
cos , , ,(2)
where:
Zo = o
o
= wave impedance in the air
K o = 2
o
= propagation constant
o = wave length in the air
h = substrate thickness
Pex ( , ) , Pey( , ) and Pez( , ) are essentially
plane-wave transfer functions of the dielectric layered
medium [2], that combine TE and TM plane-wave
modes. They depend on:
the spherical coordinates of the measuring antenna
position in the local reference system placed on the
trace.
the spatial Fourier Transform of the current density
on the trace
The obtained expression of the Electric field in far field
conditions is analytical, does not need to discretize the
trace and takes into account dielectric layers in the field
calculation.
III. EXPERIMENTAL VALIDATION
III.1 Description of the active test boards
To simplify the comparison between simulations and
measurements of the radiation of a PCB, a dedicated set
of active test boards have been developed [6]. They are
constructed using a standard four layers technology,
with a full power and full ground plane at the inside.
The component side contains all the components and
signal lines. The solder side only contains the
decoupling capacitors, the batteries holders with
batteries and a routed net for power distribution. A large
and a small version exists with respective dimensions:
295mm x 240mm and 195mm x 160mm. The layout of
the large board is shown in figure 2, that of the small
board is similar. The active test board contains an
oscillator (U6) with interchangeable clock frequency
(e.g. 5MHz, 16 MHz, 18MHz, 50MHz, 75MHz, with
the restriction that the digital logic circuitry must be able
to handle such frequencies). The oscillator drives three
4-bit counters (U1, U3, U4). The output of these
counters is connected to large busses (varying from
150mm to 250mm on the large board, 55mm to 150mm
on the small board).
3
Figure 2: Layout of the large test board
Two busses are terminated with inverters (U2, U5) of
which the output is connected to loads. The third bus
ends with a symmetric line driver (U7) of which each
symmetrical signal (pair of lines) passes through a
common mode filter (which can be shunted to inhibit its
functioning) and is directed towards a connector. All
lines have controlled impedance (75 Ohms). Several
termination schemes can be implemented: SLT (Series
Line Termination) at the begin, AC PLT (Parallel Line
Termination) at the end. All four combinations of
termination have been examined (SLT present or not
and PLT present or not). Several versions of the test
board exist, which are equipped with different, but
(standard) pin-compatible logic (LS-TTL, HCMOS,
ACT). To simplify measurements and simulations, the
board is able operate without any cabling. The power
can be supplied by unshielded rechargeable batteries
fixed on the solder side of the board by batteries
holders. They are connected to the power and ground
plane by a power net and shunted by a decoupling
capacitor.
III.2 Measurement and simulation of the active test
boards
For the results reported in this paper only the large
board has been considered and equipped with a 16MHz
clock, with ACT technology where SLT and PLT are
present.
The simulation needs technological,
geometrical and physical data which are extracted from
the layout tool used to design the board. It presents 177
components and 107 nets which are microstrip
structures. Most of nets are based on the same
configuration which is described in the figure 3 (SLT
and PLT are present).
driver receiver
Zs
5075
100 pF
75
driver receiver
Zs
5050
100 pF
75
Figure 3: Configuration of most of the nets, SLT and
PLT are present
Before to compare simulation and measurement
for RE it is important to verify the quality of the
simulated signals on PCB traces and so the quality of
models for active components. The board has been
simulated with behavioural models which include for
drivers the measured unloaded output voltage. The
figure 4 shows the comparison between the spectrum,
obtained by the means of a Fast Fourier Transform
(FFT), of the measured and simulated voltage at the
output of (U6) component.
10MHz 100MHz 1GHz
-60
-40
-20
0
20 dBV
10MHz-60
-40
-20
0
20
100MHz 1GHz
dBV
FFT on the
measured voltage
FFT on the
simulated voltage
Frequency
Frequency Figure 4: FFT on measured and simulated voltage at the
output of (U6)
The agreement is quite good until 300 MHz.
The radiated fields of the active test boards have
been measured extensively. The measurements have
been performed in the semi-anechoic rooms at Alcatel
Bell, Antwerpen (Belgium), and at CSELT, Turin (Italy)
which are normally used for EMC normative tests.
Several different positions of the board on the table have
been considered. The measurements have been
performed on the large and small test boards, with the 3
logic families and with various clock frequencies (as
long as supported by the logic). Several combinations of
line termination have been considered as well.
4
The comparisons reported on this paper are relative to
the measurements made in CSELT for which the setup is
shown in figures 5 and 6.
H=1.01 mL=10 m
h=2 m
Metal floor of semi-anechoic chamber
on board
batteries
front side of the board
Figure 5: Measurement set-up in CSELT
Figure 6: Far field measurement in CSELT
The test board is placed on a wooden table in vertical
position, at a fixed height of 1.01 m above the ground
plane of the room. At 10 m distance the receiving
antenna is placed at a fixed height of 2 m. The board is
supplied by 4 nickel cadmium batteries placed on the
solder side of the board itself. No shielding structures
(for example boxes, conducting ribbon, etc.) have been
used to limit the radiation of batteries as it is shown in
the figure 7.
Figure 7: Back of the board, presence of battery holders
without e.m. shield
The envelope of measured and simulated maximum
electric field are shown in figure 8. The measurement in
vertical polarisation has been obtained with a 120 kHz
bandwidth spectrum analyser.
108
109
0
5
10
15
20
25
30
35
40
measurement
Simulation
Frequency in Hz
Ele
ctri
c fi
eld i
n d
B
V/m
Figure 8: Comparison between the envelope of
maximum measured and simulated electric field
The envelopes of simulated and measured radiated
emissions are similar and show a minimum at around
400 MHz. But between 100 MHz and 350 MHz, the
simulation underestimates the measurement results of
5
some 15 dB. In order to investigate this difference, some
measurements in near field have been performed. The
test bench shown in figure 9 has been used to measure
the magnetic field radiated at a distance of 1cm from the
back of the board. It uses a spectrum analyser HP
8595EM, a near field probe HP 11940A and a three-axis
table for scanning the test board.
Figure 9: Three-axis scanning set-up for near field RE
measurements, the top layer of the board is shown
In figure 10 the magnetic field measured at 1
cm from the back side of the board near the batteries is
shown.
102
103
20
30
40
50
60
70
Frequency [MHz]
Hfield [dB a/m]
Figure 10: Measured magnetic field at 1 cm from the
back side of the board near the batteries
It is important to notice that the same measurement
made far away from the batteries on the back side of the
board shows emissions due to edges effects that are
much lower than the ones of the batteries themselves.
The radiation measured from the batteries is not
produced by the batteries themselves but by the power
and ground noise ( I noise) which is re-injected from
the active components over the power and ground planes
into the batteries, batteries connections and the power
routed net. These measurements show that the radiation
of the signal nets is present along the whole frequency
range, whereas the radiation due to the batteries is
restricted to the 100 MHz, 350 MHz frequency range.
This range is exactly the same where the prediction
underestimates the far field measurements. As no near
field to far field interpolation can be practically used, we
can not make any quantitative explanation but we can
reasonably justify the impact of the batteries emission
on the far field measurement, looking at its spectral
distribution in near field.
On the first designed board, due to the absence of
accessible ground plane, it was impossible to introduce a
shield. So a new version of the board has been produced
with an “EMI approach”. On the back (solder side) of
the new board a ground metallization has been brought
to the surface through several plated-through holes as it
is shown in the figure 11 and the batteries have been
mounted on it.
By this way a flexible EMI shield has been soldered as
shown in the figure 12: it covers all the batteries and
cuts the spurious radiation from the back of the board.
By this way the radiated field from the new test board
can be reasonably supposed coming only from the front
of the board (signal traces side). The figure 13 shows the
comparison between the envelop of maximum measured
radiated emissions by the original board and the new
one.
.
Figure 11: Top and bottom view of the new board
6
Figure 12: EMI shield soldered to the ground
metallization on the back of the board
108
109
0
5
10
15
20
25
30
35
40
Original board
New board
Frequency in Hz
Ele
ctri
c fi
eld
in
dB
V
/m
Original board
New board
Frequency in Hz
Ele
ctri
c fi
eld
in d
B
V/m
Figure 13: Comparison between the envelopes of
maximum measured radiated emissions by the original
board and the new one
A comparison between the radiation spectrum of the old
and new implementation shows a significant difference
of emission levels in the frequency range between 70
MHz and 350 MHz where the radiation of the batteries,
batteries connections and the power routed net of the
original board plays its role. This confirms the
hypothesis made above.
We can now compare the simulation results already
obtained above to the measurement of radiated
emissions due to the new board as shown in figure 14:
Measurement
Simulation
Frequency in Hz
Ele
ctri
c fi
eld
in
dB
V
/m
Figure 14: comparison between measurement and
simulation of the envelope of maximum radiated
emissions by the new board
The envelope of measured radiation spectrum is well
reproduced by the simulation and the gap between
measurement and simulation is less than 4 dB until 300
MHz. After this frequency the gap increases. It can be
partially due to the validity limit of models used for
active components as an example has been shown in
figure 4. This limit could be extended by the means of
Time Domain Reflectometer measurements made on a
sample of the devices. Such measurements permit to
replace the value of the output capacitance of a driver
that often suffers from uncertainty due to the dispersion
of this capacitance value by a Scattering parameter in
time domain. This topic has been described in [7].
It is important to notice that only 15 minutes on a SUN
ULTRA 1 workstation were required to simulate the
currents and RE of all the 107 nets of the board that
presents 177 components.
IV. CONCLUSIONS
An analytical method to predict RE due to
differential mode currents on PCB traces has been
presented, it takes into account effects of dielectric layer
in the field calculation and does not need to discretise
traces in short segments. It can simulate all traces of a
PCB in a reasonable computation time. The comparison
between measurements and simulations made on an
active test board shows that all couplings involved in the
measurement set-up must be well understood to be able
to explain the results of the comparison. Measurements
in near field permitted to show the radiation of the
batteries, batteries connections and the power routed net
of the board. A new design of the board was
implemented to cut this spurious effect and permitted to
validate the method used to predict the differential RE
from PCB traces.
But this experience also shows how much can be
important to include the radiation from the
power/ground distribution. The 4-ports models for
7
digital drivers used in PRESTO post-layout analysis tool
allows the simulation of Simultaneous Switching Noise
on power and ground pins of active components taking
into account the actual loads. It is then possible to
predict I noise on routed traces and on planes by the
means the Transmission Line Method (TLM) . These
features are being used through a european ESPRIT
ESD project in order to predict the RE from the power
distribution network of a board.
A good modelling of components which can avoid
uncertainty due to the dispersion of their electrical
parameters is also necessary for reliable simulations,
especially in the higher frequency ranges.
The simulation software used for signal integrity and
e.m. predictions as well the near field equipment used
for measurements are included in the THRIS [8]
environment. On the basis of results coming from this
kind of experimental validation, further improvements
of radiated field prediction will be included in the future
release of this tool.
REFERENCES
[1] E. Leroux, F. Canavero, G. Vecchi, "Prediction of
radiated electromagnetic emissions from PCB
tracks based on Green dyadics," Proc. EURO-
DAC, Brighton (UK), Sept. 18-22, 1995, pp. 354-
359.
[2] C. Felsen, N. Marcaviz, Radiation and scattering
of waves, Chp. 5, Prentice - Hall, Eaglewood
Cliffs, 1973
[3] S. Forno, S. Rochel, "Advanced simulation and
modeling techniques for hardware quality
verification of digital systems," Proc. EURO-DAC,
Grenoble (F), 1994 [4] EN55022, “Limits and methods of measurement of
radio interference characteristics of information technology equipment”, 1985
[5] Monteath, Applications of the Electromagnetic
Reciprocity Principle, Pergamon Press, 1973.
[6] R. De Smedt, W. Temmerman, “Using an Active
Test Board for Evaluation of EMC Tools on PCB
Level”, COST Action 243 EMC Workshop,
Paderborn (Germany), 7-8 April 1997
[7] E. Leroux “Conception et validation d’une
méthode numérique hybride appliquée à la
prédiction du rayonnement d’une carte
électronique connectée à son câblage”, PhD thesis,
University of Lille, june 1998
[8] P.Belforte, G.Guaschino, F.Maggioni, “A new
system for the evaluation of telecom hardware
robustness”, EMC’96 Roma