Professional Documents
Culture Documents
The general methods used to reduce EMC problems can be divided into four classes. They are proper
layout of components and cables, system grounding and bonding, shielding, and surge suppression
(diversion) and filtering.
One of the most cost-effective way of controlling EMI in a system is by proper layout of the subsystems.
This can be explained using the following example.
EMP Lightning
Conducted
Shield I
transients Strong
Sensitive (confining)
internal
circuit I source
Sensitive Weak
No shield
circuit II internal
Interference source
diverters
(filters, surge
limiters, bonding etc)
Shield I
Shield II
(excluding)
(excluding)
A two-layer shielding topology (geometry) for controlling internal and external interference sources is given
in the above figure. All the sensitive (critical) circuits are physically grouped together, as far as possible,
and provided with a shield that exclude fields external to them from affecting it. Similarly, strong internal
RT/2000 67
sources are grouped together and provided with a shield that confines the emission to within the enclosed
shield volume. The remaining weak internal sources and non-critical components are physically grouped
together without special enclosing shield. All the connection from sensitive circuits and strong internal
sources are controlled by interference diverters. All the subsystems are surrounded by an external shield
that exclude external electromagnetic disturbance (e.g., lightning). All the connections penetrating the
external shield (e.g., power, data, telephone, pipe line etc.) are provided with interference diverters.
In spite of the sensitive layout and shielding, interference fields and currents may penetrate inside a) along
insulated conductors passing through the shields, b) through openings or imperfections in the shields, and c)
by diffusion through imperfectly conducting shields. The objective of system hardening (making coupling
path inefficient) is to control these interference penetration at each shield, so that interference reaching the
sensitive circuit is within the tolerance of the circuit. In the next section we will consider the conductor
penetration of the shields and apertures in the shields in detail.
A shield has two surfaces: external and internal. If a conductor carrying interference currents is
connected to the outside of the shield, interference currents are confined to the outside of the shield and has
very little influence in the shield volume enclosed. If the conductor is connected to the inside surface of the
shield, all interference currents are available inside the shield and may couple with circuits inside the shield
volume which could potentially cause EMI.
Shield
Shield
I
I
Inside
shield
Outside Outside
shield Inside
shield shield
Current Current
density density
Fig. 4.2 Confinement of conductor current to Fig. 4.3 Conductor current injected on
the outside surface by skin effect. the inside of a shield.
In short, to preserve the integrity of the shield, interference current of the external origin must be diverted to
the outside surface of the shield as shown in Fig. 2. Several examples of the proper application of this
principle, together with some common compromises and serious violations are given in the following
Figures.
RT/2000 68
Proper Compromise Serious
violation
Outside Inside
Internal
External ground
ground
Cable shield
Metallic pipe
Insulated
Conductor Surge Arrestor
or Filter
RT/2000 69
Aperture:
The penetration of external fields through apertures that are small (compared to wavelength) is illustrated in
the following Figure.
}
Internal
cable
xternal electric field
Internal
cable
Shield
Aperture
Sometimes the aperture is a long slit (e.g., at the hinges) or may be a large opening. The fields penetrating
a small aperture depend on the aperture size. Therefore if the aperture area is subdivided into many small
areas (e.g., using a conductive mesh) the field penetration is reduced. If the aperture walls were made
thick they may behave as waveguides beyond cut-off reducing the field penetration even more. Fields
transmitted through a waveguide beyond cut-off are attenuated exponentially with distance along the guide.
Above ideas are illustrated in the following Figures.
RT/2000 70
Shield Shield Shield
When slits are present in the shield, do not lay cables parallel to the slit.
1) Ground has an impedance. At high frequencies (MHz range) it is the inductive impedance that is
dominant. Therefore at high frequencies we can not assume that currents will return along intended path
through the ground wire. Currents will return along the path of lowest impedance. It is quite possible that
some frequency components of a signal will return along one path while other frequency components of the
same signal may return along other paths. Currents returning via undesired paths may cause interference.
RT/2000 71
Undesired path
Sub system I Sub system II
Desired path
Ground wire
2) Single point vs. multipoint grounding. There are two common grounding methods that can be seen in
practice. In a single point ground system, the sub-system ground returns are tied to a single point within
that system as shown.
Source
I3
I1 I2
As we have seen before, single point ground system avoids the problem of crosstalk through common-
impedance coupling. However single point ground systems has its disadvantages. In a distributed system
the individual ground wires of the subsystems may be long and can have higher impedance or act as short
transmission lines, causing many problems. The return currents in ground wires may radiate and couple
with each other producing crosstalk. Higher frequency components will radiate and couple more efficiently
than lower frequency components. Therefore the single point ground system is not universally ideal, but
works best for low-frequency subsystems.
In a multipoint ground system a large conducting plane or grid serves as the return path and the individual
grounds of the subsystems are connected at different points on the ground plane. It is made sure that the
ground return to which individual grounds are terminated have a very low impedance between any two
points at the frequency of interest. Figure below shows an example of multipoint grounding. The
impedance between points A, B, and C are negligibly small so that there can not be any common-
impedance coupling.
RT/2000 72
Multipoint ground system require only short ground leads compared to singlepoint ground system. For the
success of the multipoint ground system it is essential that the ground plane is not contaminated by currents
from other sources. High currents on the ground plane can produce large potential difference between
sub-system ground points (e.g., A, B, C in Fig. 10) causing EMC problems. The problem of noisy
grounds can be avoided by providing dedicated ground plane for the purpose of multipoint earthing of the
subsystems.
I1 I2 I3
I1+I2+I3 I2+I3
A B C
Ground plane or grid
Typically, multipoint ground systems are used in digital subsystems where the mutual interference between
subsystems due to common impedance crosstalk and coupling between ground leads are of great concern.
Singlepoint ground systems are used in analog subsystems where low-level signals are involved. In these
cases, millivolts or even microvolts ground drops can create significant common-impedance coupling
interference problems. Single point ground systems are also typically employed in high-level subsystems
such as motor drives, where the purpose is to prevent the high level return currents from developing large
voltage drops across a common ground net.
3) Segregation of ground systems. A typical system require three separate ground systems as shown.
Low voltage/current/power subsystems are tied to a single dedicated ground point (signal ground). Within
the signal-ground subsystem, circuits may utilize singlepoint grounding, multipoint grounding or a
combination of both. The noisy ground system represents circuits that operate at high levels of
voltage/current/power. The hardware ground is connected to chassis, cabinets, racks etc. and is not
intended to carry currents unless there is a fault or for diverting ESD currents. The main purpose of
providing the distinct ground systems is the prevention of common-impedance coupling.
High Power
Digital Analog supply
power
RT/2000 73
Mains
DC Load
ZL
VS
(a)
I
Mains
DC
I Load
V ZL
S I
I I
N I
N
I
N I
N
Z1 Z 2 Z3
I
I
N
(b)
DC Mains
VS
IN
Z L Load
I
N
Z2 Z
3
(c)
Fig. 4.13. An example illustrating the problem of common-impedance coupling in an electronic circuit
and its solution. a) Electronic circuit with its ground connections b) Same circuit as in (a) showing
ground impedances (Z1, Z2, Z3), load current (I), and common-mode current (IN). c) Ground
connections that avoid the problem of crosstalk due to common-impedance coupling.
RT/2000 74
4.3 Shielding (or screening)
Shielding keep the external disturbances outside and keep internal signals inside by attenuating unwanted
electric and magnetic fields and by providing a low impedance diversion for unwanted currents. A
measure of the shielding effectiveness is the ratio of the field strength in a protected volume in the absence
of the shield to the field strength with the shield in place. However, this ratio is difficult to measure since the
fields are influenced by the presence of the shield. A more practical measure of the shielding effectiveness
(shielding factor) is
S(dB) = 20 log10S
The above definition is suitable for large structures like cabinets and shielded rooms, but, is not practical for
some configurations like shielded cables.
In principle, the problem of shielding can be studied by solving Maxwells equations with appropriate
boundary conditions. However, analytic solutions are possible only for a few idealized shield shapes
composed of well-defined homogeneous materials excited by well-defined fields. Studying these simple
cases help the engineer to understand how a shield functions and enable the engineer to design specific
practical shields whose design may not be available in standard design tables.
First we will see how a metal plate is shielding uniform plane waves. Later the theory will be extended to
near fields from electric and magnetic dipoles.
Fields are attenuated due to absorption in metal plate due to skin effect, reflections at air-metal/metal-air
interface, and due to multiple reflections in metal plate. Then the total shielding factor is given by S = SA
SR SMR, where SA, SR, SMR are shielding factors due to absorption, reflection, and multiple reflection,
respectively, taken one at a time. When expressed in dB, S(dB) = SA(dB) + SR(dB) + SMR(dB).
Absorption and reflection increase the shielding effectiveness where as multiple reflection tend to decrease
the shielding effectiveness. Multiple reflection is important only for extremely thin shields, hence they are
not considered hereafter.
Current and fields inside a metal plate due to skin effect is given by,
x
J x = J0e
x
E x = E0 e
0 x d, - skin depth (2) air metal air
x
Hx = H0 e
RT/2000 75
where J is the current density, E is the electric field, and H is the magnetic field. The electric and magnetic
fields are assumed to be incident normally on the plate. Shielding effectiveness due to absorption is defined
as,
E0 H
SA = = 0 = ed /
Ed Hd
(3)
d
S A ( dB) = 8.7 ( dB)
Absorptive attenuation, SA, inside metal is roughly 8 dB per skin depth. Skin depth of a metal is given by
2 1
= = (4)
f
Skin depth is higher for lower frequencies. Therefore absoptive attenuation is lower for lower frequencies
for same thickness of metal. Substituting (4) in (3),
S A (dB ) = 8.7d f (5)
Magnetic materials (e.g., iron) have higher absorptive attenuation than non-magnetic materials (e.g.,
aluminium).
Partial reflection and transmission occures twice, at air-metal interface, and at metal-air interface.
Ed
Ein E0 Eout
0 d
The mathematical problem of reflection and transmission of normally incident plane waves at material
boundaries is analogous to the problem of reflection and transmission at impedance discontinuities in a
transmission line. The electric field is analogous to voltage and the magnetic field is analogous to current.
At the first air-metal interface,
2Z m 2Z w
E0 = Ein and H 0 = H in (6)
Zw + Zm Zw + Zm
where Zw is the wave impedance in air which is 377 ohm for plane waves, and Zm is the wave impedance
in metal.
2Z w 2Z m
E out = E0 H out = H0 (7)
Zw + Zm Zw + Zm
RT/2000 76
For metals (<<) Zm (8)
For example, impedance of copper (Zm) at 50 Hz is 2.6x10-6 , where as at 1 GHz it increases to 0.012
. That is Zm << Zw in most situations. Therefore from (6) and (7) we can draw following conclusions.
Most of the electric field is reflected at the air-metal (first) interface itself, whereas most of the magnetic
field is transmitted through the air-metal interface. Therefore attenuation in the shield thickness is more
important for magnetic fields. For magnetic fields significant reflection occurs at the metal-air (second)
interface only.
For plane waves, or for shields in the far-field (r>>) region of the source, Zw = Z0 = 377 .
Therefore,
Z0
S A ( dB) = 20 log10
4 Zm
Z0 r Cu
= 20 log10 (10)
4 r 0
= 168 + 10 log 10 r
r f
Note that attenuation due to reflection is greatest at low frequencies and high conductivity materials, and
smallest for high frequencies and magnetic materials. Attenuation due to reflection decreases at 10
dB/decade with increase in frequency.
For shields in the near-field region (r<<) of a source, Z0 is to be replaced by the appropriate
wave impedance. We have seen before that wave impedance in the near-field region of an electric dipole
(e.g., short straight wire) is given by
E 1
ZE = (11)
H 2f r
Wave impedance in the near-field region of a magnetic dipole (e.g., a small loop of wire) is given by
E
ZH = 2f r (12)
H
Values of ZE and ZH are given below for some frequencies at a fixed distance of 1 m, calculated using (11)
and (12)
frequency (Hz) ZE () ZH ()
RT/2000 77
50 3.6x108 3.95x10-4
103 1.8x107 7.90x10-3
106 1.8x104 7.90
A short piece of wire having a voltage (electric dipole) produces a dominant electric field in its vicinity and
a small current carrying loop (magnetic dipole) produces a dominant magnetic field in its vicinity. Practical
circuits are a combination of electric and magnetic dipoles. It can be seen from the table above that ZE is
very high and ZH is very low in the near-field region. It is relatively easy to shield against high impedance
electric field where as it is difficult to shield against low impedance magnetic field. This is illustrated in the
following numerical example.
I A shielded enclosure is made of 0.5 mm thick aluminium sheet with r = 1, r = 0.6. Find the shielding
effectiveness of the enclosure for the following situation, considering only attenuation due to absorption and
reflection for a nominal frequency of 1 kHz. Use 0 = 410-7 H/m, Cu = 5.8107 S/m.
Can we use aluminium foil of thickness considerably less than 0.5 mm to get almost the same shielding
effectiveness against electric fields at 1 kHz? Justify your answer.
_________________________________________________
Solution:
1
Skin depth = = 2.7 mm
f
d d
Attenuation due to absorption S A = 20log 10 e
8.7
d = 0.5 mm
Therefore,
SA at 1 kHz = 1.6 dB
Attenuation due to absorption is the same for near-field electric sources, and near-field magnetic sources.
(Z w + Z m ) 2
SR =
4Z m
Intrinsic impedance of aluminium, Zm = 151 107
a) FAR-FIELD SOURCE
Wave impedance for far-fields, Zw = Characteristic impedance of the medium (air) = 377
RT/2000 78
377
SR
4 Zm
377
S R ( dB ) = 20 log 10 ( )
4 Zm
SR at 1 kHz = 135.9 dB
distance, r = 10 cm = 0.01 m
1
Near-field wave impedance in air for electric field source at 10 cm, Z E =
2f 0 r
ZE at 1 kHz = 1.8109
ZE
SR
4 Zm
ZE
S R ( dB ) = 20 log 10 ( )
4 Zm
SR at 1 kHz = 269.5 dB
Screening effectiveness at 1 kHz against both far and near-electric field sources is mainly due to reflection
at the first surface. Therefore enclosure made of thin aluminium foil (over some supporting structure) give
essentially similar shielding effectiveness against low frequency electric fields.
distance, r = 10 cm = 0.01 m
ZM at 1 kHz = 79010-7
( Z M +Z m )2
SR =
4 Z M Zm
( Z M + Zm ) 2
S R ( dB ) = 20 log 10
4 ZM Zm
SR at 1 kHz = 5.4 dB
RT/2000 79
i) the frequency is 1 GHz
ii) if an iron sheet of same thickness is used (r = 1000, r = 0.1). Frequency is 1 kHz.
The dominant shielding mechanism for near-field magnetic sources is the absorption loss.
However, both reflection and absorption losses are very small for low-frequency magnetic sources.
d
S dB = 8.7 = 8.7 f
To increase absorptive attenuation, use shield material with very high or . There are three possible
methods to increase low-frequency magnetic field shielding.
2. Special magnetic materials with high Special magnetic materials with high permeability (). The
magnetic fields are ducted into the magnetic material away from the protected volume.
e.g., Mumetal r=20000 at 1 kHz, r=0.03. That is, rr=600. For copper rr=1.
3. By generation of opposing flux. Induced currents in shorted turns create a magnetic flux that opposes
the main flux.
Practical shields
For shields thick enough to be self-supporting (d1 m.m) and for frequencies greater than 10 kHz, the
shielding theory predicts a shielding effectiveness in excess of 200 dB. However, in general, a shielding
effectiveness of more than 100 dB is very difficult to achieve for practical shields. Imperfections in
practical shields such as joints, rivets, corners, seams, and holes reduces the shielding effectiveness.
Shields made of thin conducting films can provide useful attenuation in most practical cases. Low
frequency magnetic fields are difficult to shield and usually determine the lower performance limit at those
frequencies. Also, cables and other conductors entering the shielded enclosure may carry unwanted
signals.
Shielded cables are used to protect the signal wires from external interference and also to provide a return
path or reference (e.g., the common coaxial cable type RG58). Consider a shielded cable above a
conductive reference or ground plane. The shield (screen) at the ends of the cable is connected to metallic
enclosures (containing electronic equipment) so that there is perfect electrical continuity are there are no
RT/2000 80
gaps through which electromagnetic fields can penetrate. The reference (enclosures) may or may not be
connected to the ground plane. When an electromagnetic field impinges on the system, a voltage Vs and
current Is develop in the transmission line system formed by the screen and the ground plane. In an ideal
situation, when the screen and enclosures are completely impervious to the electromagnetic fields, no
voltage with respect to the reference appears in the inner conductor. That is Vc=0, V0=Vs. However, in
real situations, a parasitic voltage appears between the inner conductor and ground because of
1) imperfect metallic enclosures
2) improper connection between screen and enclosure
3) imperfect cable shielding
H
Enclosure Enclosure
(reference) (reference)
Cable screening
VC VC
IS
VS V (z)
0
I 0 (z)= I S + I
In this section we will consider the imperfections in cable shields and discuss about the shielding
effectiveness of shielded cables.
External electromagnetic fields can penetrate the cable shield in three different ways.
a) Diffusive penetration of fields and currents to the inside of the cable shield. Due to the shield current Is,
(either induced by the fields or injected by some means) an axial electric field (or potential gradient) Ez is
developed on the inner shield surface.
Vc = E z dl along the length of the cable for a homogeneous screen. Within the homogeneous screen,
the azimuthal component of the magnetic field H due to the shield current Is is zero, because of symmetry.
Ez
IS
VC
b) In a braided shield, external electric and magnetic fields penetrate through the holes and couple with the
internal signal wires. Diffusive penetration as in (a) also happens.
RT/2000 81
Er
c) In a conductive spiral tape shield, the shield current flow is not parallel to the axis of the cable. The
spiral currents produce an axial magnetic field Hz inside the cable. This field induces a voltage in the inner
conductor because the conductors may not be always parallel to the magnetic field. Diffusive penetration
as in (a) also happens. Spiral tapes usually overlap each other preventing any holes.
IS
Hz
The description of shielding effectiveness as the ratio of the field outside to the field inside, as in the
case of shielded enclosure, is not convenient in the case of shielded cables. Shielded cables are generally
coaxial in structure and a transmission line (TL) is formed between the internal signal wire and the outer
shield. Additionally, another TL is formed between the shield and the ground plane, which determines the
shield current Is due to disturbance (Fig. ). The influence of the shield current on the TL formed by the
signal wire and shield can be modelled by including a distributed series voltage source and a distributed
parallel current source to the usual TL (Fig. ). The voltage and current sources are related to the properties
of the shield using the concept of transfer impedance and transfer admittance.
RT/2000 82
Ic
VC
V0 (z) V0 V0 (z)
I 0 (z) = Is + Ic
I + dz
Ez dz I
IC Z dz z
~
VC Y dz V
J(z) dz V + dz
z
dz
Transfer impedance
Consider a cable length dz, which is electrically short, and open circuited at both ends. Then signal current
Ic=0. Let Ez(z) be the axial field inside the shield due to the disturbing current on the shield. Then Ez(z)dz
is the series voltage developed over the length of cable dz. Define Ez(z)dz = ZTI0dz. From the equivalent
circuit diagram we obtain (neglecting the injected current source J(z)dz)
dV
E z ( z )dz = Z T I 0 dz = dz
dz
which implies that
1 dV
ZT = I = 0 /m (13)
I 0 dz
The transfer impedance ZT gives the open circuit voltage developed between the internal conductors and
the shield, for one ampere of the shield current, in a cable one meter long, at wavelengths greater than one
meter.
Transfer impedance is a property of the shield and always contain a diffusion component that relate the
shield current to the longitudinal electric field inside, due to finite conductivity of the shield material.
RT/2000 83
Additionally, for leaky shields such as braided shields and tape-wound shields it contains a mutual
inductance term that account for coupling through apertures or solonoidal currents in the shield.
Transfer admittance
Consider a cable of length dz, which is electrically short and inner conductor short circuited to the shield at
both ends. Then the voltage Vc between the inner conductor and shield is zero. Define J(z)dz = -V0YT dz.
From the equivalent circuit diagram we obtain (neglecting the series voltage sources),
dI
J ( z ) dz = V0YT dz = dz
dz
(14)
1 dI
Therefore, YT = V =0
V0 dz
V0 is the voltage between the internal conductors and the external structure and dI/dz is the current per unit
length flowing into the internal conductor from the external structure. The transfer admittance gives the
short circuit current induced in the internal conductor (internal conductors shorted to the shield), for one
volt applied between the shield and external structure, in a cable one meter long, at wavelengths much
greater than one meter.
The transfer admittance of metal tubular shields with no apertures is negligibly small. For shields containing
apertures, transfer admittance contains a mutual capacitance term that accounts for capacitive coupling
between the internal conductors and the external structure.
C1
V0
C2 C12
Transfer impedance characterizes the effectiveness of cable shield with respect to the disturbance shield
current Is while transfer admittance characterizes the effectiveness of the shield with respect to the
disturbance shield voltage V0.
From consideration of skin effect we know that low-frequency shield currents can penetrate (diffuse
through) the shields easily compared to high frequency currents. Therefore voltage Vc appearing between
the internal conductors and the shield (parasitic voltage) decreases with increase in frequency. For this
reason shielding performance of homogeneous shields are better at higher frequencies. However, parasitic
voltage due to magnetic coupling through holes in braided shields and due to axial magnetic field in
conductive tape wound shields increases with increase in frequency.
RT/2000 84
4.4 Diversion and filtering
Protection strategy:
Upstream Downstream
Surge
Z1
Z2 Protected Port
1. Surge protection circuit should have minimum influence in the normal operation of the protected system.
That is, series impedance should be very small (Z1<<Z2) and shunt impedance should be very large
(Z2>>ZL) for normal signal voltage and frequencies.
2. During abnormal conditions (during a surge) the series impedance should be very large (Z1>>ZL) to limit
the surge current and shunt impedance should be very small (Z2<<ZL) to divert the surge current. Note
that Z1, Z2 and ZL may be functions of frequency, voltage or current.
RT/2000 85
3. Surge protection circuits should not be damaged by the surge themselves.
During electrical overstress, voltage of the surge are larger than the normal system voltage. Therefore shunt
elements with non-linear voltage-current (V-I) characteristics can provide very low impedance during
overvoltage conditions and very high impedance during normal system voltage. The overvoltage
waveforms will have very fast rising or falling portions (very high rate of change of voltage or current).
Therefore the shunt elements are required to respond very fast, in nanoseconds. Examples of such
components are spark gaps (or gas tubes), varistors, and avalanche diodes. Electrical characteristics of
these devices will be inspected in detail in the next sections.
One of the earliest transient surge protector was the spark gap in air between two carbon blocks. A
serious disadvantage was that the carbon blocks eroded after conducting high energy transients. The
widening gap due to eroded blocks changed the electrical characteristics of the gap with time. The modern
day spark gap is between metal electrodes in a sealed tube containing a mixture of noble gases (neon,
argon, etc.). Miniature low-voltage spark gaps sealed in ceramic tubes can conduct transient current
pulses of 5 to 20 kA for 10 s without appreciable damage to the spark gap. Of all the non-linear shunt
protective devices, spark gaps have the lowest parasitic capacitance, typically between 0.5 to 2 pF.
Therefore sparkgaps can be used even in applications where the signal frequencies are in excess of 50
MHz.
Electrical characteristics:
The operation of a spark gap (gas tube) can be explained with the help of Fig. 4.2 which shows the typical
response of a low voltage gas discharge tube to an applied sinusoidal over voltage. Fig. 4.2a shows the
applied sinusoidal overvoltage and the voltage across the spark gap, Fig. 4.2b shows the current flow
through the spark gap, and Fig. 4.2c shows the voltage-current curve (V-I curve) of the spark gap. As the
voltage across the spark gap is slowly increased (Fig. 4.2a), the gap fires at voltage Vs, bringing down the
voltage (Fig. 4.2c). That is, at Vs the gap switches from the insulating state (resistance > 10 G) to the
conducting state (resistance <0.1 ). The change of state can happen within a fraction of a microsecond.
The voltage Vs (90 - 300 V) is called the dc firing voltage of the gap. Later we will see that the actual
sparkover voltage and the response time of the spark gap depend upon the rate of increase of the applied
voltage across the gap.
During the drop in voltage from Vs, the incremental resistance, dV/dI, is negative, that is, this is a negative
resistance region. The current through the gap increases (Fig. 4.2b) and the gap voltage increases slightly
to Vgl, the glow voltage. This region is called the glow region. The glow is produced by a thin layer of
excited gas atoms covering part of the cathode surface and later extending to the whole cathode surface.
Maximum current during the glow region is between 0.1 A to 1.5 A and the glow voltage is between 70 to
150 V. This is a positive resistance region. Electron-ion pairs are produced in the intense electric field
that exist between the electrodes when the spark gap is being operated in the glow region. When they get
sufficient energy they accelerate and collide with neutral atoms or ions, produce more electron-ion pairs,
finally leading to a general break down of the gap (Break down processes in gases will be studied later in
this course). During the arc phase the voltage across the gap reduces to Va (10 - 25 V) and become
virtually independent of the current. The arc current can be very high and is limited mostly by the
magnitude of the overvoltage and parameters of the circuit containing the spark gap. With decreasing
overvoltage, as in the second quarter of the applied sine wave, the current through the spark gap decreases
RT/2000 86
until it drops below the minimum value (0.01 to 0.1 A) necessary to maintain the arc. The arc is
extinguished at voltage Ve, after passing through the glow mode again. The above processes are repeated
during the negative half cycle of the sign wave. The gas discharge tube is a bipolar device. That is, its
characteristics do not depend upon the polarity of applied voltage.
The gas tube can operate either in the glow regime or in the arc regime while giving protection from
overvoltages. Both regimes are associated with a power follow current because the tube do not extinguish
unless the voltage across the tube do not fall below the glow voltage or arc voltage, as the case may be.
Sometimes thermionic emission from hot electrodes may maintain the arc even during the brief zero
crossings of a sinusoidal voltage. Prolonged follow currents can destroy a gas tube by shattering the case
or by melting the electrodes. Therefore it is essential to prevent the follow current after a surge. Follow
current is prevented by putting a varistor in series with the spark gap. More on this method will be
discussed while considering the varistor in the next section.
The sparkover voltage (Vs) and the response time (time to conduct) of spark gaps are functions of the rate
of rise of transient voltage. The spark over voltage increases and the response time decreases with
increase in dV/dt of the transient. For example, if the static spark-over voltage (or dc spark-over voltage),
usually determined by applying a low rate of rise transient (dV/dt100 V/s), of a gas discharge tube is 350
V, the impulse spark-over voltage, usually determined with a fast rate of rise transient (dV/dt=1 kV/s),
can be 750 V. The response time of the gas tube can be about 4 seconds at 100V/s rate of rise of voltage
where as it can be as small as 0.8 microseconds at 1kV/s. Gas tubes may conduct within a few
nanoseconds if the applied transient has rate of rise times about 1 MV/s.
4.4.2 Varistors
Varistors are non-linear semiconductor devices whose resistance decreases as the magnitude of the voltage
increases. Modern varistors are fabricated from metal oxides, with zinc oxide as the primary ingredient. A
typical V-I curve of a metal oxide varistor is shown in Fig. 4.3. Under normal voltages there is a small
leakage current of less than 0.1 mA and the varistor behaves like a simple high value resistor Rleak. During
overvoltage the current through the varistor increases and the voltage is clamped at a level close to the
normal voltage. This is the operating region of the varistor (Fig. 4.3) and the voltage-current relationship in
this region is given by
I = kV (16)
where is a coefficient with values between 25 and 60. The parameter k in (1) has a value extremely
small (<10-100), therefore (1) is expressed in terms of logarithms as
If (V1,I1) and (V2,I2) are two measured data points in the operating region of the varistor, the value of
can be determined as follows.
RT/2000 87
log( I 1 / I 2 )
= (18)
log(V1 / V2 )
V-I characteristics
Symbol
Operating
Region
-300 V +300
Leakage Current
region
At very large currents, usually more than 100 A, the varistor characteristics is dominated by the low value
bulk resistance Rbulk of the device. Usually varistors are fabricated in the form of discs and hence have
large parasitic capacitance values in the range of 0.2 to 10 nF. Including the inductance of the varistor
leads will complete the equivalent circuit of the varistor, which is shown in Fig. 4.4.
IDEAL RLEAKAGE
C
RBULK
Varistors are fast acting devices with response times less than 0.5 ns, if parasitic inductance due to leads
can be avoided.
RT/2000 88
Performance of varistor is affected by temperature. Excessive leakage currents can raise the temperature
of the varistor. Since the varistor has a negative temperature coefficient, the current will increase as it
become hotter, which will increase the current even further resulting finally in a thermal runaway. Varistors
are usually used in protecting electronic systems from transient overvoltages that propagate on the mains.
Power supply filters are low-pass filters commonly connected in series with the power cord of electronic
equipment to attenuate high frequency noise that is generated inside the chassis and conducted on the
power cord out of the chassis into the mains and environment. Low-pass filters may also protect
equipment from conducted high-frequency noise on the mains. The high frequency noise is usually below
the normal operating voltage and hence will not operate the non-linear surge protective devices. Filters are
not used as stand-alone devices to protect against transient overvoltages. However, filters are very useful
in attenuating high-frequency noise down stream of a non-linear protective device (spark gap, varistor).
This high-frequency noise is partly due to the remnants of the transient overvoltage, and partly due to the
action of the non-linear device itself.
V
IL ( dB) = 20 log 10 Lwo (19)
V Lw
RS +
+ VLWO RL
-
VS -
RS +
+ FILTER RL
VLW
-
VS -
SOURCE LOAD
RT/2000 89
RL
The load voltage without the filter is VLwo = V (20)
RS + RL S
RL
The load voltage with filter is VLwo = V (21)
RS + Z F + RL S
where ZF is the series impedance of the filter (ZF=jL in the case of simple low pass filter above). The
insertion loss is the ratio of (20) and (21) and is given by
ZF
IL = 20 log 10 1 + (22)
RS + RL
From (22) it is evident that the insertion loss of a filter depends on the source and load impedance, and
therefore can not be stated independently of the terminal impedance. Usually insertion loss is specified
assuming terminal impedance of 50 ohms.
A power supply filter should give protection to both common-mode (CM) and differential-mode (DM)
noise currents. The filter consists of a common-mode choke, which is two identical windings wound over
the same ferrite core, two capacitors CD1, CD2 between line and neutral on both sides of the choke , and
two capacitors CC1, CC2 between phase/neutral and ground (chassis) on both sides of the choke. The
capacitors CD1, CD2 divert the DM noise currents and capacitors CC1, CC2 divert the CM current. Usually
CC1=CC2, and is kept low (about 2 nF) to limit the leakage current to below 1 mA for safety reasons.
Otherwise there may be a shock hazard if the filter chassis is not earthed properly. Typical values for CD1
and CD2 are in the range of 0.1-0.5 F.
L
L
CD1 CD1
Protected
Source of N port
noise N
RT/2000 90