ETC AN-15

®
®
TOPSwitch Power Supply Design
Techniques for EMI and Safety
Application Note AN-15
This application note presents design techniques that reduce
conducted EMI emissions in TOPSwitch power supplies below
normally specified limits. Properly designed transformers, PC
boards, and EMI filters not only reduce conducted EMI emissions
but also suppress radiated EMI emissions and improve EMI
susceptibility. These techniques can also be used in applications
with DC input voltages such as Telecom and Television Cable
Communication (or Cablecom). Refer to AN-14 and AN-20 for
additional information. The following topics will be presented:
FCCA QP
FCCB QP
100
80
60
40
20
0
0.01
0.1
1
100
10
Frequency (MHz)
Figure 1. FCC Class A and B Limits (Quasi Peak).
Amplitude (dBµV)
40
0
0.01
0.1
1
100
10
Frequency (MHz)
Figure 2. EN55022 Class A and B Limits (Average and Quasi Peak).
80
60
40
20
0
0.01
60
120
Vfg243 QP
Vfg46 AVG
100
PI-1834-042296
Amplitude (dBµV)
100
80
20
Amplitude (dBµV)
Vfg243 QP
Vfg1046 QP
(VDE0871B QP)
EN55022A QP
EN55022A AVG
EN55022B QP
EN55022B AVG
100
PI-1624-111695
120
PI-1623-111695
120
• EMI Specifications for North America, European
Community, and Germany
• Measuring Conducted Emissions with a LISN
• Peak, Quasi-Peak, and Average Detection Methods
• Safety Principles
• EMI Filter Components
• Flyback Power Supply EMI Signature Waveforms
• Filter Analysis
• Power Cord Resonances
• Transformer Construction Techniques
• Suppression Techniques
• General Purpose TOPSwitch EMI Filters
• EMI Filter PC Layout Issues
• Practical Considerations
PI-1622-111695
120
Amplitude (dBµV)
Offline switching power supplies have high voltage and high
current switching waveforms that generate Electromagnetic
Interference (EMI) in the form of both conducted and radiated
emissions. Consequently, all off-line power supplies must be
designed to attenuate or suppress EMI emissions below
commonly acceptable limits.
80
60
40
20
0
0.1
1
10
Frequency (MHz)
Figure 3. Vfg1046 and Vfg243 Class B Limits (Quasi Peak).
100
0.01
0.1
1
100
10
Frequency (MHz)
Figure 4. Vfg243 (Quasi Peak) and Vfg46 (Average) Class B
Limits.
June 1996
AN-15
Safety is a vital issue which determines EMI filter component
selection, the transformer reinforced insulation system, and PC
board primary to secondary spacing. In fact, safety is an integral
part of the power supply/EMI filter design and is difficult to
discuss as a separate issue. Throughout this application note,
design guidance will also be presented for meeting safety
requirements in TOPSwitch power supplies.
INPUT
OUTPUT
LF
LINE
EMI Specifications
CF
The applicable EMI specification must be identified for the
intended product family and target market. In the United States,
the Federal Communications Commission (FCC) regulates
EMI specifications. Canadian specifications are similar to FCC
specifications. Figure 1 shows the conducted emissions limits
governed by FCC rules, Part 15, subpart J. Note that specification
limits are given only for quasi-peak detection methods. A
recent part 15 amendment allows manufacturers to use the
limits contained in C.I.S.P.R. Publication 22 as an alternative
when testing devices for compliance(1).
The member countries of the European Community (EC) have
established a harmonized program for electromagnetic
compatibility. EN55022 for Information Technology
Equipment is one of the first harmonized documents. EN55022
together with companion measurement document C.I.S.P.R
Publication 22 set the conducted emission limits shown in
Figure 2 for information technology products marketed to the
European Community. In fact, EN55022 limits are the same as
C.I.S.P.R Publication 22 limits. Note that class A and class B
specification limits are given for both average and quasi-peak
detection methods(2) (3).
CC
RSL
+
VSL
_
_
RSN
NEUTRAL
CF
VSN
+
CC
LF
PI-1625-111695
Figure 5. Line Impedance Stabilization Network (LISN).
Figure 3 shows the well-known and most stringent VDE 0871
specification (narrow band limits) for German markets which
has traditionally been the design target. German regulation Vfg
1046/1984 requires Information technology or Electronic Data
Processing Equipment to meet the VDE 0871 class B narrow
band limits from 10 kHz to 30 MHz. Note that specification
limits are given only for quasi-peak detection methods. When
marketing products only in Germany, there is a choice between
meeting the regulation requirements of Vfg 1046/1984 or the
new German regulation Vfg 243/1991 (as updated by Vfg 46/
1992) which has relaxed limits from 10 kHz to 150 kHz and is
harmonized with EN55022 from 150 kHz to 30 MHz. Vfg243/
1991 sets quasi-peak limits and Vfg 46/1992 adds mean or
average limits as shown in Figure 4. Figure 3 also shows
Vfg243/1991 class B quasi-peak limits to compare with
VDE0871(4) (5) (6). The EMI filter designed to meet VDE 0871
(per Vfg 1046/1984) will generally be higher cost than the EMI
2
A
6/96
filter designed to meet Vfg/243 regulation requirements.
Measuring Conducted Emissions
Details of testing apparatus and methodology are governed by
the various EMI regulations, but share the same general concept.
Conducted emissions measurements are made with a Line
Impedance Stabilization Network (LISN). Figure 5 shows the
effective filter, represented by LF and CF, inside the LISN which
passes line frequency currents but forces higher frequency
power supply conducted emission currents to flow through
coupling capacitor CC and sense resistor RS. A spectrum
analyzer or EMI receiver reads the current emission signal
magnitude as sensed voltages VSL and VSN across RSL and RSN
in dBµV.
AN-15
LISN Bonded to
Reference Plane Unit Non-conducting
Under Test Table
Load
40 cm
80 cm
80 cm
This Edge Flush Up
Against Vertical
Reference Plane
First
Pulse
80 cm
minimum
height
Steady State Peak Current
AC
0
CURRENT
Conduction Time
≅ 3 mS
PI-1626-111695
Figure 6. Typical Conducted Emissions Precompliance Test Set Up.
Figure 6 shows a typical conducted emissions pre-compliance
test setup on a wooden table at least 80 cm high constructed with
non-metallic fasteners(7). The unit under test, LISNs, and load
are all placed 40 cm from the edge of the table as shown. Six
foot cables are used between the unit under test and both the
LISN on the AC input and the load on the DC output. The LISN
and load are each located 80 cm from the unit under test with
excess cable bundled non-inductively. The edge of the table is
placed flush against a vertical reference plane at least two
meters square. The LISN is bonded to the reference plane with
a low impedance, high frequency grounding strap or braided
cable.
In applications where the power supply and load are located in
the same physical package, the cable can be omitted between
the unit under test and the load.
For design, investigation and precompliance testing, a spectrum
analyzer is highly recommended compared to EMI receivers
which are more expensive and more difficult to use. For
conducted and radiated emissions testing, the spectrum analyzer
should have a frequency range of 10 kHz to 1 Ghz, wide range
of resolution bandwidths (including C.I.S.P.R. specified
bandwidths of 200 Hz, 9 kHz, 120 kHz), built in quasi-peak
detector, video filter bandwidth adjustment capability down to
3 Hz or below for average measurements, maximum hold for
peak measurements, and an accurate and temperature
compensated local oscillator capable of centering a 100 kHz
signal in the display with insignificant frequency drift. The HP
8591EM and Tektronix 2712 (option 12)(8)are two examples of
lower cost spectrum analyzers sufficient for conducted emissions
precompliance testing.
L
AC
IN
V+
ID
CIN
L
VPI-1627-111695
Figure 7. Differential Mode Currents Charging Input Capacitor CIN.
Peak, Quasi-Peak, and Average
Detection
Power supplies operating from the 50 or 60 Hz AC mains use
a bridge rectifier and large filter capacitor to create a high
voltage DC bus from the AC input voltage as shown in
Figure 7. The bridge rectifier conducts input current for only
a short time near the peak of AC mains voltage. Actual
conduction time is typically 3 mS out of effective line frequency
periods of 8.3 to 10 mS which defines an effective “line
frequency duty cycle” of 30% to 36%. Conducted emission
currents can flow in the AC mains leads (and are sensed by the
LISN) only during the bridge rectifier conduction time. The
conducted emissions signal is actually applied to the spectrum
analyzer or receiver detector input only during bridge diode
conduction time which defines a “gating pulse” with pulse
repetitive frequency (PRF)(8)(9) equal to the AC mains frequency
(50 or 60 Hz) and “line frequency duty cycle” just defined. The
“gating pulse” effect due to bridge rectifier conduction time
causes the measured signal magnitude to change depending on
whether peak, quasi-peak, or average detection methods are
used.
A spectrum analyzer or EMI receiver displays the RMS value
of the signal(9). For example, a 100 kHz continuous sinusoidal
A
6/96
3
AN-15
Peak detection is the simplest and fastest method when measuring
conducted emissions. Resolution bandwidth is set to 200 Hz for
measurements from 10 kHz to 150 kHz and set to 9 kHz for
measurements from 150 kHz to 30 MHz. Sweep times are
relatively low. When displaying emissions in real time with no
averaging, the peaks are not constant but change in magnitude
with each measurement sweep due to the bridge conduction
gating pulse effect described above. Most spectrum analyzers
have a “maximum hold” feature which displays the highest
peak occurring over many measurement sweeps. The peak
detector measures the magnitude of the largest signal occurring
during the bridge conduction gating pulse.
The average detector is simply a low pass filter with corner
frequency sufficiently below the gating pulse repetitive
frequency or PRF. In typical spectrum analyzers, the video
filter bandwidth can be reduced to 30 Hz or below to average the
signal but the sweep time must be increased for a calibrated
measurement. For test purposes, the full conducted emissions
range starting at 10 kHz (or 150 kHz or 450 kHz, depending on
the regulation) up to 30 MHz should first be examined with a
peak detection measurement. Peak detected emissions with
insufficient margin compared to the regulation average limit
should be centered on the spectrum analyzer display with the
lowest possible frequency span per division setting before
reducing video bandwidth and performing the average
measurement sweep(10) . Figure 8 shows typical conducted
emissions from 10 kHz to 500 kHz with both peak detection and
average detection. Note that peak detection picked up an
envelope of high order harmonics from line frequency
rectification in addition to the fundamental and first three
harmonics of the 100 kHz switching frequency.
The quasi-peak detector is designed to indicate the subjective
annoyance level of interference. As an analogy, a soft noise that
happens every second is much more annoying than a loud noise
4
A
6/96
PI-1628-111695
110
100
Amplitude (dBµV)
voltage when viewed on an oscilloscope may have a peak
voltage of 1 Volt and hence an RMS voltage of 0.707 Volts. The
spectrum analyzer (with 50 Ω input) will display a value for this
100 kHz signal of 0.707 volts (or 117 dBµV or 10 dBmW)
regardless of which detection method is used (peak, quasi-peak,
or average) because the signal is continuous, narrow band, and
not modulated or gated. If the signal was broadband, modulated,
gated at a duty cycle, or in some other way not continuous, the
displayed RMS value will change with the detection method.
The measured display will then be the magnitude of an equivalent
continuous sinusoidal signal with an RMS value equal to the
RMS content of the LISN signal measured at the output of the
detector stage.
90
80
70
Peak Data
60
50
40
30
Average Data
20
100
200
300
400
500
Frequency (KHz)
Figure 8. Peak Data vs Average Data.
that happens every hour. A quasi peak-detector (actually a
calibrated, intermediate bandwidth video filter) behaves as a
leaky peak detector that partially discharges between input
signal pulses. The lower the pulse repetitive frequency (PRF),
the greater the dB differential between the peak and quasi-peak
measured response (8) (9).
Quasi-peak and average detection methods will always give a
lower measured value compared to peak detection. If a peak
detector measurement meets the average or mean specification
limit with sufficient margin, additional measurements using
average detection are not necessary. When no average limit is
specified, if the peak measurement meets the quasi-peak limit
with sufficient margin, additional measurements using quasipeak detection are not necessary. In general, when testing
TOPSwitch power supplies to the C.I.S.P.R. Publication 22,
EN55022, or Vfg 243/91(and Vfg 46/92) limits, peak measured
data usually meets the quasi-peak limit but, in some areas, may
have insufficient margin when compared with the average
limit. In this case, further measurement is necessary using
average detection.
Safety Principles
Safety principles must be examined before proceeding further
with EMI filter concepts because safety requirements place
several constraints on EMI filter design.
Virtually all equipment including computers, printers,
televisions, television decoders, video games, battery chargers,
etc., must be safety recognized by meeting the safety standard
for the intended market and carrying the appropriate safety
mark. Safety principles are very similar among the various
standards. This application note will focus on the electric shock
hazard requirements of one popular standard, IEC950(11) .
AN-15
The European International Electrotechnical Commission
Standard IEC950 entitled “Safety of Information Technology
Equipment Including Electrical Business Equipment” provides
detailed requirements for safe equipment design. Application
of IEC950 is intended to prevent injury or damage due to
hazards including electric shock, energy hazards, fire hazards,
fire, mechanical and heat hazards, radiation hazards, and
chemical hazards. IEC950 specifies the following definitions
and requirements applicable to TOPSwitch power supplies.
(This is only a partial list of the key requirements targeted
specifically at typical TOPSwitch power supply
implementations. The appropriate IEC950 section is identified
by parentheses.)
IEC950 Definitions (Applicable to TOPSwitch Power
Supplies):
(Introduction): Electric shock is due to current passing through
the human body. Currents of approximately 1 mA can cause a
reaction in persons of good health and may cause indirect
danger due to involuntary reaction. Higher currents can have
more damaging effects. Voltages up to about 40 V peak, or 60
VDC are not generally regarded as dangerous under dry
conditions, but parts which have to be touched or handled
should be at earth ground potential or properly insulated.
(1.2.4.1): Class I Equipment: equipment where protection
against electric shock is achieved by:
a) using basic insulation, and also
b) providing a means of connecting to the protective
earthing conductor in the building wiring those conductive
parts that are otherwise capable of assuming hazardous
voltages if the basic insulation fails.
(1.2.4.2): Class II Equipment: equipment in which protection
against electric shock does not rely on basic insulation only, but
in which additional safety precautions, such as double insulation
or reinforced insulation, are provided, there being no provision
for protective earthing or reliance upon installation conditions.
(1.2.8.1): Primary circuit: An internal circuit which is directly
connected to the external supply mains or other equivalent
source. In a TOPSwitch power supply, this includes the EMI
filter, discrete or common mode chokes, bridge rectifier,
transformer primary, TOPSwitch, and any components
connected to TOPSwitch such as primary bias windings and
optocoupler output transistors.
(1.2.8.5): Safety extra-low voltage (SELV) circuit: A secondary
circuit which is so designed and protected that under normal and
single fault conditions, the voltage between any two accessible
parts, or between one accessible part and the equipment protective
earthing terminal for class I equipment, does not exceed a safe
value.
(1.2.9.2): Basic Insulation: insulation to provide basic protection
against electric shock.
(1.2.9.3): Supplementary Insulation: Independent insulation
applied in addition to basic insulation in order to ensure protection
against electric shock in the event of a failure of the basic
insulation.
(1.2.9.4): Double Insulation: Insulation comprising both basic
insulation and supplementary insulation.
(1.2.9.5): Reinforced Insulation: A single insulation system
which provides a degree of protection against electric shock
equivalent to double insulation.
(1.2.9.6): Working voltage: The highest voltage to which the
insulation under consideration is, or can be, subjected when the
equipment is operating at its rated voltage under conditions of
normal use.
(1.2.9.7): Tracking: the progressive formation of conducting
paths on the surface of a solid insulating material (such as PC
board or transformer bobbin) due to the combined effects of
electric stress and electrolytic contamination on this surface.
(1.2.10.1): Creepage distance: the shortest path between two
conductive parts, or between a conductive part and the bounding
surface of the equipment, measured along the surface of the
insulation. In a TOPSwitch power supply, the most important
creepage distance is between all primary circuits and all
secondary circuits (typically 5mm to 6 mm).
(1.2.10.2): Clearance: the shortest distance between two
conductive parts, or between a conductive part and the bounding
surface of the equipment, measured through air.
(1.2.11.1): Safety Isolating Transformer: the power transformer
in which windings supplying SELV circuits are isolated from
other windings (such as primary and primary bias windings)
such that an insulation breakdown either is unlikely or does not
cause a hazardous condition on SELV windings.
(1.2.8.2): Secondary circuit: A circuit which has no direct
connection to primary power (except through properly selected
Y-capacitors) and derives its power from a transformer.
A
6/96
5
AN-15
IEC950 Requirements (Applicable to TOPSwitch Power
Supplies)
(1.4.5): In determining the most unfavorable supply voltage for
a test, the following variables shall be taken into account:
• multiple rated voltages
• extremes of rated voltage ranges
• tolerance on rated voltage as specified by the manufacturer.
If a tolerance is not specified, it shall be taken as +6% and
- 10%.
(1.6.5): Equipment intended to operate directly from the mains
supply shall be designed for a minimum supply tolerance of
+6%, -10%.
(2.1.10): Equipment shall be so designed that at an external
point of disconnection of the mains supply, there is no risk of
electric shock from stored charge on capacitors connected to the
mains circuit. Equipment shall be considered to comply if any
capacitor having a rated capacitance exceeding 0.1 uF and
connected to the external mains circuit, has a means of discharge
resulting in a time constant not exceeding 1 second for pluggable
equipment type A (non-industrial plugs and socket-outlets).
This requirement specifically applies to any EMI filter capacitor
connected directly across the AC mains which could cause a
shock hazard on the exposed prongs of an unplugged power
cord.
(5.2.2): Earth Leakage Current: Maximum earth leakage
current must not exceed the limits shown in the following table
under the most unfavorable (highest) input voltage. For class II
equipment when output is not connected to earth ground, the
test shall be made on accessible conductive parts, and to metal
foil with an area not exceeding 10 cm x 20 cm on accessible nonconductive parts.
Class Type of Equipment
II
I
I
Maximum Leakage Current
All
Hand-held
Movable
(other than hand-held)
0.25 mA
0.75 mA
3.50 mA
Table 1. Maximum Leakage Current.
(5.3.2): Electric Strength: The insulation shall be subjected for
1 minute either to a voltage of substantially sine-wave form
having a frequency of 50 Hz or 60 Hz or to a DC voltage equal
to the peak of the prescribed AC test voltage. Test voltage shall
be as specified in the following table for the appropriate grade
of insulation and the working voltage U across the insulation:
(5.4.1): Abnormal Operating and Fault Conditions: Equipment
6
A
6/96
Grade of insulation
Basic, Supplementary
Reinforced (Primary
to Secondary)
U < 130 VAC 130 < U < 250VAC
1000 VAC
2000 VAC
1500 VAC
3000 VAC
Table 2. Insulation Electric Strength.
shall be so designed that the risk of fire or electric shock due to
mechanical or electrical overload or failure, or due to abnormal
operation or careless use, is limited as far as practicable.
(5.4.6): For components and circuits (other than motors,
transformers, PC board creepage and clearance distances, or
secondary circuit electromechanical components) compliance
with the abnormal operating and fault condition requirement
(5.4.1) is checked by simulating the following conditions:
- faults in any components in primary circuits (which includes
EMI filter components, bridge rectifier, energy storage
capacitor, TOPSwitch, and all TOPSwitch connected
components);
- faults in any components where failure could adversely
affect supplementary or reinforced insulation (specifically
failure of Y2-capacitors connected between primary circuits
and secondary circuits);
- additionally, for equipment that does not comply with the
requirement of Sub-clauses 4.4.2 (Minimizing the risk of
ignition) and 4.4.3 (Flammability of materials and
components), faults in all components;
- faults arising from connection of the most unfavorable load
impedance to terminals and connectors that deliver power
or signal outputs from the equipment, other than mains
power outlets (for example: connecting a class II equipment
output terminal to earth ground will increase measured
leakage current).
The equipment, circuit diagrams and component specifications
shall be examined to determine those fault conditions that might
reasonably be expected to occur.
(In general, components designed for use between primary and
secondary circuits, rated for the full electric strength voltage,
and carrying the appropriate safety agency approvals are not
subject to the single component fault test because a short circuit
fault is extremely unlikely. Two component examples are
safety rated optocouplers and Y1-capacitors which can be
applied directly between primary and secondary circuits
operating from AC mains with rated voltages up to 250 VAC.)
AN-15
Typical AC Mains Input Voltage
Configurations
TOPSwitch power supplies are typically connected to the AC
mains in either 2-wire or 3-wire configurations. For the purposes
of EMI design presented in this application note, 2-wire and 3wire configurations are now defined.
current emissions and increase margin below the specification
limit but may not address size or cost goals of the end product.
Understanding the basics of EMI filter design and application
allows the designer to implement small, low cost, single section
EMI filters.
Z=
2-Wire AC Input
The TOPSwitch power supply 2-wire AC mains connection
may consist of one line wire and one neutral wire where the AC
mains neutral is eventually connected back to earth ground at a
local electrical panel. The 2-wire connection may also consist
of two separately phased line wires where neither is connected
directly to earth ground. The power supply SELV output may
or may not be connected directly to earth ground.
In this application note, the neutral wire will be treated as an
ungrounded AC mains or separately phased line conductor
requiring the same safety considerations as any AC mains line
conductor. In addition, the power supply SELV output return
will be assumed to connect directly to earth ground which
represents the worst case and most unfavorable connection for
safety considerations.
3-Wire AC Input
In 3-wire connections, the third wire earth ground wire will be
available for connection to EMI filter components, shields,
chassis, and enclosures. The neutral wire will be treated as an
ungrounded AC mains or separately phased line conductor
requiring the same safety considerations as any AC mains line
conductor. In addition, the power supply SELV output return
will be assumed to connect directly to earth ground which
represents the worst case and most unfavorable connection for
safety considerations.
EMI Filter Components
EMI filters are actually simple combinations of inductors or
chokes and capacitors. Series resistors, which lead to undesirable
power dissipation, are not normally used for reducing conducted
emissions.
Single-section EMI filters (one stage of common mode and
differential mode attenuation) take the least space and have the
lowest cost but require careful attention to details such as circuit
parasitics, component parasitics, and layout to meet the
specifications with adequate margin. Multiple-section filters
can also be used because one stage can be designed to overcome
the deficiencies of the other. The two section design will reduce
1
2πfC
Z= 2πfESL
Z
ESL
Actual
C
Ideal
ESR
ESR
fr
f
PI-693-031592
Figure 9. Comparison of Ideal and Real Capacitor Impedance.
Capacitors
Proper capacitor selection for EMI filters requires attention to
three key parameters: impedance characteristics, voltage ratings,
and safety specifications.
Figure 9 shows impedance characteristics for ideal and nonideal capacitor behavior. An ideal capacitor has an impedance
characteristic that decreases linearly with frequency. A real
capacitor has parasitic inductance and resistance elements
which cause the impedance to behave quite differently from an
ideal capacitor.
Equivalent series inductance (ESL) creates a capacitor self
resonant frequency fr as shown on the plot. The impedance of
the capacitor at this self-resonant frequency is determined by
equivalent series resistance (ESR). Beyond the self-resonant
frequency (fr), the capacitor actually acts like an inductor.
Capacitors with plastic film, combination plastic film/paper, or
ceramic dielectrics usually have the highest self-resonant
frequencies and are commonly used in EMI filters.
Aluminum Electrolytic Energy Storage Capacitor
Switching power supplies always have a bridge rectifier and
high voltage bulk energy storage aluminum electrolytic capacitor
to convert AC mains input voltage to high DC bus voltage
(typically 100 to 400 Volts DC) shown as CIN in Figure 7.
The impedance of this capacitor, which must always be
minimized, provides the first level of differential mode conducted
emissions filtering.
A
6/96
7
AN-15
Impedance
(Ω)
100
PI-1629-111695
500
AXIAL 33µF 12x25
10
1
0.1
0.01
100
RADIAL 22µF 10x20
RADIAL 47µF 12x25
1K
10K
100K
1M
10M
40M
the differential mode portion of the EMI filter. X-capacitors are
divided into three subclasses:
Subclass Peak Pulse IEC-664
Application Peak Impulse
Voltage
Installation
Voltage VP
In Service Category
applied before
Endurance Test
X1
> 2.5 kV
< 4.0 kV
III
X2
< 2.5 kV
II
General C < 1.0 uF
Purpose UP = 2.5 kV
X3
< 1.2 kV
-
General
Purpose
Frequency (Hz)
High Pulse C < 1.0 uF
Application UP = 4 kV
500
AXIAL 33µF 16x40
10
None
Table 3. X-Capacitor Subclass.
1
0.1
RADIAL 10µF 12x20
RADIAL 100µF 22x35
0.01
100
1K
10K
100K
1M
10M
40M
Frequency (Hz)
X2-capacitors are most commonly used in TOPSwitch power
supply EMI filters for differential mode suppression. X1capacitors can also be used but cost is higher. X3- capacitors are
not normally used.
Figure 11. 400V Aluminum Electrolytic Capacitor Impedance.
100K
PI-1631-111695
Impedance
(Ω)
100
PI-1630-111695
Figure 10. 200V Aluminum Electrolytic Capacitor Impedance.
Figures 10 and 11 show impedance of various 200V and 400V
aluminum electrolytic capacitors with radial leads (both leads
exiting one side of the capacitor can) compared with impedance
of a similar capacitor with axial leads (one lead exiting each side
of the capacitor can). Approximate dimensions are also shown
(diameter by length in mm). Radial capacitors have an impedance
characteristic that stays low up to 10 MHz while the axial
capacitors become inductive at frequencies as low as 1 MHz.
Radial capacitors should always be used and installed on end to
minimize lead length and ESL. Axial leaded capacitors should
never be used because the longer total lead length (equal to at
least one can diameter) increases ESL which increases
impedance. Note that above 1 MHz, the large axial capacitors
actually have much higher impedance (and will generate higher
conducted emission currents) than the smaller radial capacitors.
EMI Filter Capacitors
Capacitors used in EMI filters are identified by various
companies as radio interference suppressors, suppression
capacitors, or safety recognized capacitors. These capacitors
must meet the European requirement EN 132400 for safety
which defines two groups, X and Y(12) (13).
X-capacitors are used only in positions where capacitor failure
does not expose anybody to an electric shock hazard. Xcapacitors are usually connected across the AC mains as part of
8
A
6/96
Impedance
(Ω)
10K
1K
100
0.033µF
(LONG LEADS)
10
1
0.1
0.033µF
0.047µF
0.1µF
0.22µF
0.47µF
0.01
100
1K
10K
100K
1M
10M
40M
Frequency (Hz)
Figure 12. X2-Capacitor Impedance.
X2-capacitors are available from a variety of vendors including
Murata, Roederstein, Panasonic, Rifa, and Siemens. Figure 12
shows impedance plots for various sizes of X2-capacitors with
short leads and one plot for a small X2-capacitor with long
leads. Short leads should always be used to minimize impedance
and reduce high frequency conducted emission currents.
Y-capacitors are used where capacitor failure could expose
somebody to an electric shock hazard. Y-capacitors are usually
connected from the AC mains or bridge rectifier output to
SELV secondaries, chassis, shields, or earth ground. The
maximum Y-capacitor value is restricted because each
application has an allowable maximum leakage current (which
can range from 0.25 mA to 3.5 mA, depending on the AC mains
connection). There are four EN 132400 specified subclasses of
Y-capacitors:
AN-15
Y1
Type
of
Insulation
Bridged
Double Insulation or
Reinforced Insulation
Rated Test Voltages for
Voltage Quality Approval,
(VAC) Periodic and
Lot-by-Lot Testing
< 250 V
4000 VAC
Peak Impulse
Voltage VP
applied before
endurance
radiated emissions specifications. Using short leads and short
PC traces for all Y-capacitor connections is critical to meet both
conducted and radiated emissions specifications.
10M
8.0 kV
PI-1632-111695
Subclass
1M
Y2
> 150 V
Basic Insulation or
Supplementary Insulation < 250 V
1500 VAC
5.0kV
Y3
> 150 V
Basic Insulation or
Supplementary Insulation < 250 V
1500 VAC
none
< 150 V
Basic Insulation or
Supplementary Insulation
900 VAC
Y4
2.5 kV
Impedance
(Ω)
100K
10K
1K
100
4700pF
2200pF
1000pF
680pF
330pF
10
1
100
Table 4. Y-Capacitor Subclass.
In two-wire 230 VAC or universal input applications, a single
Y1-safety capacitor can be directly connected between the AC
mains or bridge rectifier output to the SELV secondary. The
single Y1-capacitor will also meet the electric strength voltage
requirement (for 230 VAC mains connected power supplies,
typically 3,000 VAC for one minute). Y1-capacitors with a
value of 1000 pF are available from Murata(14) (ACT4K-KD
series, DE1110 E 102M ACT4K-KD), Roederstein(15) (WKP
series, WKP102MCPE.OK) and Rifa(12) (PME 294 series, PME
294RB4100M). In general, Y1-capacitors are not used in 3wire applications.
Y2-capacitors do not meet reinforced insulation requirements.
In a single component failure safety investigation, one Y2capacitor may be replaced with a wire jumper to see if an electric
shock or fire hazard condition will exist. In most 2-wire
applications, a series combination of two 2200 pF Y2-capacitors
are commonly used between primary and SELV outputs so that
a short circuit failure of one Y2-capacitor creates no safety
hazard. A series connection of two Y2-capacitors is also
necessary to meet the electric strength requirement (for 230
VAC mains connected power supplies, typically 3,000 VAC
for one minute). In 3-wire applications, the Y2-capacitor may
be directly connected between AC mains or bridge rectifier
output and earth ground because the earth ground wire will
safely shunt the fault current created by a shorted Y2-capacitor.
Y2-capacitors rated at 250 VAC are available from a variety of
vendors including Murata, Roederstein, Panasonic, Rifa, and
Siemens. Figure 13 shows impedance plots for various sizes of
Y2-capacitors with short leads and one plot for a large Y2capacitor with long leads. Y-capacitors perform most of the
high frequency filtering from 10 MHz to 200 MHz. Note that
capacitor resonant frequency is usually 40 MHz or higher
unless artificially reduced with long leads or long PC traces.
Long leads and long PC traces can also cause emission currents,
though low enough to meet conducted emissions specifications,
to radiate sufficient energy from the power cord to exceed
4700pF
(LONG LEADS)
0.1
1K
10K
100K
1M
10M
40M
Frequency (Hz)
Figure 13. Y2-Capacitor Impedance.
In 115 VAC applications, a series combination of two Y2safety or two Y4-safety capacitors can be directly connected
between the AC mains or bridge rectifier to the SELV secondary.
Y3-safety capacitors are not normally used.
Safety specifications such as UL1950, UL544, and IEC950
limit the amount of fault current that can flow when a safety
ground connection has been opened or one component has
failed (Y1-capacitors, because of their construction, are excluded
from the failed component test). For example, UL1950 specifies
that information technology equipment with Class I or three
wire input (line, neutral, and earth ground), 240 VAC, 60 Hz
input must have a leakage current no higher than 3.5 mA if earth
ground is opened or one component has failed short which
restricts Y-capacitor maximum value below 0.039 µF (or
39 nF). For class II or two wire input (line, neutral, with no earth
ground), leakage current must be less than 250 µA with one
failed component which restricts Y-capacitor size to under
0.0028 µF (2.8 nF or 2800 pF) for 240 VAC, 60 Hz input.
Capacitor and input voltage tolerance must also be taken into
LINE
DUT
Leakage
Currents
NEUTRAL
GND
PI-717-032192
Figure 14. Typical Safety Measurement Setup.
A
6/96
9
AN-15
account. Figure 14 shows a typical test setup for measuring
leakage current.
Single
Layer
Windings
Inductors or chokes
Proper inductor selection for EMI filters requires attention to
three key parameters: effective impedance characteristic, current
rating, and surge current capability.
Figure 15 shows impedance characteristics for ideal and nonideal inductor behavior. Ideal inductors have an impedance
characteristic that increases linearly with frequency. Real
inductors have parasitic series resistance RS and parallel
interwinding capacitance (CW). CW creates a resonant frequency
as shown on the plot. Beyond the resonant frequency (fr), the
inductor actually behaves like a capacitor.
TOROIDAL
SOLENOIDAL
PI-708-031992
Ideal
Z= 2πfL
Z
L
CW
RS
Actual
RS
fr
f
PI-1725-121895
Figure 15. Comparison of Real and Ideal Inductor Impedance.
Power supplies have bridge rectifier input filters which draw
line frequency currents with high peak values but relatively
narrow widths as previously shown in Figure 7. A discrete filter
choke usually has a minimal effect on the peak current but must
pass the peak current without significant saturation (which
reduces effective inductance). The discrete choke must also be
rated to safely pass the higher peak value of the first surge of
current occurring when AC power is initially applied with input
capacitor CIN completely discharged.
Differential mode chokes
Differential mode chokes are simply discrete inductors designed
for EMI filters that pass line frequency or DC currents while
blocking or filtering high frequency conducted emission currents.
Differential mode chokes are usually wound on low cost
solenoidal cores of either iron powder or ferrite material as
shown in Figure 16. Toroids tend to be significantly higher in
cost but can also be used. Chokes with single layer windings
have the lowest capacitance and highest resonant frequency.
10
A
6/96
Effective inductance varies with peak differential mode choke
current flow. Refer again to Figure 7 where the bridge rectifier
and filter creates a high voltage DC bus from the AC line. AC
input current flows only during a small conduction time as
shown. Peak AC input current during normal operation
is relatively high. Differential mode chokes are designed
or selected to limit saturation at peak AC input current.
Figure 17 shows how inductance for a powdered iron toroidal
core varies with number of turns and peak current. To achieve
the desired inductance under high peak AC input current, higher
numbers of turns and/or larger choke cores are normally required.
Typical impedance characteristics for two different differential
mode chokes are shown in Figure 18. Note that the larger choke
resonates at a lower frequency and becomes capacitive. The
smaller choke has a higher impedance above 3 MHz due to the
No
Bias
Inductance
1
Z=
2πfCW
Figure 16. Diffferential Mode Chokes.
Heavy
Bias
Current
# of Turns
PI-709-031992
Figure 17. Inductance Under Current Bias.
AN-15
higher self-resonant frequency. Installing the larger choke to
attenuate the fundamental may have the effect of letting through
current components above 3 MHz.
mode choke winding. IC1 and IC2 are “common mode” currents
which may or may not be related in magnitude and phase. The
common mode choke behaves like a large inductor to common
mode currents.
IMPEDANCE vs. FREQUENCY
PI-739-032392
105
Impedance (Ω)
104
ID
103
102
1 mH
100 µH
101
100
103
104
OUT
IN
105
ID
106
PI-1633-111695
107
Frequency (Hz)
Figure 19. Ideal Common Mode Choke.
Figure 18. Typical Differential Mode Choke Impedance.
16max.
18max.
20max.
Common Mode chokes
Common mode chokes are specialized inductors designed
specifically for common mode EMI filters. The common mode
choke consists of two identical windings wound such that the
magnetic fields caused by differential mode currents cancel.
Figure 19 shows a toroidal implementation which is good for
illustration purposes but (as will be seen shortly) is not the best
choice for low-cost and practical EMI filter implementations.
Figure 19 shows three current components ID, IC1, and IC2. ID is
a differential mode current (shown also in Figure 7) which
circulates by starting at the AC mains source, flows through one
common mode choke winding towards the power supply, flows
through one bridge rectifier diode, charges the high voltage
energy storage capacitor CIN, flows back through another bridge
rectifier diode, and then flows back towards the source through
the other common mode choke winding. The magnetic fields
within the core due to the circulating differential current ID
cancel perfectly because of dot polarity. Note that the “start” of
both windings enters the core on the same side and the “finish”
of both windings leaves the core on the other side. Common
mode chokes behave like short circuits for circulating differential
mode currents such as ID which flow in through one common
mode choke winding and flow out through the other common
0.7
4min.
Differential mode chokes are usually used in EMI filters for
both differential mode and common mode filtering only for the
lowest output power levels (under 5 Watts). At higher power
levels, a properly selected common mode choke will also have
differential mode inductance for essentially no additional cost.
Two low-cost bobbin style common mode chokes simplify
EMI filter design. Figure 20 shows a typical “U-core” style
common mode choke in which the windings are wound on a
conventional bobbin. Two U-core halves are inserted into the
bobbin and secured with clamps. U-core common mode chokes
are widely available from several companies such as Tokin(16),
Tamura(17) , Panasonic/Matsushita(18), TDK(19) , and Murata(20) .
10±0.5 PI-1634-111695
Figure 20. U Core Common Mode Choke (All dimension in mm).
Figure 21 shows a newer common mode choke design with a
“spool” style two-piece bobbin. The two-piece bobbin is
snapped together around a one-piece ungapped core. A sprocket
on the bobbin engages a gear on a winding machine to spool the
wire onto the bobbin. Spool style common mode chokes are
available from Panasonic/Matsushita(18) and Tokin(16).
A
6/96
11
10.0±0.5
21.0±1.0
16.0±1.0
Common mode impedance is shown for the U-core style in
Figure 23 and the spool style in Figure 24. Also shown is
common mode impedance for a typical toroidal implementation.
Note that the toroidal common mode impedance is generally
lower than both the U-core and spool style common mode
chokes.
PI-1635-111695
One very important advantage to the bobbin style common
mode choke is an “inherent differential mode choke” due to
parasitic leakage inductance which usually eliminates any need
for additional discrete differential mode chokes. Figure 22
shows the effective common mode choke schematic consisting
of a common mode inductance in series with an effective
differential mode leakage inductance. Unlike most other
magnetic components, leakage inductance in a common mode
choke is a desirable parasitic effect which provides balanced
differential mode filtering for no additional component cost.
The common mode choke is modeled by a common mode
inductance in series with a differential mode inductance.
100K
Impedance
(Ω)
Figure 21. Spool Wound Common Mode Choke (All dimension in
mm).
1M
33 mH
18 mH
10 mH
5.6 mH
10K
1K
100
10
1
1K
Toroid
1mH
10K
100K
1M
10M
40M
Frequency (Hz)
Figure 23. U Core Common Mode Choke (Common Mode
Impedance).
1M
Impedance
(Ω)
100K
Common Mode
Inductance
PI-1637-111695
13.0±0.5
3.5±0.5
0.8
Common mode inductance of each winding is the measured
inductance of one winding with the other winding open circuited.
Differential mode inductance of each winding is equal to half
the measured inductance of one winding with the other winding
short circuited.
10K
PI-1638-111695
21.5±1.0
AN-15
33 mH
22 mH
10 mH
3.3 mH
2.2 mH
0.82 mH
1K
100
10
1
1K
Toroid
1mH
10K
100K
1M
10M
40M
Frequency (Hz)
Input
Output
Differential Mode
"Leakage"
Inductance
PI-1636-111695
Figure 22. Effective Common Mode Choke Schematic.
12
A
6/96
Figure 24. Spool Wound Common Mode Choke (Common Mode
Impedance).
Differential mode impedance is shown for the U-core style in
Figure 25 and the spool style in Figure 26. Also shown is
differential mode impedance for a typical toroidal
implementation. Note that the toroidal differential mode
impedance is quite a bit lower than both the U-core and spool
style common mode choke. With toroidal common mode
chokes, additional differential mode chokes are usually required.
For these reasons, toroidal common mode chokes are not
recommended except for the high frequency, supplemental
torodial common mode choke described below.
AN-15
100K
Impedance
(Ω)
10K
PI-1639-111695
500K
33 mH
18 mH
10 mH
5.6 mH
1K
100
Toroid
1mH
10
1
together in parallel and wound as a pair for typically 3 to 5 turns.
The toroidal core should be ferrite and “lossy” at high frequency
such as Fair-Rite 75 material. Fair-Rite toroid part number
5975001101 (with 0.5 inch OD x 0.32 inch ID x 0.25 inch
thickness) is suitable for most applications(21). This high
frequency common mode choke is usually located between
power entry and the rest of the power supply EMI filter. This
common mode choke technique can also be used on power
supply output wires.
0.1
1K
10K
100K
1M
10M
40M
INSULATED WIRES
Frequency (Hz)
Figure 25. U-Core Wound Common Mode Choke (Differential
Mode Impedance).
INPUT
Impedance
(Ω)
10K
1K
PI-1640-111695
500K
100K
33 mH
22 mH
10 mH
3.3 mH
2.2 mH
0.82 mH
100
Toroid
1mH
10
FERRITE
TOROID
OUTPUT
HIGH FREQUENCY COMMON MODE CHOKE
1
PI-1641-111695
0.1
1K
10K
100K
1M
10M
40M
Frequency (Hz)
Figure 26. Spool Wound Common Mode Choke (Differential Mode
Impedance).
Bobbin style common mode chokes can have either one or two
sections in each winding. One section per winding is lowest
cost but two sections per winding splits the winding capacitance
in half to increase resonant frequency and effective bandwidth.
The U-core common mode choke shown in Figure 20 has one
section per winding while the spool style common mode choke
shown in Figure 21 has two sections per winding. Figure 23
shows that the single section U-core style common mode
impedance is lower and resonant frequency is lower with
sharper peaking compared with the two section spool style
common mode impedance shown in Figure 24. Two sections
per winding reduce capacitance to improve common mode
impedance at high frequency.
The common mode choke must also survive the surge current
occurring when voltage is first applied to the power supply as
described earlier, as well as operate at the steady-state RMS
input current.
For reducing high frequency common mode conducted emissions
in the 10 MHz to 200 MHz range, a simple common mode choke
using a small ferrite toroid(21) and insulated wire can be wound
as shown in Figure 27 and used in addition to one of the bobbin
style common mode chokes. Both wires have thick, safety
insulated wires with different colors. The wires are held
Figure 27. High Frequency Common Mode Choke.
Flyback Power Supply EMI Signature
Flyback power supplies have a distinctive EMI signature caused
by superposition of several waveforms shown in Figure 28. The
transformer primary current IPRI, TOPSwitch Drain voltage
VDrain, diode voltage VDiode, and transformer secondary current
ISEC waveforms each generate emission currents which may
exceed the desired EMI specification limits without proper
EMI design technique.
Primary Current Waveform
Primary current IPRI begins to flow when TOPSwitch turns on.
Transformer primary current ramps to a peak value determined
by input voltage, primary inductance, switching frequency, and
duty cycle. This trapezoidal (or triangular) current waveform
is characterized in the frequency domain by a spectrum with a
fundamental at the switching frequency and harmonics
determined by the relative squareness of the waveform and
causes primarily differential mode emission currents to circulate
between the AC mains and the power supply input. This current
waveform can also create common mode emissions due to
radiated magnetic fields if the current path defined by the PC
board layout encircles a large physical area.
TOPSwitch Drain-Source Voltage Waveform
The Drain-Source voltage waveform VDrain is characterized by
high dv/dt transitions. Parasitic circuit elements (leakage
inductance, TOPSwitch output capacitance, and transformer
A
6/96
13
AN-15
capacitance) cause additional voltage peaking and ringing at
frequencies typically between 3 MHz and 12 MHz. The
TOPSwitch Drain, transformer primary, and Drain clamping
components connected to the Drain node will drive displacement
currents to earth ground through transformer capacitance or
stray capacitance. This displacement current returns
“backwards” through the line and neutral conductors back to
the TOPSwitch Drain driving node as a common mode emission
current. The displacement currents generated by the drain
voltage waveform cause spectral energy in the form of a
common mode conducted emission currents to be concentrated
at the switching frequency and 3 MHz to 12 MHz resonant
frequency (f1) of the indicated ringing voltage waveform.
Common mode emission currents will be lower with TOPSwitch
when compared with discrete MOSFET implementations
because TOPSwitch has a controlled turn on gate driver to
reduce dv/dt. Common mode emissions currents are also lower
because the TOPSwitch TO-220 tab is connected to the relatively
quiet source pin while a discrete MOSFET has the noisy drain
“transmitting” node connected directly to the tab (and heat sink)
“broadcasting antenna”.
Diode Voltage Waveform
The diode voltage waveform VDIODE is also characterized by fast
voltage changes and fast rise and fall times. Parasitic circuit
elements (transformer leakage inductance and diode capacitance)
cause additional voltage peaking and ringing at frequencies
typically between 20 MHz and 30 MHz. The diode voltage
waveform will drive displacement currents to earth ground
through transformer capacitance or stray capacitance. The
displacement currents generated by the diode voltage waveform
cause spectral energy in the form of common mode emission
currents to be concentrated at the switching frequency and
20 MHz to 30 MHz resonant frequency (f2) of the indicated
ringing voltage waveform.
Secondary Current Waveform
Secondary current ISEC begins to flow as soon as TOPSwitch
turns off. Current starts at a peak value and decreases linearly
at a rate determined by secondary inductance and output voltage.
This trapezoidal (or triangular) current waveform is characterized
in the frequency domain by a spectrum with a fundamental at
the switching frequency and harmonics determined by the
relative squareness of the waveform. Additional ringing
superimposed on the waveform is related to the drain source
voltage VDrain waveform previously discussed. This composite
current waveform can cause significant magnetic fields to
radiate if the current path defined by the PC board layout
encircles a large physical area. Spectral energy in the form of
a common mode emission current would be concentrated at the
switching frequency and 3 MHz to 12 MHz resonant frequency
(f1) of the indicated ringing current waveform.
ISEC
IPRI
IPRI
+
VDIODE
-
C
LOAD
VDRAIN
VIN
f1
ISEC
+
V
- DRAIN
VDIODE
f2
Figure 28. Examples of Typical Flyback Power Supply Waveforms Causing EMI.
14
A
6/96
PI-1724-121895
AN-15
LD
IPRI
LISN
RESISTORS
+
VSL
ISENSE
I2
LD
I4
IPRI
I1
RSL
+
VIN
CD
CD
+
CIN
VSN
RSN
I4
-
+
ESR
LD
ESR
LD
I2
ISENSE
-
I1
IPRI
IPRI
ACTUAL
MODEL
IPRI
PI-1642-111695
Figure 29. Circuit Origin for Differential Mode Emissions.
Suppression Techniques
Controlling EMI requires attention to the following areas.
• Differential mode filtering
conducting current and is replaced with a short circuit. The AC
source impedance is modeled by the effective series combination
of the 50 Ω LISN sense resistors RSL and RSN. Differential mode
filtering is performed by the LC filter consisting of differential
mode capacitor CD and two identical differential mode chokes
LD . This model is valid up to roughly 1 MHz.
• Common mode filtering
Differential mode Filter Analysis
Differential mode conducted emissions are caused by currents
circulating between the power supply and AC mains input
which means that a differential current which flows into the
power supply through the Line input wire will flow out of the
power supply through the Neutral input wire.
Most differential mode conducted emissions are caused by the
fundamental and harmonics of the triangular or trapezoidal
TOPSwitch Drain current waveform. During EMI testing,
differential mode currents generate test voltages equal in
magnitude and opposite in phase across Line LISN sense
resistor RSL and Neutral LISN sense resistor RSN.
Differential mode analysis starts by replacing the actual circuitry
with an equivalent model as shown in Figure 29. The primary
current is modeled by current source IPRI. The effective
impedance of energy storage capacitor C1 over the frequency
range of 100 kHz to 1 MHz is modeled by the Equivalent Series
Resistance or ESR. The bridge rectifier is assumed to be
PI-1738010496
• Transformer construction
The primary current switching frequency fundamental and
harmonic components IPRI(n) must be estimated, measured, or
derived by simulation. Note that measured harmonic components
are given in RMS but calculated or simulated components are
given in peak values and must be converted to RMS. A typical
harmonics envelope is shown in Figure 30 as a function of
frequency.
Fourier Coefficient
• Power cord damping
1
2 3
5 7 ...
Harmonic Number
Figure 30. Envelope of Typical Primary Current Fourier Spectrum.
A
6/96
15
AN-15
At the fundamental and harmonics of switching frequency fS,
equivalent series resistance (ESR) of bulk input capacitor CIN is
much lower impedance compared with the LD differential mode
chokes. Primary current IPRI flows almost completely through
bulk energy storage capacitor CIN which creates an effective
trapezoidal (or triangular) differential mode voltage source
proportional to ESR. Differential mode chokes and the
differential mode capacitor form a simple low pass filter to
attenuate the effective voltage source to a level below the
desired specification. Figure 31 shows the final simplified
model where the RMS source voltage for each nth current
harmonic (given in peak value) is given by:
1
VPRI (n) = ESR × IPRI (n) ×
2
(RMS)
RS
IL(n)
+
CD
VPRI(n)
_
RS
LD
PI-1643-111695
Figure 31. Simplified Differential Mode Model.
Attenuation is determined by the differential between the
magnitude of the effective voltage source in dBµV and the
desired conducted emissions specification. The voltage transfer
function H(s) is given in terms of LD, C D, and RS.
At high levels of attenuation normally required at the switching
frequency, the denominator of H(s) is dominated by the frequency
dependent terms and can be simplified as shown. Simple
algebra reveals a very useful frequency domain formula
consisting of the product of three separate terms. The first term
converts the effective ESR voltage source VPRI(s) back into
differential inductor current ID(s), the second term splits the
current between differential mode capacitor CD and LISN sense
resistors, and the third term senses the LISN current component
to create a voltage to be measured with a detector or receiver to
compare with limits in dBµV. This is a general result with
equivalent ESR voltage source VPRI(n) of each nth harmonic
shown (temporarily) in the frequency domain as VPRI(s) which
is a function of the complex frequency variable s.
A
6/96
VSN ( s) 1
= ×
VPRI ( s) 2
≈
1
L

(2 × LD × CD × s)2 +  D × s + 1
 RS

1
1
×
2 (2 × L × C × s)2 +  LD × s
D
D
 RS

VSN ( s) = VPRI ( s) ×
1
CD × s
1
×
2 × LD × s (2 × R ) +
S
1
CD × s
× RS
(Peak)
LD
16
H (s) =
For EMI filter design, only the magnitude of the most important
frequency components are examined which allows simple
magnitude expressions in terms of the harmonic integer n to be
used (rather than the complex variable s). Filter design begins
by identifying a target sense voltage VSNdBµV (n) below the
specification limits at the appropriate nth harmonic frequency.
For FCC testing, the specification begins at 450 kHz with the fifth
harmonic (n = 5) while excluding TOPSwitch 100 kHz
fundamental (n = 1) and second through fourth harmonic
frequencies (n = 2, 3, 4). For European test limits, the 100 kHz
fundamental (n = 1) and the second harmonic at 200 kHz
(n = 2) should be examined because the limit changes
significantly at 150 kHz. As an example and referring to
European EN55022 average limit for class B (Figure 2), the
average limit value is 74 dBµV at 100 kHz (n=1) and 53.5 dBµV
at 200 khz (n=2) while the quasi-peak limit values are 10 dB
higher. In most low frequency conducted emission
measurements, the measured quasi-peak value is slightly less
(1 dB to 3 dB) than the peak value. The average value, however,
can be 12 dB below the peak value which means that if the filter
is designed to meet the average limit, the quasi-peak limit will
also be met and with greater margin. In this example and for
12 dB margin overall, the peak value should be designed to touch
the average limit and average detection will provide the
remaining 12 dB attenuation. The target sense voltages are
therefore equal to the average limit or 74 dBµV at 100 kHz
(VSNdBµV(1)) and 53.5 dBµV at 200 kHz (VSNdBµV(2)). VSNdBµV (n)
is converted from dBµV to an absolute value sense voltage
VSN(n).
VSN (n) = 1.e −6 × 10
VSNdBµV (n)
20
VSN(1) is 5.01 mVRMS and VSN(2) is 473 µVRMS. Sense voltage
VSN(n) is then converted into an RMS current magnitude IL(n)
flowing through each differential mode inductor LD .
AN-15
I L (n) = VSN (n) ×
LD
1
+ ( 4 × π × n × fs × CD )2
Rs2
RS
RMS differential current IL(1) is 638µA and IL(2) is 119µA.
The target differential inductance LD can now be calculated.
VPRI (n)
LD =
I L (n) × 4 × π × n × fs
CW
CIN
RS
ESL
CW
ESR
LD
The ST202A power supply operating from 115 VAC and
delivering 15 Watts is found to operate in the discontinuous
mode with a triangular drain current waveform. Peak Drain
current IP is 0.8 A and duty cycle is 0.3. C6 (0.1 µF) is
differential capacitor CD. ESR of input capacitor C1 is 0.375 Ω.
From simulation, calculation, or measurement with the power
supply connected to the LISN but without an EMI filter, the
equivalent source voltage fundamental VPRI (1) is 59.3 mVRMS
and second harmonic VPRI(2) is 43.0 mVRMS. Differential
inductance LD is found to be 74µH in each leg for attenuation of
the fundamental but the second harmonic requires a higher
inductance value of 144µH in each leg to achieve the desired
attenuation because the EN55022 specification is more stringent
at 200 kHz. The higher inductance value is used in the design.
Note also that different combinations of L and C are possible but
the LC product will remain the same. Note also that, in common
mode chokes, total measured differential inductance is twice
the value calculated for each leg (288µH in this example).
Peak load current normally limits the size of discrete chokes
to between 100µH and 1 mH (especially in mains applications
with peak-charging capacitive input filters). Practical discrete
chokes are cost effective only at the lower output power levels
(5 Watts and below). Single discrete chokes attenuate the
differential mode but have little effect on common mode
emission currents. These limitations for discrete, differential
mode chokes can be overcome by selecting a common mode
choke with parasitic leakage or differential inductance equal to
or greater than the differential choke inductance value calculated
above. (Note: with a common mode choke, measure inductance
of one winding with the other winding shorted for total leakage
or differential inductance. The effective differential inductance
in each leg is half the measured value.)
Filter effectiveness decreases as parasitic elements of the filter
components themselves become significant. The effective
circuit model above 1 MHz is shown in Figure 32. Note the
additional ESL terms in both energy storage capacitor CIN and
differential capacitor C D. Note also the shunt winding
capacitance CW across each differential mode filter choke LD.
As the frequency increases, the parasitic components begin to
dominate, reducing filter effectiveness. Fortunately, the
ESL
CD
PI-1860-050796
Figure 32. High Frequency Model of the Differential Mode Filter.
harmonics of the trapezoidal (or triangular) TOPSwitch Drain
current waveform are also decreasing above 1 MHz, which
tends to offset the degradation in filter performance. Above
1 MHz, current emissions which exceed the desired specification
are usually common mode emissions caused by either ringing
waveforms identified earlier or resonances caused by parasitic
components themselves.
Physical component layout becomes increasingly critical above
1 MHz. Improper layout can lead to increased capacitor ESL.
It is also possible for noise voltages or currents to couple around
the EMI filter directly into the mains.
Common Mode Filter Analysis
Common mode conducted emissions are caused by common
mode currents that do not circulate between the AC mains and
power supply input. Balanced common mode currents flow
simultaneously in power supply line and neutral input wires
such that common mode line current is equal in magnitude and
in phase with common mode neutral current. Unbalanced
common mode currents flow in either power supply line or
neutral input wires separately. Common mode conducted
emissions are caused by TOPSwitch Drain Voltage VDRAIN and
output Diode Voltage VDIODE as shown in Figure 33.
TOPSwitch Drain voltage VDRAIN drives displacement current
through various stray parasitic capacitance terms. CS1 is stray
TOPSwitch Drain capacitance to earth ground. COSS is
TOPSwitch output capacitance. CBD1 through CBD4 are the
effective capacitance terms across each bridge diode. CAC is the
capacitive coupling across the AC mains input (which is very
low when testing with LISNs). Note that secondary is shown
connected directly to earth ground. Transformer capacitance is
distributed but can be modeled with the following six discrete
capacitance terms:
CW1: Winding capacitance from “noisy” or switching side
of the transformer primary to “noisy” side of the secondary.
A
6/96
17
AN-15
CBD1
V+
CBD2
CW3
C
VDIODE
W5
ESL
VAC
CAC
CIN
CW6
VDRAIN
+
CW4
C
W1
-
C
W2
CBD3
V-
CBD4
ESR
COSS
CS2
CS1
PI-1644-111695
Figure 33. Circuit Origin for Common Mode Emissions.
CW2: Winding capacitance from “noisy” or switching side
of the transformer primary to “quiet” side of the secondary.
CW3: Winding capacitance from “quiet” side of the
transformer primary to the “noisy” or switching side of the
secondary.
CW4: Winding capacitance from “quiet” side of the
transformer primary to the “quiet” side of the secondary.
(This is actually a “good” stray capacitance term which is
usually augmented with an additional, Y-capacitor to return
displacement currents back to the driving source).
CW5: Winding capacitance across the primary.
CW6: Winding capacitance across the secondary (CW5 and
CW6 combine to cause a transformer resonant frequency of
400 kHz to 2 MHz above which each winding impedance is
capacitive rather than inductive).
The TOPSwitch Drain node directly drives displacement current
into each of the following stray capacitance terms: CS1 , CW1,
CW2, COSS , and CW5. Each displacement current (ICS1, ICW1, ICW2,
ICOSS, and ICW5) must eventually return to the driving node
(TOPSwitch Drain pin). Each current splits many times but
some fraction of each displacement current may flow through
the power supply AC input conductors and be measured as
18
A
6/96
common mode emission currents as follows:
ICS1: Splits into earth ground component and secondary
component. Secondary current component returns to the
TOPSwitch Drain through transformer capacitance. Earth
ground component returns up from ground into the neutral
wire (and is sensed by the LISN), AC couples into the Line
wire (and is sensed by the LISN), flows back through bridge
diodes (either superimposing on line frequency current
during bridge diode conduction or through effective bridge
diode capacitance when diodes are not conducting), to V+
and V- bus wires, and returns to the TOPSwitch Drain pin
through CW5 and COSS .
ICW1: Splits between CW3 returning to primary, CS2 to
earth ground, and CW6. Splits again between CW2 and earth
ground. Earth ground components returns through neutral
and line wires (as explained above) and is sensed by the
LISN.
ICW2: Splits between CW6 and earth ground. Earth ground
component returns through neutral and line wires (as
explained above) and is sensed by the LISN.
ICOSS: Splits between CIN (note that CIN Equivalent Series
Inductance or ESL will choke off high frequencies) andV-.
V- component flows out the bridge rectifier, down through
line and neutral wires (and is sensed by the LISN) to earth
AN-15
ground, up to secondary, and couples back to TOPSwitch
through transformer winding capacitance.
ICW5: Splits between CIN (with ESL) and V+. V+ component
flows out the bridge rectifier, down through line and neutral
wires (and is sensed by the LISN) to earth ground, up to
secondary, and couples back to TOPSwitch through
transformer winding capacitance.
The ferrite bead or toroid(21) should have an effective impedance
of 100 Ω in the 15 to 25 MHz range. The bead is placed over
the safety ground wire between the enclosure power entry
connector and the internal safety ground attachment point of the
enclosure. The toroid is installed in similar fashion but can
accommodate up to 5 or 6 turns of the insulated safety ground
wire.
Transformer Construction
Superposition of all these different displacement currents will
lead to some cancellation but there will always be “leftover”
high frequency current components measured as common
mode conduction emissions. The asymmetries in various
parasitic capacitance terms also explain how common mode
emission currents can become “unbalanced” creating net current
flowing only in the line or neutral wire. A similar analysis can
be performed using the output rectifier anode as the driving
point voltage source. For obvious reasons, common mode
emission currents are best measured because analysis is quite
difficult.
Common mode filters require relatively high values of inductance
because safety standards restrict common mode Y-capacitor
size to limit leakage current as previously discussed. Common
mode chokes between 10 mH and 33 mH are used in most
applications because inductance normally required is unaffected
by the circulating differential mode current. Discrete chokes
can also be used in some low power applications if the peak
current is taken into account and a discrete choke is placed in
each leg for balanced high frequency impedance.
Physical component layout becomes increasingly critical above
1 MHz. Improper layout can lead to increased capacitor ESL.
It is also possible for noise voltages in close proximity to the
EMI filter to couple around the filter directly into the mains.
Common mode capacitors must have extremely short traces
connecting directly to the transformer pins and to each other as
well.
Power Cord Damping
Applications with 3-wire power cords require special attention.
A six foot power cord can be modeled as a transmission line
with distributed inductance and capacitance, characteristic
impedance of approximately 100 Ω, and little damping which
leads to a sharp, well defined resonance, typically between 15
and 25 MHz. This resonance can amplify existing common
mode emission currents to levels in excess of the desired limit.
A small, lossy ferrite bead or toroid placed over the earth ground
lead wire reduces the resonant peak by adding series damping.
Flyback transformers use gapped ferrite cores which may have
fringing fields as shown in Figure 34. Gaps should be confined
to the center leg of either one or both core halves so that the
fringing field can be effectively shielded by the windings. End
gaps “leak” magnetic flux due to the fringing field which can
produce common mode emissions.
END GAP
WITH
FRINGING
FIELD
NO
END GAP
PI-743-032392
Figure 34. End Gap Magnetic Flux Leakage.
Proper transformer construction techniques are necessary for
reducing common mode emissions. Figure 35 shows a typical
insulated wire wound transformer cross section. The transformer
primary connects between the relatively quiet high voltage DC
bus and the noisy TOPSwitch Drain pin (which has the high
voltage switching waveform). When the primary is wound with
two layers, the primary half with the dot mark is connected to
TOPSwitch which is then buried or shielded under the primary
half connected to the high voltage DC bus as shown. One layer
of 2 mil tape separates the two primary halves to reduce
capacitance and high frequency ringing. Another layer of tape
separates the primary winding from the insulated wire wound
secondary. The combination of tape and insulation thickness
reduces capacitance between primary and secondary which
reduces common mode emission currents. One more layer of
tape separates the secondary from the primary referenced bias
winding.
A
6/96
19
AN-15
BIAS
{
TAPE
}
TAPE
SECONDARY
(INSULATED)
PRIMARY
connected to each core half (manganese zinc ferrite normally
used in 100 kHz flyback transformers is conductive) and the foil
ends are electrically connected to create a shorted turn. The flux
band can usually be left floating without making additional
bobbin changes to meet reinforced insulation requirements for
safety. The flux band (and core) may also be connected back to
primary or to secondary but bobbin construction must be
examined to ensure creepage distance is sufficient to meet
reinforced insulation requirements for safety.
PI-1646-111695
Figure 35. Insulated Wire Wound Transformer Cross Section.
Figure 36 shows a typical margin wound transformer cross
section. Each winding is placed between symmetric margins as
shown while the safety insulation extends beyond the margins
up to the walls of the bobbin flange. The split primary
construction shown reduces leakage inductance in higher power
applications. The primary half connected to TOPSwitch is the
first layer followed by one wrap of 2 mil polyester film tape for
basic insulation. The bias winding is wound next in a single
layer. The bias winding is usually just a few turns but wound
using up to three parallel wires to cover more of the bobbin
width and effectively shield the noisy TOPSwitch Drain
connected primary half. Three wraps of 2 mil polyester film
tape (3M 1298 or equivalent UL recognized tape) provide the
necessary reinforced insulation for safety as well as reducing
capacitance between primary and secondary to minimize
common mode emission currents. Secondary is now wound
between margins followed again by three wraps of 2 mil
polyester film tape for reinforced insulation. The second half
of the primary is wound followed by three wraps of 2 mil
polyester film tape as final insulation.
SECOND PRIMARY HALF
REINFORCED INSULATION
SECONDARY
REINFORCED INSULATION
MARGINS
(BOTH
SIDES)
BIAS
FIRST PRIMARY HALF
PI-1647-111695
Figure 36. Margin Wound Transformer Cross Section.
In some applications, a copper foil “flux band” over the outside
of the completed transformer as shown in Figure 37 may reduce
some common mode emissions. The copper foil wraps one
complete turn over the exposed (but insulated) windings and
each core end gap. Foil width is cut to fit between bobbin
flanges while maintaining required creepage distance for
reinforced insulation. For best effect, the foil is electrically
20
A
6/96
TRANSFORMER FLUX BAND
COPPER
FOIL
STRAP
PI-1648-111695
Figure 37. Transformer Flux Band.
In most TOPSwitch power supplies, a transformer shield is not
necessary because TOPSwitch has controlled turn on which
limits high voltage dv/dt and reduces common mode emission
currents. For those few applications where further reduction in
common mode emission currents is desired, Figure 38 shows
proper shield placement within the transformer. The shield
intercepts interwinding capacitive displacement currents and
returns them to the primary circuitry. Figure 38 shows the
shield connected to V+ but the shield can also be connected to
V- if more convenient for construction or layout reasons. Note
that safety insulation or creepage distance is required between
the primary connected shield and SELV secondary outputs.
The foil shield width is selected to fit between primary safety
margins. Length is precut for one full turn with slight overlap
at the ends. The termination lead wire is soldered to the copper
foil shield in the center (equidistant from each end). Tape
insulation is usually applied to the copper foil shield before
placing on the transformer. The ends must be insulated such
that the foil shield does not form a shorted turn inside the
transformer (compared with the external “belly band” described
earlier which has a shorted turn but is physically located outside
the transformer).
Refer to AN-18 for more information on transformer
construction.
AN-15
TRANSFORMER SHIELD PLACEMENT
Insulated Wire Transformer
Margin Transformer
}
SECOND PRIMARY HALF
SECONDARY
BIAS
SHIELD
PRIMARY
SECONDARY
MARGINS
TAPE
SHIELD
BIAS
FIRST PRIMARY HALF
PI-1649-111695
Figure 38. Transformer Shield Placement.
General Purpose TOPSwitch EMI
Filters
attenuate differential mode emission currents. C7 (Y1-safety
capacitor) and the common mode inductance of common mode
choke L2 attenuate common mode emission currents. Note that
C7 can be replaced by two series connected Y2-safety capacitors,
each with twice the value shown.
2-Wire AC Input
A typical TOPSwitch power supply and EMI filter for 2 wire AC
input applications is shown in Figure 39. X-capacitor C6 and
the differential mode inductance of common mode choke L2
T1
8
D2
UG8BT
7.5 V
R1
39 Ω
1
L2
22 mH
VR2
1N5995B
6.2 V
2
7
3
C6
0.1 µF
X2
C3
120 µF
25 V
U2
NEC2501-H
D1
UF4005
C1
33 µF
400 V
R2
68 Ω
C2
680 µF
25 V
VR1
P6KE200
BR1
400 V
L1
3.3 µH
C5
47µF
RTN
D3
1N4148
4
DRAIN
F1
3.15 A
L
SOURCE
CONTROL
U1
TOP202YAI
C4
0.1 µF
C7
1.0 nF
Y1
N
J1
PI-1650-111695
Figure 39. Typical 2-Wire TOPSwitch Power Supply and EMI Filter.
A
6/96
21
AN-15
3-Wire AC Input
A typical EMI filter for 3 wire AC input applications is shown
in Figure 40. X-capacitor C6 and the differential mode
inductance of common mode choke L2 attenuate differential
mode emission currents. C7 (Y1-safety capacitor), the common
mode inductance of common mode choke L2, and small, lossy
ferrite toroid L3 attenuate common mode emission currents(21).
L3 also damps power cord resonances as previously described.
T1
8
D2
UG8BT
7.5 V
R1
39 Ω
V+ 1
L2
22 mH
VR2
1N5995B
6.2 V
2
7
3
V-
C6
0.1 µF
X2
C3
120 µF
25 V
U2
NEC2501-H
D1
UF4005
C1
33 µF
400 V
R2
68 Ω
C2
680 µF
25 V
VR1
P6KE200
BR1
400 V
L1
3.3 µH
C5
47µF
RTN
D3
1N4148
C8
0.1 nF
4
DRAIN
F1
3.15 A
L
SOURCE
CONTROL
U1
TOP202YAI
C4
0.1 µF
C7
1.0 nF
Y2
N
L3
J1
PI-1651-111695
Figure 40. Typical 3-Wire TOPSwitch Power Supply and EMI Filter.
22
A
6/96
AN-15
Enhanced EMI filter design for Video Applications and
Reduced Radiated Emissions
Figures 41 and 42 show typical EMI filters for 2-wire input
video applications (such as television set-top cable and satellite
decoders). These techniques can also reduce radiated emissions
by keeping high frequency conducted emission currents (30
MHz to 200 MHz) out of the power cable which may act like an
antennae and broadcast radiated emissions.
D2
UG8BT
T1
8
V+
7.5 V
1
1000 pF
CA
1000 pF
500 V
L2
22 mH
R2
68 Ω
C3
120 µF
25 V
U2
NEC2501-H
VR2
1N5995B
6.2 V
2
C1
33 µF
400 V
7
3
V-
C6
0.1 µF
X2
R1
39 Ω
C2
680 µF
25 V
VR1
P6KE200
D1
UF4005
BR1
400 V
L1
3.3 µH
C5
47µF
RTN
D3
1N4148
4
DRAIN
F1
3.15 A
L
SOURCE
CONTROL
U1
TOP202YAI
C4
0.1 µF
C7
1.0 nF
Y1
N
J1
PI-1653-111695
Figure 41. Typical 2-Wire TOPSwitch Power Supply and EMI Filter for Video Applications and Reduced Radiated Emissions.
Figures 41 and 42 both show a low cost, ceramic, 1000 pF
capacitor across output diode D2 to help control dv/dt and diode
ringing voltages. Higher output voltages may require reducing
the size of the capacitor and increasing the voltage rating.
In Figure 41, another 1000 pF, 500 V capacitor (CA) is added
in parallel with input energy storage capacitor C1 for high
frequency bypass in the 30 MHz to 200 MHz range. This
additional capacitor should be placed directly from TOPSwitch
source pin to transformer V+ pin.
Figure 42 shows an alternate circuit using three Y2-capacitors
to balance the high frequency common mode impedance to V+
and V-. Note that two Y2-capacitors (CA and C7) are essentially
in series and effectively provide high frequency bypass across
C1.
A
6/96
23
AN-15
D2
UG8BT
T1
8
L1
3.3 µH
7.5 V
R1
39 Ω
V+ 1
1000 pF
C2
680 µF
25 V
VR1
P6KE200
C3
120 µF
25 V
U2
NEC2501-H
D1
UF4005
BR1
400 V
L2
22 mH
VR2
1N5995B
6.2 V
2
C1
33 µF
400 V
7
RTN
CA
3
V-
C6
0.1 µF
X2
R2
68 Ω
C5
47µF
D3
1N4148
2.2 nF
Y2
C8
2.2 nF
Y2
4
DRAIN
F1
3.15 A
L
SOURCE
C7
2.2 nF
Y2
CONTROL
U1
TOP202YAI
C4
0.1 µF
N
J1
PI-1654-111695
Figure 42. 2-Wire TOPSwitch Power Supply and EMI Filter with Three Y-Capacitors For Video Applications and Reduced Radiated
Emissions.
Alternative Filter Approach Without Common Mode Choke
An alternative filter for lower power (below 5 watt) applications
is shown in Figure 43. This filter splits the high voltage energy
storage capacitor to create a filter. Peak current in L2 and L3 is
approximately half the peak current in bridge rectifier BR1 due
to capacitor C1. Differential mode attenuation is provided by
C1, C4, L2, and L3. The secondary is AC coupled back to the
primary by Y1-safety capacitor C7. Note that C7 can be
replaced by two series connected Y2-safety capacitors with
twice the value as shown.
L2
V+
1 mH
1
8
D1
BZT03-C200
200 V
1.5 W
D4
1N5822
L1
(Bead)
5V
C2
330 µF
25 V
C3
150 µF
25 V
5
BR1
400V
C1
2.2 µF
400 V
D2
UF4005
C4
2.2 µF
400 V
4
U1
D
C6
0.047 µF
X2
L
N
1 mH
V-
3
T1
T1200
C5
47 µF
10 V
D3
1N4148
Figure 43. Low Power (Below 5W) TOPSwitch Supply Using Discrete Normal Mode Chokes.
A
6/96
C7
1.0 µF
Y1
S
L3
24
R1
22 Ω
2
C
TOP
210
S
RTN
PI-1652-111695
AN-15
metal enclosures, plastic enclosures with a conductive coatings
on the internal surfaces, stamped and formed metal shapes,
deep drawn cans, and metal foil. Low-impedance connections
to the shield are important. Long wires, which degrades
performance, must be avoided.
Enclosure Shielding
Many applications, such as cellular phone battery chargers,
printers, portable computer adapters and chargers, and video
games are packaged in plastic enclosures with no additional
shielding. Conducted and radiated emissions are controlled
with proper selection of EMI filter components, careful
transformer construction, and tight PC layout practices.
Figure 44 shows a typical 3 wire implementation with the
enclosure shield connected directly to the third wire earth
ground. The enclosure is AC coupled with Y2-capacitor C7
back to primary circuitry. The enclosure is connected to
secondary through low voltage, ceramic capacitor C8 or may be
directly connected depending on the system configuration.
Electrical safety is maintained under single component failure
conditions such as a short circuit failure of C7 or open circuit
failure of the connection to third wire earth ground. The third
wire earth ground connection to the enclosure safely shunts
fault current to provide protection if C7 fails short. C7 safely
limits fault current flow to less than 3.5 mA (IEC950 limit for
three wire, 250 VAC) if the third wire earth ground fails open
circuit.
Some applications, such as desktop computers and other
information technology equipment, have increased sensitivity
to conducted and radiated emissions and have a conductive
enclosure connected to the AC mains third wire earth ground.
Other applications, including distributed neural networks and
consumer electronic devices such as VCRs and TV set-top
decoders, also have a conductive enclosure but are powered
from 2 wire AC input and have no earth ground connection.
The enclosure forms a conductive shell which intercepts and
returns displacement currents back to the primary circuitry as
shown in Figures 44-46. Practical implementations include
T1
8
V+
D2
UG8BT
7.5 V
R1
39 Ω
1
U2
NEC2501-H
D1
UF4005
L2
22 mH
7
3
V-
C6
0.1 µF
X2
C3
120 µF
25 V
VR2
1N5995B
6.2 V
2
C1
33 µF
400 V
R2
68 Ω
C2
680 µF
25 V
VR1
P6KE200
BR1
400 V
L1
3.3 µH
C5
47µF
RTN
D3
1N4148
C8
0.1 µF
4
DRAIN
F1
3.15 A
L
N
SOURCE
CONTROL
U1
TOP202YAI
C4
0.1 µF
C7
1.0 nF
Y2
L3
J1
PI-1657-111695
Figure 44. Typical 3-Wire TOPSwitch Power Supply and EMI Filter with AC Coupled SELV Potential Shielded Enclosure.
A
6/96
25
AN-15
Figure 45 shows a typical 2 wire implementation with the
enclosure shield connected to secondary RTN which places the
enclosure at SELV potential. PC board creepage distance
between primary circuits and the SELV shielded enclosure
must meet reinforced insulation requirements. The enclosure is
AC coupled with a small Y1-capacitor (C7) back to primary
circuitry. Another common approach is to use a series
combination of two Y2-safety capacitors which meets electrical
safety requirements because each capacitor will safely limit
fault current to under 250 µA (IEC950 limit for two wire, 250
VAC) if the other capacitor fails short.
T1
8
D2
UG8BT
7.5 V
R1
39 Ω
V+ 1
VR2
1N5995B
6.2 V
2
C1
33 µF
400 V
7
3
V-
C6
0.1 µF
X2
C3
120 µF
25 V
U2
NEC2501-H
D1
UF4005
BR1
400 V
R2
68 Ω
C2
680 µF
25 V
VR1
P6KE200
L2
22 mH
L1
3.3 µH
C5
47µF
RTN
D3
1N4148
4
DRAIN
F1
3.15 A
L
SOURCE
CONTROL
U1
TOP202YAI
C4
0.1 µF
C7
1.0 nF
Y1
N
J1
PI-1656-111695
Figure 45. Typical 2-Wire TOPSwitch Power Supply and EMI Filter with SELV Connected Shield.
26
A
6/96
AN-15
Figure 46 shows a typical 2 wire implementation with the
enclosure shield at primary potential (usually connected to the
TOPSwitch Source pin). This technique is used for partial
enclosures (which may actually be the TOPSwitch heat sink)
which are safely isolated from the SELV output voltages. PC
board creepage distance between primary connected shield and
the circuits connected to SELV output voltages must meet
reinforced insulation requirements. The enclosure is AC coupled
with a small reinforced insulation, Y1-capacitor (C7) back to
SELV output circuitry. Another common approach is to use a
series combination of two Y2-safety capacitors which meets
electrical safety requirements because each capacitor will safely
limit fault current to under 250µA (IEC950 limit for two wire,
250 VAC) if the other capacitor fails short.
T1
8
D2
UG8BT
7.5 V
R1
39 Ω
V+ 1
L2
22 mH
VR2
1N5995B
6.2 V
2
7
3
V-
C6
0.1 µF
X2
C3
120 µF
25 V
U2
NEC2501-H
D1
UF4005
C1
33 µF
400 V
R2
68 Ω
C2
680 µF
25 V
VR1
P6KE200
BR1
400 V
L1
3.3 µH
C5
47µF
RTN
D3
1N4148
4
DRAIN
F1
3.15 A
L
SOURCE
CONTROL
U1
TOP202YAI
C4
0.1 µF
C7
1.0 nF
Y1
N
J1
PI-1655-111695
Figure 46. Typical 2-Wire TOPSwitch Power Supply and EMI Filter with Primary Connected Partial Shield.
A
6/96
27
AN-15
EMI Filter Layout Issues
Filter layout is extremely important to obtain the desired
attenuation. Poor layout practice can cause conducted emissions
to actually couple around the filter components directly into the
AC mains conductors or cause radiated emissions.
Keep power stage and output components away from the EMI
filter to prevent coupling around the filter. The best approach
is to place the EMI filter at one end of a rectangular power
supply shape and place the output at the other end as shown in
Figure 47. Square power supply shapes should be avoided if
possible since the power stage and the output components will
be in close proximity to the EMI filter, allowing noise to couple
directly into the mains.
(a)
WRONG
(b)
RIGHT
PI-275-081090
Figure 48. Bending the Bus to Minimize Resistive Effects.
EMI
Power
Stage Output
been described in detail. The EMI filter must attenuate the
emissions below the specification limit. Implementing a
successful EMI filter is an iterative process. The basic steps
include:
(a)
RIGHT
• Identify and attenuate the differential mode fundamental
EMI
Power
Stage
(b)
WRONG
• Identify and attenuate the common mode fundamental
• Identify other emissions over the spec limit.
Output
PI-745-032392
Figure 47. Power Supply Layout to Minimize Noise Coupling.
Capacitor lead length must be minimized as much as possible
to reduce ESL. This includes the traces on the PC board leading
up to the capacitor pads. Y-capacitor lead lengths and trace
lengths are the most critical because the Y-capacitors couple
high frequency currents (10 MHz to 200 MHz) back to primary
circuitry. Figure 48 shows the right and wrong way to route PC
traces to capacitors.
• Determine whether each emission is differential mode or
common mode.
• Use average or quasi-peak measurements on peak emissions
to verify that the emission actually has insufficient margin
compared to the EMI standard.
• Determine whether each emission is coupling around or
passing through the EMI filter.
• Change the filter design or control the circuit source to
attenuate each emission below the specification limit.
Locate the differential mode filter capacitor across the AC input
conductors as close as possible to the power entry point.
• Go back and check the earlier emission levels to make sure
a change did not cause a different problem to occur.
Practical Considerations
Differential mode Versus Common mode
The first frequency sweep for EMI conducted emissions on a
power supply with no EMI filter will usually produce a spectrum
as shown in Figure 49. The fundamental is outside the
specification limit as well as some of the harmonics. Each
Successful EMI filter design begins with knowledge of the
switching power supply noise sources generating differential
mode and common mode conducted emissions which have
28
A
6/96
AN-15
harmonic is composed of both differential mode and common
mode emissions.
40
Using Splitters
Splitters combine the output signals of the two LISNs to
determine if a specific emission is common mode or differential
mode. Two splitters are required: an in-phase splitter (Mini
Circuits Lab ZSC-2-2) where VOUT is the sum of the two LISN
signals and a 180 degree out-of-phase splitter (ZSCJ-2-2)
where VOUT is the differential between the two LISN signals(22).
The splitter setup is shown in Figure 51.
20
Splitter units are also available which allow switching between
differential mode and common mode tests.
PI-746-032392
80
60
ID
0
0
1
2
L
Frequency (MHz)
+Vn
+Vcb
Figure 49. Typical Conducted Emissions Data without EMI Filter.
In Figure 50, a differential mode component with magnitude
70 dBµV is shown relative to a common mode component with
magnitude 50 dBµV. 20 dB differential between the two
components is actually an order of magnitude between the
absolute values of the two components. The signals will add
and superimpose if the phasing is correct but the overall effect
on the measured signal level is slight (10% increase, or less than
1 dB).
To
Power
Supply
Icb
Splitter
G
In
Spectrum
Analyzer
dB/µV
Conducted Emissions (dBµV)
will have to be employed to reduce this harmonic any further.
Icu Icb
f
-Vn
+Vcb
+Vcu
N
To
Power
Supply
70 dB(µV)
ID
40 dB(µV)
Figure 51. Separating Differential Mode from Common Mode
Using a Splitter.
30 dB
DM Filter
Differential-mode
Common-mode
PI-1727-121895
Figure 50. Superimposed Common Mode and Differential Mode
Harmonics.
A differential mode filter with 30 dB attenuation at the harmonic
frequency of interest will not attenuate the measured peak by
30 dB. The differential mode component will be attenuated
from 70 dBµV to 40 dBµV, but the 50 dBµV common mode
peak will now dominate the measurement. Further differential
mode attenuation will have no effect on the measured harmonic
because the signal is common mode. Common mode filtering
PI-1728-121895
Differential Mode Splitter Measurement
Differential mode emission current (ID) circulates from the
power supply through the first LISN sense resistor (producing
an in phase sense voltage), through the second LISN sense
resistor (producing an out of phase sense voltage), and back to
the power supply. The output voltage of the in-phase splitter
will have no differential mode component because the opposite
phased sense voltages effectively cancel. The output voltage of
the 180 degree out-of-phase splitter will have differential mode
components 6 dB higher than those measured directly at the
LISN as the sense voltages are now effectively in phase and
sum.
Balanced Common Mode Splitter Measurement
Balanced common mode currents (ICB) are defined as currents
with similar amplitude and phase that flow from ground through
A
6/96
29
AN-15
Unbalanced Common Mode Splitter Measurement
Unbalanced common mode currents (ICU) flow from ground
through either LISN sense resistor. Unbalanced common mode
currents are found when the EMI filter does not have balanced
impedance in each leg or when noise from the power path
returns from the AC mains asymmetrically through one side of
the EMI filter (typically caused by asymmetric parasitic
capacitance). The output voltage of the in-phase splitter will
have unbalanced common mode components equal to those
measured directly at the LISN because there is no cancellation.
The output voltage of the 180 degree out-of-phase splitter will
also have unbalanced common mode components equal to
those measured directly at the LISN for the same reason.
frequency (above which the choke behaves like a capacitor) and
effective Q. Identify multi-resonant behavior due to multiple
layer winding.
Spectrum
Analyzer
dBµV
each LISN sense resistor (producing sense voltages with the
same phase) and through the AC input to the power supply. The
output voltage of the in-phase splitter will have balanced
common mode components 6 dB higher than those measured
directly at the LISN because the in phase sense voltages
effectively add. The output voltage of the 180 degree out-ofphase splitter will have no balanced common mode components
as the sense voltages are now effectively out of phase and
cancel.
f
L
LISN
EMI
FILTER
G
N
POWER
SUPPLY
&
LOAD
LISN
GROUNDED BOX
PI-749-032392
The results of using the two splitters on each of the three types
of emissions is shown in Table 5.
LISN
OUTPUT
IN
PHASE
180° OUT
OF PHASE
Differential (VD)
0
VD + 6 dB
Balanced
Common-mode (Vcb)
Vcb + 6 dB
0
Unbalanced
Common-mode (Vcu)
Vcu
Vcu
Table 5. Splitter Signal Levels.
EMI Filter Component Measurements
No EMI filter component is perfect. At some frequency all
components “give up” their basic characteristic to the effects of
parasitics.
Measure all capacitors. Identify specifically the self-resonant
frequency (above which the capacitor looks like an inductor)
and effective Q.
Measure all chokes. Identify specifically for the self resonant
30
A
6/96
Figure 52. Typical Conducted Emissions System Test Set-Up with
Grounded Box.
Spatial Coupling
As power supplies get smaller the EMI filter gets physically
closer to circuitry acting as noise generators. High dv/dt
voltage waveforms and high di/dt current loops generate fields
which may spatially couple around the EMI filter and induce
emission currents directly in the mains. Noise currents which
couple around the filter must be distinguished from the noise
currents which are passing through the filter.
One way to separate conducted emission currents is to place the
power supply power circuitry and load within a grounded box
as shown in Figure 52. The EMI filter is connected between the
enclosed power supply and AC mains. The box will contain the
fields, allowing conducted emission currents to be directly
measured. This is especially effective when analyzing the
fundamental. (Note that this technique is for investigative
purposes only and must not be used for final test data).
The spatial coupling emissions can be reduced by containing
the fields at their sources with local shields. Local shields over
primary power circuitry such as the flyback transformer, primary
damper, clamp diode, and TOPSwitch can be used to contain
fields. Local shields can also be used over secondary circuitry
such as output rectifiers. Shields can also be applied around the
EMI filter although the preferred approach is to contain the field
AN-15
at the source. Heat sinks can also be used as shields.
Another type of shield is a conductive plate approximately the
same size as the printed circuit board. This plate can be
connected to earth ground, primary reference, or secondary
reference (depending on the safety insulation system chosen).
This is attractive for applications connecting to the third wire
ground but without a conductive enclosure.
Lossy Beads
Small beads can be used in circuit leads to damp or eliminate
high frequency ringing. Ferrite beads from Fair-Rite(21) are
available in a variety of shapes. These beads feature low
impedance at low frequency for minimal effect on the current
waveform but have high impedance at high frequency with
significant parallel resistance to damp and reduce ringing
voltage waveforms.
Grounding
Some applications connect the output voltage to earth ground.
Others have no connection at all to earth ground. It is important
to identify the earth ground connection expected for each
application and to test in that configuration. EMI testing should
be performed in both grounded and ungrounded configurations.
Power Cord
The power cord resonance previously described can interfere
with conducted emissions testing. Switch between two power
cords of different lengths to separate power cord resonances
from other conducted emissions.
Miscellaneous Test Tricks
Terminate opposite LISN with 50 ohm terminator. The LISN
sense impedance is actually determined by the termination and
will change if not properly terminated.
Warm up equipment including Device Under Test (DUT) for at
least 1 hour before testing so results will be repeatable.
Make sure analyzer sweep speed is low enough to capture the
peaks of each harmonic. The bridge rectifier conducts current
(both power and emission currents) for a short time compared
with the full line cycle which effectively “pulse width modulates”
the emission currents. Slower sweep speeds will collect enough
data to accurately measure the peak of each current emission. A
peak hold test can also be used to fill in the peaks in a few
sweeps.
Peak measurements take the least amount of time but
specifications are given in quasi-peak or average limits. Both
quasi-peak and average measurement techniques give lower
readings when compared to the peak value. If the peak value
meets the average or quasi peak limit specification, there is no
need to take further data with the average or quasi peak
methods. To save test time, use the quasi-peak and average test
methods only when the measured peak value is close to or
exceeds the target specification.
Recommended Step-by-Step Procedure
1) Determine differential mode fundamental (and low frequency
harmonics).
2) Calculate and select filter X-capacitance and target differential
inductance. Select bobbin style common mode choke with
sufficient differential mode inductance and AC current rating
(use discrete chokes only for low power, under 5 Watts).
3) Measure impedance versus frequency for each component.
Select components with resonant frequencies that do not coincide
with waveform ringing frequencies in the power supply.
4) Use a nominal value AC source through the LISNs to
provide power to the power supply. Use the 180 degree splitter
to extract the differential mode fundamental current component.
Measure the fundamental with slow sweep speed and measure
peak value using maximum hold. Compare the measured
differential mode fundamental with the calculated value. Use
average or quasi-peak detection as required to properly compare
measured reading with limits of the chosen standard. Increase
X2-capacitance or select common mode choke with higher
differential inductance if necessary.
5) Examine entire frequency range for differential mode
components close to or in excess of the specification limit.
Make measurements on both Line and Neutral LISN. Pay
special attention to frequency ranges around measured
component resonances and identified circuit ringing frequencies.
Use average or quasi-peak detection as required to properly
compare measured reading with limits of the chosen standard.
Modify differential mode filter design if necessary.
6) Use in-phase splitter to extract the balanced common mode
fundamental and low frequency harmonic current components.
For 2-wire applications, place the largest value Y1-safety
capacitor (subject to leakage current limitations but typically
1 nF) to output return. Two series connected Y2-safety capacitors
(typically each with 2.2 nF value) can also be used. For 3-wire
applications, place the largest value Y2-safety capacitor (subject
to leakage current limitations but typically 1 nF up to 33 nF)
from the power supply high voltage return to earth ground and
then connect a 0.1 uF low voltage ceramic capacitor from earth
ground to secondary return. Y-capacitor leads must be very
short to attenuate high frequency current emissions. Measure
balanced common mode fundamental and low frequency
harmonics. Increase size of common mode choke if necessary.
Select the smallest, widest bandwidth common mode choke
(with sufficient RMS current ratings) that attenuates the balanced
common mode fundamental to the desired level. Measure
impedance versus frequency for each component. Select
components with resonant frequencies that do not coincide with
A
6/96
31
AN-15
waveform ringing frequencies in the power supply.
7) Examine entire frequency range for balanced or unbalanced
common mode components close to or in excess of the
specification limit. Make measurements on both Line and
Neutral LISN. Pay special attention to frequency ranges around
measured component resonances and identified circuit ringing
frequencies. Use average or quasi-peak detection as required to
properly compare measured reading with limits of the chosen
standard. Modify common mode filter design if necessary.
8) Remove splitters. Measure signals from both Line and
Neutral over the entire frequency range. Emissions above 1
Mhz are usually common mode. Solve spatial coupling effects
using additional Y-capacitors (see Figures 41 and 42), improved
transformer construction (see Figures 35 and 36), higher
bandwidth two section common mode choke (see Figure 21),
additional high frequency common mode choke (see Figure 27)
or shielding techniques (Figures 44-46).
9) Perform final test with secondary connected to ground and
also with secondary isolated from LISN ground.
BIBLIOGRAPHY
(1)
FCC Harmonizes Its ITE Rules with C.I.S.P.R.
Requirements, Compliance Engineering, pp 7, January/
February, 1994.
(9)
(2)
European Standard EN55022, “Limits and Methods of
Measurement of Radio Interference Characteristics of
Information Technology Equipment”, Cenelec, 1994.
(10) Elliott Laboratories, 684 W. Maude Avenue, Sunnyvale,
CA, 94086 (408-245-7800).
(3)
“Limits and Methods of Measurement of Radio
Interference Characteristics of Information Technology
Equipment”, C.I.S.P.R Publication 22, 1993.
(4)
Dash, D. and Straus, I.; “EMC Regulations in Germany”;
Compliance Engineering 1994 Reference Guide, pp. 8596, Compliance Engineering, Boxborough, MA.
(5)
Regulation Vfg 243/1991, “Radio Interference
Suppression of Radio-Frequency Equipment for
Industrial, Scientific, Medical (ISM) and Similar Purposes
and Equipment used in Information Processing Systems;
General License”, German Federal Minister for Post and
Telecommunications.
(6)
Amending Regulation Vfg 46/1992, German Federal
Minister for Post and Telecommunications.
(7)
“Meeting the EC Emissions Requirement”, Handbook of
EC EMC Compliance, Compliance Design Incorporated,
1993, pp 7-15.
(8)
32
M. Engelson, “EMI Applications using the Tektronix
2712 Spectrum Analyzer”, Application Note, Tektronix,
1993.
A
6/96
M. Engelson, “Measures of EMC, A Review of Basic
EMC Measurement Techniques and Standard Practices”,
EMC Test and Design, November/December 1990.
(11) IEC1950, “Safety of Information Technology Equipment
Including Electrical Business Equipment”, Second
Edition, International Electrotechnical Commission, 1991.
(12) Evox Rifa, “Capacitors 1995”, Catalogue, Evox- Rifa
Inc., 100 Tri-State International, Suite 290, Lincolnshire,
Illinois 60069, (1-708-948-9511).
(13) European Standard EN 132400, “Sectional Specification:
Fixed Capacitors for Electromagnetic Interference
Suppression and Connection to the Supply Mains”,
(Assessment Level D), Cenelec, 1994.
(14) Murata Electronics, North America, Inc., “Disk Ceramic
Capacitor Catalog no. C-OS-C 1991”, Murata Erie North
America, 2200 Lake Park Drive, Smyrna, Georgia
30080, (1-800-831-9172).
(15) Roederstein, "EMC Radio Interference Suppression
Components", 1991, (704-872-8101).
(16) Tokin, “EMC Line Filters Vol. 2 CD-07JE”, Tokin
America Inc., 155 Nicholson Lane, San Jose, CA 95134
(1-408-432-8020).
AN-15
(17) Tamura , “Tamura Common Mode Coils for AC Line &
EMI Filtering, 0.1 to 50 Mhz”, Electronic Engineers
Master Catalogue (EEM), 1995, Tamura Corporation of
America, P.O. Box 892230, Temecuca, CA 92589,
(1-909-699-1270).
(19) TDK Corporation of America, "TDK EMI Prevention
Components", EVE-005B, 1993, (708-803-6100).
(18) Panasonic Industrial Company, “Panasonic Line Filters”,
Digikey Catalog 961, January-February 1996 (1-800344-4539). Also “Panasonic Inductors/Transformers 94/
95” #9404155S1, Panasonic Industrial Co., 2 Panasonic
Way (7H-3), Secaucus, New Jersey 07094 (1-201-3484630).
(21) Fair-Rite, “Fair-Rite Soft Ferrites” Catalogue, 12th edition,
Fair-Rite Products Corporation, P.O. Box J, 1 Commercial
Row, Wallkill, NY 12589, (914-895-2055).
(20) Murata Erie, "EMI/RFI Filter Catalog", E-06-A, 1993,
(1-800-831-9172).
(22) Mini-Circuits Laboratories, 13 Neptune Ave., P.O. Box
350166, Brooklyn, NY 11235, (718-934-4520).
A
6/96
33
AN-15
NOTES
34
A
6/96
AN-15
NOTES
A
6/96
35
AN-15
Power Integrations reserves the right to make changes to its products at any time to improve reliability or manufacturability.
Power Integrations does not assume any liability arising from the use of any device or circuit described herein, nor does it
convey any license under its patent rights or the rights of others.
PI Logo and TOPSwitch are registered trademarks of Power Integrations, Inc.
©Copyright 1994, Power Integrations, Inc. 477 N. Mathilda Avenue, Sunnyvale, CA 94086
WORLD HEADQUARTERS
Power Integrations, Inc.
477 N. Mathilda Avenue
Sunnyvale, CA 94086
USA
Main:
408•523•9200
Customer Service:
Phone: 408•523•9265
Fax:
408•523•9365
AMERICAS
For Your Nearest Sales/Rep Office
Please Contact Customer Service
Phone: 408•523•9265
Fax:
408•523•9365
JAPAN
Power Integrations, K.K
Keihin-Tatemono 1st Bldg.
12-20 Shin-Yokohama 2-Chome, Kohoku-ku
Yokohama-shi, Kanagawa 222 Japan
Phone: 81•(0)•45•471•1021
Fax:
81•(0)•45•471•3717
ASIA & OCEANIA
For Your Nearest Sales/Rep Office
Please Contact Customer Service
Phone: 408•523•9265
Fax:
408•523•9365
36
A
6/96
EUROPE & AFRICA
Power Integrations (Europe) Ltd.
Mountbatten House
Fairacres
Windsor SL4 4LE
United Kingdom
Phone: 44•(0)•1753•622•208
Fax:
44•(0)•1753•622•209
APPLICATIONS HOTLINE
World Wide
408•523•9260
APPLICATIONS FAX
Americas
408•523•9361
Europe/Africa
44•(0)•1753•622•209
Japan
81•(0)•45•471•3717
Asia/Oceania
408•523•9364