Design Techniques for EMC Part 3 - Filtering and Suppressing Transients (second part)

Note: This has been split in three, the final part will appear in Issue 68

This is the third in a series of six articles on basic good-practice electromagnetic compatibility (EMC) techniques in electronic design, to be published during 2006. It is intended for designers of electronic modules, products and equipment, but to avoid having to write modules/products/equipment throughout - everything that is sold as the result of a design process will be called a ’product’ here.

This series is an update of the series first published in the UK EMC Journal in 1999 [1], and includes basic good EMC practices relevant for electronic, printed-circuit-board (PCB) and mechanical designers in all applications areas (household, commercial, entertainment, industrial, medical and healthcare, automotive, railway, marine, aerospace, military, etc.). Safety risks caused by electromagnetic interference (EMI) are not covered here; see [2] for more on this issue.

These articles deal with the practical issues of what EMC techniques should generally be used and how they should generally be applied. Why they are needed or why they work is not covered (or, at least, not covered in any theoretical depth) - but they are well understood academically and well proven over decades of practice. A good understanding of the basics of EMC is a great benefit in helping to prevent under- or over-engineering, but goes beyond the scope of these articles.

The techniques covered in these six articles will be:

  1. Circuit design (digital, analogue, switch-mode, communications), and choosing components
  2. Cables and connectors
  3. Filtering and suppressing transients
  4. Shielding
  5. PCB layout (including transmission lines)
  6. ESD, surge, electromechanical devices, power factor correction, voltage fluctuations, supply dips and dropouts

Many textbooks and articles have been written about all of the above topics, so this magazine article format can do no more than introduce the various issues and point to the most important of the basic good-practice EMC design techniques. References are provided for further study and more in-depth EMC design techniques.

Table of contents for this article

3. Part 3 - Filters and transient suppressors

In the last Issue 66, September 2006

3.1 Introduction
3.2 Designing and selecting filters
3.2.1 How filters work
3.2.2 Imperfections in the basic filter circuits
3.2.3 The importance of the RF Reference
3.2.4 Differential-mode (DM) and Common-mode(CM)
3.2.5 Maximising impedance discontinuities
3.2.6 Using soft ferrite cores

In this issue

3.2.7 Issues with wound components (inductors,transformers, chokes, etc.)
3.2.8 Specifying and designing filters
3.2.9 Problems with real-life supply impedances
3.2.10 Problems with real-life switch-mode converter input impedances
3.2.11 Damping filter resonances that cause gain
3.2.12 Filters and safety
3.3 Filter installation issues
3.3.1 Input and output conductors
3.3.2 Skin effect and the flow of surface currents
3.3.3 The synergy of filtering and shielding
3.3.4 Assembly and installation techniques for filters that penetrate shields
3.3.5 Designing to prevent corrosion
3.3.6 Filters connected in series
3.4 Types of overvoltage transients and surges

In the next issue

3.5 Protecting from surges
3.5.1 Protecting insulators from surge overvoltages
3.5.2 Protecting conductors from surge overcurrents
3.5.3 Protecting electromechanical contacts from surges
3.5.4 Galvanic isolation is the best defence against surges
3.5.5 Surge suppression with filters
3.5.6 Suppression with surge protection devices (SPDs)
3.5.7 Types of Surge Protection Device (SPD)
3.5.8 Characteristics and comparisons of SPD types
3.5.9 Minimising the inductance in series with SPDs
3.5.10 Rating SPDs
3.5.11 Combining SPDs
3.5.12 A hierarchy of surge protection
3.5.13 Protecting SPDs
3.5.14 Equipment reliability and maintenance issues
3.5.15 Surge protection products
3.5.16 ’Earth lift’ problems in systems
3.5.17 Data needs error detection/correction

3.6 References
3.7 Acknowledgements

3. Part 3 - Filters and transient suppressors

3.2.7 Issues with wound components (inductors, transformers, chokes, etc.)

Three main issues of concern for wound (inductive) components are the control of their stray magnetic fields, the variability of their parameters with current and temperature, and their resistivity.

Closed magnetic circuits are preferred, to reduce stray fields

If the path of the magnetic field includes air, the component can be a significant cause of emissions, and can also pick up ambient magnetic fields and cause noise in the circuit. As a result, the EMC of wound components benefits from having a closed magnetic circuit, such as a ferrite bead, cylinder or toroid. This is important for filters, and also for chokes and transformers in switch-mode power converters, whether AC-DC, DC-AC (e.g. inverters) or DC-DC.

For example, an inductor wound on a rod core is essentially the same as a ferrite rod antenna in a typical AM radio. Many types of low-current inductors (up to about 1mH) or ferrite suppression components are available as axial components, essentially ferrite rods with a winding on them. While they can be perfectly acceptable in some applications, in others they can pick up noise from ambient fields and cause or suffer from crosstalk and/or emissions or immunity problems - so it is generally more reliable to use types based on ferrite cores that are beads, cylinders or toroids, which have closed magnetic circuits. Figure 3K illustrates this issue.

This can be a significant problem for DM chokes carrying higher currents, because they have large magnetic fluxes and rely on air gaps to prevent their cores from saturating. Where an air gap is unavoidable, solutions include applying a shield over the component to contain its stray fields, or else use a core made of iron or iron oxide powder in an epoxy binder, or a similar distributed-air-gap material.

Close proximity of a shield reduces the component’s inductance - and the shield might become saturated from the high field strength and no longer provide shielding - so the shield should not be too close to the core or the windings. Iron powder or similar cores have no discrete air gaps, relying instead on hundreds of millions of microscopic air gaps between the magnetic particles in their cores, with much smaller stray magnetic fields as a result.>/p> Slide 10

Figure 3K Stray fields from air gaps in the cores of wound components are bad for EMC

Allowing for the variation of parameters due to DC and/or LF currents, and temperature the wall of the enclosure, with the filter’s body electrically bonded directly to the shielding surface of the wall, otherwise both the filtering and shielding performances will be degraded by stray coupling around the filter. The type of filter required is often called a bulkhead-mounting, or through-bulkhead filter, because it fits through the metal wall (bulkhead) that it is mounted upon. The shielded enclosure considerably reduces the stray coupling between the filter’s input and output. It may even be necessary to fit a conductive EMC gasket around the aperture in the shield where the filter is mounted, for the maximum possible filtering and shielding. Issues of filtering with shielded enclosures are covered in more detail below.

As the DC and/or low-frequency current in a filter increases, the inductance value of all the series inductors falls, affecting the filter performance and resonant frequencies. Also - for all soft ferrites - the frequency at which the peak impedance occurs increases. Sufficient levels of current will saturate the magnetic circuit, causing the inductance/impedance to fall to practically zero. This is a common cause of the differences between simple calculations or simulations and real-life filter performance.

Variations in temperature have a similar effect. Above about 25øC, the inductance of a magnetic core reduces, falling more rapidly as the ’Curie Point’ is approached, and zero above that point. The Curie Point depends upon the type of material, but is generally between 100 and 200øC. An experiment that showed how the performance of a mains filter could be reduced by as much as 20dB, by variations in current loading and ambient temperature that remained within the operational ratings of the filter, is described in [7].

So if designing a filter, get all the appropriate graphs of current and temperature dependency from the core suppliers and take this effect into account during the filter design for the full range of currents and temperatures they are required to operate over. And if choosing a filter from a supplier, make sure you understand what its minimum performance will be over the range of currents and temperatures (simultaneously) that it will experience in operation.

A particular problem is AC-DC mains power supplies that do not meet EN/IEC 61000-3-2 Class D. They draw their mains currents as peaks that are many times higher than their rated or measured RMS supply current. These peaks will degrade the attenuation 100 times per second (120 times/second in 60Hz countries) and might even cause momentary saturation of filter inductors and seriously compromise the filtering performance achieved.

CM chokes aim to balance the send and return currents so there is no net magnetisation of their cores by the wanted DM currents. But because no windings can be perfect, there is always some imbalance, which manifests itself as a DM choke in series with the CM choke. The resulting imbalance currents can saturate the cores of CM chokes, especially because their cores are made as small as possible to save space and reduce cost.

If designed carefully, CM chokes always run cool. Saturated inductors run warm, and if used on power frequencies they may be heard to hum or buzz, or felt to vibrate, both of which are clues to possible errors in design.

Resistivity of ferrite cores

All ferrites are ceramics, but they are not insulators - different types of material have differing resistivities. So if the insulation on the winding wire is inadequate for the stresses and strains imposed by winding it on the core, the core can ’short-out’ the windings, or at least provide a parallel resistance that affects performance. If the winding is carrying a hazardous voltage, failed insulation can result in a ’live’ core that can be a safety hazard to service personnel, or even to a user.

There is always stray capacitance between a winding and its core, and because ferrites are ceramics their dielectric constant is high so the stray C is increased. RF voltages on the windings will therefore induce RF currents in the core, and because ferrite is conductive, these currents can flow out of the core into other conductors via stray capacitance or resistive contact.

Service personnel could get shocks and RF burns from the cores of wound filter components or transformers that are handling large amounts of RF. These stray RF core currents are CM, and so can cause significant problems for EMC. It is important either to insulate the cores to help inhibit the flow of stray CM currents, or else to provide a connection to the core that returns the stray current to the appropriate part of the circuit.

The most appropriate part of the circuit to return the core current to is not necessarily the ’earth’, ’ground’, ’chassis’, ’frame’ or ’0V’. All currents flow in loops, and for good EMC the loop areas must be minimised - so the correct technique is to figure out where the stray RF core currents originally came from, and return them back to their source by the path that encloses the smallest area. For the filters in ’typical’ electronic products there is usually no need to worry about the RF currents and voltages associated with ferrite cores. But control of core currents can be important for achieving EMC in the ferrites associated with switch-mode power conversion (flyback chokes, transformers, etc.) because switch-mode waveforms contain a great deal of energy in the RF spectrum. But this article is about filters so will not go any further into switch-mode EMC design issues.

Stray capacitance and its effect on high frequency impedance

Stray capacitances between the input and output of a choke limits its high-frequency impedance, by acting as a ’bypass’ in parallel with the choke. This effect can be seen in the impedance versus frequency curves of all of the ferrite chokes shown in Figure 3J - as frequency increases their impedance increases, but eventually a point is reached when their impedance starts to decrease as frequency continues to increase. In this region the impedance of a choke is dominated by the stray capacitances between its windings, and also between its input and output terminals and/or leads.

For this reason it is very important for high-frequency performance to keep the input and output terminals of series filtering elements (such as chokes or resistors) - and any leads, circuits or PCB traces attached to those terminals - as far apart from each other as practical. For good performance at 100MHz and above it can even be important to shield the input circuit from the output circuit - a topic discussed in 3.3.3 below.

Surface mounted ferrite beads can achieve high impedances at very high frequencies, because their parasitic capacitances are so small. But it is easy to ruin their performance above 100MHz by routing their input and output traces near to each other, increasing the stray input-output capacitance, so PCB layout is very important (see Part 2 of [8] for more on PCB design and layout techniques for EMC filters).

To obtain increased attenuation, it is tempting to wind the conductor(s) several times around the same magnetic core to increase its impedance. But this might not be as effective as required at higher frequencies, because of the increase in the stray capacitances created by the extra windings. The higher the number of windings on a core, the closer their proximity to each other, and the higher the stray capacitance between them. This effect is very strong with multilayer windings, especially if they are pile-wound - as this winding technique does not control where the windings lie with respect to each other, from one unit to another.

Chokes with multiple winding layers, including pile-winding, can achieve very high impedances at lower frequencies, and in some applications this is all that is required, for example when suppressing the RF noises emitted by a phase-angle controlled triac in a lighting dimmer, where significant levels of attenuation may only be needed up to a few MHz. Interestingly, at least one manufacturer of such lighting dimmers has found that chokes with layered or piled windings do not always provide the same attenuation performance when the chokes are reversed in the circuit. This is because of the complex nature of the stray capacitances in such a choke.

To improve the impedance at higher frequencies with highly-wound or multilayer chokes, it is better to use sectional winding techniques to reduce the overall input-output stray capacitance. One example of this type of choke is the ’super-toroid’ winding shown in Figure 3L. The aim of this technique is to split the winding in half - each half being wound on different portions of the toroid to reduce the stray capacitance coupling between them and increase the impedance of the choke at higher frequencies. Notice that each half of the winding is wound in the opposite direction to the other half - but because these halves are wound in different directions their fluxes do not cancel out, they add together to maximise impedance.

Slide 11

Figure 3L Winding a ’super-toroid’

A super-toroid wound with a single conductor is a DM choke, whereas one wound with a multi-filar twisted cable, as shown in Figure 3L, is a CM choke (see later).

Another way to improve the impedance at higher frequencies is to string a number of ferrite tubes or toroids in series along a cable. However, it is possible that their impedances and stray capacitances could interact to create resonances that will defeat the aim of this technique. So when using this technique it is best to simulate all such designs, using reasonably accurate estimations for the stray ’components’ (or extract them from a 3-dimensional field simulation of the physical structure), or simply to build them and test their transfer function with instruments.

One way to reliably achieve a choke that has a very low stray interwinding capacitance and hence the very highest series impedance that the ferrite material is capable of, is to string a number of ferrite tubes or toroids along a conductor, with all of the ferrites touching each other. Instead of multiple windings around each ferrite core, the conductor is only passed once through the centre of each core. Of course this makes a very large or long device overall.

3.2.8 Specifying and designing filters

When specifying a filter prior to design or selection, it is necessary to know the spectrum of the wanted signals, so that the filter’s response can be tailored to pass the wanted signals whilst impeding unwanted noises (interference). It is easy to specify 50 or 60Hz mains filters, because the RF noises to be filtered are at a much higher frequency than the wanted 50 or 60Hz sinewaves. But it is not so easy to specify a filter when the signal and noise spectra overlap, as they do for most digital signals and interconnections.

However, most emissions and immunity problems are caused by CM noises, whereas wanted signals are DM - so we can use CM filtering on noise that is within the signal’s spectrum without attenuating or distorting the signal. Of course, nothing is perfect, so CM filtering will have some effect on the wanted DM signals, and it is part of the design to make sure that the attenuation and/or distortion of the DM signals are within acceptable limits.

The key component for CM filtering is the CM ’choke’, which generally uses a soft ferrite core with the send and return conductors for the DM signal wound together on the choke’s ferrite core, usually wound bi-filar, tri-filar, etc. with the input and output leads/pins on opposite sides of the core, to reduce the stray input-output capacitance that limits the performance at the higher frequencies.

When a magnetic circuit is wrapped around both (or all) of the send and return conductors associated with a signal or power circuit, it will only attenuate CM currents. The magnetic fluxes created by the send and return current paths of the wanted DM signals cancel out, so they experience no effect from the magnetic circuit. In practice there is always some leakage inductance, caused by imbalances in the send/return windings, hence there is always some DM attenuation. This inevitable leakage can be turned into a benefit by providing both CM and DM filtering in one component. In ordinary CM chokes, the DM leakage is not controlled, and can vary considerably, but some EMC filter component manufacturers (e.g. Murata) offer ranges of CM chokes with specified DM impedances.

Some aspects of CM choke filtering are shown in Figure 3M.

Slide 12

Figure 3M Common-mode (CM) filtering with CM chokes

The cancellation of DM flux in CM chokes (sometimes called ’current-balanced’ chokes) allows large inductance values (e.g. milliHenries) to be achieved with small components, whereas DM chokes of the same physical size achieve much lower inductances (e.g. microHenries) and must become physically larger as DM currents increase.

Of course we want to design so as to have confidence in passing EMC tests the first time - to avoid delays caused by iterations and retesting - but for volume-manufactured products we also do not want to add too much cost by over-engineering. Designing CM filters creates some difficulties here, because CM voltages and currents are mostly unknown in both frequency and amplitude until the product is EMC tested.

Computer-aided EMC simulation packages are now available that allow CM effects to be predicted. To give sufficiently accurate results the simulator must model all of the stray (parasitic) parameters of the components (which can be done in any circuit simulator), but it must also solve the fields associated with the components, conductors and other physical structures, in three dimensions. These simulators require powerful computers with a great deal of memory, such as modern 64-bit PCs, and the software packages can cost as much as a top-of the range luxury car, but they can be very effective if sufficient effort is put into learning how to use them, modelling the components and digitising the actual physical design. Computer simulation offers the possibility of designing EMC right first time with the lowest component cost - or at least getting much closer to the optimum EMC design more quickly.

In the absence of an accurate computer simulation, it is best to design prototypes for a range of filter component or packaged filter options, and keep a stock of all the filter components or packages that could be needed, from the lowest-cost to the highest-specification (often available as free samples from their manufacturers). In the case of PCB-mounted filters, ’universal filter pad patterns’ can be developed that can accommodate from zero-ohm links through individual resistors or ferrite beads to CM chokes, plus two or three-terminal capacitors, to create a wide range of filter types including RC, LC, Tee, and π filters.

EMC pre-compliance testing begins with the lowest-cost filter options fitted, and if these turn out to be inadequate different components or higher-specification filters are fitted until the tests are passed. An alternative and arguably better approach starts off with the highest-specification filtering, to make sure no other problems exist - and when the tests are passed the filters are then reduced in cost to what works well enough. For products containing digital and switch-mode electronics, it is usually enough to use pre-compliance emissions tests - low emissions usually being a sign of adequate immunity, but only the full suite of emissions and immunity tests will prove whether adequate filters have been employed.

Such approaches help avoid the ’no room for the filter’ problem, which could require a complete redesign of the product. Most engineers soon gain experience with their own technologies and applications, permitting smaller-sized filter pad patterns - but the resulting design rules must be considered afresh whenever construction, ICs, or technology changes (e.g. due to a die-shrunk microprocessor, see Part 1 of [6]).

It is best if filter performance provides an ’engineering margin’ of at least 6dB beyond the emissions or immunity requirements of the test standards, to allow for device and assembly tolerances, measurement accuracy, etc. If the EMC tests are less accurate than those achieved by nationally accredited laboratories (e.g. by [9]) it is best to allow even greater margins for error, remembering that EMC test repeatability at a given test laboratory is usually no better than +/-4dB, and repeatability between test laboratories that have been accredited by the same accreditation body is often no better than +/-10dB.

Some types of interconnections, such as ribbon cables using single-ended signalling, suffer from high levels of DM emissions and immunity, and it might prove impossible to achieve adequate EMC by using filtering alone - because the necessary filtering attenuates the wanted signals by too much. In such cases it may be necessary to use shielded interconnections (see Part 1 of [6]) to attenuate frequencies that can’t be filtered sufficiently without damaging the wanted signals. Sometimes it is most cost-effective or necessary to use filtering and shielding at the same time.

Slide 13

Figure 3N Typical single-stage filters, using CM and DM filtering techniques

Figure 3N shows how CM and DM filtering techniques are combined in three examples of simple mains filters. These are called single-stage filters, because they only use one inductive element. Because mains-powered products must isolate the hazardous mains voltages from any touchable parts of its body for safety reasons, their CM emissions will generally have a high source impedance, so they are firstly attenuated with capacitors between phase and the RF Reference, and subsequently with a CM choke, to maximise the impedance discontinuities for the CM noises (see Figure 3E).

Because products draw power from their phase and neutral conductors, their DM emissions tend to have low impedance. So they are firstly attenuated by a DM choke (maybe the leakage inductance of a CM choke), and subsequently by a capacitor between the phases, to maximise the impedance discontinuities for the DM noises.

The low leakage (often called ’medical’) filter shown in Figure 3N has no capacitors between phase and earth, so that it may be used in medical applications where very low earth leakage currents are required to protect patients (maximum 50 or 60Hz leakages for some medical products can be as low as 10μA). This type of filter tends to rely on larger, higher-impedance CM chokes, and its lack of ’earthy’ capacitors can sometimes make it useful in applications other than medical.

Figure 3N also shows an example signal filter using a CM choke in an LC filter type. Military signal cable filters tend to rely on C-only and π types, probably because most traditional military equipment has a substantial and well-engineered RF Reference Plane between items of equipment (die-cast metal boxes electrically bonded by multiple mounting bolts directly onto metal surfaces in metal-bodied vehicles) and between items of equipment and their electrical power sources (generators and/or batteries).

Some civilian applications (such as digital telecommunications exchanges, computer rooms, and semiconductor manufacturing) achieve or approach such high-performance RF References in systems and installations, but most domestic, commercial, and industrial products are used in systems and installations that rely on networks of green/yellow insulated ’safety earth’ wires for their reference, which of course have very high impedances and resonances at RF frequencies. The most predictable signal filters in such applications tend to be R, L, RC, LC, or Tee types (using soft ferrites for any Ls). These types of filters impose lower levels of RF currents on the RF Reference than C-only or π filters, reducing the RF potential differences between different parts of the RF Reference and helping to reduce CM emissions as a result. As military vehicles use new materials such as carbon fibre, their RF Reference Planes suffer higher impedances, and they may find R, L, RC, LC, or Tee filters more cost-effective than C or π.

The use of a CM choke as shown in Figure 3N, instead of a series of individual ferrite beads, can allow substantial CM filtering to be achieved at frequencies as low as 150kHz, whilst allowing wanted (DM) signals of 15MHz or more to pass through unattenuated. There are inevitable tolerances between individual components of the same type, so (all else being equal) the CM attenuation of a filter that uses a single CM choke for all the conductors in a cable will give superior CM attenuation than one using a row of individual ferrites. However, CM chokes are often more costly than the equivalent number of single ferrites, and careful PCB layout and component choice can achieve a design that can be fitted with a CM choke if the single ferrites aren’t adequate, with just a few minutes with a soldering iron, instead of a PCB design iteration.

Slide 14

Figure 3P Some typical 2-stage mains filters using CM and DM filtering techniques

Mains filters with more than two stages are frequently used, and Figure 3P shows two typical designs. The multiple filter stages provide more impedance discontinuities for DM and CM noises, hence more attenuation. Standard filter products are available with up to four stages, and with ratings of up to 100 Amps, single or three-phase. These are the types of filters that are often required for use with high-power variable-speed AC motor drives, uninterruptible power supplies (UPSs), or other powerful switch-mode power converters.

3.2.9 Problems with real-life supply impedances

The CM and DM impedances of the public AC mains supply can vary from about 2Ω to 2,000Ω depending on the frequency and time of day. All filters that use inductors and capacitors are resonant circuits, with their resonant frequencies and hence their attenuations depending critically on their source and load impedances. But most filter datasheets are based upon CISPR17 measurements taken with 50Ω source and load impedances, which means that their specifications are always better than their real-life performance.

Single-stage filters are very sensitive to source and load impedances, and have a resonant peak that provides gain, rather than attenuation, when operated with certain source and load impedances. Single-stage filter gain usually pops up in the 150kHz to 10MHz region and can be as bad as 20dB, so it is possible that fitting a mains filter with a good (50Ω/50Ω) specification can actually increase emissions and/or worsen susceptibility. It is not uncommon to have a switch-mode power supply or inverter motor drive with excessive emissions between 200kHz and 1MHz, and fit a low-cost filter with a data sheet that shows sufficient attenuation to pass the tests - only to find that the filter increases the emissions!

To avoid this situation, only consider filters whose manufacturers specify both CM (sometimes called ’asymmetrical’) and DM (sometimes called ’symmetrical’) performance, for both matched 50Ω/50Ω and mismatched sources and loads. Mismatched figures are taken with 0.1Ω source and 100Ω load, and vice versa, using the CISPR17 test standard that is also used for 50Ω/50Ω tests. Drawing a line that represents the worst-cases of all the different datasheet curves results in a filter specification that can generally be relied upon - providing the filter is not overloaded or overheated (see earlier) and is installed correctly (see later). Figure 3Q shows an example of estimating a filter’s worst-case attenuation curve.

Slide 15

Figure 3Q Deriving reliable estimates of real-life mains filter attenuation from manufacturer’s data

Mains filters with two or more stages (some examples are shown in Figure 3P) have at least one internal circuit node, and these have impedances that do not depend as much on the source or load impedances. As a result, when installed correctly and not overloaded or overheated, they are more likely to provide real-life attenuation that approaches their 50Ω/50Ω datasheet specifications. Of course, they are larger and cost more than simpler single-stage filters, so if they might be required - the design should allow sufficient room.

Multi-stage filters still suffer from gain with some combinations of input and output impedances, but when designed correctly adding more stages reduces the maximum gain and the frequency at which it occurs. For example, some manufacturers’ data sheets show that their two-stage mains filters have gains of up to 10dB at frequencies between 10 and 100kHz.

Figure 3Q and the above discussion concerned mains filters, but exactly the same resonant gain issues arise for signal filters that use Ls and Cs, at frequencies below about 30MHz. Signal filters almost always use soft ferrites, which are resistive in the upper part of their frequency range and so do not cause resonances there - but they are inductive in the lower part so can cause resonances at those frequencies. Unfortunately, few if any signal filter manufacturers provide attenuation figures for source or load impedances other than 50Ω - so if RC, resistive Tee or p filters are not good enough and inductive components must be used, it might be best to experiment.

Experiments should take place at an early design stage, using sample filters and an RF signal generator and oscilloscope or spectrum analyser (if you don’t have a network analyser), to see what attenuation can realistically be expected when used with the actual source and load impedances in your application. The differences between data sheet figures and real-life attenuation can be dramatic, and you do not want to discover such interesting effects during the EMC testing of a supposedly finished design!

3.2.10 Problems with real-life switch-mode converter input impedances

Switch-mode power converters (e.g. switch-mode regulators, switch-mode amplifiers, DC/DC converters or DC/AC inverters) can have a negative dynamic input resistance at their power input terminals. This can interact with the impedance of an input filter, resulting in instability and even oscillation that can destroy the switch-mode converter and/or other equipment connected to the same power source. The problem is mainly caused by the series inductance in the filter, which can be compensated by using a larger value of capacitance connected across the input terminals, and also by damping the inductance (see 3.2.11 below).

[5], written in 1973, describes these issues, and defines the conditions required for oscillation to take place. It also describes methods for preventing the instability and/or oscillation. Many more references can be found on the Internet, by searching with appropriate terms.

3.2.11 Damping filter resonances that cause gain

Slide 16

Figure 3R Damping the filter resonances that cause gain

All LC filter circuits resonate, and at resonance they have a gain that depends upon the amount of loss in the circuit. Filter resonances can amplify surge overvoltages, as discussed later in 3.5.5, increasing their potential to cause damage.

Loss is caused by resistance, but most L and C filter components are designed to have low internal resistances to minimise their internal heating - so the resonant gain can be high. Figure 3R shows how resistors can be added to filter circuits to provide resonant ’damping’, to reduce the resonant gain. Such circuits cannot eliminate resonance entirely in any practical filter, but in some applications they can tame the resonances sufficiently well.

Although a mains filter is shown in Figure 3R, the resistive damping techniques described can also be used for differential and single-ended signal filters.

3.2.12 Filters and safety

Class 1 products have a protective (safety) earth conductor connected to their metal structures, which are also their RF References, and mains filter capacitors connected to the filter’s earth/ground cause leakage currents in the safety earthing/grounding system that can be dangerous. The maximum limits for these currents should be no larger than those specified by the appropriate safety standard(s) for the type of product concerned, e.g. IEC/EN: 60950, 61010-1, 60601-1, 60335-1, 60204-1, etc. Typically: double insulated products (no protective earth connection) must have <0.25mA, Class 1 protectively-earthed portable products must have <0.75mA and fixed ones must have <3.5mA. Class 1 protectively-earthed industrial fixed products might be permitted to have earth-leakage currents up to 5% of the rated phase current - when specified warning labels are fitted - but at the other extreme patient-connected medical products can be limited to <0.01mA.

In systems and installations the earth leakages from numbers of filters can build up to create large earth currents that can be very dangerous indeed. Tens of amps of leakage current is not unusual in the main protective earth terminal of a modern office building, due to its very many PCs and PC monitors, each with their mains filters leaking up to 3.5mA.

Mains filtering is an area where EMC requirements can often come into conflict with safety needs, and of course safety must always come first. So always take the relevant safety standards into account when designing or selecting mains filters, remembering that most X and Y capacitors have tolerances of +/-20%.

Mains filters sold for 50Hz use may generally be used on supplies from DC to 400Hz with the same performance (but check with their manufacturers). Also remember that the earth-leakage currents caused by filter capacitors connected to the earth/ground will increase as the supply frequency increases, so filters that meet the relevant safety standards at 50Hz might not comply at 60Hz, and may be decidedly dangerous on 400Hz.

Capacitors connected between the phases and RF Reference should always be approved to all relevant safety standards for both the application and the voltage. They will usually be Y1 (for double insulated products) or Y2 (for Class 1 products with an earthed protective bonding network). Capacitors between phases should also be safety approved, e.g. types X1 and X2, more to prevent fire hazards than to prevent shocks.

It is always best to use mains filters (or components) for which third-party safety approval certificates have been obtained and checked for their authenticity, filter model and variant, temperature range, voltage and current ratings, and the application of the correct safety standard. Forged safety approval certificates are not unknown, even from manufacturers who might be expected not to run such risks, so I always recommend that certificates are checked with their issuing Approvals Bodies, who in my experience are always happy to help, to make sure they are not forgeries.

3.3 Filter installation issues

Real-life filter performance is totally dependant on how they are installed, especially on the impedance of the RF Reference and the impedance of the method used to electrically bond the filter to its RF Reference. Not only should these impedances be much lower than that of the shunt capacitors in the filters, they should also allow the internal and external CM surface currents to find their optimum return paths. This section discusses these issues, and the practical installation guidance that results.

3.3.1 Input and output conductors

Stray RF coupling between the conductors associated with their unfiltered and filtered sides easily degrades filter attenuation. This problem is generally worse at higher frequencies, because the impedances of stray capacitances and stray mutual inductances reduce as frequencies increase, increasing the amount of stray coupling bypassing the filter. Many engineers have been very surprised by the ease with which high frequencies will bypass (’leak around’) a filter, given half a chance.

In an unshielded enclosure, filters should be positioned as near to the point of entry of the cable as possible. The maximum possible separation distances should be maintained between the filter’s external and internal cables, and between all of the conductors associated with the circuits on either side of the filter. Conductors in air should be spaced at least 100mm apart, more if they are routed in parallel for more than a few centimetres. Closer spacings might be acceptable for PCB traces and components - but only if they are much closer to the PCB’s RF Reference Plane than the spacing between them. Filter input and output conductors should never, ever, be bundled together, or share the same cable or cable route, unless they are each well shielded. See [5] for how to shield conductors effectively.

Where the enclosure is shielded, it is essential to mount the filter in

3.3.2 Skin effect and the flow of surface currents

Where the frequencies to be attenuated are not very high, it could be acceptable to use a few direct bonds, or a few millimetres of wire or braid to provide the electrical bonding to the RF Reference, providing the impedance of the bonding method is much less than that of the filter’s shunt capacitors at the highest frequency of concern. But to understand how to assemble/install filters correctly for good RF performance at high frequencies, we need to understand ’skin effect’.

All RF currents travel as surface currents, because all conductors have a skin effect that effectively causes them to shield their inner depths from RF currents. Figure 3S illustrates the general principle, and shows that as the frequency increases, the current is constrained to flow closer to the surface, increasing the current density at the surface of the conductor. One skin-depth is the depth into the conductor by which the current density has decreased to 1/e of what it was - about 0.368. By two skin-depths into a conductor the current density has reduced to (1/e)2, or 0.135, by three skin-depths it has reduced to 1/e)3, or 0.05, and so on.

Slide 17

Figure 3S Skin effect: examples of cross-sectional current density in a metal sheet

Figure 3S gives the formula for calculating one skin-depth δ, where μ0 is the permeability of free space (4π.10-7 Henries per metre); μR is the (dimensionless) relative permeability of the conductor material (most common conductors, such as copper, aluminium and tin, have a μR of 1.0) and σ is the conductivity of the conductor material in mho/metre. Copper has a nominal volume resistivity ρv of 1.72.10-8 Ω-m, giving it a nominal conductivity of 58.106, so one skin-depth in nominal copper is given by δ = 66/√f (δ is given in millimetres when f is in Hz). For example, at 160MHz: one skin-depth is 0.005mm, so 0.05mm below the surface of a copper conductor, the RF current density is 0.0025 of the density at the surface, an attenuation of 52dB.

Figure 3T shows graphs of skin-depth versus frequency for some common materials, to save having to find out the values of their conductivity and calculate δ. Mild steel is shown as an example of a ferromagnetic material (nickel is another), and to show that their high values of μR result in smaller skin-depths, but also that their permeability is frequency-sensitive and disappears above some critical frequency.

Slide 18

Figure 3T Graph of skin-depth (d) for copper, aluminium, and mild steel

[10] contains information on the material properties of a wide range of conductors, for calculating skin depth, and also a great deal of other useful information for designers. [11] is a useful source for information on skin-depth.

Above a few tens of MHz most conductors and metal items (such as the cases of filters) are several skin-depths thick, so RF currents travel as surface currents in them. Taking this phenomenon into account in the design of a filter’s assembly/installation is essential for the achievement of good emissions and/or immunity performance.

Figure 3U shows how providing a continuous metal bond between a filter and the shielding enclosure of a product ensures that the external CM noise currents do not enter the enclosure and cause interference and immunity problems, and the internal CM noise currents remain inside the enclosure and do not escape to cause emissions problems. Figure 3U shows a simple capacitor filter, but the principle applies to all types of filters.

As a result, the optimum way to bond a filter to its RF Reference Plane, for the best performance at the highest frequencies, is what is often called ’360ø direct metal-metal contact’ - meaning that the filter’s metalwork and the RF Reference Plane are in direct contact with each other all around the periphery of the filter (hence the term 360ø).

Commercial and industrial conducted emissions standards generally only measure up to 30MHz, and at such low frequencies it is often sufficient to bond a filter to an enclosure with a single direct metal-to-metal connection between the filter’s case and the enclosure. Where the filter is only required for low frequencies, e.g. below 1MHz, it may even be possible to use a very short length of wire or braid to connect its metal case to the enclosure metalwork, plus of course the enclosure will not need to be a proper shield either (enclosure shielding will be covered by Part 4 of this series). But there is a synergistic relationship between filtering and shielding, discussed in more detail in the following section.

Slide 19

Figure 3U Filter RF bonding and the skin effect

Filters that employ capacitors connected between power or signal conductors and the RF Reference depend upon the RF Reference - and their connection to it - having a much lower impedance than the filter capacitors, at all of the frequencies to be attenuated. The connection between the capacitors and RF Reference should be very short and direct, less than one-hundredth of a wavelength long at the highest frequency to be attenuated, and should also have a very low inductance. This usually means that wires or even braid straps cannot be used to electrically bond filters to the RF Reference Plane, except for low frequencies (say, below 1MHz).

Figure 3V, which is taken from [12], shows the sorts of bad effects that even a short length of interconnecting wire can have on a standard single-stage mains filter even when measured with 50Ω/50Ω source and load impedances - its best possible case. If the 10mm wire were replaced with at least one direct metal-to-metal bond, performance at 30MHz and above would improve dramatically.

It is acceptable to fit green/yellow wires of any length to mains filters, for safety reasons, as long as there is also at least one direct metal-to-metal electrical bond between the filter’s metal case and the product’s RF Reference. When a mains filter’s metal-to-metal bonds have been designed to maintain a very low impedance over the lifecycle of the product, there is no need for a green/yellow ’safety earth’ wire as well - but safety inspectors are generally much more reassured when they can see a green/yellow bonding wire with anti-vibration anti-corrosion connections at both ends. (But, as discussed above, it would be a mistake to assume that the green/yellow safety wire was adequate for achieving the filter’s EMC performance.)

Slide 20

Figure 3V Comparison of two lengths of filter earth-bonding wire 3.3.3 The synergy of filtering and shielding


Some mains filter manufacturers only design and specify their filters to provide attenuation over the frequency range of the conducted emissions tests (typically up to 30MHz for commercial and industrial products), to keep costs low. Unfortunately, if such filters have poor attenuation above 30MHz, they will degrade the shielding effectiveness (SE) of a shielded enclosure above that frequency by permitting RF signals to leak out via the filtered cables - resulting in problems for both emissions and immunity.

It does not matter what is the ostensible purpose of a conductor, e.g. mains or DC power, audio, whatever - if its filtering and/or shielding provides less attenuation than is required for the shielded enclosure, it will degrade the SE of the enclosure. The filtering and/or shielding of cables used for audio, mice or keyboards are often ignored when they exit a shielded enclosure. The assumption is usually that the signals they carry will not cause a problem for EMC. But this overlooks the fact that all conductors or whatever type or signal designation always behave as ’accidental antennas’ (see [5]), very readily picking-up EM noises on either side of a shielded barrier and retransmitting them on the other side - unless specifically prevented from doing so by the application of shielding and/or filtering.

If good high frequency shielding is required, all unshielded cables that enter the enclosure (including mains) must be filtered with good attenuation at the highest frequency of concern for shielding purposes. So where shielding is required up to 1GHz (for example), only employ filters with data showing good attenuation up to at least 1GHz. Few mains filters intended for commercial and industrial equipment specify attenuation above 100MHz, so additional high-frequency filtering might be needed. However, some filter manufacturers (e.g. EMC Solutions) specify their filters up to 1GHz.

3.3.4 Assembly and installation techniques for filters that penetrate shields

As discussed above, the performance of shielded enclosures can easily be degraded by RF noise that ’leaks’ out along the cables that enter and exit the enclosure. The shielding/filtering synergy issues discussed above are vital considerations when high levels of shielding or filtering are required (e.g. >40dB) at frequencies >100MHz.

The design of shielded cables was covered in [5], and the design of shielded enclosures will be covered in Part 4 of this series. This section discusses how filters should be installed in shielded enclosures so that they do not permit RF noises to pass through them that could compromise the SE of the enclosure.

Slide 21

Figure 3W Three-electrode (’three-terminal’) and feedthrough capacitors

Figure 3W shows an example of a ’feedthrough’ capacitor specifically designed for use where unshielded conductors penetrate a shielding enclosure, which could be a product’s enclosure or an internal shielded volume. Another, higher current style of feedthrough capacitor was shown in Figure 3A. Feedthrough capacitors have three terminals, for input, output and ’ground’. The signal to be filtered enters at one side of its electrodes and exits at the other, having to pass the ground electrode as it does so. The middle ’ground’ terminal connects directly to the shield, using a 360ø electrical bond so that the internal and external surface currents stay separated on either side of the shield, as shown in see Figure 3U, allowing the shield to function correctly. If designed correctly, the shielded enclosure prevents stray coupling between the capacitor’s input and output terminals, and also provides the filter with an RF Reference Plane with negligible impedance at the highest frequency of concern, all of which helps the filter employing the feedthrough capacitor to achieve the best performance it is capable of.

When used as (or in) filters, traditional feedthrough capacitors such as the ones shown in Figures 3W and 3A provide much better attenuation, at much higher frequencies, than is possible by using ordinary two-terminal capacitors. Traditional feedthrough capacitors are soldered or screwed into a shield wall and connected to the circuits on either side by wire conductors. They are often used between shielded compartments within RF equipment, e.g. to filter the DC power that passes between the RF, IF and digital sections of an RF receiver or spectrum analyser.

Traditional feedthrough filters, such as those shown in Figure 3W, are also available as ’filter pins’ in some standard connectors, such as some D-types and military circular connectors. (Note that not all connectors with built-in filters use feedthrough filter pins, some use discrete components on miniature internal PCBs, which will not achieve as good an attenuation at the highest frequencies.)

Traditional feedthrough filters are not favoured for modern volume-manufactured products because of their high component cost, and the high cost of their manual assembly and the assembly of the wires they connect to. Volume-manufactured products prefer to use SMD components automatically assembled on PCBs - but since a true feedthrough capacitor cannot be automatically assembled, three-terminal capacitors have been developed to fulfil this purpose.

Figure 3W includes an example of a three-terminal capacitor intended for SMD assembly processes, and Figure 3X shows an example of how it is used in conjunction with PCB shielding. The capacitor is aligned with the shield wall so that its input and output terminals are shielded from each other by the PCB-mounted shielding-can, and the capacitor’s centre ’ground’ terminal is soldered directly to a guard trace that follows the wall of the shield-can and connects it to a PCB plane (almost always 0V) with a wall of via holes.

Slide 22

Figure 3X Overview of PCB shielding and filtering

The gaps that are cut out of the shield-can’s wall for the bodies of the filters are known as ’mouseholes’ (for reasons that should be obvious to anyone who enjoys ’Tom and Jerry’ cartoons). Three-terminal capacitors and the filters that use them cannot be as good as proper 360ø shield-bonded feedthrough types, because there will always be some stray coupling through the mouseholes in the shield. But careful control of the maximum dimensions of the mouseholes, and of the spacing between the via-holes connecting the shield wall to the PCB plane that provides the shield’s sixth side, can nevertheless achieve excellent performance. For more details on this, see Part 2 of [8].

Figure 3Y shows attenuation of a three-terminal SMD π filter assembled on a PCB, and the effect of adding a PCB-mounted shielding-can in the manner shown in Figure 3X. Without the PCB’s shield-can fitted, the filter performance is quite respectable at about 50dB at 100MHz, but above that frequency it falls off at 20dB per decade, so that it is only about 30dB at 1GHz, and it would presumably be about 10dB at 10GHz.

But the addition of the shielding-can reduces the stray coupling bypassing the filter very considerably, and also allows correct separation of internal and external surface currents and provides an RF Reference Plane that has a much lower impedance over the frequencies measured. The result is an attenuation of around 70dB at 100MHz, and a more-or-less flat attenuation that maintains about 65dB up to 1GHz - easily 35dB more attenuation than was achieved without the shield. It is not clear what the performance with the shield-can would be above 1GHz, but as there is no sign of any roll-off even at 1GHz it is likely that an attenuation of at least 45dB would be achieved at 10GHz.

Slide 23

Figure 3Y The synergy of filtering and shielding

As described in section 2.6 of [5], shielded cables exiting a shielded PCB region require shielded connectors or glands that are electrically bonded to the shield-can’s wall by mechanical fixings, soldering or gasketting that makes multiple connections around its periphery - preferably full 360ø bonding.

The experiment whose results are shown in Figure 3Y reveals two important things:

a) Filters that must provide significant levels of attenuation at frequencies above 100MHz, must employ shielding techniques as well. They will not be able to achieve the required performance otherwise.
b) Modern digital ICs produce large amounts of CM and DM noise at frequencies above 1GHz, and products supplied to the USA already have to comply with FCC emissions limits above this frequency. The EN standards used to achieve a presumption of conformity with the EMC Directive for products supplied to Europe will soon be changing to include emissions and immunity requirements above 1GHz - at least to 2.7GHz and maybe higher. To comply with these requirements using low-cost SMD PCB assemblies will require the use of shielding wherever GHz frequencies need to be filtered.

There are now many suppliers of PCB-mounted shielding-cans that can be used with three-terminal filters, and they have many types that can be automatically assembled like any other SMD component. Part 2 of [8] has more details on these shield-cans, and also describes a number of different PCB layouts appropriate for filtering off-board connectors. An example of one of these layouts is shown in Figure 3Z.

Slide 24

Figure 3Z Example layout for an unshielded off-board connector, using Tee filtering

Where filters must penetrate the shield of a product’s overall enclosure, and PCB-mounted components are not suitable, more traditional feedthrough or ’bulkhead-mounted’ filters in metal cases are the best. A point to watch out for is whether the metal cases of such filters are seamless - good filters are enclosed in what are actually well-shielded enclosures themselves. Filters that have metal cases with apertures, seams or gaps in them give poor attenuation at high frequencies regardless of what their data sheet says, because they compromise the attenuation of the shielded enclosure they are assembled/installed onto.

’Chassis mounted’ filters include types with screw terminals, spade or blade terminals, or flying leads (Figure 3A shows some examples of chassis mounted filters with spade terminals) and cost less than proper bulkhead or feedthrough types, but cannot be assembled to shields so as to reduce stray coupling between their inputs and outputs. The result is that they are not as effective as feedthrough or bulkhead mounting types at higher frequencies, especially above about 10MHz. Their performance can be maximised by mounting them with multiple direct metal-to-metal bonds to an RF Reference Plane that is a shielded enclosure wall, or at least a very large metal plate, plus routing their input and output cables very close to the RF Reference Plane and keeping them and any circuits or components they connect to very far apart.

However, their performance can be significantly improved by the use of what is known as the ’dirty box’ shielding technique illustrated in Figure 3AA. This figure shows a shielded enclosure, and an example of the correct installation of a traditional high-performance feedthrough filter. It also shows an example of an IEC 320 appliance mains inlet connector with an internal filter. The important issue with such inlet filters is that they should have seamless metal bodies that make a direct metal-to-metal connection to the wall of the shielded enclosure.

Many manufacturers have fitted mains connectors with built-in filters, relying on their mounting screws and green/yellow safety earth wire to make the necessary electrical bonds, and have found the EMC performance to be almost useless. As discussed above, the length of the green/yellow safety wire is simply too long, and a problem with most built-in filter connectors is that their mounting screws bear onto plastic mouldings, so they don’t provide any metal-to-metal connections. The correct way to install such filters is to ensure that an area of the enclosure’s shield wall is free from paint or anodising, and has a highly conductive surface that will be pressed firmly against the filter’s metal body when it is assembled. Sometimes it may even be necessary to bond the bodies of such filters 360ø to the shield wall all around the perimeter of the filter’s metal case, requiring high surface conductivity for the metalwork on both sides of the gasket, and protection from corrosion (see below).

Slide 25

Figure 3AA Mounting filters in the walls of shielded enclosures (examples shown are all power filters)

When chassis-mounted filters are applied to cables entering or exiting a shielded enclosure, the portion of the cable that enters the enclosure to connect to the filter degrades the attenuation of the filter by causing stray coupling to its other terminals. This portion of cable also degrades the SE of the enclosure by acting as an accidental antenna (see [5]), especially at higher frequencies.

To maximise the high-frequency performance of such filters and prevent degradation of the enclosure shielding, such filters should be installed using the ’dirty-box’ method illustrated in Figure 3AA. The Dirty Box is a five-sided shielded cover that fits over the filter and the external cable entry, within the overall shielded enclosure. It must have metal-to-metal bonds at multiple points between its walls and the wall of the shielded enclosure, spaced apart by much less than λ/10 at the highest frequency to be controlled, and covering the entire perimeter of the Dirty Box’s walls. Conductive gaskets might help reduce assembly time by reducing the number of fixing screws, or might even be necessary to achieve sufficiently good bonding to the enclosure wall.

The filter is mounted inside the Dirty Box, with its input and output conductors kept as short and as far apart from each other as possible, to reduce their stray coupling - but even so the higher frequencies will still couple between them. If the resulting high-frequency stray coupling is problematic and cannot be reduced by careful cable routing within the Dirty Box, soft-ferrite CM chokes and/or high-frequency feedthrough filters may be needed on either (or both) the input and output cables, fitted at the point where they enter or exit the Dirty Box.

'Shielded room' filters are also available, and although intended for EMC test chambers (as shown in Figure 3B) they can be used for shielded equipment cabinets as well. These are essentially screw, spade or blade terminal filters with two Dirty Boxes, one over the input terminals and their conductors, and one over the output terminals and their conductors, to minimise the stray coupling between input and output.

Conduit fittings are usually provided for the filtered side of room filters, to provide shielding for their conductors whilst they enter the shielded room or enclosure. Where the conduit enters the shielded room or enclosure it must electrically bond 360§ at the shield wall, as illustrated in Figure 3AB. Shielded cables may be used instead of conduits, as long as they bond 360° at both ends, to the filter's case and the shielded room or enclosure wall using appropriate glands or connectors.

Slide 26

Figure 3AB Mounting 'room filters' to the walls of shielded rooms

Figure 3AC shows an overview of shielding and filtering at the level of the final system or installation.  Where an electrical/electronic product has an overall shielded enclosure, all of the conductors that enter or exit that enclosure must be shielded, and/or filtered, at the point where they enter/exit the enclosure. There are no exceptions to this rule, whatever the purpose of the conductors, including safety earth wires: metal armour or draw-wires for cables, fibre-optics, or hydraulic hoses; metal pipes for gasses or liquids; metal ductwork for cables, air-conditioning, etc. Conductors permitted to be connected directly to the shield wall should be so connected, using 360° bonding techniques just as if they were cable shields (see [5]).

Unshielded conductors that are not directly bonded to the enclosure at point of entry/exit must be filtered, taking into account all of the techniques discussed above concerning the synergy of filtering and shielding.

Slide 27

Figure 3AC Shielding and filtering at installation level

3.3.5 Designing to prevent corrosion
All metal-to-metal bonds associated with filters (and shielding), and all conductive gaskets, must be designed to provide low impedance for the anticipated lifecycle of the product, despite the mechanical, climatic, biological, chemical and other physical environments the product is exposed to. This generally means choosing metals, platings and gasket materials that resist oxidation, and it also means ensuring that the materials in contact are sufficiently close in the galvanic series so that they don't suffer unduly from galvanic corrosion. IEC 60950 is a safety standard but provides some useful guidance on these issues, and there is also a lot of information available freely on the Internet.

Effective 'vapour-phase corrosion inhibition techniques' are claimed to have been developed in recent years, by Cortec Corporation (, and should be investigated, especially where corrosion is a significant problem.

3.3.6 Filters connected in series
It sometimes happens that a product is supplied with mains filtering, but its RF emissions are too high (or its immunity too low) for the equipment, system or installation it is used in. This is often a problem where a large number of identical or similar devices are used in one product or system, for example a number of low-power inverter motor drives in one industrial cabinet. Each product may meet the relevant emissions limits individually, but when a number are all operating at once the aggregate of their emissions might exceed the permitted limits.

In such situations it is tempting to simply add another mains filter, which would then appear in series with the mains filters already fitted in the products. Often a single-stage filter is chosen because the filtering requirements are only modest. The gain problems that can occur with filters with 'mismatched' source/load impedances, especially single-stage types, were discussed earlier - but sometimes connecting filters in series can result in resonances that are not present in any of the filters when they are tested individually. So adding the extra filter can sometimes create worse emissions or immunity than before.

Solutions include replacing the original filters in the products with ones that achieve higher performance, or experimenting with different types of additional filters to find ones that work well when connected in series with the filters in the products. If the circuits of the filters involved (product and additional) are known, circuit simulators such as Spice should be able to predict resonance problems in advance, and guide the choice of appropriate devices.

3.4 Types of overvoltage transients and surges
There are many types of transient overvoltage phenomena that can cause damage or interference due to overvoltages and/or overcurrents, for example:

Electro-static discharge (ESD).    'Personnel ESD' has risetimes under 1ns, peak voltages typically in the region +/-8 to +/-115kV when humidity is 20% or more (but maybe up to +/-35kV in some circumstances). The initial ESD charge-transfer impulse is often a unidirectional impulse waveform lasting a few tens of nanoseconds, but resonances mean that conductors generally experience a damped oscillatory waveform that can last for milliseconds. The energy content of an ESD event is low, although sufficient to damage sensitive semiconductors. There are other kinds of ESD phenomenon, including machine ESD, and furniture ESD. As well as conducted transients, ESD events generate very intense pulsed electric and magnetic fields, up to kV/m and kA/m in close proximity.  Design techniques specifically for suppressing ESD will be discussed in Part 6 of this series.

Fast Transient Bursts.   Caused by arcs and sparks at electromechanical contacts or degraded insulation, their conducted noise is characterised by a broadband spectrum with an upper frequency and risetime that depends on the length of conductor between the spark and the observation point. Within a metre or two of a spark, the upper frequency of the conducted noise can exceed 300MHz and the risetime be less than 1ns.
Typical peak voltages on mains cables can be up to +/-2kV, maybe +/-4kV in industries where high-powers are switched, but in exceptional circumstances they can be much higher. The energy content of a burst is quite low, although sufficient to damage sensitive semiconductors. This phenomenon is suppressed by the application of the techniques described in all six parts of this series, taking the characteristics of the fast transient bursts into account.

Nuclear and High-Altitude Electromagnetic Pulse (NEMP and HEMP).   Originally only a concern for military and telecommunication infrastructure organisations, the possibility of 'electronic warfare' or 'electronic terrorism' in the commercial and industrial sectors is making it a more mainstream issue. A number of standards in the IEC 61000-5 series are being written to cover these phenomena, how to test for them, and how to design to protect from them. Essentially, suppression requires the application of the various techniques described in all six parts of this series, taking the characteristics of the EM threats into account.

Slide 28

Figure 3AD Some types of overvoltage transients and surges

Surges.   Surges are generally fairly slow but powerful EM phenomena, with risetimes typically in the range 50ns -10&#956;s, caused by lightning and also by the sudden release of energy stored in reactors (e.g. the energy in the rotor of a motor, released as a surge when the motor is switched off). They are generally measured in +/-kV, and can last from milliseconds to as long as seconds in some unusual cases. There are many surge voltage and current waveshapes associated with different situations, including: 'unidirectional', 'damped oscillatory' and 'ring wave'.

The surge test standard used most commonly in commercial/industrial applications for surge testing is EN/IEC 61000-4-5, which uses unidirectional surge waveshapes and assumes (presumably on the basis of real information) that line-to-line surges in AC power cables have a source impedance of 2&#937;, and line-to-ground surges have 12&#937; . It also assumes that surge currents flowing in power and earth/ground conductors induce CM surges into any long conductors (whether power, signal, data or control) with a source impedance of 42&#937;, and earth-lift surges due to currents flowing in the common earth/ground structure have an impedance of 2&#937;. Induction and earth-lift surges are usually only considered significant where cables are longer than 10 metres, but in some circumstances much shorter cables could be vulnerable.

The low surge impedances mean that, for example, a +/-2kV mains voltage surge line-to-line has an associated surge current of around +/-1kA. The voltage and current risetimes differ due to the inductance of the circuits, but the resulting peak powers are typically measured in MW, and the total energy associated with a surge is typically measured in tens of Joules (enough to vaporise the iron wire in metal-clad resistors, and blow the terminals off their ends). As a result, the main problems caused by surges is actual damage to devices, and even to conductors (e.g. PCB traces) by overvoltage, overcurrent or over-dissipation - but it should not be forgotten that they also cause errors in signals and data.

Power frequency overvoltages.  There are various types of faults in power distribution systems, and they can cause the 'earth' or 'ground' voltage of an item of equipment to suffer various levels of voltage at the powerline frequency (and its harmonics, in the case of distorted AC power waveforms). In the (usual) case of earthed/grounded neutrals, the neutral experiences the same overvoltage. The overvoltages are typically up to 50% of the power systems phase-neutral voltage, but in some cases can approach 100%. Their durations are measured in fractions of a second, or even seconds, so although their voltages are not generally as high as the other overvoltages listed above, the total energy delivered can be very large indeed. Faults in high-voltage power systems, can be a very significant indeed, so it is lucky that they are quite rare.

These overvoltages can also be caused when one of the other types of overvoltage listed above cause power current to flow to earth/ground, e.g. via a spark or the operation of a line-to-ground SPD. This is often called a follow-on SPD current. Overvoltages can also be caused if a power cable is accidentally shorted to a conductor associated with the circuit to be protected, sometimes called a 'power-cross', which can last for minutes, hours, or even be continuous.

Other types of overvoltage.  Figure 3AE shows the "ITI (CBEMA) Curve". This curve and its Application Note [13] describe an AC input voltage envelope for 120V 60Hz power systems, which typically can be tolerated (no interruption in function) by most Information Technology Equipment (ITE). The ITI Curve and its Application Note are <strong>not</strong> intended to serve as a design specification for products or AC distribution systems but this if often how it is employed.

Slide 29

Figure 3AE The ITI (CBEMA) Curve

3.6 References
[1] Keith Armstrong, "Design Techniques for EMC", UK EMC Journal, a 6-part series published bimonthly over the period February - December 1999. An improved version of this original series is available via the "Publications & Downloads" page at
[2] The Institution of Electrical Engineers (IEE), Professional Network on Functional Safety, "EMC and Functional Safety Resource List", via the "Publications & Downloads" page at
[3] Arthur B Williams, "Electronic Filter Design Handbook", McGraw Hill, 1981, ISBN 0-07-070430-9
[4] John R Barnes, "Robust Electronic Design Reference Book, Volume I", Kluwer Academic Publishers, 2004, ISBN: 1-4020-7737-8
[5] Sokal, N. O., System Oscillations From Negative Input Resistance at Power Input Port of Switching-Mode Regulator, Amplifier, DC/DC Converter, or DC/AC Inverter, IEEE Power Electronics Specialists Conference (PESC) 1973 Record, pp. 138-140.
[6] Keith Armstrong, "Design Techniques for EMC, Part 2 - Cables and Connectors", The EMC Journal, May and July 2006, available from
[7] Keith Armstrong, "Design Techniques for EMC, Part 0 - Introduction and Part 1 - Circuit Design and Choice of Components", The EMC Journal, January 2006 pp 29-41, plus March 2006 pp 30-37, available from
[8] F Beck and J Sroka, "EMC Performance of Drive Application Under Real Load Condition", presented at the Industrial Forums in EMC Zurich 2001, and also a Schaffner EMV AG application note dated 11th March 1999. It was also presented by W L Klampfer at the 8th International Conference on Electromagnetic Interference and Compatibility, INCEMIC 2003, ISBN: 81-900652-1-1, publication date: 18-19 Dec. 2003.
[9] Keith Armstrong, "Advanced PCB Design and Layout Techniques for EMC", an 8-part series published in the EMC & Compliance Journal, March 2004 - November 2005. An improved version of this series is available via the "Publications & Downloads" page at
[10]The United Kingdom Accreditation Service,
[11]John R Barnes, Robust Electronic Design Reference Book, Volume II, Appendices, Kluwer Academic Publishers, 2004, ISBN 1-4020-7738-6
[12]RF Caf‚, Skin Depth,
[13]Tim Williams and Keith Armstrong, "EMC for Systems and Installations", Newnes 2000, ISBN 0 7506 4167 3, especially chapter 8,, RS Components Part No. 377-6463
[14]ITI (CBEMA) Curve and Application Note:
[15]"PCB Layout: The Impact of Lighting and Power-Cross Transients", Milton Hilliard, Compliance Engineering, January/February 2003 pp 25-30, available from the archives at
[16]MIL-STD-275 "Printed Wiring for Electronic Equipment, Revision: E, Dated: 31 December 1984", available via:
[17]Akihiko Yagasaki, Characteristics of a Special-Isolation Transformer Capable of Protecting from High-Voltage Surges and Its Performance", IEEE Trans. EMC, Vol. 43, No. 3, August 2001, pp 340-347
[18]Daniel Dunlap, Protection of SPD Products, ITEM 2000, pp 148-157, visit and search by 'Daniel Dunlap'.
[19]Keith Armstrong, The Benefits of Applying IEC 61000-5-2 to Cable Screen Bonding and Earthing, IEE Seminar entitled: "EMC - its nearly all about the cabling", The IEE, Savoy Place, London, January 22nd 2003.

3.7 Acknowledgements
I am very grateful to the following people for suggesting a number of corrections, modifications and additions to the first series published in 1999 [1]: Feng Chen, Kevin Ellis, Neil Helsby, Alan Keenan, Mike Langrish, Tom Liszka, Tom Sato, and John Woodgate. I am also indebted to more recent input from Richard Marshall.

Eur Ing Keith Armstrong CEng MIEE MIEEE
Partner, Cherry Clough Consultants,, Member EMCIA 
Phone:  +44 (0)1785 660 247, Fax:  +44 (0)1785 660 247,