Design Techniques for EMC Part 3 - Filtering and Suppressing Transients

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 has been split in two, second half will appear in Issue 67).

 

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 part of the article

3.Part 3 - Filters and transient suppressors
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

3. Part 3 - Filtering, and suppressing transients

Slide 1

Figure 3A There are many types of filters (such as these examples from Schaffner)

3.1 Introduction

Filters are used to attenuate unwanted frequencies travelling along conductors, and are characterised by attenuation versus frequency curves. Transient suppressors, such as surge protection devices (SPDs), attenuate unwanted voltage surges travelling along conductors, and are characterised by graphs of voltage ’let-through’ versus time.

Incorrect use of filters or SPDs can make a product’s emissions or immunity worse than if they were not used at all. More expensive filters or SPDs are not necessarily the best. You cannot in general choose a filter or SPD from a distributor’s catalogue, by simply checking its ratings, performance and intended application, and expect it to provide the benefits you need for your product.

Many books have been written on filter design, such as Arthur B Williams’ [3]. No doubt there is a more modern edition available, but filter design has not changed much over the years. There are also now a number of circuit simulators that run on PCs and can be used to simulate filters. This article will not go into poles and zeroes and that sort of detail - instead it will describe the things which need to be taken into account so that filters designed using textbooks, circuit simulators such as Spice, or chosen from catalogues, stand a chance of performing as required, and avoid unpleasant and/or costly experiences.

Filter design or selection is not a ’black art’, but nevertheless it is difficult to predict exactly what performance a given filter will achieve when installed in a product, especially at frequencies above 100MHz, so it is often necessary to experiment with different options to find the most cost-effective. Planning and designing for such flexibility from the start of a project is very worthwhile, and an example of what John R Barnes [4] calls "wiggle room" and I call "anti-Murphy design". My approach is based upon the well-known Murphy’s Law - I find that designers who try to anticipate the surprises that Murphy might have in store for them reach their design targets and timescales more reliably, and the resulting products have a lower overall cost of manufacture because they have not had to have filters, or larger filters than were hoped for, squeezed in somehow at the end of a project when compliance tests were failed.

Slide 2

Figure 3B Examples of high-performance ’room filters’ (at BSI’s EMC test lab, Hemel Hempstead)

In general, this article considers filters that are fitted at the boundary between a product and its external electromagnetic (EM) environment. Filters used inside an item, for example between a switch-mode power supply and a sensitive analogue circuit, will share most if not all of the same considerations - because filters always separate two areas or zones that should not be allowed to crosstalk or otherwise freely intermingle their signals.

3.2 Designing and selecting filters

3.2.1 How filters work

Ignoring all the poles and zeroes in the filter textbooks: filters work by creating an intentional discontinuity in the characteristic impedance of a current path, reflecting radio frequency (RF) energy away from a protected circuit, or absorbing the RF energy (converting it to heat) - rather like a shield does, as will be described in Part 4 of this series.

The greater the discontinuity, the greater the attenuation. So if the source impedance of an unwanted signal (noise) is 100Ω and we put a 1k%#937; impedance in series with it, only about 10% of the signal gets through the high impedance - an attenuation of around 20dB. A similar effect can be created by instead connecting the 100Ω noise to the ’RF Reference’ via an impedance that is much lower than 100Ω: for example, 5Ω would provide an attenuation of around 26dB.

Filters use electronic components such as resistors (R), inductors (L), and capacitors (C) to create the desired impedance discontinuities over the ranges of frequencies of concern. R, L, or C components can be used as filters on their own, but combining them gives better attenuation. LC types can give better attenuation than RC types, and are often used in power circuits because of their lower losses, but all LC filters are resonators that can produce gain at some frequencies, so they need to be carefully designed, taking their actual source and load impedances into account, to ensure attenuation over the desired range of frequencies. RC types generally provide more reliable filter performance.

Slide 3

Figure 3C Different types of simple single-line filters

A number of basic schematics exist for low-pass filters based on R, L and C, and are shown in Figure 3A. There are high-pass equivalents, and band-pass or notch filters can also be achieved with passive components like these - but low-pass filters are mostly used for EMC, so this is the type shown in Figure 3A and discussed in this article.

Simple inductive filters (chokes, ferrites, etc.) have no RF Reference connection, so are especially useful where no RF Reference exists, or if it exists but does not have a structure that provides a low enough impedance at the highest frequencies of concern. Unfortunately, such very simple filters are generally unable to achieve very high attenuations - typically between 3 and 20dB, depending on the frequency.

Capacitors can also be used on their own as very simple filters (by creating a ’high-to-low’ impedance discontinuity), or as part of a more complex filter circuit that includes inductors and/or resistors. But the effectiveness of a capacitor in a filter depends upon the impedance of the RF Reference it is using as its ’ground’, and also upon the impedance of the interconnection between the capacitor and the RF Reference (e.g. wire leads, PCB traces). As a result, manufacturer’s data sheet figures for capacitive filters are rarely achieved in real-life because they were tested with RF References that were solid copper sheets covering entire bench-tops, and so had a lower impedance than is usually possible in real life.

Many a well-designed and expensive filter has had its performance wasted by being connected to a poorly performing RF Reference, or by being bonded to an excellent Reference by a short length of wire instead of the direct metal-to-metal contact that was needed.

An example of a common use of RCR filters is to connect computer boards to displays via flexible circuits, to reduce the emissions from the ’flexi’. The resistor values in these filters are often chosen as much for transmission-line matching (see section 2.7 of [5]), as they are for filtering.

Filters must pass the wanted signals/power, while attenuating unwanted ’noise’. So filter specification must begin with knowledge of the full spectrum of the wanted signal or power. It is common these days for the spectrum of a wanted signal to contain very high frequencies that are not needed, but cause EMC problems, due to the very fast switching edges of modern digital and switch-mode devices. Analogue signals are also polluted with such noise, due to stray coupling from digital and switch-mode circuits nearby. These very high frequencies can be removed by filtering and/or shielding, and it is good EMC practice to remove them at their sources, rather than wait until they have polluted many more conductors, and this was discussed in section 1.1.2 and Figure 1B of [6].

Active filters can be designed, based upon operational amplifiers (opamps), using feedback techniques to achieve remarkable attenuations. But above some frequency, the phase-shifts inherent in all opamps convert the attenuation of feedback circuits into amplification. So unless you have the experience and skills to really know what you are doing, and unless you are using op-amps with gain-bandwidth products measured in many GHz - always use passive filters based on Rs, Ls and Cs to control frequencies above 1MHz.

3.2.2 Imperfections in the basic filter circuits

All components have imperfections, and these were discussed in section 1.8.1 of [6]. These imperfections prevent us from designing cost-effective filters quickly and easily. For example: resistors lose attenuation at high frequencies due to their stray parallel capacitance. Inductors lose attenuation when their stray capacitance causes them to self-resonate, and at higher frequencies. Capacitors suffer from self-inductance, causing them to self-resonate and lose attenuation too.

RC filters are the most predictable EMC filters, as they do not resonate strongly. Values of R over the range 1Ω to 10kΩ are commonly used in EMC engineering, with C values typically less than 100nF. RC filters are mostly used where a DC or low-frequency signal from a low source impedance is connected to a high impedance load: the R is connected to the source side, the C connected to the load side, as shown in the lower part of Figure 3E (below), where they can provide high attenuation at low cost.

LC, inductive Tee and inductive p filters can provide higher attenuation with lower losses than filters using resistors, but are resonant circuits, sensitive to their source and load impedances, so need more careful design.

3.2.3 The importance of the RF Reference

The RF Reference is the node on a circuit’s schematic that we define as our reference voltage when designing an RF circuit or measuring its performance. For the most cost-effective EMC, all circuits (digital, analogue, switch-mode, etc.) should now be designed using RF techniques, and this was discussed in Parts 0 and 1 of this series [6].

It is common practice to call the RF Reference ’earth’ or ’ground’, although it might instead be called ’chassis’ or ’frame’ in some applications, and in circuits it is usually the same structure as the 0V power supply distribution so it is often called 0V. But all these terms are potentially misleading, because what matters in EMC engineering is the impedance of the conductor structure that is being used as the reference for the RF signals or noises, at the frequencies to be controlled. The RF Reference is very important indeed, for all filters that are more than simple series impedances. For filters to function as desired, the impedance seen by the return currents as they flow in the RF Reference must be much less than the impedance of any filter elements connected to that Reference.

So, if we are using a 10nF capacitor in an RC filter to shunt the RF noise to ’ground’, and we want the RC filter to operate as close to its theoretical performance as possible up to 100MHz, we should realise that the reactive impedance of the capacitor (assuming a self-inductance of 1nH) at 100MHz is approximately 0.65Ω (almost all of which, incidentally, is due to its self-inductance). To create a ’ground’ structure that has an impedance of much less than 0.65Ω at 100MHz is quite difficult, because a 10mm length of 1mm diameter wire or 1mm wide PCB trace has an impedance of about 6.3Ω at that frequency. Increasing the diameter of the wire, or the width of the PCB trace, reduces the impedance but not by a great deal - 10mm length of 4mm diameter wire or a 4mm wide trace would still be around 3.2Ω.

A great many earths, grounds, chassis, frames, and 0V systems are made of wire or PCB trace conductors, and designers assume that because they are labelled ’earth’, ’ground’ or ’0V’ they actually are at earth, ground or 0V potential - but in fact they have such high impedances at RF that they have significantly different potentials at various points on their structures, depending on the RF currents flowing in them. Above a few tens of MHz, the only conductive structures that can achieve a low enough impedance to be useful as an RF Reference for circuits, and especially for filters, are metal areas or planes - which is why RF References are quite often called RF Reference Planes.

The circuits that use a plane as their RF Reference must be located much closer than one-tenth of a wavelength (λ/10) to it, ideally λ/100 or even less - at the highest frequency to be controlled. This helps prevent the connections to the plane from behaving as resonating antennas with impedances possibly in the hundreds of Ω. At 1GHz this would mean a maximum spacing of 30mm, and better EMC would be achieved by being much closer than that, ideally 3mm or less.

Where a circuit is shielded by placing it in a metal box, it can use one or more walls of the box, and/or the rear, base and top as its RF Reference. Generally, this would still be called an RF Reference Plane, despite that fact that they are different sides of a metal box. An important consideration in the design of the structure of an RF Reference is that ’surface currents’ must be able to flow freely where they will, all over the area being used. Surface currents are discussed below in the section on Skin Effect.

Many electronic engineers are familiar with the idea of ’single point earths/grounds’ - sometimes called ’star earths’ or ’star grounds’. In such designs the voltage reference is a single physical point, and everything that needs to use it connects to it by a conductor (a wire or PCB trace). Analysing these conductors in terms of impedance, or in terms of their likelihood of becoming resonating antennas, as discussed in the above paragraphs, quickly shows us that single-point or star conductive structures are no use to us for EMC - their conductors are simply too long.

The continued shrinking of semiconductor feature sizes means that even commonplace digital glue-logic devices (e.g. HCMOS) now generate significant noise emissions at frequencies up to 1GHz, and modern FPGAs and microprocessors can be very much worse for EMC - both in level and frequency - than such glue logic ICs. To stand any chance of controlling such frequencies requires lengths of wire, PCB traces or via holes, that are no more than a few millimetres long, preferably <1mm. Using flat braid straps instead of round conductors simply raises the useful frequency by a little, but not by enough to control hundreds of MHz. So single-point earthing/grounding techniques are now only of interest to students of the history of technology, regardless of the type of power or signals involved.

All circuits and interconnections now suffer from RF noise that is coupled into them from digital, switch-mode and/or wireless circuits inside the same product, and they also suffer from RF noise coupled from nearby cables and ambient EM fields in their environments. These coupled noises can cause any circuit or interconnection to be source of RF emissions, and/or a victim of interference, and this is true even for DC instrumentation and low-frequency analogue signals such as audio. So these days, all electronic designers should use EMC design techniques, to achieve the functional performance or reliability they need, and/or to pass EMC compliance tests for emissions and/or immunity.

Despite the fact that the design of the RF Reference Plane is crucial to cost-effective EMC design, many engineers (me included) still tend to refer to ’earth’ ’ground’ or 0V, thereby often leading to confusion and miscommunication with people who think a length of wire can be a useful part of an ’earth’ structure as long as it has green/yellow insulation. So it is important to look beyond the terms that are being used to identify the physical structure that will be used as the RF Reference Plane, or to create it if it does not yet exist.

3.2.4 Differential-mode (DM) and Common-mode (CM)

Wanted signals are always DM: they flow along the ’send’ conductor, and flow back along the ’return’ conductor(s). In single-ended signalling, all the return currents share a common conducting structure, usually the 0V of the DC power distribution system. In balanced (or ’differential’) signalling there is a dedicated conductor for the return current path as well as for the send path, and for good signal quality and EMC the two are routed together as a twisted pair.

However, unavoidable imbalances in the physical realisations of interconnections in PCBs and cables cause CM noise voltages and currents to arise, as shown by Figure 3D. CM currents flow out on both send and return conductors at the same time, and return via another route, often the protective (safety) earth structure and/or the mains power distribution network. CM noise currents and voltages are typically measured in mA and mV, whereas DM power or signal currents are generally several orders of magnitude larger. But the very much larger loop areas associated with CM noise currents and voltages makes them more important for EMC than the DM signals whose imbalances caused them. Above about 1MHz most unwanted emissions are mostly CM.

A great deal of RF interconnect design is concerned with making cables and PCB traces that have better balance, to reduce the ’longitudinal conversion loss’ (LCL) that converts the wanted signal energy into unwanted CM noise. The lower the LCL, the further the wanted signal will propagate with an acceptable quality, or the higher the frequency that can be sent with acceptable quality over the same distance - hence the computer networking industry’s progress from Cat 5 to Cat 6 and eventually to Cat 7 cables for Ethernet. Each increase in Category is accompanied by better balance, resulting in better LCLs at higher frequencies and reduced generation of CM noise for a given type of Ethernet signal.

Because of the existence of DM and CM signals and noises, we need to be able to apply filtering techniques to both of them. Below 1MHz, we are more likely to be concerned just with filtering DM signals and noise. But at higher frequencies we can use DM filtering to reduce the amounts of RF present in conductors, so as to reduce the amount of CM noise currents and voltages created by their imbalances. We can also use CM filtering to reduce the amounts of CM noise present.

Slide 4

Figure 3D Differential-mode (wanted) signals versus common-mode ’noise leakage’

3.2.5 Maximising impedance discontinuities

As mentioned earlier, to design effective filters we must maximise impedance discontinuities, at the frequencies of concern for emissions and/or immunity, and Figure 3E tries to demonstrate this concept for single-ended signals. Capacitors are used in conjunction with the RF Reference Plane (see above) to create low impedances, applied in shunt, whilst resistors or inductors are used to create high impedances, applied in series.

When the source and load impedances seen by a current (DM or CM) are both low - also taking into account the impedances of their current loops including their return paths - a ’Tee’ filter (with either R or L) is preferred. When the source and load impedances seen by a current (DM or CM) are high - also taking into account the impedances of their current loops including their return paths - a π (’Pi’) filter (with either R or L) is preferred. When the source impedance for a current (DM or CM) is low, and its load impedance is high (or vice-versa) - also taking into account the impedances of their current loops including their return paths - an RC or LC filter (with the R or L connected to the low impedance side) is preferred.

For low-power circuits with low-frequency wanted signals and high impedance loads, it is often possible to replace the inductors in these simple circuits with resistors of between 100Ω and 10kΩ to save cost and even sometimes achieve higher attenuations over wider frequency ranges.

Slide 5

Figure 3E Maximising impedance discontinuities to improve attenuation

When the source impedance of the noisy current (DM or CM) to be filtered is low, fitting a simple C filter will increase the noise currents flowing in the circuit, increasing the magnetic (H) field emissions, and will also increase the noise voltages across the circuit’s 0V plane, increasing its CM electric (E) field emissions. Preventing DM or CM noise currents from increasing is another reason why we always follow a low impedance source with a series resistor or soft ferrite choke (to create an RC, LC or Tee filter).

For balanced (differential) signals the RF Reference in Figure 3E is replaced by the return conductor for the conductor pair - but only for DM filtering. For CM filtering we need two circuits as shown in Figure 3E - one for the send conductor and one for the return conductor, both of them connecting their capacitors to the RF Reference Plane. CM filtering can also benefit greatly from the use of ’CM chokes’ - described later and shown in Figures 3L, 3M and 3N.

3.2.6 Using soft ferrite cores

All inductors (L) suffer from RF resonances, and are only effective in filters at frequencies not far above their first (parallel) resonance (see section 1.8.1 of [6]). But so-called ’soft ferrites’ behave resistively at RF, and the resulting lack of RF resonances helps make filters that use them have better and more predictable performance. For example, a typical small ’soft ferrite’ bead a few millimetres in diameter will have around 1μH of inductance and 0.1Ω of resistance at DC, but around 80Ω of real resistance (not inductive reactance) at frequencies from 30MHz to 1GHz or more. Some leaded soft ferrites are available with resistances of over 1kΩ at 100MHz, but a much wider range of surface mounted device (SMD) soft ferrites is available with resistances up to 1kΩ or more at selectable frequencies from 30MHz to 2GHz.

Soft ferrite components are known by a variety of names, including ’RF suppressers’, ’Interference suppressers’, ’Suppression chokes’, and ’Shield beads’. Figures 3F and 3G show some of the cable-mounted soft ferrite parts available. A very wide range of PCB-mounted soft ferrite components is also available, but not shown in these figures. Figure 3G includes a standard VGA cable, showing the standard soft-ferrite CM choke that all VGA cables are required to have at each end, for the products they interconnect to meet emissions regulations in Europe and the USA (at least).

The bottom right-hand-side of Figure 3G shows a toroidal soft ferrite core used as a CM choke, in this case with ’41/2 turns’ of cable wrapped around it. As will be described below, the attenuation of a filter at the highest frequencies is governed by the stray coupling between its input and output conductors - so it is important that the input and output conductors of a CM choke, such as the one in this photograph, are as far apart from each other as possible. This results in the winding format that can clearly be seen in the Figure - only half the core is wound and the input and output cables are on opposite sides from each other, the perhaps odd description of it as having ’41/2 turns’ is a common way of making clear that input and output conductors are on opposite sides.

Slide 6

Figure 3F A wide variety of soft ferrite cores is available (these examples of toroidal and cylindrical types are from Philips)

Slide 7

Figure 3G Some more examples of soft ferrite cores

A useful soft ferrite component is a cylinder split lengthways and held in a plastic clip-on housing, and some examples of this are also included in Figure 3G, for round cables as well as for flat cable styles. Such split ferrites are very easy to apply to cables (and to remove if found to be ineffective), and EMC engineers tend to carry many of these around with them, using them for the diagnosis, isolation and curing of EMC problems, both DM and CM. A ferrite cylinder clipped around an entire cable or cable bundle, including all the send and return conductors, is a CM choke, but if clipped over just a send or return conductor it is a DM choke.

Slide 8

Figure 3H Choosing soft ferrites

Choosing soft ferrites involves checking that their impedance is as high as required over the frequency range for which significant attenuation is required. Soft ferrite components always have impedance versus frequency curves that are smooth and not discontinuous, whereas the curves for inductors will show one or more discontinuities (changes in slope from positive to negative, or vice-versa, that occur at a point) that reveal the presence of self-resonances.

Some data sheets only provide impedance data for a portion of the frequency range you are concerned with. But it would be a mistake to assume anything about the impedance they will achieve in the frequency range for which no data is provided - always make sure you have manufacturers data on the impedance over the entire frequency range that you wish to control.

An aspect of choosing soft ferrites that is often overlooked, is that their impedance versus frequency curves vary with their DC and/or LF current. Typical data sheet curves assume zero current in the device, but as the current increases the frequency at which the peak impedance occurs will also increase, and the peak impedance may reduce.

Often, when an emissions or immunity test is failing at some frequency (e.g. a clock harmonic at 228MHz), a soft ferrite will be chosen that has a very high impedance close to this frequency, and it will be added in series with traces on the PCB that are thought to be the cause of the problem. But the DC or LF current in those traces could make the frequency of the peak impedance increase by enough that the actual impedance achieved at the problem frequency is not high enough to provide significant attenuation and pass the test. Instead, the currents in the traces and the frequency/current variation of the type of devices to be used should have been taken into account, to select a device that would have its peak impedance at the problem frequency when the trace current is passing through it.

Slide 9

Figure 3J A wide range of soft ferrite devices is available (these examples from the Murata BLM21 series, 0805, typically rated 100 - 200mA)

Several manufacturers offer ranges of soft ferrite RF suppression components, and are continually adding to them. Recent additions include SMD parts rated at 3A continuous current at DC and low frequencies; yet provide 1kΩ or more around 100MHz. Other recent additions include parts that provide impedances of 1kΩ or more over the range 100MHz to 2GHz.

Curves such as those in the two top graphs in Figure 3J are most suitable for filtering low-frequency signals. The two bottom curves show devices that have been tailored for filtering unwanted harmonics from digital waveforms, whilst leaving sufficient lower-frequency harmonics to create digital waveforms that have rise and fall-times fast enough to reliably meet the maximum skew requirements of the receiving devices. CM ferrites tend to have impedance versus frequency curves similar to the top left-hand graph, since there is no need for any CM currents at any frequencies.

When simulating filters or other circuits using soft ferrite components, a simple device model cannot be used - the parameters are frequency-dependant and current-dependant, and may also be temperature dependant, and should be modelled as such to achieve any accuracy in the simulation over a range of frequencies. Some circuit simulators may be unable to handle models with such complex parameters.

3.5 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 http://www.cherryclough.com
[2] The Institution of Electrical Engineers (IEE), Professional Network on Functional Safety, "EMC and Functional Safety Resource List", via the "Publications & Downloads" page at http://www.cherryclough.com
[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] Keith Armstrong, "Design Techniques for EMC, Part 2 - Cables and Connectors", The EMC Journal, May and July 2006, available from http://www.compliance-club.com.
[6] 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 http://www.compliance-club.com.
[7] 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.
[8] 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 http://www.cherryclough.com
[9] The United Kingdom Accreditation Service, http://www.ukas.com
[10]Tim Williams and Keith Armstrong, "EMC for Systems and Installations", Newnes 2000, ISBN 0 7506 4167 3, especially chapter 8, www.newnespress.com, RS Components Part No. 377-6463

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.

Eur Ing Keith Armstrong CEng MIEE MIEEE Partner, Cherry Clough Consultants,
www.cherryclough.com, Member EMCIA
Phone: +44 (0)1785 660 247, Fax: +44 (0)1785 660 247,
Email: keith.armstrong@cherryclough.com www.cherryclough.com