Filtering--Interfaces: Attenuating noise

The purpose of a filter in the context of EMC is to prevent interference from travelling either into or out of equipment via its interfaces. This reduces conducted coupling directly, and also helps to reduce radiated coupling if the interference is radiated to or from the cables that connect to the interfaces.

---Variations of low pass filter -- (a) basic elements (b) L filter

The low-pass filter

Typically, unwanted interference is at a higher frequency than the wanted signal that is connected via the interface. This is not universally true; in some cases the wanted signal may be a radio frequency and all disturbances both above and below this frequency should be rejected. Or, the wanted signal may be wideband, such as video or network data signals, and the interference to be rejected may occupy the same part of the spectrum. In general though, the interference frequency is above that of the wanted signal. This then calls for a low pass filter to provide the desired attenuation characteristic.

The conventional low pass filter is built from two elements: series inductance or resistance, and parallel capacitance. The minimum configuration (a) is one or other of these, and combining them gives the L-filter (b), the T filter (c) or the pi filter (d). The choice of inductance or resistance for the series element is usually determined by the wanted signal current that has to be passed: power filters will typically be unable to stand more than a few ohms resistance, but signal filters might easily be able to cope with kilohms. Resistance has the advantage that it absorbs interference energy and does not contribute to resonances, and is of course cheaper and smaller than inductance.

Inductance on the other hand can provide high RF impedance with little DC or low frequency loss.

Choice of component types for a filter is determined by the frequency range that has to be attenuated. Non-ideal components have parasitic impedances that limit their effectiveness -capacitors have unwanted stray inductance, inductors have stray capacitance. High frequency performance demands that these parasitics are minimized, and this puts restrictions on the type of construction. In general, any single component can only achieve a maximum of 40-50dB of attenuation before its parasitics limit its performance, hence high performance filters must use multiple stages.

Capacitors create a low impedance to attenuate the interference frequencies, and are therefore most effective in a high impedance circuit. Inductors create a high impedance to the interference, and are most effective in a low impedance circuit.

Differential versus common mode in filters

At least as important as the configuration of the circuit, is the mode of interference that is being suppressed. The differences between differential and common mode. A filter must be designed to deal with the mode that is most relevant to the interference that is present; conversely, a filter that is attenuating the wrong mode will be ineffective, no matter how good it is. (This is actually quite a useful diagnostic device!) The way that an L-configuration filter would be wired to attenuate (a) differential mode and (b) common mode interference on a signal pair. The differential mode interference is present between the conductors of the pair, and therefore the capacitor must be placed in parallel across the conductors and the inductor is in series with just one conductor. By contrast, the common mode interference appears on both conductors together, returning via the ground/earth connection external to the desired circuit. Therefore, the differential mode filter configuration would not affect this mode at all- the capacitor is across two conductors which are carrying the same interference voltage and is therefore invisible to the interference, and the inductor is shorted by the lack of impedance in the opposite conductor.

--- Differential mode and common mode configurations

To make a common mode L filter, the series impedance must be present in all the signal conductors, and the capacitance must bridge between each signal conductor and the ground/earth reference. Separate series inductors or resistors in each line can be used for common mode attenuation and will similarly be effective in differential mode. But it is also possible to implement all the inductors on a common core and wire them such that only the common mode current is affected. This has two crucial advantages:

++ magnetic flux due to differential mode current cancels in the core, so a given core size can handle much larger signal or power currents and inductances without saturating, allowing higher attenuation;

++ the windings are invisible to the signal current; therefore a filter can be built which attenuates in-band common mode interference without affecting the desired signal, which is in differential mode. This is of particular relevance to wideband signals.

The same convenient effect does not occur with the capacitors, which appear across each line and ground/earth. This can have a serious effect on wideband circuits and often means that either such capacitors must have a very low value, or they cannot be used at all. For low frequency, low impedance signal and power circuits no such problem exists. Most of the commonly-used filtered D type connectors use this principle: each pin has an in-line capacitor connected between it and the connector shell, which is ground/earthed to the zone boundary.

The most important aspect of the common mode filter is that a connection to an appropriate ground/earth point must be available for the capacitors. If it is not, then capacitors cannot be used and only series chokes or resistors are useable. Even if an ground/earth connection is possible, the capacitors' effectiveness will be compromised if they are not connected via a low impedance to a good-quality RF ground/earth. This normally means a direct, short bond to the chassis of the equipment being filtered.

A common mode filter using multi-winding chokes and low value capacitors can be designed to have negligible effect on differential mode currents, but you can instead design a filter to act in both differential mode and common mode as well, if this is desirable. Higher value parallel capacitors to ground/earth will increasingly contribute to differential mode attenuation. The leakage inductance of the common mode choke will also appear in series with the differential mode circuit. Careful choke design and construction can optimize this leakage inductance value so that it provides the desired level of differential attenuation without contributing to choke saturation by, or attenuation of, the wanted differential currents.

Source and load impedances

The performance of any filter depends heavily on the impedance seen at its terminals.

There are four relevant impedances for a simple single phase mains filter:

++ differential mode (symmetrical) at the mains port;

++ common mode (asymmetrical) at the mains port;

++ differential mode (symmetrical) at the equipment port;

++ common mode (asymmetrical) at the equipment port.

All these impedances will be complex and frequency dependent in real life, but most filters have their performance specified by tests done with 50~ source and load impedances, which leads us straight to a very important point- filter specifications are optimistic when compared with their performance in reality.

Consider a typical supply filter, installed between the AC power supply and an AC- DC converter typical of the DC power supply of an electronic apparatus. The impedance of the AC supply can vary by as much as from 2 to 2,000~ depending on the loads that are connected to it, the nature of the supply transformer and the wiring to the point of connection. The impedance is complex and is both time- and frequency- dependent. Research in the 1970s collected a wide variety of data on the domain within which the impedance might fall: an indicative diagram is shown, which gives the possible boundaries of the impedance domains for the residential public mains supply in the complex plane at three discrete frequencies, 10kHz, 100kHz and 30MHz.

--- Impedance domains for the mains power supply

Looking from the filter into the equipment, the impedance of the AC-DC converter circuitry appears as a low impedance when the rectifiers are turned on near the peaks of the supply waveform, and as a high impedance at all other times. The situation is far from being the matched 50 -ohm / 50 -ohm set-up used to measure filter attenuation.

Filter specifications employ 50 -ohm source and load impedances because most RF test equipment uses 50 -ohm sources and loads and cables, and because the main specification standard (CISPR 17) requires this. For most practical uses of filters the specifications obtained by this method are at best optimistic, and at worst misleading.

Filters made from inductors and capacitors are resonant circuits, and their performance and resonance can depend critically on their source and load impedances.

An expensive filter with excellent 50/50 -ohm performance may actually give worse results in practice than a cheaper one with a mediocre 50/50 -ohm specification.

The problem of resonant gain

The most sensitive to source and load impedances are supply filters with a single stage.

They can easily provide gain, rather than attenuation, when operated with source and load impedances other than 50 -ohm. This gain usually appears in the 150kHz to 1MHz region and can be as much as 10 or 20dB, leading to the possibility that fitting an unsuitable mains filter can increase emissions and/or worsen susceptibility.

Filters with two or more stages are able to maintain an internal circuit node at an impedance which does not depend very much on the source and load impedances, so they are better able to provide a performance at least vaguely in line with their 50/50 -ohm specification. Of course, they are larger and cost more.

The easiest way to deal with the source/load impedance problem is only to use filters whose manufacturers specify differential mode (symmetrical) performance for both "matched" 50/50 -ohm and "mismatched" sources and loads. Regulations requires that mismatched figures are taken with 0.1 -ohm source and 100 -ohm load, and vice versa. Draw an attenuation versus frequency curve consisting of the worst-case figures from each of these various curves, and use this as the filter's specification. An example of this graphical procedure is shown.

--- Deriving reliable filter attenuation figures from manufacturers' data When filters are chosen using this technique to try to meet the predicted or actual requirements their performance can be at least as good as expected. When the 50/50 -ohm figures alone are used to predict filter performance the result is often disappointing.

The problem with this worst-case method using mismatched 0.1 -ohm/100 -ohm filter data is that the real impedances are somewhere between the worst case and the standard 50 -ohm termination. The effect is that the filter will almost certainly be over-specified, especially if performance below 1MHz is the main concern. The performance will also be different depending on the dominant interference path, common mode (asymmetric) or differential mode (symmetric), since the filter's attenuation is different for each mode.

If the data on impedance domains and coupling paths are actually known, then it is possible to analyze the effectiveness of any given filter using suitable software, to optimize its design. This would be reasonable if the cost of the filter was a significant concern in the overall project budget. For system use though, this is rarely the case, and the measurements needed to determine the impedances are not likely to be cost effective. In this situation, using worst case manufacturer's figures for an off- the-shelf unit is a reasonable approach.

Layout and installation

Incorrect filter construction or mounting technique can easily compromise radiated emissions and immunity. Poor shielding can easily compromise conducted emissions and immunity. The correct way to view filtering and shielding is as a synergy, with each one complementing the other.

Locating the filter

A filter is rarely located anywhere else except at a zone boundary (see the discussion of zoning). This is for two reasons:

++ the filter is part of the protection offered by the zone barrier. Locating it at a distance from the barrier would allow cables between the filter and the barrier to breach this protection;

++ a filter needs a high-integrity ground/earth reference for good high-frequency operation. The zone barrier, which will usually be either a shielding wall in a cabinet or chamber or the entry to an ground/earthing mesh, provides this directly.

A large metal plate (at least lm x lm) bonded to the ground/earth structure at the Single Point of Connection (SPC) to a zone can also serve as such a reference.

An example of the application of these principles is given in the next section. In addition to this general rule, filters should be located as near as possible to the apparatus which is expected to be the source or victim of disturbances, to minimize the impedance of the connection. If, as a compromise, the filter is positioned outside the protected area or apparatus, the wiring between the filter and the protected area should be twisted and positioned close to the ground/earthing structure.

--- Filter location

--- Mounting supply filters in shielded enclosures

Filter construction and mounting

The higher the frequency, the more a filter is compromised by RF leakage from its unfiltered side to its filtered side. Many engineers have been surprised by the ease with which RF will leak around a filter.

Where an external cable to be filtered enters a shielded enclosure or room, the filter should be fixed into the metal wall at the point of cable entry and RF bonded to the metalwork all around the aperture it fits into. Through-bulkhead filters are the best, since they maintain the integrity of the shield, but are often expensive to purchase and install. For mains supply filters, the IEC-320 inlet is the most common commercial style of bulkhead filter for up to 10 amps, single phase, 230Vrms.

Because of commercial pressures, most mains filter manufacturers only specify their parts over the frequency range of the conducted emissions tests (up to 30MHz). The filter becomes progressively less effective above 30MHz and can compromise the shielding integrity of shielded enclosures, possibly causing problems with radiated electromagnetic disturbances. Good layout inside the enclosure will minimize high- frequency coupling onto the internal (filtered) supply and hence control this possibility.

For high powers most commercially available filters use screw-terminal block or Faston connections, making bulkhead mounting impossible. How to mount a screw-terminal filter using the "dirty-box" method. This encloses the filter in an individually shielded, segregated box within the main shielded enclosure. The general clean/dirty box arrangement is here; the dirty box segregation is applied to just one filter. The filter input and output cables in the dirty box must be very short and far away from each other, but even so high frequencies may still leak across and ferrite sleeves () may be needed on either or both cables if performance above 30MHz appears to be compromised.

Filters sold as "room filters" generally deal with the problem of leakage due to cables by enclosing the filtered side in the metal filter box and passing them through their mounting base via a standard circular conduit fitting. Such filters are intended to be mounted directly on the external metal wall of a shielded enclosure (of any size) with only their filtered output appearing on the inside of the enclosure. This construction effectively shields the unfiltered from the filtered cables and so allows the filter to function effectively up to the highest frequencies. This is the method used by most of the supply filters intended for EMC test chamber applications.

The ground/earth connection

All commercial filters are housed in metal bodies of some sort, with the body forming the filter's ground/earth connection. An IEC inlet filter with a metal body installed in a shielded enclosure can only give a good account of itself at frequencies above a few tens of MHz if its body has a seamless construction and its body is RF bonded to the shielding metalwork. The same is true for any other metal-bodied filter.

The reason is that any inductance due to a wired ground/earth connection will turn the filter into a high-pass configuration. The greater the inductance - that is, the longer the wire

- the lower the frequency at which this becomes significant. This section discusses the impedance of ground/earthing wires in greater detail. In the context of filters, the measured effect: about 25dB difference at 15MHz; of two different lengths of ground/earth wire to chassis on the attenuation of a simple single-stage mains filter.

--- Comparison of filter ground/earth bonding Bonding the case directly to the chassis ground/earth is the only sure way to realize a filter' s published performance. The often-provided separate tag to the case is for safety purposes, not for use as an EMC connection.

Wiring to filters

Filtered and unfiltered cables must be kept rigorously segregated as they are always at least one Class of cable apart (see cable classes and segregation). The rules on segregation generally advise between 150 and 300mm separation distance between adjacent classes, and 600mm or more between the most sensitive and most noisy classes. Of course, this may not be possible at the filter terminals themselves, since the filter body may be smaller. But in situations where there is no screen across the filter, input and output cables should be carefully dressed away from each other as they leave the filter terminals until the required separation distance is reached- screened cable, or conduit, may be needed close to the filter to help prevent leakage from the filtered to the unfiltered cable and vice-versa.

If a filter must be installed in-line in a cable tray or conduit acting as a PEC, then all the cables in that conduit must be filtered, otherwise coupling to and from the unfiltered cables will compromise the high frequency attenuation.

On no account should the installation technician be allowed or encouraged to loom together a filter's input and output wiring. This desire for neatness is going too far and will render an expensive filter almost worthless!

Next: Mains filters

Prev: Installing cable systems

top of page  Article Index   Home