Following some basic precautions and working with the filter manufacturer can solve a variety of filter problems..

Mark Carter

Electromagnetic interference (EMI) is all around us. EMI is created naturally from everyday sources such as lightning, rain, and even strong winds. Cosmic noise emitted by celestial bodies (e.g., solar flares) within and outside our solar system can cause EMI problems on Earth.

There are also man-made EMI sources to contend with. These sources include televisions, power transmission lines, ignition systems, fluorescent lighting, radar transmitters, electric car chargers, and computing devices, just to name a few. Man-made EMI sources challenge the equipment designer and component engineer to find a solution to keep the signals coming to or going from their black box clean and usable. In most cases, a feed-through filter is selected and placed in the signal path to shunt any interference to ground. Sometimes, this seemingly simple solution is actually the start of just another series of problems.

Aside from inherent manufacturing defects, most feed-through filter problems are caused by improper installation, improper device selection for the application, or incomplete specifications.

Figure 1. Bulkhead-mount filters and terminal configurations.

This article addresses some common maladies that can compromise EMI filter performance in an application. The basic screw-neck header or bulkhead-mount configuration (see Figure 1) is referenced for these discussions. These filters are passive devices with bidirectional effects, capable of passing low-frequency line signals of high current (e.g., 400 Hz and 20 A) through the center conductor while attenuating higher-frequency signals (e.g., from 30 kHz to 1 GHz) to ground through a discoidal, multilayer ceramic capacitor.

Unlike conventional leaded capacitors, the discoidal capacitor's coaxial configuration provides two unique advantages. It prevents radiation from the input from coupling directly with the capacitor output. This construction also has inherently low self-inductance. This combination provides excellent shunting of EMI at frequencies approaching 1 GHz. The addition of magnetic elements (wire-wound coils, toroids, or beads) placed in series with the capacitor increases the impedance of the line, making the filter even more effective.

Figure 2. Bulkhead mounting.

When housed in a hermetically sealed tube with feed-through axial terminals and a threaded post or neck acting as the case terminal to ground, EMI filters can be extremely rugged devices capable of operation in extreme environments. This construction allows the filter to be mounted through a bulkhead (see Figure 2) so that the shielding integrity of an enclosure is not compromised. The feed-through terminals, isolated from the case by either compression glass seals or a potting compound, are typically leads or a solder lug. This construction allows the filter to reduce electrical noise going to or coming from a device. Filters do not differentiate between interference and other signals generated either inside or outside of the enclosure.

Filter Selection

Feed-through filters consist of either a capacitor (C only) or a combination of capacitors and magnetic elements (LC, Pi, T, or LL). Each type of filter fits a particular application requirement. Proper selection is critical. The most economical and reliable solution is to select the filter with the fewest internal parts. This filter will naturally cost less than one with a more complex internal assembly, and it will also have fewer things that can go awry in the field. The following sections describe each of the basic filter types and provide typical attenuation response characteristics based on the Dearborn Electronics (Longwood, FL) Type JX EMI filter series.

Figure 3. C-only filter schematic.

C-Only Filters. C-only filters, or filters that consist solely of capacitive elements (see Figure 3), are best suited for filtering high-frequency signals on lines with very high impedances. The attenuation of these devices increases in steps of 20 dB per decade from the filter's cutoff frequency up to the frequency at which the attenuation is at least 60 dB.

LC Filters. LC filters (see Figure 4) are best suited for applications in which there are large differences between line and load impedances. These devices consist of a capacitive element, in the same manner as the C-only filter, and an inductive element connected in series with the capacitor, between the input and output terminals. Usually, it is best to install the filter so that the inductive element faces the lower impedance. This means that, in some applications, it is desirable to have the capacitive element close to the threaded or screw-neck header end of the filter.

Figure 4. An LC filter schematic (L2 threaded neck).

In other cases, the reverse setup would be desirable, with the inductive element located on the threaded or screw-neck header end. Flexible manufacturers provide both styles in their catalogs, indicating the two different configurations. For example, in one EMI filter catalog, the designation "L1" indicates that the inductive element is located on the threaded end, whereas "L2" indicates that the capacitive element is located on the threaded end. The attenuation of these filters increases in steps of 40 dB per decade from the filter's cutoff frequency up to the frequency at which the attenuation is at least 70 dB.

Pi Filters. Pi filters consist of three elements (see Figure 5). A series inductive element is positioned between two capacitors connected to ground. Pi filters are best suited for applications in which the input and output impedances are of similar value and high levels of attenuation are required. These filters typically increase attenuation by 60 dB per decade from the filter's cutoff frequency to the frequency at which the attenuation is at least 80 dB.

Figure 5. Pi filter schematic.

T Filters. The T filter (see Figure 6) is also a three-element device, but there are two series inductors connected between the input and output terminals on each side of a single capacitor connected to ground. T filters perform in much the same manner as Pi filters, increasing attenuation in steps of 60 dB per decade from the filter's cutoff frequency to the frequency at which the attenuation is at least 60 dB. Select this filter type when both the input and output impedances are low.

LL Filters. Internally the most complicated device, the LL filter (see Figure 7) consists of two feed-through capacitors connected between line and ground, with two interspersed inductors connected in series with the capacitors between the input and output terminals. It can be thought of as two LC filters in series. These filters increase in attenuation in steps of 80 dB per decade from the filter's cutoff frequency to the frequency at which the attenuation is at least 80 dB.

Figure 6. T filter schematic.

Operating Current

Another important consideration in filter selection is the operating current. Do not skimp on this requirement. It directly affects the heating of the central contact of the ceramic capacitor from the heat generated in service through the conductor wire due to its resistance. In some cases, the current rating may be provided by the part number.

A very common problem occurs in ac applications. One frequently hears comments like, "The part exploded!" "It was all burned up inside!" or "The parts are shorting out in just a few days!" Many times, the user has simply divided the dc voltage rating listed in a catalog by two, as a rule of thumb, and assumed that this is the ac operating voltage at the maximum dc operating temperature. Never do this. (If you have been doing this, you have been lucky and, eventually, your luck is going to run out.)

Figure 7. LL figure schematics (threaded neck).

This is one of the most common application mistakes made with any electronic component. The result is that strange things begin to happen in ac applications that would never occur in dc applications. Two of the most nefarious are corona and heat rise. Always make certain that the filter selected for an ac application has an ac voltage rating. Pay particular attention to the ac operating frequency and the ac maximum operating temperature.

X7R Dielectric Characteristics

Most EMI filters use X7R ceramic formulations for the capacitor dielectric. This is a ferroelectric material with certain inherent disadvantages. Over time, this dielectric will experience capacitance loss due to intercrystalline aging. This capacitance loss occurs logarithmically when temperatures go below the Curie point of the formulation. The applied voltage, ac or dc, also affects capacitance when the voltage level is low in comparison with the device's rated voltage. When this happens, a polarizing effect takes place that can drop the capacitance value by as much as 50% from the original value. The X7R dielectric also exhibits a large change in capacitance with temperature. The capacitance can drop by as much as 10% from the original value at high (85°C) and low (–55°C) temperatures. If the filter is suddenly not achieving the response it once had at frequencies below 100 kHz, these X7R characteristics may help to explain why.

Another undesirable attribute of the X7R dielectric, particularly in ac applications, is its loss characteristics or dissipation factor (DF). Typically, this can be as great as 0.02 absolute at 1 kHz, to greater than 0.04 absolute at 100 kHz at 25°C. This loss characteristic is directly related to the self-generated heat rise inside the device during ac operation. This heat rise must be added to the ambient operating temperature to have an accurate indication of the filter's actual temperature during operation. You may think that you have an ac filter operating at 125°C, when in actuality it may be running at closer to 160°C.

This can cause serious problems because it places unexpected stresses on the dielectric. The first of these is mechanical stress. Imagine equipment operating in the winter, or at high altitude, that is suddenly turned on—known as a cold-start situation. The differences in the thermal coefficients of expansion between the ceramic capacitor, the metal housing, and the center conductor are quite large. This places unnecessary stresses on the fragile ceramic layers, which can lead to cracking and delamination. The breakdown strength of the dielectric layers is also affected by temperature. In time, the stresses will lead to a premature failure such as an electrical short.

Unfortunately, this root cause of the failure, cracks in the ceramic dielectric layers, is frequently misdiagnosed during failure analysis. During analysis, a user may review a cross-section of a failed EMI filter and find microscopic cracks in the layers. The common conclusion is that the cracks were there from the beginning when, in fact, the cracks were actually caused by the heating and cooling during ac operation. Consequently, the user may then mistakenly impose a rigorous microscopic analysis of the ceramic capacitors on the manufacturer prior to assembly. The result is the perpetuation of the problem at a higher per-unit cost, passed on by the manufacturer.

Nonferroelectric Ceramic Dielectrics

The shortcomings of X7R dielectrics can be avoided by using nonferroelectric ceramic materials. Nonferroelectric ceramic dielectrics have the following characteristics:

• They do not exhibit the time dependence of the dielectric constant and therefore do not experience a drop in capacitance from intercrystalline aging.

• They have constant capacitance regardless of the applied voltage.

• They experience virtually no change in capacitance with temperature.

Figure 8. Type JX internal heating (at left) and Type JC internal heating (at right).

This means that the nonferroelectric dielectric is neither subjected to mechanical stresses caused by the differences in thermal expansion nor subjected to voltage stresses at temperatures exceeding that of the ambient environment. The result is a filter with a longer life expectancy and greater reliability. These high-performance filters are available in all of the common filter configurations, C, LC, Pi, and T, and in the same housing styles as the Type JX series of EMI filters.

Installation

Mount the filter as close as possible to the power line's egress from the device being filtered. The filter's input and output leads should be physically separated to provide the greatest amount of electrical isolation possible. At any point of penetration through the device's electrical shield, make sure that the shield's continuity is maintained. A lot can go wrong with a feed-through filter from the very start that has nothing to do with the application. Improper soldering is the first suspect.

A soldering iron that is held on the center conductor for too long during installation could cause several different problems. First, it may cause the ceramic capacitor to develop microcracks in the dielectric layers because the heat is conducted down the center conductor to the ceramic capacitor. Over time, these tiny cracks can accumulate debris and contaminants and, in conjunction with the applied electric field, develop conductive paths. The conductive paths then lead to an ever-increasing level of leakage current until an electrical short occurs.

Second, the center conductor of the ceramic capacitor may actually be desoldered, removing the capacitor from the filter circuit completely. This of course will drastically affect the attenuation response of the filter.

Third, if hermeticity is required to protect the integrity of the filter construction inside the housing, desoldering the seals of the filter will lead to premature failure. This is especially true in many avionics applications in which constant changes in atmospheric pressure and temperature create large amounts of condensation. Moisture is quickly absorbed by the ceramic material. This will eventually eat away at the internal electrodes, leading to a loss of capacitance or a high DF.

The remedy to these problems is simple. Train production operators to solder according to the IPC/EIA J-STD-001C Joint Industry Standard. If the problem persists, contact the filter manufacturer for recommendations. Typically, this means using a hotter soldering iron for as short a time as possible, rather than a cooler iron applied for a longer period. The solder alloy used during installation is also important. It cannot have a higher melting point than the solder that the manufacturer used to seal the device. Usually, a 60% tin and 40% lead solder alloy should be used during installation.

The filter's metal case must make direct, low-resistance contact with the metal chassis, cabinet, or ground plane. The terms snug and tight have different meanings to many people, and this applies to the personnel on the production floor as well. This terminology may work well for nuts and bolts, but it does not work for feed-through filters. When installing feed-through filters, it is important not to overtorque the threaded or screw-neck header.

If too much torque is applied when the mounting hardware is installed, the torsional stresses applied down through the center conductor may actually twist the internal components of the filter. This can literally rip off the center conductor, reduce the overall center-conductor cross-sectional area, or damage the magnetic components inside the case. Avoid this potential problem by providing production operators with a torque wrench.

The most common thread size for feed-through filters is 1/4 in.-28. Never torque these terminals more than 48 in.-lb. A useful reference for other thread sizes can be found in Table XIV of MIL-PRF-15733, the general military specification for radio-frequency interference filters and capacitors.

Never bend the terminals on a filter. Bending can transmit the same types of forces down the center conductor as applying too much torque to the threaded or screw-neck header. If a special orientation of the terminals is required, ask the manufacturer to supply the finished feed-through filter in that configuration.

After a filter has been selected and installed, how will you know if there is a problem? Some simple electrical tests and visual examination can answer the question quickly. Remove the filter and visually examine it. Do any of the solder joints of the filter construction show signs of desoldering, especially the terminals and the outside diameter of the compression glass seals? Is the insulating glass or the potting compound cracked? Is the screw-neck header twisted and misshapen? If any of these conditions are detected, then there may have been a problem during installation, or the part may have been operating at a temperature that was too high.

Does the capacitance from the center conductor and the case meet the minimum required value? If it is only about half of the requirement in a Pi-configured filter, then one of the capacitors has failed. When electrified at the rated voltage, does the filter meet the maximum leakage current (or the minimum insulation resistance) specified? If not, then the ceramic capacitor may be headed to an early failure as an electrical short.

Specifications

The user should supply the following specification information to the filter manufacturer:

• Minimum capacitance value at a specified frequency.

• Working dc voltage.

• Working ac voltage and frequency.

• Current rating.

• Operating-temperature range, ac and dc.

• Dc resistance.

• Maximum DF at a specified frequency.

• Minimum insertion-loss requirements, both with no load and under load.

• Dielectric withstanding voltage.

• Vibration requirements.

• A mechanical outline drawing with finish or plating.

• Seal test requirements.

One specification requirement that is commonly overlooked, at times with dire consequences, is a transient-voltage requirement, such as from a lightning-strike phenomenon. If the filter is subjected to this event in the application environment, it is very important that the requirement is conveyed to the filter manufacturer so it can be taken into account during part selection or design.

Regarding environmental testing, it is completely acceptable to include temperature cycling and burn-in requirements in the specification. They provide a convenient method to cull out filters with infant mortality. However, if it becomes necessary to revise the specification repeatedly to increase the length of the burn-in or the burn-in voltage, or to impose percent-defective-allowed (PDA) requirements, then the filter is probably incorrect for the application. This is the time to sit down with the supplier, spread the application cards on the table, and work together to find the best solution.

Conclusion

Feed-through filters can solve a variety of signal-processing problems when they are properly specified, selected, and installed. Providing production operators with the proper training and tools will avoid many problems later during final equipment testing and in the field. Always select the simplest filter to solve the EMI problem; this will be the most economical and reliable solution.

Be wary of ac applications, and never fall into the 1Ž2-dc-voltage trap. It is best to consult with the manufacturer when selecting parts for these applications. Communicate to the filter manufacturer, in the filter specification, the application requirements in as much detail as possible. If these steps are followed, then solving EMI problems will be as easy as one, two, three.

Acknowledgments

The thermographs were provided courtesy of Charles Stroo and John Slusarek of Rockwell Collins (Cedar Rapids, IA).

Mark Carter is chief scientist at Dearborn Electronics Inc. (Longwood, FL). The author can be contacted via e-mail at dearborn92@att.net.