Filter components

Component self-resonance

Filter components, like all others, are imperfect. Inductors have self-capacitance, capacitors have self-inductance. This complicates the equivalent circuit at high frequencies and means that a typical filter using discrete components will start to lose its performance above a certain frequency - which can be below 1MHz for large components, up to 50MHz for small leaded components and into 100s of MHz for surface mount parts. The complementary stray reactance resonates with the main component to cause “self resonance”. The larger the components are physically, the lower will be the self resonant frequency (SRF).

For capacitors, as the frequency increases beyond capacitor self-resonance the impedance of the capacitors in the circuit actually rises, so that the insertion loss begins to fall. This can be countered by using special construction for the capacitors. Similarly, inductors have a SRF beyond which their impedance starts to fall. Particular forms of winding geometry will reduce the internal winding capacitance and raise the SRF: the start and finish of the winding should be physically separate, and multiple winding layers should be treated particularly carefully. To obtain good performance at high frequencies, components must be chosen for a high SRF.

Several capacitor manufacturers offer software applications which give the frequency related data for their products. The usual frequency resonance formula can be used to determine SRF if you know both the component marked value and its parasitic:

F = 1/(2 · π · √ (L · C)) (1)

The common mode choke

A common mode choke is most easily described when it has two windings, one in each leg of a two wire circuit. The two windings are nominally identical and on the same core, which is often though not invariably toroidal, to minimise magnetic flux leakage.

The sense of the windings is such that differential currents, in which the “go” current in one wire is equal and opposite to the “return” current in the other, each create a magnetic flux in the core, but because they are equal and opposite the two fluxes cancel, leaving no net magnetic flux. Thus since the core is invisible the differential mode inductance is very small, being dominated by the residual difference between the windings, known as the leakage inductance.

By contrast the flux from common mode currents in the wires adds in the core, and therefore the full inductance of the choke is presented to common mode signals. So, the magnetic permeability of the core has maximum effect for common mode currents and negligible effect for differential mode currents

Chokes used in this way are sometimes known as “current-compensated” chokes, since as well as being invisible to differential signals, they can carry large values of low frequency or DC current without fear of core saturation and loss of inductance. Alternatively, by designing in a suitable amount of imbalance, the differential mode inductance can be tailored to provide some DM attenuation as well as CM, at the cost of a reduced current rating.

Common mode chokes can be constructed with multiple windings for filtering interfaces with many lines. As long as the fluxes due to the differential mode currents sum to zero in the core, the choke will operate correctly. Thus one winding on the core must be inserted in series even with the 0V rail, if this carries differential return currents. Crosstalk between the circuits will be determined by the balance of the windings, and their inter-winding capacitance.

In general, depending on the construction configurations of common-mode chokes, there are two types: sectional winding common-mode chokes (a) and bifilar winding common-mode chokes (b). One notable distinction between the two types is observed in their leakage inductance, with bifilar types achieving very low leakage inductance. Consequently, it's evident that bifilar types are particularly advantageous for high-frequency applications. Picture Courtesy: Wurth Elektronik

Ferrite chip impeders

A very common and widely available part using ferrite is the surface mount chip impeder(often referred to as ferrite beads or beads). It is made using the same principle as the multilayer chip capacitor but with a different interconnection of the metallization layers, so that they form a series rather than parallel connection, and with ferrite instead of ceramic dielectric. Ferrite chips come in the same size packages as ceramics (0402, 0603, 0805, 1206 etc) and can be used just as liberally. Typical applications are:

series impedance in power rails, to segment the power supply to different parts of the circuit

impedance in series with each wire of an interface port, to attenuate the RF interference passing through the interface

in clock lines, to reduce the higher order harmonics without too much degradation of the clock waveform edges

In fact, these parts can be used anywhere you would like to use a resistor, but can’t because of the DC or low frequency resistance that is introduced. Important parameters which determine their use are current rating and high frequency impedance, conventionally specified at 100MHz for comparison purposes. Typical general purpose parts exhibit 100 – 600Ω at 100MHz, but a wide range is available either side of this. With a range of parts for a given impedance at this frequency, the trade-off is generally that increased impedance below 100MHz results in a lesser impedance above it, or vice-versa. As with all ferrites (compare with cable sleeve ferrites for instance), their impedance has both resistive and inductive parts at high frequency and you need to check the manufacturer’s curves to apply them precisely; but for power supply decoupling, for instance, the impedance either side of the ferrite is so low that their selection is not critical. One aspect to be aware of is that because of their construction, they make good fuses: the current rating must be respected, including possible surge currents that might occur on power-up or power-down. And any DC bias current, as in a power distribution application, will reduce the available impedance, perhaps by several times, so again you need to check the data for this effect.

Three-terminal, X2Y and feedthrough capacitors

The quest for lower self-inductance for filter capacitors has led to some special types of construction.

Three-terminal

The unwanted lead inductance can be put to some use if the capacitor is given a three-terminal construction. The ground connection remains as a single terminal, but the signal line enters and exits through separate terminals. This construction can be implemented as either a leaded or surface mount part.

The lead inductance now forms a T-filter with the capacitor, greatly improving its high-frequency performance. Lead inductance can be enhanced by incorporating a ferrite bead on each of the upper leads. The ground terminal of the capacitor still has some residual inductance which limits its performance, and the ground connection must be made within the shortest possible distance. Also, because the signal or power line current passes through the device, its current rating must be respected. Even so, the 3-terminal configuration can extend the effectiveness of a small ceramic capacitor from below 50MHz to upwards of 200MHz and reduce its dependence on terminating impedances, which is particularly useful for interference in the VHF band.

X2Y

The X2Y component is another (but different) 3 terminal EMI chip device, which can be used either as a balanced line filter or a low-ESL decoupler.

When used in balanced line applications, this unusual design provides simultaneous line-to-line and line-to-ground (both differential and common mode) filtering, using a single ceramic chip. For unbalanced applications, it provides ultra low ESL (equivalent series inductance). It can replace several conventional parts, for balanced and unbalanced lines, twisted pairs and DC motor interfaces.

The component has a reduced inductance when compared to that of a conventional capacitor. This results from the particular internal electrode structure which inherently reduces the inductance by using the cancellation effect of opposing currents in close proximity (this principle is widely used elsewhere in EMC design - see for instance decoupling capacitor vias). The capacitance line to ground (common mode) is closely matched due to the symmetry within the design. As the device includes line to ground capacitance for both lines, any temperature, ageing and voltage effects will have an equal influence on both lines, therefore maintaining balance. Because the part is not a lead-through device, the current rating limitations of a standard 3 terminal chip do not apply.

Feedthrough

Any leaded capacitor is still limited in effectiveness by the inductance of its connection to the ground point. For the ultimate performance, and especially where penetration of a screened enclosure must be protected into the microwave region, then a feedthrough construction is essential.

Here, the ground connection is made by the outer body of the capacitor being screwed or soldered directly to the metal screening or bulkhead. Because the current can flow out from the central conductor over 360º around it, there is effectively no inductance associated with this terminal and the capacitor performance is maintained well into the GHz region. The inductance of the through lead can be increased, thereby creating a π- or T-section filter, by separating the ceramic metallization into two parts and incorporating a ferrite bead within the construction.

Feedthrough capacitors are available in a wide range of voltage and capacitance ratings but they are generally expensive and their cost increases with size. They are rarely used for EMC purposes in low-cost commercial products.