small cell EMI filter selection - silver threaded feedthrough capacitor mounted through an aluminium panel, showing the ratings that matter

October 12, 2026

Small Cell EMI Filter Selection: Which Feedthrough Capacitor Ratings Matter

Choosing a feedthrough capacitor for a small cell looks like a long datasheet exercise, but the decision really comes down to four numbers: the insertion loss you need, the current the interface carries, the capacitance that delivers that loss at the frequencies you care about, and the voltage the line has to survive. Get those four right and the rest, package and mounting style, follows. Get the capacitance wrong on its own and the part will still measure correctly on a bench while doing very little where the noise actually is.

Insertion Loss Is the Number Everything Else Serves

Insertion loss is defined as the ratio of the voltage across the load without the filter to the voltage across the load with it, expressed in decibels:

IL(dB) = 20 log (E1 / E2)

That figure only means something if you know the source and load impedance it was measured at, which is why feedthrough capacitor datasheets quote insertion loss for a 50 ohm system. Real circuits are rarely 50 ohm, so the published curve is a comparison tool rather than a promise. Where the impedances are known, the relationship is:

IL(dB) = 20 log [ 1 + ZsZl / Zt(Zs + Zl) ]

with Zt the transfer impedance of the capacitor. The practical takeaway is that attenuation improves when the source and load impedances are high relative to the capacitor’s impedance, which is why a feedthrough capacitor placed at a high impedance interface outperforms the same part in a low impedance one.

Deciding How Much Attenuation You Actually Need

Small cells sit in the same spectrum they are trying to use, so the filtering requirement is set by the radio band rather than by a generic EMC target. Attenuation in the cellular bands matters most, and the 450 MHz region upward is where a badly filtered interface starts coupling into the radio.

A practical target for an interface like this is at least 30 dB of attenuation across the relevant mobile band, measured in the interface’s own impedance rather than in 50 ohm. Below that threshold, cable-borne common mode current re-radiates from the enclosure and the radio hears it directly.

Typical published performance for parts in this class runs across a wide range, and comparing them tells you most of what you need:

Insertion lossCapacitanceVoltageCurrent
20 dB at 100 MHz1 nF200 V DC5 A
40 dB, 0.3 to 3500 MHz4.3 µFlow voltage2 A
70 dB at 1 GHz1.2 µF80 V10 A
30 dB, 4 to 6 GHz22 pF50 V1 A

Note how the last row reaches useful attenuation in the microwave bands with only 22 pF, and how the 4.3 µF part reaches its 40 dB only at low voltage. Capacitance is not a proxy for performance.

Capacitance and Self-Resonance

This is where most selection mistakes happen. A feedthrough capacitor attenuates by shunting noise to the enclosure wall, and its effectiveness at any frequency depends on its impedance there. Impedance falls with capacitance, which suggests a bigger part is always better, until self-resonance gets in the way.

Every capacitor has an equivalent series inductance, and at some frequency that inductance resonates with the capacitance. Above the self-resonant frequency the part behaves as an inductor and its impedance rises again, so attenuation in the band above resonance gets worse, not better. A larger capacitance pushes the self-resonant frequency down, which can put the useful attenuation band below the noise you are chasing.

Feedthrough construction helps here because the centre conductor passes straight through the ceramic, which keeps the series inductance far lower than a conventional leaded or SMD capacitor of the same value. That is the whole reason these parts are used on enclosure walls. It does not, however, make capacitance irrelevant: pick a value whose self-resonance sits above the interference band you are filtering, rather than picking the largest value that fits.

Current Rating Follows the Interface

The current rating has to cover the worst case the interface carries, not the average. For a small cell, the interfaces divide neatly:

  • PoE input, which carries both data and power on the same pair, typically under an amp per port but with a surge requirement on top
  • DC power feed, which may carry several amps
  • Antenna feeder, where current is small but RF voltage can be significant
  • Control and alarm lines, usually a few hundred milliamps at most

Parts in this class are commonly rated 1 A to 10 A, with the higher figures tied to the high power threaded and tubular ranges. Choose against the interface, and if a line can see a fault current, size for that rather than for normal operation. A capacitor that fails open leaves the interface unfiltered; one that fails closed on a power line is a much worse outcome.

Voltage Rating and the Surge Case

Voltage is usually the easiest of the four to set, provided the surge case is included. Working voltages on these interfaces are modest, but a feedthrough capacitor on an outdoor or tower-mounted unit can see induced surges, and the part sits between the conductor and the enclosure ground, which is exactly where a surge wants to go.

Rated voltages of 50 V, 100 V and 200 V DC cover most small cell interfaces. Where a line leaves the enclosure, size the capacitor for the surge the installation standard requires, and remember that the higher capacitance values are generally only available at the lower voltage ratings. That constraint often decides the selection on its own.

C-Type or Pi-Type

A C-type feedthrough is a single capacitor from the conductor to the wall. It has the fewest parts, the lowest series inductance and the highest self-resonance, which makes it the natural choice where the noise band is high and the required attenuation is moderate.

A Pi or T network adds inductance either side of the capacitor, trading a higher component count and a lower self-resonant frequency for more attenuation at lower frequencies. Where the interference is broadband and starts below the cellular band, the extra elements are what buy the low end. Where the problem is a microwave band, the C-type part is usually the better tool.

Dielectric Choice Sets the Cut-Off

The ceramic dielectric determines how much the capacitance moves over temperature and applied voltage, and since the cut-off frequency depends on capacitance, the filtering moves with it. C0G, also written NP0, holds its value tightly and keeps the attenuation curve where it was designed. X7R and the Y5 series give more capacitance in the same body but drift with temperature and bias, which shifts the cut-off in service.

For a filter that has to hold a specified attenuation over an outdoor temperature range, choose the dielectric first and the capacitance second. Where the requirement is simply to knock down a band with margin to spare, a wider dielectric is usually acceptable.

Putting the Four Numbers Together

A worked case. A small cell has a PoE input carrying 0.7 A, an operating voltage of 48 V with a surge requirement of 200 V, and a radio that has shown desensitisation from cable-borne noise across 700 MHz to 2.7 GHz.

The attenuation target sets at least 30 dB in that band, which rules out very large capacitance because self-resonance would land inside the band. The voltage requirement sets a 200 V rating, which limits the available capacitance. Within those two constraints, a C-type threaded feedthrough in the low nanofarad range, with a C0G dielectric so the curve does not drift, covers the band. The current rating only has to clear 0.7 A, so a 5 A part has ample margin. The four numbers resolved to: 30 dB minimum, 5 A, a few nanofarads, 200 V.

small cell feedthrough capacitor selection - threaded feedthrough capacitors of different sizes on a workbench mat

Xuansn manufactures the threaded, tubular and high power feedthrough capacitors used on interfaces like these, across the threaded feedthrough capacitor range, with insertion loss curves measured in 50 ohm and available on request. If you are working through the four numbers for a specific interface, send the band, the current and the standing voltage to coco@xuanxcapacitors.com, or contact us here, and we will match a part to them.

FAQ

Is a bigger capacitance always better for EMI filtering?

No. Larger capacitance lowers impedance but also lowers the self-resonant frequency. Above that point the part behaves inductively and attenuation falls away, so an oversized capacitor can be worse than a correctly sized one in the band you are trying to filter.

What insertion loss should a small cell interface have?

Around 30 dB or more across the cellular band the radio actually uses, measured at the interface impedance rather than in a 50 ohm test fixture. The exact number should come from the radio’s own desensitisation measurements rather than a generic target.

Why do feedthrough capacitors attenuate better than ordinary capacitors?

Because the conductor passes straight through the ceramic, the series inductance is far lower than a leaded or surface mount part of the same capacitance. Lower inductance means a higher self-resonant frequency and useful attenuation further up the spectrum.

Do I need a Pi filter instead of a C filter for a small cell?

Only if the interference extends well below the cellular band. A C-type part is simpler, has higher self-resonance and handles the microwave bands better. A Pi network adds attenuation at lower frequencies at the cost of a lower self-resonant frequency.

Leave A Comment