September 19, 2026
5G Base Station EMI: Why Small Cells Are a Different Filtering Problem
Filtering for 5G base station EMI was designed around the cabinet. The radios have since moved onto ceilings and lamp posts, power now arrives over the data cable, and the fan is gone. Those three changes move where the noise escapes, and with it the place where a feedthrough capacitor earns its keep.
The Numbers Behind the Squeeze
Small Cell Forum’s 2026 forecast, built from a survey of 103 deployers, puts small cell deployments and upgrades on a 9.1 percent CAGR between 2024 and 2031. Annual shipments reach 9.8 million units by 2031, and cumulative shipments over the period total 67.2 million.
| Metric | Figure |
|---|---|
| Deployments and upgrades, 2024 to 2031 | 9.1% CAGR |
| Annual shipments by 2031 | 9.8 million units |
| Cumulative shipments to 2031 | 67.2 million units |
| 5G Standalone share of deployments | 11% in 2024 to 89% by 2031 |
| Indoor share of cumulative deployments | 54% |
Two things follow from that shape. Most of these radios end up indoors, mounted on a ceiling or a wall close to the people using them. And they arrive in numbers that make a single compliant unit only the starting point, because what a receiver downstream experiences is the sum of everything near it.
A Small Cell Is Not a Small Macro Cell
The comparison that matters has less to do with power rating than with where the power comes from and what happens to the heat.
| Macro site | Small cell | |
|---|---|---|
| Power delivery | Dedicated DC feed, routed separately from signal | Power over Ethernet, sharing the data cable |
| Cooling | Forced air, cabinet volume to spare | Sealed, convection only |
| Enclosure | Rack cabinet with few penetrations | Shoebox-sized, with connectors and vents taking a large share of the surface |
| Dominant escape path | Apertures and seams | The cable itself |
The power change is the one that catches people out. A macro site feeds DC down a dedicated pair and keeps signal on fibre, so the two paths can be filtered separately. A small cell often receives everything over one Cat5e or Cat6 run. Power over Ethernet Type 4 carries up to 90 W, which on a 48 V rail means roughly 2 A spread across the pairs, and that same cable is the one carrying the Ethernet signalling.
Cooling matters for a duller reason: a sealed enclosure running warm has no way to shed the heat, and capacitor life falls quickly with temperature. A part with a comfortable margin in a fan-cooled cabinet may sit far closer to its limit behind a ceiling tile.
Where the Noise Leaves the Box
Disturbance leaves a radio unit by two routes. It radiates through openings in the chassis, or it conducts out along cables. In a large cabinet the first route dominates. In a small cell the second one does.
The reason is common mode current. Signal currents travel in opposite directions on a pair and their fields largely cancel. Common mode currents travel the same direction on every conductor, including the shield, and the whole cable then behaves like a single antenna. A patch cable a few tens of centimetres long approaches a wavelength at 1 GHz, which is where a poorly terminated line radiates most effectively.
| Source inside the unit | Where it tends to leave |
|---|---|
| Switching supply, 100 kHz to a few MHz | Power pair and its shield, in common mode |
| Ethernet PHY clock and its harmonics | Data pairs and the cable shield |
| Power amplifier harmonics | Antenna feed and chassis seams |
| Digital control bus | Control lines leaving the enclosure |
Every one of those paths crosses the enclosure wall at some point. That crossing is the design problem, and it is the reason a feedthrough capacitor belongs at the bulkhead rather than on the board behind it.
Why a Feedthrough Capacitor, and Why There
Shielding a box is straightforward until something has to pass through the shield. Cut a hole for a cable and the hole radiates; the better the rest of the enclosure, the more that one opening matters.
A feedthrough capacitor solves both halves at once. The conductor runs straight through the middle of the capacitor, and the counter electrode is contacted concentrically to the case, so the part mounts into the shielding wall and the line passes through it. Noise on that line is bypassed to the chassis at the exact point where it would otherwise have escaped.
The concentric geometry also keeps parasitic inductance low, which is what separates a feedthrough part from an ordinary capacitor at high frequency. A surface mount capacitor with its pads and traces reaches self resonance and becomes inductive above it, so attenuation falls away. The coaxial construction pushes that limit much higher. Published curves for one common threaded part, the TDK B85321A2945A501, show 25 dB at 10 kHz, 40 dB at 100 kHz, 90 dB at 1 MHz and 100 dB from 10 MHz through 1 GHz. The exact figures depend on the part, but the shape of the curve is typical.
Mounting is not optional detail. The part has to seat into a threaded hole or bushing with solid metal-to-metal contact around its full circumference. A feedthrough capacitor held in place by a loose nut, or sitting on a painted surface, loses much of what it was bought for.
Xuansn makes threaded feedthrough capacitors in the M3 to M5 range that screw directly into a bulkhead, along with high power feedthrough capacitors for lines carrying larger currents.
Specifying One for a Radio Unit
| Parameter | What to check on a compact radio |
|---|---|
| Capacitance | Higher values attenuate more at low frequency but raise leakage current, which matters on a mains-referenced line |
| Current rating | Cover the full PoE load, not the nominal draw; Type 4 reaches roughly 2 A total |
| Voltage rating | Sized above the highest transient on the line, not just the 48 V rail |
| Insertion loss | Read the curve at the frequencies your unit actually emits, which for a switching supply means the low end matters as much as the high end |
| Thread and body size | Has to match the wall thickness and the space around the connector |
| Temperature rating | Derate for a sealed enclosure running above ambient |
The mounting style guide covers how threaded, solder-in and press-in parts differ in installation, and the introduction to feedthrough capacitors explains the construction if this is the first time you are specifying one.
What Comes Next in This Series
This is the opening piece in a short series on EMI filtering for 5G radio equipment. The next one looks at where filtering sits in a typical small cell power architecture, from the PoE entry point through to the radio front end, and which of those points actually justify a bulkhead part.
Common Questions
Can an MLCC replace a feedthrough capacitor?
Not at the bulkhead. An MLCC mounted on the board filters noise that has already entered the enclosure, and its pad inductance limits how well it works at the frequencies a radio emits. A feedthrough part filters at the point where the line crosses the shield.
Does a feedthrough capacitor still work at mmWave frequencies?
The concentric construction keeps attenuation useful well into the GHz range, and published curves for threaded parts often extend to 1 GHz and beyond. How far it holds depends on the specific part and on the quality of the bulkhead contact.
How does the part need to be grounded?
Through its own body, into the chassis. That is what the threaded mounting is for, and it is why metal-to-metal contact matters more than torque figures. Large capacitance values also bring leakage current requirements, so check the applicable safety specification for your application.
Need feedthrough capacitors for a 5G radio design? Xuansn Capacitor manufactures threaded, solder-in and high power feedthrough capacitors for EMI filtering – our engineers will reply within 24 hours with feasibility and a quote.
✉ coco@xuanxcapacitors.com | ☎ +86-769-8166 8821 | Send your specification