Why Cables Radiate: Shield Termination and the Pigtail Problem
Why cables act as antennas, how a cable shield works, and why a pigtail ruins it: transfer impedance, 360° termination, one end vs both ends, bench checks.
When a product fails radiated emissions, the cable is usually the antenna and the board is the source. In the 30 to 300 MHz range a 1 m cable radiates far more efficiently than anything on a typical PCB, and a shielded cable only helps if its shield is terminated properly.
Why a cable becomes an antenna
The signal current on a cable flows out on one conductor and back on another, so its fields largely cancel. The problem is common-mode current, which flows in the same direction on all conductors, including the shield, and returns through stray capacitance to the surroundings. It is driven by any voltage between the circuit reference at the connector and whatever the cable sees as its counterpart: the chassis, a second cable or the ground plane of the test site. Typical sources are ground voltage drops caused by fast return currents on the PCB, clock crosstalk onto I/O traces and a shield tied to a noisy circuit ground. Very little current is needed: in our earlier post on common-mode current, about 8 µA on a 1 m cable at 30 MHz reached the CISPR 32 Class B limit at 10 m.
Resonant cable lengths
A cable driven against the product's ground behaves roughly like a monopole and radiates most efficiently near the frequency where its length is a quarter wavelength, f = c / (4L). With c = 299 792 458 m/s, that is 149.9 MHz for 0.5 m, 74.9 MHz for 1 m, 50.0 MHz for 1.5 m and 37.5 MHz for 2 m. A cable between two units behaves more like a dipole, with its first resonance near a half wavelength. Insulation, nearby metal and routing shift these values. A broad emission peak that moves when the cable is rerouted or replaced with one of a different length points to the cable as the antenna.
How a shield works: transfer impedance
A shield does not stop common-mode current; it gives that current a path on its outer surface back to the chassis, away from the inner conductors. How well a cable achieves this is described by its transfer impedance Z_T, in Ω/m: the voltage induced between the inner conductors and the shield per metre of cable for each ampere flowing on the shield. A lower Z_T means less coupling, both for emissions (internal noise leaking out) and for immunity (external currents leaking in).
At low frequencies Z_T is close to the DC resistance of the shield. At higher frequencies a single braid gets worse because the magnetic field leaks through its openings, while a solid tube keeps improving thanks to the skin effect; foil-plus-braid and double-braid constructions are used when a low Z_T is needed at high frequencies. But Z_T describes only the cable. The connection between the shield and the chassis is in series with it, and in practice it is often the largest term.
The pigtail problem
A pigtail is the shield gathered into a wire, or a drain wire, that runs to a connector pin, a screw or a solder tag. Shield current must flow through it to reach the chassis, so its inductance sits in series with the shield path, and the voltage across it, V = I · 2πf · L, appears between the shield and the chassis. That voltage drives the outer surface of the shield like an unshielded conductor, so the cable radiates again; in an immunity test the same impedance lets external current couple into the circuit.
The inductance of a straight round wire is approximately L ≈ 0.2 · l · (ln(2l/r) − 0.75) µH, with length l and radius r in metres. A 25 mm pigtail of 1 mm diameter wire gives about 19.3 nH, with a reactance of 1.21 Ω at 10 MHz, 3.63 Ω at 30 MHz, 12.11 Ω at 100 MHz and 36.33 Ω at 300 MHz. Just 1 mA of shield current at 100 MHz therefore produces about 12 mV between the shield and the chassis, right on the outside of the cable. Shortening the pigtail to 10 mm (5.9 nH, 3.69 Ω at 100 MHz) helps but does not solve the problem; a 360° connection, where the current spreads around the full circumference over a very short path, has far lower inductance.
Termination methods compared
| Termination | How it is made | High-frequency performance | Notes |
|---|---|---|---|
| 360° backshell | Metal or metallised backshell clamps the braid all round; connector shell bonded to an unpainted panel | Best | Needs conductive contact all round; plastic backshells do not shield |
| Cable gland or saddle clamp | Exposed shield clamped to the chassis at the cable entry by an EMC cable gland or a metal P-clamp | Very good | Common for motor and drive cables; clamp at the entry point, not deep inside the cabinet |
| Short pigtail to chassis | Shield gathered into a short lead screwed to the chassis | Fair; degrades as frequency rises | Keep it as short as possible; mainly for low-frequency interference |
| Drain wire to a connector pin | Shield carried through a pin to the PCB ground | Poor; can be worse than no shield | Shield current flows into the circuit ground (the pin 1 problem) |
| Floating shield | Shield not connected at one or both ends | Poor at RF | The open shield end can act as an antenna |
One end or both ends?
The rule of connecting a shield at one end only comes from low-frequency analogue work: with both ends connected, a mains-frequency potential difference between two grounds drives current through the shield and can couple hum into a sensitive signal. At radio frequencies the rule fails. Once the cable is no longer short compared with the wavelength (a 1 m cable is already λ/10 at 30 MHz), the open end of a shield connected at one end behaves like an antenna and offers no return path for high-frequency common-mode current. For emissions and RF immunity, terminate the shield at both ends, 360° where possible. If low-frequency ground loops are a real concern, use a hybrid termination: one end direct, the other through a capacitor that is open at mains frequency and close to a short circuit at RF, provided its mounting inductance is kept low.
Connectors and the pin 1 problem
The shield should meet the chassis exactly where the cable enters the enclosure, so that its current returns along the outside of the box and never crosses the circuit. The pin 1 problem, a term from professional audio, describes the opposite: the shield is wired through a pin to the circuit ground on the PCB, so shield current flows through the board's ground and turns into noise or emissions. Board-mounted USB, Ethernet or D-sub connectors show the same fault when the shell is soldered to a PCB ground that is bonded to the chassis far away or through a long wire. Bond the board ground to the chassis right at the I/O connectors with short, wide connections, and remove paint or anodising at the contact areas. Unshielded cables have nothing to terminate, so put common-mode filtering, such as a common-mode choke, at the connector.
Bench checks
- Move or reroute the cable while watching the spectrum; if the peak changes, the cable is part of the antenna.
- Measure common-mode current with a current probe around the whole cable before and after each change, so that termination options are compared with numbers.
- Try a better termination quickly by wrapping copper tape with conductive adhesive, or fitting a metal clamp, from the exposed braid to the chassis.
Common mistakes
- Treating DC continuity as proof of a good termination. A 25 mm pigtail reads almost 0 Ω on a meter but over 12 Ω at 100 MHz.
- Clamping onto the non-conductive side of a foil shield. Many foils are metallised on one face only; the clamp or drain wire must contact the metal side.
- Running the pigtail or unshielded conductors several centimetres inside the enclosure, which couples internal noise straight onto the cable.
- Using a ferrite to cover up a poor termination. It may pass at one frequency, but the margin is often lost when the cable or layout changes.
Calculators used in this post
More posts
Results and summaries are for reference. For certification and test reports use the latest official standard text and calibrated instrument data.