Shielding and Bonding

Ask most people how to reduce interference and they will suggest shielding: put the circuit in a metal box and the problem will be contained. The advice is not wrong, but it is incomplete in ways that lead to expensive disappointments. A metal enclosure can provide 80 dB of shielding effectiveness — or almost none. The difference comes down to how the shield is built, how cables enter it, and how it is bonded to everything else.

Bonding — connecting metal enclosures, chassis, cable shields and ground references together with low-impedance connections — is the other half of the story. Even the best-shielded circuit will radiate and pick up interference if it is connected to poorly bonded cables that act as antennas. The cables are usually the weak point, not the enclosure. Treating shielding and bonding as one system, rather than two separate problems, is what produces quiet equipment and a quiet workbench.

The Purpose of Shielding

Electromagnetic shielding serves two purposes that are essentially mirror images of each other: preventing electromagnetic fields from entering a region (shielding from outside interference), and preventing electromagnetic fields from leaving a region (containing interference from a noise source). In practice you often want both — you want a sensitive input stage protected from external interference while the noisy parts of the same equipment, such as a switching supply or a processor, do not radiate into it.

Shielding works on both electric fields (E-fields) and magnetic fields (H-fields), but by different mechanisms and with different material requirements. Electric field shielding — the Faraday cage effect — works by redistributing charge on the outer surface of a conductive enclosure in response to an external electric field. The charge redistribution creates an internal field that exactly cancels the external field. Any conductive material works as an electric field shield, and the shield does not need to be thick to be effective against electric fields at low frequencies.

Magnetic field shielding is much harder. At low frequencies (below about 1 kHz), magnetic fields penetrate most materials easily, and effective low-frequency magnetic shielding requires high-permeability materials such as mu-metal (a nickel-iron alloy). However, at RF frequencies — the range where most interference problems occur — the skin effect transforms the situation dramatically. At RF, even ordinary copper and aluminum become effective magnetic shields because the induced eddy currents in the material create an opposing field that cancels the external magnetic field within the skin depth.

This is why simple aluminum project boxes and steel enclosures provide excellent shielding at HF and above, while they provide almost no magnetic shielding at audio frequencies. The shielding mechanism at RF is different from the shielding mechanism at audio — understanding this distinction prevents you from wasting money on exotic materials for RF work.

A well-bonded workbench: instruments and the project enclosure are strapped to one ground bus with short, wide conductors, and cables are bonded or filtered where they enter the enclosure.

View Larger

How Shielding Works: Skin Depth

The skin effect is the tendency of high-frequency current to flow in a very thin surface layer of a conductor rather than through its entire cross-section. As frequency increases, the depth of this surface layer — the skin depth — decreases. The skin depth formula is:

The practical implication is important: you do not need exotic materials or thick walls to achieve excellent RF shielding. A simple aluminum project box, a tin-plate enclosure, or even heavy aluminum foil provides more shielding attenuation from a few megahertz to the gigahertz range than any practical noise source requires — provided the enclosure has no gaps, slots, or poorly bonded seams. The shielding is limited by the imperfections in the construction, not by the material properties of the metal itself.

Apertures and Their Effect on Shielding

A perfect continuous metal enclosure provides near-infinite shielding at RF frequencies (limited only by material properties and wall thickness). A real enclosure has joints, seams, ventilation holes, panel meters, display windows, and connector openings. Each one of these is an aperture — a gap in the conducting surface — and each aperture can compromise the shielding effectiveness depending on its size relative to wavelength.

The key rule is this: a slot or gap becomes an effective radiator (or receiver) of electromagnetic energy when its length approaches half a wavelength at the frequency of interest. A slot antenna is a classic example — a half-wavelength slot in a conducting plane radiates efficiently. The same physics that makes slot antennas work makes slots in shielded enclosures dangerous to shielding performance.

The practical threshold is a slot length of λ/20 (one-twentieth of a wavelength). Below this length, the slot radiates or receives negligibly. Above this length, shielding effectiveness begins to degrade significantly. At HF frequencies (let us use 14 MHz, a wavelength of about 21 meters, as an example): λ = 300/14 ≈ 21 meters. λ/20 at 14 MHz is about 1.05 meters. Even a large slot in an HF enclosure would need to be over a meter long to cause significant problems at 14 MHz. This is why commercial radios rarely need extreme attention to HF shielding of the enclosure itself.

The situation changes radically at higher frequencies. At 433 MHz, a common frequency for remote controls and sensor links, λ = 0.69 m and λ/20 = 35 mm (about 1.4 inches). Now a 50 mm slot in an enclosure — perhaps a gap around a panel-mounted display — significantly compromises shielding. At 2.4 GHz, λ/20 is only 6 mm, so even the gaps between screws along a seam matter. This is why equipment that must be shielded at UHF and microwave frequencies uses closely spaced screws, conductive gaskets or conductive tape along every seam, while the same precautions are unnecessary below about 30 MHz.

Ventilation holes are a common source of shielding compromise. A single large ventilation hole degrades shielding much more than the same total area distributed as many small holes, because the shielding-compromising factor is proportional to the longest dimension of the aperture rather than its area. A ventilation array of 3 mm holes (spaced 6 mm apart) provides good ventilation while keeping each aperture well below λ/20 at VHF. A single 50 mm ventilation hole in the same enclosure would degrade VHF shielding significantly.

Panel meters, LCD screens, and other display windows require special treatment in well-shielded enclosures. Glass and plastic are transparent to RF. A panel meter window that provides a 60 × 80 mm opening in a shielded enclosure should be covered with fine conducting mesh (wire mesh with spacing much less than λ/20) to maintain shielding while allowing visual access.

Cable Penetrations: The Weak Point

In practice, the most significant source of shielding compromise in any enclosure is not the seams or ventilation holes — it is the cables. Every cable that enters or exits a shielded enclosure is a potential antenna that bypasses the shielding entirely. An unfiltered cable can deliver interference from outside directly to the circuits inside, no matter how perfect the metal walls of the enclosure are. Similarly, an unfiltered cable can conduct noise from a circuit inside directly to the outside environment.

The principle is that shielding works by keeping electromagnetic energy outside the enclosure — but if that energy is converted to conducted energy on a wire and brought inside through a hole in the shield, the shield has been completely defeated. This is not a subtle effect. A single unfiltered cable through a shielded wall can completely negate the benefit of even a very well-constructed shield.

The correct approach for cable penetrations depends on the cable type:

Coaxial cables: The coax shield must be bonded directly to the enclosure at the entry point, not just passed through a hole. A bulkhead connector mounted in the metal wall — a BNC, SMA or N-type chassis connector — bonds the shield to the enclosure all the way around at the exact point of penetration. Currents on the outside of the shield terminate at the wall and do not continue inside, while the signal inside the coax passes through unaffected.

Multi-conductor signal cables: Use feed-through filter capacitors at the entry point. These are capacitors with one lead connected to the conductor passing through the wall and the other lead bonded to the wall itself. They shunt RF energy to the wall (ground) without affecting the DC or audio-frequency signals on the cable. For critical applications, pi-section feed-through filters (inductor between two shunt capacitors) provide even better rejection.

USB, Ethernet, and control cables: These are harder to filter because they carry high-frequency digital signals that cannot be heavily filtered without corrupting the data. For USB, a ferrite common-mode choke at the entry point provides some suppression without affecting the differential-mode data signal. For critical shielding applications, fiber optic or galvanically isolated interfaces eliminate the cable penetration problem entirely by breaking the conductive path.

Bonding: Connecting Everything Together

Bonding is the act of creating a low-impedance electrical connection between two metal objects. In equipment and on a workbench, good bonding connects enclosures, chassis, cable shields and circuit ground references so that no significant potential difference exists between them at the frequencies that matter.

Why does a small potential difference matter? Because any potential difference between two connected pieces of equipment drives current along the cables connecting them, and those cables are also antennas. A cable carrying that current radiates, and the current flowing in its shield or ground conductor adds noise to the signals it carries. Even millivolts of RF difference between two chassis can produce noticeable interference on the audio, sensor or data cables between them.

The cure is to keep everything at the same potential by providing a low-impedance connection between them. If there is no potential difference to drive a current, no interfering current can flow.

A bond is only effective if its impedance is low at the frequencies of concern. This is where many grounding arrangements fail. A long piece of wire, even 2 mm (#12 AWG) copper, has self-inductance that makes its impedance significant at radio frequencies. A wire 1 meter long has roughly 1 µH of inductance; at 14 MHz its reactance is 88 ohms, and at 100 MHz it is over 600 ohms. Equipment “connected” through such a wire can still be at very different RF potentials.

The solution is short, wide, flat conductors rather than long wires. A flat strap has much lower inductance than a round wire of the same length, and the shortest possible connection matters even more. A common rule of thumb for RF bonding straps is a length no more than five times the width — a 50 mm (2 in) strap no longer than 250 mm (10 in). Inside equipment, the best bond is direct metal-to-metal contact: a circuit board ground fastened to the chassis at several points with short standoffs, or a connector shell screwed straight to the panel.

Ground Loops and How Bonding Prevents Them

A ground loop exists whenever there are two or more conductive paths between two pieces of equipment — typically a signal cable and a separate path through the mains safety earth, a shared power supply or another cable. The paths form a closed loop. Any changing magnetic field threading the loop induces a voltage in it, and any difference between the earth potentials at the two ends drives current around it. That current flows through the signal cable's shield or ground conductor and appears as noise in the signal.

The classic example: a computer and an audio amplifier are connected by an audio cable and both are plugged into mains outlets with safety earths. The two earths are connected through the building wiring, which carries small leakage and induced currents, so they are at slightly different potentials. The loop formed by the audio cable shield, the two mains earth connections and the building wiring picks up hum and drives it through the audio cable's ground. The result is the familiar 50 or 60 Hz hum and its harmonics in the speakers, often with buzz from the computer's own power supply mixed in.

The same thing happens on the workbench: a mains-earthed oscilloscope connected to a circuit that is also earthed through a USB cable to a computer forms a loop, and the noise it picks up shows on the scope trace even though it is not really in the circuit under test.

There are several cures. Bonding the equipment together with a short, low-impedance path, close to the signal cable, means the induced current flows mostly in the bond rather than in the signal cable. Balanced (differential) connections reject the voltage that remains. Breaking the loop in the signal path — with an audio isolation transformer, a USB isolator or an optical link — removes the problem entirely. What you must never do is break the loop by removing the mains safety earth.

RF Bonding vs Safety Earth

The difference between the safety earth and an RF bond causes confusion even among experienced engineers. They serve different purposes and are implemented differently, even though they are usually connected together.

The safety earth is the green or green-and-yellow conductor in a mains cord. Its purpose is entirely about personal safety: it connects exposed metal to the earth conductor of the building wiring, so that if a live conductor touches the chassis because of a fault, the fault current flows to earth and trips the fuse or breaker rather than flowing through a person. Electrical codes require it on all equipment that is not double insulated. It runs wherever the mains wiring runs, often tens of meters, before reaching the service panel.

An RF bond is a short, low-impedance connection between chassis, cable shields and ground references. Its purpose is to keep them at the same potential at high frequencies. It must be short and wide, and it cannot follow the mains wiring through the walls, because the inductance of a long path makes it useless at RF.

The two systems work together: the safety earth handles fault current and low-frequency safety, and RF bonds handle high-frequency noise. Adding RF bonds never replaces the safety earth, and a safety earth conductor is never a substitute for an RF bond.

Common Bonding Mistakes

Knowing the right approach makes common mistakes easy to recognize:

Long wire bonds. A piece of ordinary hookup wire between two chassis looks like a bond but is not an effective one at RF. Its self-inductance leaves the two chassis at different RF potentials. Use the shortest, widest connection possible.

Pigtail shield connections. A cable shield twisted into a pigtail and soldered to a ground pin has enough inductance to make the shield almost useless above a few megahertz. Terminate shields all the way around, with a metal connector shell or a cable clamp.

Daisy-chain grounding. Connecting equipment in a chain, each grounded only through the next, creates a long, high-impedance path, and the equipment at the end of the chain is effectively not bonded at RF. Use a star or a ground plane: each item connected by its own short path to a common point or a common metal surface.

Painted or anodized joints. Paint and anodizing are insulators. Enclosure panels that are screwed together over painted or anodized surfaces may have no reliable electrical contact at all. Mask the paint at mating surfaces, use conductive finishes such as chromate conversion coating, or use serrated washers that cut through to bare metal.

Circuit ground connected to the chassis in the wrong place. If a circuit board's ground connects to the enclosure only at one point far from the cable connectors, noise currents from the cables must flow across the board to reach the chassis. Bond the circuit ground to the chassis at the connectors, where the cables enter.

Bonding Products, Racks and Vehicles

Products and enclosures. In a well-designed product, every connector shell is bonded directly to the metal enclosure, the circuit board ground connects to the enclosure at the connector panel, all panels are bonded to each other along their seams, and cables entering the box are filtered or shielded at the point of entry. Together these make the enclosure, the cable shields and the circuit ground one continuous shield.

Equipment racks and workbenches. In a rack, each chassis should be bonded to the rack rails with conductive mounting hardware or short straps, and the rack bonded to the building's protective earth. On a workbench, a ground plane — a sheet of copper-clad board or aluminum under the circuit being tested — gives a low-impedance reference to which instruments and test circuits can be bonded with short leads.

Vehicles. In cars, boats and RVs, the metal body or a common ground bus is the bonding reference. Electronics should connect their ground directly to the battery negative or a main ground point with short, heavy conductors, and noise sources such as alternators, fuel pumps and ignition systems should be suppressed at the source. Many vehicle noise problems are caused by corroded or loose ground straps between the engine, body and battery.