Twisted pair inside metal enclosure: when it helps

Twisting a short signal-and-return pair inside a grounded metal enclosure can reduce magnetic pickup, but the improvement depends on the actual return path. If the conductors already run together, remain short and stay away from interference sources, the difference may be negligible. For a millivolt-level transducer operating at 120 kHz, first establish where return current flows. Then reduce the loop area and compare noise under controlled conditions. The grounded enclosure and the twisted pair address different coupling paths; neither makes the other unnecessary.

Why the return path decides the benefit

Magnetic pickup depends on the loop formed by the outgoing signal path and its return. Two nearby conductors form a smaller loop than a signal wire returning through a distant enclosure connection.

Twisting the pair reverses the orientation of successive small loops. In a reasonably uniform magnetic field, the induced voltages in those loops tend to cancel. Cancellation becomes less predictable near a small transformer, inductor or high-current conductor, where the field changes sharply with position.

The important question is not whether two wires look twisted. It is whether they carry the signal and its corresponding return current.

Consider a transducer connected to a coaxial bulkhead connector. The signal lead reaches the center contact, while the return eventually reaches the connector shell. Three arrangements behave differently:

• Dedicated signal and return leads: Both conductors run together from the transducer to the connector. Twisting can reduce their magnetic pickup.
• Signal lead with an enclosure return: The signal crosses the enclosure, while return current flows through the casing. Twisting the signal with an unrelated wire does not control that loop.
• Dedicated return plus casing connections: The return lead and casing provide parallel paths. Current divides according to their frequency-dependent impedances, so the twisted pair may carry only part of the return.

Do not remove an intentional transducer-to-case connection or protective bond merely to force a preferred routing arrangement. Follow the transducer's documented grounding requirements. For the broader grounding decision, use grounded versus floating metal enclosures for low-noise signals.

How much loop area can matter

For a simple loop in a uniform sinusoidal magnetic field, induced voltage increases with frequency, field strength and loop area. With the field perpendicular to the loop, the approximate relationship is:

Induced voltage RMS = 2 × pi × frequency × magnetic flux density RMS × loop area

This is an illustrative coupling calculation, not a prediction of voltage at the receiver. Circuit impedance, loading and field orientation also affect the measured result.

Take an approximately 2 in [50 mm] internal run with conductors separated by 0.4 in [10 mm]. Using the metric dimensions, the rectangular loop area is 500 square millimetres. A uniform 1 microtesla RMS field at 120 kHz would induce approximately 377 microvolts RMS around that loop.

Reduce the separation to 0.04 in [1 mm], keeping the same length. The area becomes 50 square millimetres, and the calculated induced voltage falls to approximately 38 microvolts RMS.

Those numbers compare two untwisted loops. Twisting can add cancellation, but connector terminations and untwisted ends leave residual loop area. A nearby, nonuniform field also limits cancellation.

The example explains why even a short connection can matter beside a strong interference source. It does not mean every enclosure contains a field of that strength.

Also distinguish the signal frequency from the interference frequency. A transducer operating at 120 kHz can encounter interference elsewhere in the spectrum. Interference near its operating frequency is especially troublesome because filtering may also remove the wanted signal.

Low-frequency magnetic shielding is a separate design question. Use enclosure shielding for 50 Hz noise when mains-frequency fields are the suspected source.

What a wiring comparison can prove

A controlled comparison can establish whether the internal wiring contributes meaningfully to the measured noise. It cannot establish the enclosure's shielding performance across all frequencies or operating conditions.

Make wiring changes only after isolating hazardous energy and verifying absence of voltage. Keep protective bonding and machine safety functions intact throughout the investigation.

Use the same transducer excitation, receiver settings, cable, load and mechanical conditions for each comparison. Keep the enclosure closed whenever its shielding is part of the test. Moving a lid or external cable between measurements can overwhelm the effect being investigated.

Compare these configurations where the documented circuit arrangement permits them:

• The existing routing, recorded before changes.
• The same signal and return conductors routed closely together.
• A twisted pair following the same route, without unnecessary extra length.
• The preferred pair moved farther from the suspected interference source.

Record the desired signal amplitude and the unwanted component separately. A reduced overall waveform amplitude is not automatically an improvement; it might indicate signal loading or changed transducer excitation.

Use measurement equipment with suitable sensitivity, bandwidth and electrical ratings. A long probe ground lead can introduce its own pickup loop. Use an appropriately rated differential or isolated measurement method when required by the circuit. Never disconnect an instrument's protective earth to obtain a quieter trace.

A repeatable improvement above the measurement variation supports keeping the change. No detectable improvement means only that twisting was not significant under those test conditions. It does not prove that the enclosure blocks every magnetic field.

Where the pair belongs inside the panel

Route the signal and its return together from the transducer terminals to the coaxial connector. Keep the separation small at both ends, rather than twisting the middle and leaving wide loops at the terminations.

Avoid adding length merely to obtain more twists. A short, close pair can be preferable to a longer twisted route that passes an interference source.

Give particular attention to these nearby components:

• Switching-power-supply transformers and inductors.
• Drive output wiring and motor conductors.
• Relay coils, contactors and switched inductive loads.
• High-current supply and return paths that separate around the enclosure.
• Switching nodes with fast voltage transitions near high-impedance signal inputs.

There is no universal separation distance that guarantees a millivolt-level noise target. Source current, switching speed, receiver impedance and geometry all matter. Move sensitive wiring away from the source as far as the layout reasonably allows, then verify the result.

Where a crossing is unavoidable, a near-right-angle crossing can reduce parallel exposure. It does not compensate for routing directly over a transformer or inductor.

Support the pair so closing the door cannot pull it apart or press it against power wiring. Follow conductor bend limits and avoid stressing fine transducer terminals with tight twists.

If the signal terminates on a PCB, continue the return-path review through the connector and board interface. The distinction between circuit reference and chassis bonding is covered in grounding a PCB inside a metal enclosure.

What twisting will not fix

Twisting does not turn a single-ended transducer connection into a balanced interface. A balanced receiver can reject some common-mode interference, but only when the complete source, cable and receiver arrangement supports that behavior.

Twisting also does not replace an electric-field shield. Keeping conductors together can help equalize coupling, but unequal impedances can convert coupled interference into a differential error.

A grounded conductive enclosure can reduce electric-field coupling. Its magnetic attenuation depends on material properties, wall thickness, frequency and construction, not simply the presence of a ground connection.

Check the coaxial transition as carefully as the internal pair. A poor connector-shell connection, a damaged shield or an external return-current problem can dominate the result. Twisting a short internal lead cannot repair those faults.

Do not confuse signal return, cable shield and protective bonding. They can meet at documented connection points, but their functions remain different. Protective bonding must not depend on a delicate signal wire or coaxial cable shield.

Enclosure seams and finishes can also affect shielding continuity. Use how paint affects EMI shielding when coatings interrupt intended conductive contacts. For interference at higher frequencies, RF leakage through enclosure holes and seams addresses a different limitation that twisting cannot solve.

What to specify on the drawing

Specify the electrical path and physical route, rather than writing only “twist wires.” The drawing should make the intended current loop unambiguous.

Choose the permissible noise from the measurement error budget. For example, a 10 microvolt unwanted component is 1% of a 1 millivolt signal when both use the same amplitude convention. That is an illustrative ratio, not a universal transducer specification.

If the layout needs a different enclosure footprint, use the stainless enclosure size finder to check available space. Leave room for connector access, strain relief and separation from interference sources. Material selection alone does not establish shielding performance.

For a short internal run, the practical default is simple: keep the real signal and return together, preserve required bonds, and avoid nearby interference sources. Retain twisting when it is easy to implement correctly or when a controlled comparison shows a useful improvement.

Enclosures for this job

Catalog sizes that suit the job this article answers, picked from its subject; open one for its full specifications.

Questions