Earthing and bonding for control panels and instruments

The short answer

Protective earth is a safety conductor: it carries fault current until the protective device operates, reaches every exposed conductive part of the enclosure, and has its continuity proven by test under IEC 60204-1 and IEC 61439-1. Functional or instrument earth is a signal reference: it carries no fault current by design and exists so every analogue and serial circuit agrees where zero is. A compliant installation has only one earth system, so the two are bonded together, but they meet at exactly one point. Join them at two and you have built a loop, and the current circulating in it flows through your measurement.

Two jobs, one system

The protective earth requirement is impedance, low enough that fault current operates the device in the required time, with continuity that survives the panel being opened, transported and worked on. The functional earth requirement is potential equality: every reference point sits at the same potential and no stray current passes through it.

What is never acceptable is a separate earth electrode for the clean earth. IS 3043 and IEC 60364 both require all earths at an installation to be bonded. An isolated instrument electrode sits at a different potential from the rest of the plant during a fault, which makes the instrument enclosure live relative to everything a person can touch at the same moment. Clean earth means a separate bar with a single tie, not a separate rod.

Why a second bond makes a loop

Two points on a plant earth system are not at the same potential. A large motor start, a fault elsewhere on the board, or harmonic current circulating in the PE network routinely puts hundreds of millivolts to a volt or two between points thirty or fifty metres apart. Bond a screen at both ends and that voltage appears across the screen, the screen carries current, and coupling between screen and pair puts a fraction of it in series with your signal.

The scale is worth working out once. A 4-20 mA signal across a 250 ohm input is 1 V to 5 V, so the span is 4 V and one per cent of span is 40 mV at the card. You do not need much of a volt to move a reading by a per cent. That is why the classic symptom is a loop that is steady all night and wanders when the plant starts up.

Screens: one end, and which end

Earth the screen where the receiving circuit's reference already lives, and at the far end cut it back, insulate it and tape it rather than cutting it flush, so the next person can see the decision was deliberate.

For a two-wire 4-20 mA transmitter powered from the panel, the reference is the panel 0 V, so the screen lands on the instrument earth bar in the panel and the field end floats. The exception is a sensing element already earthed in the field, such as a grounded-junction thermocouple or an instrument whose screen is bonded internally to its case. That circuit is already referenced at the field end and must not be earthed again in the panel. The rule is not "always the panel end", it is one point, at the reference.

Pair screens and any overall screen land on the same bar, and must not touch each other, the enclosure or the armour on the way. Armour is protective earth and is bonded at both ends through the glands. The screen is not the armour.

Referencing a 4-20 mA loop

A current loop is referenced to 0 V in exactly one place, normally the negative of the loop supply, which is usually the analogue card's common. Earth the transmitter negative in the field as well and earth now sits in parallel with the return conductor: part of the signal current goes home through the earth system, the card reads low, and the error moves with plant conditions.

Two practical points. On a non-isolated analogue input card every channel shares one common, so a single field earth on one channel disturbs every channel. Channel-to-channel isolation removes the constraint, group isolation does not, and it is worth knowing which you have before designing. And on a HART loop, screen-loop noise lands in the 1200 to 2200 Hz band where the FSK signal lives, so the first symptom is often a handheld that cannot see the device. For the diagnostic sequence, see 4-20 mA loop troubleshooting.

The panel: bars, bonds and continuity

• A copper PE bar bolted to the backplate, one conductor per bolt, star topology. No daisy-chaining several conductors under one screw.
• A separate instrument earth bar on insulators, collecting screens and analogue commons, tied to the PE bar by a single labelled conductor. That conductor is the star point.
• The 24 V control supply 0 V bonded to PE at one place, normally at the power supply, on its own terminal group rather than mixed with analogue commons.
• Doors, hinged gland plates and swing frames bonded with a flexible strap. IEC 61439-1 requires effective continuity of the protective circuit to every exposed conductive part, and hinges and slide rails are not that path.
• Serrated washers or a scraped land under every bonding point, because a powder coat is a good insulator.
• DIN rail is not an earth conductor. Earth terminal blocks that clip to it are fine, and the rail still needs its own bonding conductor.

IEC 60204-1 verifies the protective bonding circuit by injecting at least 10 A at 50 or 60 Hz from a PELV source between the PE terminal and points of the bonding circuit, then comparing the voltage drop against limits tabulated by conductor cross-section and length. The standard specifies a real current because a loose ring lug, or a screw that never broke through paint, passes a continuity buzzer at a few milliamps and fails at 10 A. Test every door, gland plate and earthed device body, record it, and test again after transport and installation, because that is when straps go missing.

Earthing a VFD, and the one case where both ends are earthed

A drive's earth is a high-frequency component. What has to get back to the inverter is common-mode current at the switching frequency and its harmonics, from tens of kilohertz to tens of megahertz, and at those frequencies conductor inductance dominates while resistance is irrelevant. A long round conductor is a poor bond however well it measures on a DC meter. Use a wide flat braid, as short as the layout allows, from the drive's earth stud to a bare unpainted mounting plate, and bond that plate to the PE bar.

The screened motor cable is then glanded 360 degrees at both ends: an EMC gland at the drive, and an EMC gland or full circumferential clamp at the motor terminal box. A pigtail defeats it, because a hundred millimetres of wire is a significant impedance at ten megahertz.

This does not contradict the single-end rule, because the two screens do opposite jobs. A signal screen is earthed once to keep unwanted current out of a millivolt-scale circuit. A motor cable screen is a deliberate return path that brings the inverter's common-mode current straight back to the inverter, and a return path needs a connection at both ends to exist at all. The circulating current a two-end bond causes is meaningless on a cable already carrying 415 V and tens of amps. The corollary is the useful part: pigtail the drive screen and that current still gets home, it just routes through the plant earth, the panel and your instrument cables on the way.

What goes wrong

Analogue readings that drift when plant load changes. Almost always a screen earthed at both ends, or a transmitter negative earthed in the field. Before rewiring, clamp a milliamp-capable clamp meter around the screen at the panel gland, because current in a conductor that should carry none is the whole answer. Or lift the screen off the bar and measure between it and the bar: volts means a loop.

Communication errors that change when somebody touches a cable. RS-485 with the screen bonded at both ends, or with no signal common at all, so the two ends' references drift apart until the drivers run out of common-mode range. Touching the cable shifts a marginal pigtail or changes capacitance to earth, which is why the fault appears to follow the technician. RS-485 needs three conductors: A, B and a common.

Nuisance RCD trips on drives. EMC filter capacitors and the capacitance of a screened motor cable produce a standing earth leakage current before anything is wrong, scaling with cable length and switching frequency, so several drives behind one 30 mA device trip on the sum. More seriously, a drive can produce smooth DC residual current, which can blind a type AC or type A device so it fails to trip when it should. Where a drive sits on an RCD-protected supply, use a type B device unless the drive manufacturer states type A is sufficient for that model, give each drive its own device, keep the motor cable short, and reduce switching frequency if leakage is marginal.

A thermocouple that jumps when a heater switches. Grounded-junction thermocouple, and the compensating cable screen earthed in the panel as well. Pick one.

What to learn next

Earthing decisions only survive if the wiring discipline around them does, so read control panel wiring standards for segregation and labelling, then VFD parameter setup, since the settings that govern leakage current and emission are the other half of the drive earthing problem.