A panel can be wired correctly to the schematic and still produce a flowmeter reading that wanders whenever a motor accelerates. The cause is rarely a faulty component. It is the path the drive gives its switching currents back to the source, and that path is decided by the layout, not by the circuit diagram.
A customer once sent us a panel with a complaint that reads like a riddle. The line ran perfectly on manual. In automatic, the analogue level signal from a tank jumped by a few per cent every time the mixer drive ramped up, a servo axis logged encoder faults roughly once a shift, and a residual current device tripped about once a week for no reason anyone could reproduce. Three symptoms, three different suppliers blaming each other, and a schematic in which every wire was exactly where it belonged.
None of it was a component fault. All three symptoms came from the same thing: the high-frequency current a variable-frequency drive pushes out of its output terminals had no proper way home, so it found its own way through the panel, through the signal wiring and through the earth system. That path is not on the circuit diagram. It is decided by the physical layout, by the cable types and by how the shields are terminated, which is why a panel can be wired correctly and behave badly.
What a drive actually puts on the motor cable
A modern drive builds its output from voltage pulses switched by IGBTs with edges of a few kilovolts per microsecond. The motor sees an average that behaves like a sine wave; the cable sees the edges. Every one of those edges drives a current into the capacitance between the motor cable conductors and their shield, and between the motor windings and the motor frame. A shielded motor cable has something like 150 to 300 picofarads per metre, so twenty metres of cable is several nanofarads, and a steep edge into a few nanofarads is an ampere-scale pulse with energy at frequencies in the megahertz.
That current is not a fault current. It is a normal, designed-in consequence of switching, and it must return to the drive from which it came. If a low-impedance route back exists, it takes it, stays within a metre or two of wiring and bothers nobody. If it does not, the current returns through whatever is available: the machine frame, the conduit, the PE conductor shared with the sensor supply, the shield of a measurement cable, the encoder cable of the neighbouring servo. Each of those paths has impedance, and current through impedance is voltage. That voltage appears exactly where it is least wanted, as a shift in the reference of an analogue input or as noise across the receivers of an encoder.
The single most important detail: how the shield is terminated
The route home is the shield of the motor cable, and it only works if the shield is bonded at both ends over its full circumference. A pigtail, where the braid is gathered into a short wire and landed on a terminal, has enough inductance to be close to an open circuit at megahertz frequencies, and it is the defect we find most often. The correct termination is an EMC gland or a shield clamp that presses the braid all the way around against a clean, unpainted metal surface, at the drive end and again at the motor terminal box. A symmetrical shielded cable, with three phase conductors and three symmetrically arranged protective conductors inside the shield, is worth its extra cost on any cable of length, because it gives the common-mode current a balanced path and reduces the voltage impressed on the motor frame.
Signal cables follow a different rule, and confusing the two causes its own problems. A measurement cable carrying a slow analogue signal is normally bonded at one end only, at the cabinet, so that no power-frequency current can circulate in the shield between two points of the plant that sit at different earth potentials. Fast signals, encoder and fieldbus cables are bonded at both ends, because for them the high-frequency screening matters more than the earth loop, and their transceivers are designed for it. When a cable must be earthed at one end for hum and at both for high frequency, the usual compromise is a hard bond at the cabinet and a small capacitor at the far end.
The panel is part of the circuit
Inside the enclosure, the mounting plate is not a shelf. It is the high-frequency bonding plane that ties the drive, its filter, the shield clamps and the PE bar together with the shortest possible paths. That means bare metal where components mount, paint scraped or contact washers used under every stud, and short flat straps rather than long round wires wherever a bond carries high frequency, because a strap has a fraction of the inductance of a wire of the same cross-section. The drive, its EMC filter and the cable entries belong close together; a filter connected to its drive by a metre and a half of wire routed across the panel has already lost most of its effect.
Separation inside the trunking follows from the same logic. Motor cables, braking resistor cables and mains cables run in their own duct; signal, encoder and fieldbus cables run in another, ideally 200 mm away, and where they have to cross they cross at a right angle rather than running side by side. A metre of parallel run inside a duct couples more noise than any filter later removes. The rule survives in the field only if the person building the panel understands why it exists, which is why we draw the routing on the layout rather than leaving it to the wiring technician on the day.
Bearing currents and the motor end
The same common-mode voltage that upsets signals also appears across the motor bearings. When the oil film breaks down, the discharge pits the race, and over months the pitting develops into the regular fluting pattern that turns a quiet motor into a noisy one and then into a failed one. The defence, in order of cost: proper symmetrical shielded cable with 360 degree terminations, which removes much of the driving voltage; then a common-mode choke on the drive output; then a shaft earthing ring to give the current a path that does not go through a bearing; and on larger machines an insulated bearing at the non-drive end. Motors that fail bearings every year on a drive and never on a direct-on-line starter are not suffering from bad bearings.
Long motor cables bring a second effect. The cable behaves as a transmission line, and the mismatch at the motor end reflects the switching edges, so the motor can see voltage peaks approaching twice the DC bus. Standard inverter-duty insulation tolerates this over normal distances, but beyond the length the drive manufacturer states, a dv/dt filter or a sine filter belongs in the design. The same length increases leakage current, and leakage current is what trips residual current devices that were never chosen for the job: where a drive is involved, a Type B device is required, because the residual current is not a clean alternating waveform that a Type A device can see correctly.
Compliance is an installation property, not a part number
IEC 61800-3 classifies drive installations into categories C1 to C4 according to the environment they are intended for, and drive makers state which category their built-in filter achieves and under what conditions: a maximum motor cable length, a shielded cable, a specified switching frequency. Those conditions are part of the declaration. A machine that uses an unshielded motor cable, or doubles the stated length, or raises the switching frequency during commissioning to make a motor quieter, is no longer inside the conditions under which anyone tested anything, whatever the drive datasheet says.
This matters commercially as well as technically. When a machine is exported into the European market, its EMC declaration rests on the installation being what the documentation says it is. We therefore write the cable types, the maximum lengths, the termination method and the switching frequency into the machine manual, and we check those points during the factory acceptance test rather than discovering at a customer site that somebody substituted an unshielded cable because it was in stock.
The checks that find these faults quickly
When a panel arrives with symptoms like the ones in the first paragraph, the sequence is short. Look at the motor cable first: is it shielded, is the shield clamped over its circumference at both ends, is there a pigtail anywhere. Look at the routing: does the motor cable share a duct with signals. Look at the mounting plate: is it painted under the drive and under the shield clamps. Measure the resistance from the drive body to the PE bar and from the PE bar to the machine frame; anything much above a tenth of an ohm is a bond to remake, and a low resistance reading still does not prove a low high-frequency impedance if the connection is a long thin wire. Then reproduce the symptom while the drive ramps, because a fault that appears only during acceleration has already told you where it lives.
In the case of the panel with three symptoms, the repairs were unglamorous: a symmetrical shielded motor cable replacing an unshielded one, two EMC glands, the analogue signal moved out of the power duct and converted from a voltage signal to a 4 to 20 milliampere loop, and a Type B residual current device. The analogue reading stopped wandering, the encoder faults stopped, and the nuisance trips stopped. Nothing in the schematic changed at all, which is exactly the point.
— GANI Manufacturing engineering team