IMDHMIControls LibraryShop main store

Troubleshooting

Fix a 4-20 mA output ground fault with a signal isolator

Panel wiring and electrical Published 6 min read

Short answer

The 109 V DC between the two signal negatives is a reference-potential fault, not a difference to bond away. Break the loop with a 4-20 mA signal isolator rated for continuous mains working voltage, and leave the PLC's low-voltage domain and the driver's mains-referenced circuitry unconnected.

On this page
  1. What the measurement proves
  2. Why you cannot reference one to the other
  3. Why the standalone controller survived
  4. The fix: a repeating signal isolator
  5. Sizing the barrier
  6. Commission it and prove the fix
  7. Questions asked

Stop trying to reference the PLC ground to the induction driver — there is no reference to be had. The 109 V DC you measured between the two PID signal negatives is a reference-potential fault: the driver's 4-20 mA input sits at a mains-related potential, your analog output sits on the PLC's low-voltage common, and a direct wire passes the difference straight through the output stage. The fix is a signal isolator between them, with a barrier rated for continuous mains working voltage, not for the offset you happened to read. Bonding the enclosures will not work: you traced the driver and its control board has no chassis connection. What follows sorts the evidence, sizes the barrier, and gives the commissioning checks.

What the measurement proves #

Evidence Value What it tells you
PLC supply 24 V DC from its own DIN-rail supply The analog output is referenced to the PLC's 0 V domain, and to nothing else.
Driver supply Mains, L and N only No protective earth reaches the control board; there is no bondable reference point on the driver.
Your measurement DC reading between the two PID signal negatives A reference-potential fault, confirmed. It is one day's reading, not a design limit.
Standalone PID controller Drove the same input successfully, also on L and N Proves the input accepts a current-loop command. Proves nothing about its isolation.

Across the 24 V output stage, 109 ÷ 24 = 4.5: the offset is about 4.5 times the working voltage that stage was built for, and the driver's input circuitry saw the same event from its side. That is your catastrophic short, on the first connection. Treat the driver input as non-isolated from its mains circuitry until a datasheet says otherwise — you have none that claims isolation, and every measurement points the other way.

Why you cannot reference one to the other #

There are only two ways to force a shared reference here, and both fail:

  • Bond the enclosures. You traced the driver: the control board has no connection to its chassis, and the board is fed by L and N with no earth. A chassis bond never reaches the signal reference, so the offset survives it.
  • Tie PLC 0 V to the driver's signal negative. This lifts your whole low-voltage system — every card in the rack, every field wire, the panel door — to a mains-related potential. It turns a signal problem into a shock and damage problem.

IEC 61010-1, the safety standard for measurement and control equipment, requires the separation between a mains circuit and a low-voltage signal circuit to be insulation rated for the mains working voltage and its transients. A 24 V analog output channel is a signal connection, not that insulation.

Work on the driver's input terminals only with the driver isolated at its breaker, and treat every terminal on that board as mains-referenced until proven otherwise.

Why the standalone controller survived #

It is fed L and N from the same class of supply as the driver, so its 4-20 mA output stage lived in the same mains-referenced domain — or floated close enough to it that only a small difference ever appeared across it. Its success is evidence about the command format, not about the driver input's isolation. It is also your wiring template: something already drove that input cleanly, and an isolator with an active output reproduces exactly that.

The fix: a repeating signal isolator #

Order an isolator that repeats the loop one-to-one — 4-20 mA in and out — with a galvanic barrier, an externally powered active output stage, and a 24 V DC auxiliary supply input you feed from the PLC's DIN-rail supply. The input loop lands on the PLC analog output, the output loop sources current into the driver's PID input, and no copper crosses the barrier.

PLC AO+ / AO-  ->  isolator IN+ / IN-  |  barrier  |  isolator OUT+ / OUT-  ->  driver PID in+ / in-

Three wiring rules: each loop carries exactly its own two wires; nothing bridges an input terminal to an output terminal; neither loop gets earthed to make the sides match.

When a drive and its controller share one 24 V domain, none of this is needed — wiring a 1769-OF2 analog output straight to a PowerFlex 753 input is the normal case, and yours is the exception that isolation exists for.

Sizing the barrier #

Inputs, each with its source:

  • Driver supply: 230 V AC nominal, 50 Hz — your installation.
  • High-side supply tolerance: +10%, giving the worst steady line voltage the barrier can sit across.
  • Standing offset: 109 V DC — your meter, both units powered, signal pair disconnected.
  • Class figure for DIN-rail 4-20 mA isolators: 1500 V AC dielectric withstand, as published by signal-conditioner makers for this class.

Rule in words: the barrier's continuous working-voltage rating must exceed the largest steady potential that can appear across it. Because the driver input must be treated as mains-referenced, that potential is the mains working voltage — the measured offset is only where the two domains happened to sit on the day you took the reading.

Arithmetic:

  • Worst steady mains: 230 V AC × 1.10 = 253 V AC.
  • Sizing from the measurement instead: that worst-case figure ÷ 109 V = 2.3, so a barrier bought for the reading alone runs 2.3 times over rating at high line.
  • Check against the class: 1500 V AC ÷ 253 V AC = 5.9 — the class has the headroom, but 1500 V is a one-minute dielectric test level, not a continuous rating.

Result: the barrier needs a continuous working-voltage rating at least equal to that worst-case high line.

Decision: order the 300 V AC continuous-working-voltage class — it clears the requirement with margin. Reject any isolator whose datasheet quotes only a dielectric test figure with no continuous working voltage. If the driver input is later proven isolated, the rating is simply generous, not wasted.

Commission it and prove the fix #

  1. Isolate the driver at its breaker and remove every remnant of the old direct link.
  2. Mount the isolator on the PLC-side DIN rail and wire its auxiliary supply from the PLC's supply.
  3. Prove the PLC channel survived the original event before anything else touches it: command 20 mA into a 250 Ω test resistor and read 0.020 A × 250 Ω = 5.0 V across it; command 4 mA and read 1.0 V. If either reading sags, the channel took damage — and on the input side of a rack, detecting a failed analog input channel is the same discipline.
  4. Wire PLC AO+ and AO− to isolator IN+ and IN−. Nothing else shares that loop.
  5. Wire isolator OUT+ and OUT− to the driver's PID input + and −. Connect neither driver terminal to PLC 0 V or to earth.
  6. Restore the driver, command the low end of the range, and read the driver-side loop current in series: expect 4.0 mA. Ramp to full scale and confirm the driver tracks the whole range.
  7. With everything running, measure between PLC 0 V and the driver's input negative. Expect a standing offset near your original reading. That potential is now across the barrier, where it belongs — it is the fix working, not a fault.

Questions asked

Can I tie the PLC 0 V to the driver's signal negative through a resistor instead of fitting an isolator?

No. A resistor limits the fault current but still ties your whole low-voltage domain to a mains-referenced potential, so the rack, the field wiring and anyone touching a PLC terminal sit at that offset. It also adds a voltage drop that varies with loop current, bending your scaling. Only a galvanic barrier passes the current value while keeping the two references apart.

Why did the standalone PID controller drive the same input without damage?

Because it drew L and N like the driver, its output hardware already lived in a mains-referenced world, or floated close enough that whatever difference existed stayed small. That tells you the command format works, not that the input is isolated — and it is exactly the arrangement an isolator with an active output recreates on your behalf.

What isolation rating does the isolator actually need?

A continuous working voltage above the worst-case high line: the 230 V supply plus its 10% tolerance. The 1500 V AC figure quoted for most DIN-rail isolators is a one-minute dielectric test, not a continuous rating; find the datasheet line that states working voltage. The 300 V AC class clears the requirement with margin.

Should the isolator be loop-powered or externally powered?

Externally powered, with an active output that sources the driver-side loop, mirroring the controller that already worked. A loop-powered isolator takes its operating power as a voltage drop across the input loop, which only suits your case if the PLC output card has that compliance to spare. Check the card's maximum load specification before choosing one.

After the isolator is fitted, should the offset reading disappear?

No, and do not chase it. A meter between PLC 0 V and the driver's input negative still shows a standing offset near the original figure, because the two domains still sit apart. The difference is that the potential now rests across the isolation barrier instead of through your output stage, which is exactly where it belongs.

Keep reading

  • Troubleshooting

    Detect a failed analog input channel and report it to SCADA

    Read the raw input word and the module's per-channel diagnostics before you touch the SCP scaling, then alarm channel health as its own tag that names the channel number. A failed channel then reaches SCADA as a fault with an identity, not as a plausible scaled level, so you know whether to carry a card, a transducer, or wire.

    Controllers and I/O

  • Selection guide

    Add 4-20 mA analog inputs when the CompactLogix rack is full

    Swap the installed four-channel analog input card for an eight-channel one when channel isolation matches; that adds four inputs against a need of three, with no rack, power-supply or enclosure change. If isolation rules the swap out, reposition the existing power supply to free an end slot. When the cabinet itself is full, add a POINT I/O analog module or IO-Link converters over EtherNet/IP.

    Controllers and I/O

  • Troubleshooting

    Fixing 4-20 mA scaling on a Quantum AVI030 analog input

    On the Quantum AVI030 analog input, the raw range follows the channel resolution: 12-bit reports 0-4095 counts and 16-bit maps 4-20 mA to 6400-32000 counts. Neither 0-64000 nor a mismatched 0-4095 will scale or hold; set the block's raw limits to match the configured resolution and verify with 4, 12 and 20 mA test currents.

    Controllers and I/O