How to Run a 4-20 mA Loop Check That Catches Swapped Transmitters

A loop check proves the whole path, not the math

Analog scaling review checks the numbers. A loop check checks the physical path: field instrument, wiring, marshalling, input module, PLC scaling, SCADA tag, and the HMI display. Scaling can be perfect while the loop is still wrong because two transmitters are swapped in the marshalling cabinet.

The point of a loop check is simple. Inject or apply a known signal at the field end, then confirm the same value appears everywhere it should, with the correct tag name. Do it once, carefully, before the process is live and a wrong reading can cause a real action.

Decide where you inject the signal

There is no single right injection point. Each one proves a different length of the path.

Injection point	What it proves	What it skips
Field transmitter (apply real process)	Whole loop including sensor	Slow, needs process medium
Transmitter terminals (loop calibrator, mA source)	Instrument to SCADA, minus sensor	Sensor element and process seal
Marshalling terminal	Cabinet wiring to SCADA	Field cable and instrument
Input module terminal	PLC and SCADA only	All field wiring

Prefer injecting as far into the field as practical. A calibrator at the transmitter terminals catches swapped cables and wrong marshalling that an input-module injection will happily hide.

Test at multiple points, not just one

A single mid-scale check passes on a reversed or wrongly ranged loop. Use at least three points and add the fault points.

0% (4 mA): confirms zero and live-zero handling.
50% (12 mA): confirms the tag tracks and is not frozen.
100% (20 mA): confirms full scale and no display clamp.
Underrange (3.6 mA): confirms broken-wire and bad-quality behavior.
Overrange (21 mA): confirms the loop is not silently clamped to full scale.

The 3.6 mA and 21 mA points are not arbitrary. NAMUR NE43 sets the fault bands for a 4-20 mA signal: a downscale failure is signalled at ≤ 3.6 mA and an upscale failure at ≥ 21 mA, with the 3.8–20.5 mA window reserved for valid measurement. If the input card is configured to NE43, drive the calibrator to exactly those edges and confirm the card flips to fault, not that it merely reads a low or high number. A card that treats 3.7 mA as a valid 0% is misconfigured, and mid-scale testing will never show it.

Record the actual reading at each point at every layer. "Looked right" is not a record. A number you can compare later is.

Confirm the tag identity, not only the value

The most expensive loop-check miss is a correct value on the wrong tag. When you inject 12 mA and see 50% somewhere, prove it is the point you think it is.

Have one person at the field device and one at the HMI, talking on radio.
Wiggle only the loop under test and watch that specific tag move.
Read the full tag name and description back before signing.
Check that the point does not also move a neighbor tag, which means a wiring cross.
Confirm the HMI faceplate, alarm list, and trend all reference the same tag.

Two transmitters swapped between adjacent loops will each read plausibly all day. Only a deliberate one-loop-at-a-time check finds it.

Watch the fault and quality behavior

A loop check is the best time to prove that bad signals look bad. Commission these states on purpose while you have the calibrator connected.

Open the loop and confirm the input shows underrange or bad quality, not a frozen last value.
Drive overrange and confirm the HMI shows a fault indication, not a clamped 100%.
Confirm alarms based on the point behave as designed at each limit.
Verify the historian records the bad quality, not a smooth interpolated line.
Check that maintenance bypass or simulation does not hide the fault indication.

A valid process zero and a dead loop must not look identical on the screen. If they do, fix it before startup, not after an operator trusts a flat line.

Handle smart transmitters and digital paths

Not every analog point is a raw 4-20 mA loop anymore. HART, Modbus, and OPC UA paths need the same discipline with different injection.

For HART, put the transmitter in fixed-current (loop test) mode from the handheld or DD tool and step it 4-8-12-16-20 mA, rather than back-driving the loop with a mA source. Remember HART needs 230–600 Ω of loop resistance to communicate; a bench setup with no 250 Ω sense resistor will read current fine but drop the digital link.
For Modbus or gateway points, write a known value at the source and confirm word order and scaling along the way.
For OPC UA, confirm the browsed node and the SCADA tag map to the same physical instrument.
Do not assume a digital value is correct because it looks reasonable; still test low, mid, and high.

The injection method changes. The rule does not: prove the whole path with a known value and the right identity.

Keep a signed loop-check record

The record is the deliverable, not a formality. For each analog loop, capture:

Tag name, description, and loop or drawing reference.
Instrument range and calibration date.
Injection point and method used.
Reading at 0%, 50%, 100%, underrange, and overrange, at field and SCADA.
Fault and alarm behavior observed.
Date, technician, and witness.

This sheet is dull on commissioning day. It becomes valuable the first time a transmitter is replaced, a cabinet is rewired, or an operator questions a reading and someone needs to know the loop was ever proven at all.

Common failure modes

Injecting at the input module and missing swapped field cables.
Testing one mid-scale point on a reversed loop.
A correct value landing on the wrong tag name.
Overrange silently clamped to 100% with no fault indication.
A dead loop showing a frozen last value that looks like a valid zero.
Modbus or 32-bit float points with the wrong word order proven only at mid-scale.
No signed record, so nobody can say later whether the loop was ever checked.