Pressure Transmitter Reads 68% When the Gauge Says 60%: Scaling, Noise, and Fault Detection

Work backwards from the gauge. 60% of 250 kPa = 150 kPa. Signal current = 4 mA + (0.60 × 16 mA) = 13.6 mA Raw counts = (13.6 ÷ 20) × 32767 = 22 282 counts

Apply the PLC's actual (wrong) scaling: 22 282 ÷ 32767 × 250 kPa = 170 kPa → 170 ÷ 250 = 68 % - exactly the reported trend. The logic is fine; the scaling is not.

Why it happens: the engineer scaled the full 0–20 mA span instead of the live-zero span. The zero offset error is the 4 mA point: 4 ÷ 20 × 32767 = 6553 counts, which the block interprets as 50 kPa - a 20% of span error that shrinks toward the top end.

Correct scaling:

• 4 mA (6553 counts) → 0 kPa
• 20 mA (32767 counts) → 250 kPa
• Gain check: 13.6 mA → (13.6 − 4) ÷ 16 × 250 = 150 kPa (60%) ✓

Practical note: most platforms have a built-in '4–20 mA' input mode that applies the live-zero offset in hardware/configuration; verify which mode the channel is actually in rather than layering a second correction in the scaling block.

Diagnosis. The symptom - noise appearing when a large 480 V motor starts - points to inductive coupling from the feeder into the unshielded, tray-shared signal pair, and/or a ground-loop/common-mode problem. The transmitter is loop-powered from a shared 24 V supply, so the current loop itself is a single-ended circuit with a 250 Ω burden in the PLC input; any common-mode voltage between the field and the panel appears as an input error.

Correct fixes, in order:

1. Replace the unshielded pair with shielded twisted pair and terminate the shield at one end (the panel/PLC ground), with the field end left open or capacitively terminated.
2. Physically separate the signal cable from the motor feeders - a separate tray, conduit or at minimum a barrier, and right-angle crossings only.
3. Bond the panel and machine frames to the same grounding electrode system so there is no potential difference for the common-mode voltage to drive through the loop.
4. If separation is impossible, fit a signal isolator or a 4–20 mA loop repeater at the panel, and use a shielded instrument enclosure with a local ground reference.
5. Check the 24 V supply loading and the loop burden: five transmitters at ~21 mA each is negligible current, but the total voltage budget must leave the transmitter its minimum operating voltage (24 V − 0.25 kΩ × 0.021 A ≈ 18.8 V at the transmitter, which is fine - confirm against the transmitter's minimum).

Tempting shortcut to reject: adding PLC filtering or a digital average. That hides the symptom, adds lag to a control loop, and leaves an intermittent common-mode problem that will eventually produce a bad bump or a spurious trip. Filtering is acceptable only as a temporary measure with a documented work order for the real fix.

A live-zero loop carries its own diagnostics - a healthy 4–20 mA transmitter should never be at 0 mA.

Configure under-range and over-range fault detection:

• < 3.6 mA - failed transmitter, broken wire, or lost loop power. Flag as a hard fault, drive the operator display to a fault state (not 0 kPa), and force any control output to a safe state.
• 3.6 to 3.8 mA / 20.5 to 21 mA - out-of-range warning; the process has moved outside the calibrated span.
• > 21 mA - over-range or transmitter electronics fault.
• 0 mA - broken loop or blown supply fuse; distinguish from a legitimate 'zero pressure' reading, which is 4 mA.

Implementation: use the card's raw-count fault thresholds rather than the scaled block, latch and alarm the condition, and require operator acknowledgement. Tie the fault to the control logic - for example, hold the last good value with a timer and then trip to a failsafe condition, rather than letting a 0 mA signal read as 0 kPa and be interpreted as 'tank empty'.

Also verify the high side: the gauge and the transmitter should be compared at both a low and a high point after calibration; a single-point calibration check will not reveal a span error, which is exactly what the 20%-of-span scaling mistake above produced.