PID deviation alarm trips at the wrong scale: enter counts

The threshold follows the PV and SP format

The PID chapter of the CPU user manual states it in one line: "the data format is the same as the PV and SP". The same chapter positions the yellow deviation thresholds as the operator's early warning, and it documents the rule for 12-bit and 15-bit PV and SP formats. Your loop runs a 16-bit analog input card on a 4-20 mA signal, scaled 13,107 to 65,535 counts for a 0 to 300 unit span, so the loop compares in counts up to 65,535. The number in the PV Yellow deviation register, V2017 in this setup, has to be a count value too. That is the whole problem: the HMI stored 50, the loop read 50 as 50 counts, and the comparison ran in a scale the entry never intended. Check the rule, the formats and the alarm definitions in the PID chapter of your CPU's user manual.

Symptoms and what each points to

• The alarm trips almost immediately. A threshold of 50 reads as 50 counts, about 0.29 units on this scale, so the yellow alarm sits on while the process is barely off setpoint.
• PID View prints PV and SP between 13,107 and 65,535. The loop receives raw counts from the analog card, so the threshold belongs in counts too.
• The screen shows 50 but the trip point is wrong anyway. What the HMI displays and what lands in the register are two different numbers until you have read the register back; why the scale value differs from the PLC analog input covers that gap.
• The alarm never trips at all. The mix is reversed: PV and SP run in engineering units and the threshold was written as counts, so 8,738 on a 0 to 300 scale is unreachable.

Worked example: converting 50 units to counts

Take the setup as scaled: 13,107 to 65,535 counts across a 0 to 300 unit range, 4-20 mA, one loop.

1. Count span: 65,535 - 13,107 = 52,428 counts.
2. Counts per unit: 52,428 / 300 = 174.76 counts per unit.
3. Threshold: 50 x 174.76 = 8,738 counts. Write 8,738 into the PV Yellow register.
4. The 13,107 offset cancels. Deviation is PV minus SP and both carry the offset. A setpoint of 200 units is 13,107 + (200 x 174.76) = 48,059 counts; a process value of 250 units is 13,107 + (250 x 174.76) = 56,797 counts. The difference is 56,797 - 48,059 = 8,738 counts, the same 50 units. Wherever the setpoint sits, only 174.76 matters.
5. The wrong entry, quantified: 50 / 174.76 = 0.29 units, so a threshold of 50 trips at about a third of one unit of deviation.
6. What moves the number: only the two spans. Respan the loop to 0 to 400 units and the factor becomes 52,428 / 400 = 131.07 counts per unit, so 50 units is 50 x 131.07 = 6,554 counts. Write the factor next to the loop documentation, because any respan changes it.

The fix: one scale for the whole loop

Pick one representation and keep every value the loop touches on it.

Counts in the loop. Leave PV and SP as the card delivers them and convert each threshold once: counts = units x 174.76 on this scale. A 50-unit warning becomes 8,738 counts. Give the operator units anyway: set the HMI numeric entry's scaling so typing 50 writes 8,738 to the register, or take the threshold off the screen and enter it from the programming software. Entry scaling changes only what gets stored; the loop reads the register. Fix an HMI numeric entry that writes wrong values walks through the read-back check.

Engineering units in the loop. Add one scaling rung that converts the analog counts to 0 to 300 before the PID instruction, point PV at the scaled value, and keep the SP and the yellow and red thresholds in units. Then 50 means 50 and no conversion is needed at the screen. The counts-per-unit math is the same one used when scaling a 4-20 mA transducer for an HMI display.

Mixing the two is what tripped the alarm wrong: one loop, one scale.

Check the alarm before it goes back in service

1. Open PID View, the loop monitor in the PLC programming software, and read PV, SP, the deviation and the threshold register. In this setup the threshold register is V2017.
2. Confirm what format the operands print in. Counts between 13,107 and 65,535 mean a count threshold; values on a 0 to 300 scale mean the threshold stays in units.
3. Write the converted threshold, 8,738 here, then read the register back and compare the stored value against what the HMI showed.
4. Move the PV a known amount off the SP. Ten units is 10 x 174.76 = 1,748 counts; the yellow alarm should come in at that deviation and clear when the PV returns.
5. Drive the PV above and below the SP once each so you know which deviation alarm, high side or low side, your logic actually uses.

Run these checks with the process under control before anyone relies on the alarm again; a threshold that trips at the wrong point hides a real excursion while it looks armed.

If the floor needs to see the alarm

PID View is an engineering seat, so the alarm state also has to reach the people at the line, on a board of its own. Our shop-floor screens and andon-style boards are built for that kind of floor display, and live factory data display: signage or dashboard? covers what belongs on it and how it gets there.

Recommendation

Recommendation: Keep PV and SP in raw counts, convert each deviation threshold once with the counts-per-unit factor, and let the HMI's entry scaling handle the operator's units: on the 0 to 300 scale, a 50-unit yellow warning is 8,738 counts. Read the threshold register back in PID View and test the trip point above and below the setpoint before the alarm goes back in service. If you respan the analog range, recompute the factor first; the same 52,428-count span across 400 units drops it to 131.07 counts per unit.

Questions