ISA CCST Level III common mistakes
ISA CCST Level III · Understand the traps and how to avoid them.
On a transmitter reading off-scale high, immediately pulling and bench-replacing the transmitter before measuring anything in the loop. The replacement reads the same, and the real fault - an open equalizer on a three-valve manifold, a plugged impulse line, or a lost reference leg on a DP level installation - is still there.
Bench work feels productive and the transmitter is the visible suspect, so the technician never establishes where in the chain (process tap, impulse line, transmitter, wiring, input card, DCS scaling) the error actually originates.
What to do instead
Read the process with an independent standard, compute the expected mA from the calibrated range, then measure mA at the transmitter terminals and again with a loop calibrator injected at the input card. That splits the fault into field-side versus system-side before any part is disturbed. Finish by verifying the manifold valve lineup and blowing down the impulse lines.
Checking a device at 0 % and 100 % only, and passing it against the nameplate reference accuracy (say ±0.05 % of span) while ignoring temperature effect, static-pressure effect, and the uncertainty of the calibrator being used.
Reference accuracy is the easiest number to find on a datasheet, and a two-point check is quick. Candidates conflate reference accuracy with total installed error and never think about test uncertainty ratio.
What to do instead
Run a five-point up-and-down check (0, 25, 50, 75, 100 % and back). Compare against a tolerance built from total probable error at the actual process conditions, not the headline number, and keep the calibrator at least 4:1 better than the device under test. Record as-found and as-left values at every point.
Treating a DP flow signal as linear - seeing 25 % of DP span and answering 25 % flow - or converting a 12 mA signal to 50 % of the measured variable when the range is suppressed or the square-root extraction is done inside the DCS rather than in the transmitter.
The 4–20 mA scale is linear in the transmitted variable, and the location of the linearization (transmitter versus control system) is easy to lose track of under exam time pressure.
What to do instead
Write the transfer function before calculating: flow % = square root of DP %. Check the instrument datasheet and loop drawing for where extraction is performed so it is applied once and not twice. For suppressed or elevated-zero ranges, work in percent of calibrated span first, then convert to engineering units.
Answering a tuning scenario by moving proportional band and integral together, judging a new tuning set after a single oscillation, and tuning a loop that is still in manual or whose valve is sticking.
Most hands-on tuning is self-taught trial and error. Without a repeatable procedure, each adjustment confounds the previous one and the candidate cannot explain why the numbers were chosen.
What to do instead
Stabilize the loop in manual, confirm the valve actually responds, then change one term at a time and observe for at least three to four dead times. Compare the observed oscillation period with the reset setting and with the loop dead time: a period that tracks the reset setting points at integral, a period tied to dead time points at gain. Then apply a named method (closed-loop ultimate gain/period, or lambda/IMC with a chosen closed-loop time constant).
Attributing a limit cycle to the controller and retuning, when the controller output trend shows a flat-topped staircase and the measured variable a sawtooth of constant amplitude whose period does not change when tuning is altered.
On a trend an oscillation looks like a tuning problem, and the valve is the last element inspected. Confirming stick-slip requires a deliberate step test that candidates skip.
What to do instead
Put the loop in manual, step the output 5–10 % up and down, and record actual valve travel against output. Deadband on reversal plus a jump on breakaway indicates stiction. Correct it mechanically (packing, positioner, actuator), then retune only after the valve moves freely.
Mixing up primary and secondary loops in a cascade - answering scaling and windup questions as if the primary drove the valve directly - and failing to state the speed relationship and the bumpless-transfer condition.
Candidates recall plant loops from memory instead of drawing the block diagram, so the setpoint path and the initialization sequence get blurred.
What to do instead
Sketch it every time: primary PV into primary PID, primary output as the secondary setpoint, secondary PID output to the valve. Check that the secondary is roughly three to five times faster, that the primary output limits match the secondary setpoint range, and that bumpless transfer means secondary in cascade with primary in manual. For feedforward, state where the wild variable enters and the lead-lag applied.
Signing off a loop check after a continuity or bump test at the marshalling cabinet only, with no confirmation that the DCS tag, engineering units, range, alarm limits, graphic, and final element all respond to the real stimulus.
Continuity is quick and easy to demonstrate, while a true end-to-end check demands coordination with operations and time in the loop folder.
What to do instead
Work from the loop folder: inject at the sensor or simulate at the transmitter, confirm the indicated value and alarm annunciation at the operator station, stroke the final element through 0/25/50/75/100 %, verify fail-safe action and output direction, and initial and date each step.
Handing a loop over to operations with temporary bypasses, forced blocks, or simulator jumpers still in place, or with a control valve left in manual at a fixed output.
Start-up pressure pushes the crew toward getting the unit running, and bypasses are tracked informally or verbally rather than on a controlled list.
What to do instead
Keep a written bypass/override log with owner, reason, and required removal date. Walk down every force and jumper during the pre-startup check, confirm the valve in auto, prove fail-safe on loss of signal, air, and power, and obtain the operations sign-off before the loop goes into service.
Accepting a verbal scope change from the client and absorbing it into the crew's work without a written change order, updated schedule, or revised baseline - then finding the hours are gone at closeout.
Changes usually arrive as casual requests in the field. The supervisor wants to be helpful and no formal mechanism is invoked, so the added work is invisible to the estimate and the schedule.
What to do instead
Document the request the same day, price labor and materials against the original estimate, issue a change order, and obtain written authorization before the work starts. Then re-baseline the schedule and track approved versus pending changes as separate line items.
Setting or keeping a calibration interval because 'that is what the sheet says,' with no review of as-found data - or the opposite, quietly extending intervals to save labor without any out-of-tolerance analysis or documented justification.
Intervals are often inherited from a previous site or a vendor recommendation, and no one connects the as-found results already being collected to the interval decision.
What to do instead
Base the interval on instrument criticality, manufacturer guidance, and the historical as-found failure rate. If a meaningful share of as-founds exceed tolerance, shorten the interval; extend only with documented out-of-tolerance analysis, a defined review date, and records retained per the plant's quality and regulatory requirements.