During commissioning of a new unit, every I&C engineer meets the same acceptance question: command-to-feedback deviation on a control valve — how many percent is acceptable? Are the standards the same for SIPOS, Rotork, Fisher, and ABB? A discussion collected 40-plus replies and sorted out all three layers: the code limit, current reality, and field rules of thumb.
Per the commissioning and acceptance standards for thermal control, a deviation of no more than 3% between command and feedback is acceptable. This is the acceptance floor and the basis of acceptance tests. But peers stress: the code is a floor, not a target — "the rule is 3%, but smaller is always better"; "some actuators have non-adjustable dead bands — the dead band alone is 3%, and automatic control simply cannot run well on that."
The current-reality numbers in the discussion are remarkably consistent:
- Modern intelligent actuators are accurate — "basically within 1%, with screen displays reading to 0.1%";
- Imports: within 1% — "SIPOS almost zero," "Rotork under 1%," "imports 0.5%, domestic 2%" are the most repeated summaries;
- Domestic brands: 2–3% is generally acceptable, with good batches reaching 2%;
- Some plants tighten the internal standard: "our plant requires control valves within 2%, and critical duties like feed-pump scoop tubes within 1.5%" — tightening on important loops is common practice;
- Rotork dead band is settable internally, factory default about 2%, adjustable per loop; one plant tuned a Rotork valve to "about 1%" and warned "too small hunts" — a dead band is not "the smaller the better"; too small invites frequent cycling and even oscillation.
The discussion separates the two clearly:
- Deviation: the steady-state difference between command and feedback — a measure of commissioning and calibration quality;
- Dead band: the range of small command changes the actuator ignores — a design parameter, not an "error";
- What commissioning must truly satisfy is the loop's control quality: "do not fixate on command and feedback — the purpose is the controlled variable. If the loop meets its control-quality requirement, even 5% can be acceptable; conversely, a 0.5% dead band looks tiny, but if each valve stroke disturbs the controlled variable heavily, that valve does not suit the loop."
1. At 0% command the feedback should read near 0% — "command 0%, the valve stops at 3% — doesn't that mean it cannot shut? That cannot count as acceptable": shutoff margin depends directly on zero-point accuracy;
2. Loops to be run in automatic deserve tighter limits than manual ones — "3% will never run automatic well: you command +2% and nothing moves";
3. Local and remote displays must agree — a valve with zero deviation locally but deviation in the remote reading points to signal transmission or range configuration, not the actuator's setting;
4. For pneumatic valves, mind where the feedback is mounted — "pneumatic errors are smaller on ABB, Siemens; integral types are small; if the feedback sits in a separate feedback box, the error grows, tuning gets hard, and 3% is already decent";
5. Typical sources of excess deviation: the close seat not installed correctly, limits not set, dead band not configured — check these three before blaming the hardware.
Remember the acceptance yardstick like this: the 3% code limit is the floor; within 1% is normal for imported intelligent units; 2–3% acceptable for domestic; important loops tightened by plant standards (e.g., 1.5% for scoop tubes); evaluate dead band and deviation separately, and let loop control quality have the final word. When commissioning finds excess deviation, work through "close-seat installation → limit setting → dead-band configuration" and the problem almost always converges.
