What a Positioner Adds

Why this matters

A modulating valve that lands short of where it was told to go gets blamed on the controller, on the sensor, or on "the loop," and a tech who does not know what a positioner does will either add one to a problem it cannot solve or pull one off a valve that needed it. The device is small and the misunderstanding is expensive: a positioner bolted onto a badly sized valve makes the hunting worse, not better, and a positioner removed from a high-friction valve turns a working system into one that misses setpoint every time the load reverses.

The other half is diagnostic. On a valve that already has one, the positioner's own output is a live readout of how much force the valve is stealing, and it will tell you the packing is tightening long before anyone complains.

Before you touch a valve in service

Stroking a modulating valve moves process fluid. Do not change a valve's signal on a system in service to watch it travel: either the valve is isolated, the line depressurized and drained through a vent or drain point you have watched run dry, and the isolation verified at the valve body, or the stroke is scheduled with the person who owns that process and who knows what a swing in flow will do downstream. On a spring-return actuator, the spring is stored energy under 29 CFR 1910.147 whether or not the air is off: isolate and bleed the instrument air supply, confirm zero at the actuator gauge, and do not split the actuator case without the manufacturer's spring-compression procedure, because a compressed diaphragm spring released by loosening the case bolts will launch the case.

What the actuator does when nothing is watching it

A spring-opposed diaphragm actuator is a force balance and nothing more. Air pressure on the diaphragm pushes one way, the spring pushes the other, and the stem stops wherever those two forces are equal. The signal chain in front of it (typically a 4 to 20 mA control signal into a current-to-pressure converter, which puts out a 3 to 15 psi pneumatic signal) is calibrated so that the signal span covers the actuator's bench range, meaning the pressure band over which the stem travels from one end of its stroke to the other with no load on it.

That works exactly as designed as long as the only two forces in the balance are the air and the spring. They never are. Packing friction resists motion in whichever direction the stem is trying to move. Process pressure pushes on the plug, and on an unbalanced trim that force changes as the plug approaches the seat. Neither of those appears in the signal, so neither appears in the position. The stem does not go where the controller asked. It goes where the forces cancel.

The give-away is that the error reverses sign with direction of travel. Approaching a given signal from below, the valve sits short. Approaching the same signal from above, it sits long. That signature is the whole reason positioners exist, and it is worth naming: this is a linear spring-opposed diaphragm actuator with a bench range spanning full travel. On a double-acting piston actuator with no spring there is no bench range to compare against, and on a rotary actuator the available torque varies through the arc, so the same friction produces a position error that is worse at some angles than others.

What the positioner actually is

A positioner is a second control loop, nested inside the first, and that is the most useful way to hold it in your head. It takes two inputs, the incoming demand signal and a mechanical or electrical feedback of actual stem position, compares them, and drives its output pressure hard in whichever direction closes the gap. It has its own supply connection, usually at a higher pressure than the signal, and it will spend that supply freely.

So the actuator without a positioner is a signal-to-force device: give it a signal, get a force, and let the forces argue about position. With a positioner it becomes a signal-to-position device: give it a signal, and the positioner keeps buying force until position matches. The friction did not go away. Somebody is now paying for it on purpose, and you can watch them pay.

What it does not add, and this is the part that gets missed

The positioner controls stem position. Every problem that is not a stem position problem survives it intact.

• It does not add capacity. A valve that is undersized is wide open and still short of flow. The positioner will confirm, accurately, that the stem is at 100 percent.
• It does not fix wrong trim characteristic. If an equal-percentage plug is doing a job that wanted linear, the relationship between position and flow is still wrong. The positioner makes position follow signal faithfully and the process still behaves oddly at one end of the range.
• It does not restore rangeability lost to a worn seat. Below the point where the trim stops controlling, position keeps changing and flow does not. The positioner is not lying, it just is not measuring the thing that matters.
• It does not create tight shutoff. Seat leakage class is a property of the trim and the seating force, not the feedback loop. A positioner can hold the stem hard against the seat only up to the force the actuator has.
• It does not fix an oversized valve. In fact it sharpens the problem: an oversized valve does its whole job in the bottom fraction of travel, and now that fraction is followed precisely, so small controller moves produce large flow moves.

There is a general rule buried in that list. A positioner corrects errors whose cause is force. It cannot correct errors whose cause is geometry, sizing, or trim condition.

The cost of nesting a loop inside a loop

Two loops sharing one final element can fight. The design rule for cascaded loops is that the inner loop has to settle substantially faster than the outer one, and the common guidance is a factor of roughly three or more in response speed. The exact multiplier matters less than the direction: if the positioner is slow, sticky, or tuned for high gain on a valve with a large actuator volume, the outer process loop starts issuing corrections while the inner loop is still moving, and the two chase each other. What you see is a valve that oscillates around its target with the controller in automatic and sits still with the controller in manual.

That test is the cheap one and it is worth running first. Put the controller in manual and hold the output constant, without changing the value, and watch the valve. A valve that still cycles with a fixed input has an inner-loop problem: positioner tuning, a loose feedback linkage, or a supply that cannot keep up. A valve that goes quiet in manual and hunts in automatic has an outer-loop problem, and the sibling article on why a loop hunts is where that goes.

Worked example: reading what the packing is taking

A globe valve, air-to-open, spring-opposed diaphragm actuator, bench range 3 to 15 psi across full travel, no positioner. The controller is asked for 40 percent.

Signal into the actuator: 3 + (0.40 x 12) = 7.8 psi.

Now run a reversal check with the line isolated and drained, so the stroke moves nothing but the stem. Bring the signal up from 0 percent to 40 percent and read the travel indicator: 33 percent. Bring it down from 100 percent to 40 percent: 47 percent. Same signal, two positions, a spread of 14 percentage points of travel.

Convert that spread into the pressure it represents. Full travel is 12 psi of bench range, so 14 percent of travel is 0.14 x 12 = 1.68 psi total, or about 0.84 psi on each side of the demand. That is what the packing and the seals are taking out of the balance in each direction, expressed in diaphragm pressure. Nothing about the controller, the sensor, or the wiring is involved.

Now fit a positioner and repeat. The spread collapses to the positioner's own deadband, which on a functioning unit is a small fraction of a percent of travel. The valve now goes to 40 percent from either direction. Here is the number that matters afterwards: gauge the positioner's output pressure while it holds 40 percent, reading the gauge without altering the signal so the valve does not move while you are working at it. If the packing were free, holding 40 percent would take something near the bench-range value, 7.8 psi. Suppose it reads 10.5 psi. The difference, 10.5 - 7.8 = 2.7 psi, is the force the packing and the process load are consuming, converted back into diaphragm pressure by the same 12 psi bench range.

Trend that one number across visits. When the pressure needed to hold a given position climbs while the position stays perfect, the valve is getting stiffer and the positioner is hiding it. That is the failure mode: a positioner is very good at making a deteriorating valve look healthy right up until the supply pressure runs out of headroom, at which point the valve stops reaching the ends of its travel and everyone treats a months-old packing problem as a sudden failure.

Two conditions on that comparison, both in the same breath as the number. It holds for a spring-opposed diaphragm actuator whose bench range is known and whose spring has not taken a set, since a sagged spring shifts the whole reference. It does not transfer to a double-acting piston actuator, which has no bench range to compare against; there you watch the differential between the two ports instead, and the arithmetic above does not apply.

Checking your own read

Three checks, and they fail in different directions.

Verify the feedback is actually measuring the stem and not the linkage. A feedback arm with a worn pin or a loose set screw reports the demand back to the positioner faithfully while the stem sits somewhere else, and the positioner then holds a wrong position with total confidence. Compare the travel indicator on the valve body against the positioner's reported position at both ends and at mid-travel. Disagreement anywhere means the loop is closed around the wrong thing.

Verify the supply has headroom. Positioners work by having more pressure available than the signal range needs. If supply is barely above the top of the bench range, the positioner has nothing left to spend on friction, and it will look fine at low demand and fail at high demand. Read supply pressure with the valve at its hardest working point, not at rest.

Verify the bench range against the nameplate rather than assuming 3 to 15. A valve re-springed at some point in its life, or one specified with an offset range to get seating force, will make every pressure comparison in this article read wrong by a fixed amount, and the error looks exactly like friction.

References

• Instrument Society of America and general trade-standard practice for control valve terminology, bench set, and seat leakage classification
• 29 CFR 1910.147 for control of hazardous energy, including stored spring and pneumatic energy in actuators
• Manufacturer documentation for actuator spring compression and disassembly procedures
• See related: What an Actuator Does and How Far It Actually Went; Why a Control Loop Hunts; Why Throttling a Valve Changes More Than Flow