Control valves, positioners and final elements

Why does my control valve hunt in AUTO but hold steady in MANUAL?

MANUAL breaks the feedback loop. An oscillation that stops there is being sustained by the controller working on the valve — almost always integral action ramping against stiction, or gain and reset set wrong for the loop dead time.

In MANUAL the output is a fixed number and nothing feeds back. The valve holds still no matter how much friction it has. MANUAL is a divider, not a fix. Read the shape of the trend to see which cause you have.

• Stiction. Controller output ramps as a triangle or sawtooth. The process variable snaps between two levels in a near-square wave. The valve is stuck. Integral action ramps the output until static friction lets go, the valve jumps past where it needed to be, and it starts over.
• Gain too high. Output and process variable both come out rounded and near sinusoidal, at a period close to the ultimate period. On a dead-time-dominant loop that runs about 2 to 4 dead times. On a lag-dominant loop it stretches out.
• Reset too fast. Same rounded shape, longer period, commonly past 6 dead times. Lengthen the reset time before you touch gain. On a controller scaled in minutes per repeat that is a larger number; in repeats per minute, a smaller one.

One test settles the argument. Cut the controller gain in half. A limit cycle driven by valve resolution takes its amplitude from the valve and its period from the controller, so if the amplitude holds and only the period changes, the valve is the cause and no amount of retuning will remove it. And if the valve hunts with the controller in MANUAL, the loop was never the cause. Look at the positioner, a volume booster bypass screwed down too far, or a real disturbance arriving from upstream.

What is control valve stiction and how do I tell if my valve has it?

Stiction is the gap between the static friction holding the stem and the lower sliding friction once it breaks loose, and you find it in MANUAL by stepping the signal in small increments and watching where actual stem travel stops following.

The number you want is resolution: the smallest same-direction signal change that produces a corresponding change in travel. Start mid-travel in MANUAL. Step the signal in 0.25% increments in one direction and note the first step the stem answers. Reverse, and note how much signal it takes to start it back. That reversal figure is dead band, the range the input can be varied through, on reversal, with no travel. It carries friction and lost motion together, so it is never smaller than the resolution. The performance classes commonly written into valve specifications:

• 0.1% — tight control, the class strong-acid and strong-base pH loops need.
• 0.4% — fast loops such as compressor surge and pressure control.
• 1% — basic control. Covers most flow and level loops and many temperature loops.
• 2% — loops where variability does not matter.

Those are specification practice, not a legal limit. What governs is how much variability the loop can absorb. Measure travel at the stem, with a scale or a dial indicator. Do not read the positioner's own feedback number — it reports its travel sensor, and that is not necessarily the closure member. The usual sources: packing compression, seat and seal contact near the closed position, piston actuator O-rings, backlash in rack-and-pinion linkages, and shaft windup on a rotary valve, where the shaft twists before the disc turns. Step-response test procedures are ANSI/ISA-75.25.01 and ISA-TR75.25.02.

How do I stroke a control valve, and what should I record at each point?

Stroke it in MANUAL at 0, 25, 50, 75 and 100% going up and again coming down. At every point record three numbers: signal in mA, loading pressure at the actuator in psi, and actual stem position off the travel scale.

Three numbers per point is the whole method. Signal is what was asked for. Loading pressure is what the transducer and positioner delivered. Stem position is what the valve actually did. Take the loop out of service first. A valve in MANUAL still moves the process, and a full stroke on a live line is a process upset with a technician's name on it.

On a 4-20 mA signal, 4 mA is 0% of signal and 20 mA is 100% of signal. That is not the same as 0% and 100% open. On a fail-open assembly, 4 mA is wide open. Take every point twice, once from below and once from above; dead band only appears on the reversal. Note the loading pressure where the stem first moves off the seat, then compare it to the bench set stamped on the actuator nameplate — not to the 3-15 psi operating range. The two are often different on purpose. First movement at 7 psi is normal on a 6-15 psi bench set. On a 3-15 psi bench set it is friction or seat load.

My control valve won't reach full travel or won't shut off tight — what do I check?

Put a gauge on the actuator loading line and read the pressure at the end of travel that fails. Loading at supply with the valve still refusing to go is a mechanical fault. Loading below supply means the I/P, the positioner or the air supply is not delivering.

Bench set is the spring range with no process load on the plug. Seating force is what is left after travel has compressed the spring. Take the difference between the two ranges and multiply by diaphragm area. An actuator with a 6 to 15 psig bench set, a 3 to 15 psig operating range and 100 square inches of diaphragm area carries 3 psi of precompression, so 300 lbf of seat load. Shutoff class is a force requirement, not a preference. A metal-seated Class IV is typically specified at 40 pounds per lineal inch of port circumference up through a 4-3/8 inch port and 80 pounds per lineal inch above that; Class VI metal seats run around 300 pounds per lineal inch.

A valve that leaks a little through the seat may be doing exactly what it was bought to do. Under ANSI/FCI 70-2 (IEC 60534-4), Class II allows 0.5% of rated capacity, Class III 0.1%, Class IV 0.01%, and Class V 0.0005 mL per minute per inch of port diameter per psi of differential. Class VI is a bubble count, and a 4 inch port is allowed 11 bubbles per minute, tested at 50 psid or the maximum operating differential, whichever is lower. Check the class on the specification sheet. Then supply pressure, travel stops and the actuator-to-stem connection. Condemn the trim last.

Why does my control valve sound like gravel is running through it?

That is cavitation. The liquid drops below its vapor pressure at the vena contracta, and the bubbles collapse back to liquid as pressure recovers downstream. Every collapse is an implosion against the trim.

Sound narrows the candidates. Pressure confirms them. Cavitation pops and rattles like gravel or rocks in the body; light cavitation sounds more like frying bacon. Flashing is a steady hiss, closer to sandblasting. Aerodynamic noise on a gas service is a broadband roar with no popping in it.

Then confirm with numbers. Get P1, P2 and the flowing temperature, look up the vapor pressure at that temperature, and work in absolute pressure throughout. The service cavitation index used in ISA-RP75.23 is sigma = (P1 − Pv) / (P1 − P2). Lower sigma is more severe service. The incipient and damage sigma values for the specific valve come from the manufacturer's data, not from a universal threshold. Damage shows up just downstream of the restriction — plug, seat, cage — as rough cinder or slag rather than smooth wear. Noise counts on its own terms. 85 dBA is the OSHA 8-hour action level, 90 dBA the permissible exposure limit, and a hard-cavitating valve commonly runs past 100 dBA. That is a hearing-conservation number, not a damage measurement. Sigma is what tells you whether the trim is being eaten. New trim of the same design at the same pressures will cavitate again. Stage the pressure drop or raise P2.

What is the difference between cavitation and flashing in a control valve?

Both start the same way. Pressure at the vena contracta falls below the liquid's vapor pressure and bubbles form. Downstream pressure decides which one you have: if P2 recovers above vapor pressure the bubbles collapse and that is cavitation, and if P2 stays at or below vapor pressure they remain vapor and that is flashing.

Compare P2 to Pv at the flowing temperature, both absolute. Nothing else decides it. The damage says the same thing after the fact. Cavitation damage is localized and rough, a cinder or slag texture concentrated near the vena contracta. Flashing damage is smooth, shiny and polished, and it spreads across the body outlet and the downstream pipe.

What matters in the field is which one the valve can do something about. Cavitation it can eliminate. Stage the pressure drop through multiple restrictions and the vena contracta stays above Pv. Flashing it cannot, because the two-phase flow exists downstream regardless of trim. Hardened materials, an angle body that directs the flashing stream into the outlet, and an expanded downstream pipe are damage management, not a cure.

Choked flow begins at ΔP = FL²(P1 − FF·Pv) with FF = 0.96 − 0.28√(Pv/Pc). FL is where valve style shows up. Single-ported globe bodies run about 0.85 to 0.90. V-notch and segmented ball valves about 0.55 to 0.65 at 90 degrees. High-performance butterfly valves about 0.50 to 0.60 wide open. Take the exact figure from the manufacturer's coefficient table for that size and that travel, because FL moves with both. A low-FL valve recovers more pressure, so it cavitates and chokes at a smaller drop than a globe valve in the same line.

My I/P converter has 12 mA input but no output air pressure — what do I check first?

Check supply pressure at the transducer itself. A 4-20 mA to 3-15 psi unit at 12 mA should be putting out 9 psi, and to do that it needs a supply about 5 psi above the top of its output range, which is roughly 20 psi.

Work it in this order and stop at the first number that is wrong.

• Supply. Gauge the transducer's own supply port, not the header. The general rule is 5 psi (0.3 bar) above the upper range limit of the output — about 20 psi for a 3-15 psi output, about 35 psi for 6-30 psi. The nameplate maximum is commonly around 50 psi (3.4 bar).
• Filter and restrictor. Pull the filter cap and confirm both are open and clean. A plugged restriction orifice produces exactly this symptom. Drain the filter regulator dripwell while you are there.
• Current at the terminals. Break the loop and read it in series, and only where breaking it is permitted. In a classified area that is a hot work or intrinsic safety question first. An open, a reversed polarity or a shorted pair is the usual finding. Burden is rarely the cause. Most I/P transducers drop only a few volts at 20 mA, equivalent to roughly 75 to 250 Ω, and some low-burden designs are nearer a 4 V offset plus 40 Ω. The manual gives the number for the model on the bracket.
• Range switches. If the unit is set for 12-20 mA split range, then 12 mA is 0% and 3 psi is the correct output, not a failure.
• Force it. Drive the input to 20 mA and then up to the unit's rated maximum input, commonly about 24 mA. Output should build to near supply. Do not exceed the rated input. Now remove the input: output should fall to near zero, typically under 2 psi (0.14 bar). If it does not fall, the vent and exhaust passage is blocked.

If output builds at the transducer but the valve does not move, the fault is downstream: tubing, positioner, actuator. A transducer that will not respond on good clean supply is where the field repair ends. Current I/P transducers are built as replaceable modules rather than bench-repairable assemblies, and the instruction manual lists which parts are actually available.

What supply air pressure and air quality does a control valve need?

Regulate the supply about 5 psi above the top of the instrument's output range — roughly 20 psi for a 3-15 psi signal, 35 psi for 6-30 psi — and hold the air to ISA-7.0.01 for dryness, particle size and oil content.

ANSI/ISA-7.0.01 sets three numbers. Pressure dew point at least 10 °C (18 °F) below the lowest temperature any part of the air system sees. Particles no larger than 40 micrometers. Oil content under 1 ppm, with no corrosive contaminants. ISO 8573-1 is the other classification you will see quoted on a specification sheet. A filter regulator with a 5 micrometer element ahead of the transducer is standard practice and is what most valve manufacturers specify.

Size the header and tubing for stroking, not for bleed. A modern low-bleed I/P consumes on the order of 4 to 6 scfh steady state, but filling an actuator takes a few hundred scfh. Tubing or a solenoid that chokes that flow shows up as slow stroking time, not as a pressure fault. On any transducer without internal supply regulation, supply pressure moves the output, so a drooping or unregulated supply reads at the panel as a calibration error. The specification sheet states the supply pressure effect as a percent of output span per psi of supply change.

Wet or dirty air is a utility problem, not an instrument problem. Replacing a transducer that plugged on rust and condensate buys only the time until the next slug arrives.

How do I stop a control valve packing leak - can I just tighten the packing nut?

On a conventional packing set, yes. Tightening is the right first move. But you buy the seal with friction, and past the point where the leak stops you are trading a small leak for stiction that comes back as a hunting loop.

Tighten one flat at a time — 1/6 turn on a hex nut. Stroke the valve fully between adjustments. Stop the moment the leak stops, and do not add one more for luck. Published increments range from 1/12 to 1/4 turn depending on the maker, and the valve's own instruction manual governs. Every turn past the seal point drives resolution up. A loop with a 0.4% resolution budget can have all of it spent on a packing nut.

Live-loaded packing is not adjusted by feel. Belleville stacks are compressed to a height or torque given in the valve instruction manual. Run them down past that and you have defeated the point of live loading, which is holding a constant load as the packing relaxes. Material sets the friction you are working with. PTFE V-ring is rated to about 450 °F at very low friction. Graphite runs to 1000 °F and beyond, with graphite ribbon and filament sets commonly rated 1200 °F in non-oxidizing service, at high friction. Low-friction graphite systems are rated below plain graphite, so read the packing system's own table instead of assuming graphite means 1200 °F. Every one of these ratings is packing box temperature, not process temperature.

Know when the wrench is the wrong tool. A scored stem will not seal at any torque. Packing already run down to its stops has no compression left to give. Both need the bonnet opened and the stem pulled. In fugitive emission service the EPA Method 21 leak definition for valves is 500 ppm above background under 40 CFR 60 subpart VVa, subpart GGGa and 40 CFR 63.168, while the older 40 CFR 60 subparts VV and GGG define a valve leak at 10,000 ppm. Consent decrees and state permits frequently set a lower number than any of these, so the site's own LDAR plan is the document that governs. A certified low-emission packing set that is leaking needs a repack to the qualified procedure, not a field adjustment. And a packing leak on a live line in toxic or high-pressure service is a scheduled job with the process isolated.

What is the difference between air-to-open and air-to-close control valves?

Air-to-open means increasing air pressure on the actuator opens the valve and the spring closes it; air-to-close means increasing air pressure closes it and the spring opens it.

The action you get is the actuator and the body together, not the actuator alone. A Fisher 657 is a direct-acting actuator: air to the upper diaphragm casing pushes the stem down. Bolt it to a push-down-to-close plug and you have air-to-close; bolt the same actuator to a push-down-to-open body and you have air-to-open. A Fisher 667 is reverse-acting, with air to the lower casing pushing the stem up. Standard operating ranges for both are 3-15 psig and 6-30 psig.

The field check takes a minute. With a direct-acting positioner, an air-to-open valve is closed at 4 mA and open at 20 mA, and an air-to-close valve does the reverse. Watch the stem, not the DCS faceplate.

Keep the two actions separate in your head. The actuator and body decide what loading pressure does. The positioner decides what the mA signal does. A reverse-acting positioner on an air-to-close assembly makes the valve open on rising mA, but the assembly is still air-to-close on loading pressure — and loading pressure is what decides where it goes when the air is lost. On a split-ranged pair, a transducer set for 4-12 mA or 12-20 mA still puts out the full 3-15 psi over its half of the signal, so each valve gets a full stroke from half the range.

Fail open vs fail closed - how do I tell which way my control valve fails?

Isolate the process, bleed the air off the actuator, and watch which way the stem goes. That is the only answer you can trust. The nameplate, the P&ID and the loop sheet all describe intent, not what is currently bolted on.

The paperwork gives you the expected answer. P&IDs mark it FC, FO or FL, or with an arrow showing the fail direction. On a spring-return actuator, air-to-open is normally fail closed and air-to-close is normally fail open. Two things break that shortcut.

• A double-acting piston actuator with no spring fails in place, not open and not closed, unless a lock-up valve, trip valve or volume tank is fitted — and only if that device still works.
• Signal failure is not air failure. Lose the 4-20 mA signal and full supply air is still there, so the positioner decides where the valve lands: a direct-acting positioner exhausts and the spring takes it to the air-fail position, a reverse-acting one drives toward full loading and the opposite end, and some positioners are configured to hold last position. Lose supply air and the spring always decides. The same valve can land in two different places depending on which one you lost.

Fail direction is a safety function. Verify it with the process isolated and the line depressurized, then write the result on the loop sheet. During a trip the next person needs that answer and will not have time to prove it.

Do I need a positioner, or can the I/P drive the actuator directly?

An I/P alone can drive an actuator, but it is open loop. It delivers a pressure and never learns where the stem went. A positioner closes a loop around actual travel, and that is what buys resolution.

Without one, stem position is a force balance: loading pressure against spring rate, packing friction, and process force on the plug. Every pound of friction becomes position error. Unbalanced plug force changes with process pressure, so the same 9 psi puts the stem in a different place on a different day. With a positioner, travel feedback drives the loading pressure to whatever it takes. That is why friction shows up as dead band on a positioned valve and as offset on a bare one.

Fit one on anything in the 0.4% resolution class or tighter, on rotary valves where shaft windup adds lost motion, on unbalanced plugs at high pressure, and on any actuator too large to fill through a transducer's maximum output flow in the stroking time the loop needs.

The case against a positioner is dynamic, not economic. A positioner is an inner loop, and on a fast loop — flow, liquid pressure — a positioner that is not several times faster than the process degrades the response instead of improving it. That is why fast flow loops were historically run on a bare transducer. A current digital positioner is fast enough that the objection rarely holds and the resolution argument wins, but check the positioner's own step response before deciding either way. A positioner also brings its own dynamics. A positioner and a volume booster fighting each other will oscillate at the actuator with the controller in MANUAL, and the fix is the booster bypass adjustment, not the controller.

When this page runs out

When these checks stop narrowing it down, the next step is a full valve signature — travel, friction and seat load recorded against signal on the assembly itself. That means taking the valve out of service. Cut trim, a damaged seat or a scored packing bore is valve shop work, and a shop with the bench, the spare trim and the OEM parts will do it faster and better than a field crew with the bonnet open on a pipe rack. If the valve is a certified low-emission assembly, or it is the final element in a safety instrumented function, the repair has to follow the qualified procedure and the proof test, and Bridges Industrial will say so rather than adjust it in the field. What field instrumentation work covers well is everything upstream of the flange: the signal, the air, the transducer, the positioner, the calibration, and the loop driving all of it.