- Control loop performance — the ability of a process control loop to maintain the controlled variable at its setpoint with minimum deviation and stable behaviour — is jointly determined by the controller tuning, the process dynamics, and the mechanical performance of the final control element, and poor valve mechanical performance (deadband, hysteresis, and nonlinearity) cannot be compensated by controller tuning alone.
- Valve deadband — the range of controller output change that produces no change in valve position — is the most damaging mechanical characteristic for loop performance, and reducing deadband through positioner upgrade and packing maintenance is typically more effective than controller retuning in improving poorly performing flow and pressure loops.
- The ISA-5.9 control valve performance characterisation standard provides a systematic framework for quantifying valve mechanical performance parameters, enabling objective specification and acceptance testing of control valve assemblies for precision control applications.
How Valve Mechanical Behaviour Affects Control Loop Performance
A process control loop consists of three primary elements: the sensor (measuring the process variable), the controller (calculating the correction needed), and the final control element (applying the correction to the process). Most control engineers focus their improvement efforts on the controller — tuning PID parameters, adjusting setpoints, modifying control logic. But in practice, the final control element — the valve — is responsible for a disproportionate share of poor loop performance, and addressing valve mechanical issues is often more effective than controller tuning in achieving better steady-state and dynamic performance.
The valve’s contribution to loop performance can be characterised by four key mechanical parameters: deadband (the range of input signal change that produces no output change), hysteresis (the difference in output position between increasing and decreasing input signal at the same input value), resolution (the smallest input change that produces a measurable output change), and nonlinearity (deviation of the actual flow characteristic from the specified inherent characteristic). Each of these parameters degrades loop performance in a different way, and each has different physical causes in the valve-actuator-positioner assembly.
Deadband: The Dominant Performance Limiter
Deadband is the range of controller output values over which the valve position does not change. If the controller increases its output signal from 50% to 52%, and the valve does not move until the signal reaches 52% (but then moves normally above 52%), the deadband is 2%. In a well-tuned PI controller, a 2% deadband causes the controller to overshoot the setpoint by approximately the deadband width, then oscillate at the deadband boundary in a pattern called limit cycling.
Limit cycling driven by valve deadband is the most common cause of persistent oscillation in pressure, flow, and level control loops. The oscillation has a characteristic triangular waveform on the controlled variable and a characteristic square wave on the controller output — the controller drives in one direction until the process responds, then reverses when the setpoint is exceeded, then drives in the other direction until the process responds again. This cycling continues indefinitely and consumes actuator travel, packing life, and valve seat integrity, while providing poor control performance.
Reducing deadband requires addressing its physical causes: packing friction (the dominant cause in most valves — the packing must be overcome before the stem moves), seat friction (particularly in butterfly valves where the disc bears heavily on the seat near closure), and positioner mechanical deadband (the positioner’s own internal deadband, which sets a floor below which no valve can perform better). Upgrading from a direct-acting pneumatic positioner (typical deadband 1–3%) to a digital smart positioner (typical deadband 0.05–0.2%) is frequently the single highest-impact action available for improving loop performance.
Loop Tuning Interaction With Valve Performance
PID controller tuning parameters — proportional gain (Kp), integral time (Ti), and derivative time (Td) — must be matched to both the process dynamics and the valve mechanical performance. The correct tuning for a valve with low deadband and fast response is different from the correct tuning for the same process with a high-deadband, slow-response valve — and applying the tuning appropriate for a good valve to a poor valve typically causes oscillation, while applying conservative (slow) tuning appropriate for a poor valve to a good valve leaves response speed on the table.
The practical consequence: when a valve is upgraded (packing replaced, positioner upgraded, or valve replaced with a better-specified unit), the controller tuning must be reviewed and likely tightened. The existing conservative tuning that was set to stabilise a high-deadband valve will be too slow for the improved valve, leaving achievable performance unrealised. This retune step is consistently overlooked in maintenance practices — the valve is fixed, declared working, and the loop left at the old tuning, capturing only a fraction of the available performance improvement.
ISA-5.9 Control Valve Performance Characterisation
ISA-5.9 — “Control Valve Response Measurement from Step Inputs” — defines standardised test procedures for measuring control valve performance parameters including step response, deadband, hysteresis, and resolution. The standardised test procedure allows performance parameters measured at different facilities or by different technicians to be directly compared, providing a basis for performance specification and acceptance testing.
The step response test applies input signal steps of defined amplitude (typically 1%, 3%, 5%, and 10% of span) from several starting positions (typically 10%, 30%, 50%, 70%, and 90% open) and measures the time for the valve to reach 63% and 98% of the commanded position change. The results characterise the valve’s dynamic response across its travel range and reveal position-dependent behaviour — many valves respond faster at mid-stroke than near the seat, where friction is higher.
Valve Gain Variation and Its Effect on Loop Stability
The flow gain of a control valve — the relationship between valve travel change and flow change — varies throughout the stroke for all real valves. Equal-percentage trim has intentionally varying gain (low gain at small opening, high gain at large opening) that compensates for the pressure drop characteristics of the installed system. But superimposed on this designed variation are unintended gain variations: nonlinearity in the trim Cv curve relative to the specified characteristic, changes in process gain as the operating point moves, and the effect of valve mechanical characteristics (friction, deadband) on the effective gain seen by the controller.
The gain product of the controller, valve, and process must be kept approximately constant across the operating range for consistent loop stability. If the valve gain increases by a factor of 3 between 30% and 70% open (as can occur in poorly characterised trim), the controller that is correctly tuned at 30% opening will be overly aggressive at 70% opening, causing oscillation at higher flows. Gain scheduling — adjusting the controller gain as a function of valve position or process operating point — can compensate for gain variation, but only if the gain variation is well-characterised and the control system supports gain scheduling logic.
Diagnosing Poor Loop Performance: A Systematic Approach
When a control loop performs poorly — excessive oscillation, slow response, sustained offset — the diagnosis should follow a systematic sequence that separates valve issues from controller issues from process issues. The starting point is always data: a time-trend of the controller output and the process variable over an extended period (24–48 hours of normal operation) reveals the loop behaviour pattern more clearly than any single snapshot.
Limit cycling at the valve deadband boundary is identified by a regular oscillation in the process variable with a square-wave controller output — the controller is hitting the deadband and reversing. The oscillation period is related to the process time constant and the deadband width. If the controller output shows small, frequent changes and the process variable oscillates regularly, deadband is almost certainly the cause. The fix is to reduce the valve deadband (upgrade the positioner, reduce packing load) or to add a small deadband to the controller (accepting slightly larger steady-state deviation to prevent the limit cycle).