Please note: This year, I want to help us laugh and learn by making one of my humorous books available to download for free over each of the next six columns. The first book that got us going down this route is How to become an instrument engineer—the making of a prima donna.
Control valves are used to manipulate flows in most control loops. Control valve selection and specification is critical to achieving good control loop and process performance. Here’s a look at some of the essential aspects of the best control valves. Let’s start with what I think is the missing knowledge that’s important for the control valve’s dynamic response. I wrote about it in Annex A for the ISA-TR75.25.02-2000 (R2023) technical report titled, “Control valve response measurement from step inputs.”
The absence of valve response requirements, the need to fill in a leakage class on valve specification forms, an emphasis on minimizing cost and in some cases pressure drop, and a perception that excess capacity is good for future capability may lead an engineer to think that valves typically designed for on-off service are a good option for throttling control because of lower cost, tighter shutoff and lower pressure drop. Often, valves designed for on-off service employ actuators and assemblies including linkages and shaft connections with severe inherent limitations that greatly reduce control-loop performance. This annex provides the sources, consequences, fixes and examples of valve response nonlinearities to understand the ramifications of such a decision. It concludes with examples of specifications and tests to help a good throttling valve meet application performance objectives. A broad view of nonlinearities is taken to include anything that changes the valve’s response.
Get your subscription to Control's tri-weekly newsletter.
The knowledge presented gives guidance, including examples of specifications and tests in Table A.1, to improve loop performance and shouldn’t be taken as requirements. The goal is to make suppliers and users aware of the impact of valve response on loop performance, so better decisions are made about selecting throttling control valves. The need for tight shutoff can be met by a separate on-off valve coordinated with the throttling valve.
Valve specifications must emphasize achieving the best response metrics, including response time, deadtime, resolution, lost motion, travel gain, flow gain and installed rangeability. Response metrics must be based on the change in effective flow coefficient reflecting the actual movement of the internal closure member. Due to lost motion in positioner readback, actuator shaft to stem, and stem to internal closure member (e.g., ball or disk) connections, readbacks of valve position aren’t necessarily representative of actual closure member positions. Consequently, bench tests need a travel indicator attached to the actual closure member. Since what we’re really interested in is the change in effective flow coefficient, and because process temperature and pressure can affect resolution and lost motion, response tests done with a precise low-noise flow measurement in a pilot or actual plant or flow lab can provide the most representative response metrics.
The time to 86% of the final valve response (T86) for a step change in signal is critical for many loops. This response time often increases with actuator size and step size due to slewing rate. The response time greatly increases for small step sizes for many pneumatic positioner and actuator designs, particularly as signals reverse direction. The deadtime part of the response time increases for these positioner designs and systems with significant resolution limits and lost motion often aggravated by higher friction. Constant speed actuators (e.g., electric and electro-hydraulic) can result in a fast T86 for small steps and a slower T86 for larger steps. The deadtime part of the response time is most detrimental, especially in terms of the peak error for a load disturbance because a control loop can’t begin a correction until the valve starts to respond.
The response time is critical for compressor surge control and most pressure control loops. The large T86 response time for small signal reversals can cause a limit cycle when the longer T86 response time significantly slows the overall control loop step response time. In general, the T86 valve step response time should be less than 10% of the desired closed-loop time constant for self-regulating processes or arrest time for integrating processes.
Using a volume booster on the positioner output, with booster bypass opened just enough to stop position hunting by enabling the positioner to see part of the actuator volume that’s much larger than the booster volume, can make valve response faster without causing oscillations. Using volume boosters , instead of positioners mistakenly advocated for fast processes, can cause seriously unsafe instabilities.
For pneumatically actuated valves, the valve response deadtime for a step change in signal is a combination of pre-stroke deadtime and the deadtime due to positioner sensitivity limits interacting with friction induced dead band. The pre-stroke deadtime depends on actuator volume and fill-and-exhaust rates. It’s applicable when moving from an end-point cutoff.
Such deadtime can be estimated by the fill and exhaust factors for an actuator type and volume that’s divided by the corresponding () flow coefficients exemplified. During mid-travel reversals, deadband induced from positioner sensitivity and friction is inversely related to step size, and can increase dramatically for small signal changes. Higher friction forces require a larger change in actuator pressure to reverse direction and more deadtime. In general, the valve deadtime should be less than 10% of the total loop deadtime.
There are additional sources of deadtime due to gradual changes rather than step changes in controller output. The gradual change can be approximated as a ramp, and the additional deadtime can be estimated as the lost motion and resolution divided by the average ramp rate in the controller output. For a reversal in direction of controller output, the additional deadtime occurs for the deadband, which is the sum of resolution and lost motion. For steps continuing in the same direction, the additional deadtime is the result of resolution.
For pneumatically actuated valves, the stair-step response due to resolution is often the result of the difference between static and dynamic friction of piston seals, stem packing and valve seat or seal components, which can be worse due to wear and corrosion. Movement doesn’t start until the force exceeds the static friction. The movement of the internal closure member (e.g., plug, disk or ball) jumps and doesn’t stop because the dynamic (sliding) friction is less than the static friction. This leads to a stair-step response. Clearance between the gear teeth on piston-actuator, rack-and-pinion connections worsens resolution caused by differences in static and dynamic friction. The hole pattern of a “drilled hole valve cage” can cause resolution issues of the flow coefficient and the process response. A non-zero resolution causes a limit-cycle if there’s one or more integrators anywhere (e.g., PID, positioner, process). The limit-cycle amplitude for a self-regulating process is the open-loop gain multiplied by the resolution. The open-loop gain is the product of the valve-travel gain, valve-flow gain, process gain and measurement gain. Steep installed flow characteristics, oversized valves and sensitive processes such as pH, as well as narrow measurement spans, can result in extremely large amplitudes in the limit-cycle. Low-friction packing and minimizing rubbing of seats and seals, particularly near-closed positions, can greatly improve resolution.
Lost motion is the magnitude of the percent offset between the percent-position and percent-signal input after a reversal of input signal minus the initial offset. Lost motion can be estimated as the deadband minus the resolution. Major sources of lost motion are friction, backlash and shaft windup. Lost motion from friction is proportional to friction forces and inversely proportional to I/P and/or positioner gain. Backlash is often due to play in linkages seen in piston link-arm and Scotch-yoke actuators, and in pinned or keylock shaft-to-stem and stem-to-ball or disc connections for rotary valves. Lost motion is also caused by shaft windup in rotary valves, when the piston or diaphragm actuator shaft twists before it moves and increases with friction in packing and seat or seal. A reversal in signal requires a reversal in twist, causing lost motion. Lost motion causes a limit cycle if there are two or more integrators anywhere in the control loop. Lost motion in rotary valves can be greatly minimized by converting linear actuation to rotary motion using a zero-clearance drivetrain (e.g., lever arms with rod end bearings), clamped actuator to valve stem connections including clamped splined connections, large stem diameters, and zero-clearance stem to flow element connections (e.g., stems integrally cast with ball or disk or plug).
Valve travel gain is the final change in closure member position divided by the step change in signal, which are both expressed in percent of full scale. If the positioner has a characterization of the input signal, then the change in output of the characterizer is used as the change in input signal. Valve travel gain is particularly affected by resolution and lost motion for signal changes slightly larger than the resolution or lost motion since the change in closure member position is reduced. Using integral action in the positioner may improve travel gain but the ability to help the PID reject fast load disturbances is reduced by the need to decrease the positioner gain and set an integral dead band to reduce limit-cycles from resolution. The positioner can be thought of as a secondary loop, where fast immediate response to demands of primary loop is most important. Errors in the primary loop from offsets in the secondary loop can be quickly eliminated by feedback control correction of positioner signals.
The flow gain (product of travel gain and valve flow gain) contribution to the open-loop gain is the final change in flow in engineering units divided by the step change in percent signal. It’s affected by travel gain, input characterization in the positioner and installed flow characteristic. The result is often a severe nonlinearity especially for small signal changes due to travel gain nonlinearity, and for a low valve pressure drop to system pressure drop ratio due to installed flow characteristic nonlinearity.
The flow gain nonlinearity can be reduced by a more precise valve (better resolution and less lost motion), minimal excess capacity, and a more linear installed flow characteristic. Given a precise and properly sized valve and a well-known and constant installed flow characteristic, signal characterization can greatly reduce the flow gain nonlinearity. The PID gain can then generally be increased since it’s no longer set to deal with the steepest slope (highest gain) of the installed flow characteristic. The higher PID gain can decrease the deadtime from resolution and lost motion by increasing the rate of change of the PID output signal. The increased change in signal on the flatter portions of the installed flow characteristic also helps reduce the dead time from resolution and lost motion by the magnification of the change in signal by the signal characterizer. Furthermore, identifying open-loop gain depends less on step size due to less local changes in flow gain making tuning more accurate.
For installations with a low valve pressure drop to system pressure drop ratio (e.g., < 0.1), inherent flow characteristics develop severely distorted installed flow characteristics. The distortion results in linear inherent flow characteristics approaching a quick-opening flow characteristic with a large flow gain amplifying resolution and lost motion effects, and 50% of maximum flow reached below 20% valve position. The distortion results in an equal percentage inherent flow characteristic with nearly zero flow gain below 5% valve position. There’s a severe loss of linearity for a linear inherent flow characteristic and severe loss of installed rangeability for both characteristics.
A more useful term than “inherent rangeability” is installed rangeability, which is the maximum controllable flow divided by the minimum controllable flow. The minimum controllable flow is the deadband that-s the corresponding flow on the installed flow characteristic near the closed position. For example, if the deadband is 0.4%, the minimum controllable flow would be the flow from the installed flow characteristic at 0.4% valve position. The resulting installed rangeability raises awareness about the consequences of trying to select valves with large capacity, tighter shutoffs and lower prices that appear to use less energy. Rangeability is greatly improved in valves that are more precise and optimally sized with the valve-to-system pressure drop ratio greater than 0.25 for an equal percentage inherent flow characteristic, and a valve-to-system pressure drop ratio greater than 0.5 for a linear inherent flow characteristic.
Control valves are used to manipulate flows in most control loops. Control valve selection and specification is critical to achieving good control loop and process performance. Here’s a look at some of the essential aspects of the best control valves. Let’s start with what I think is the missing knowledge that’s important for the control valve’s dynamic response. I wrote about it in Annex A for theISA-TR75.25.02-2000 (R2023) technical report titled, “Control valve response measurement from step inputs.”
The absence of valve response requirements, the need to fill in a leakage class on valve specification forms, an emphasis on minimizing cost and in some cases pressure drop, and a perception that excess capacity is good for future capability may lead an engineer to think that valves typically designed for on-off service are a good option for throttling control because of lower cost, tighter shutoff and lower pressure drop. Often, valves designed for on-off service employ actuators and assemblies including linkages and shaft connections with severe inherent limitations that greatly reduce control-loop performance. This annex provides the sources, consequences, fixes and examples of valve response nonlinearities to understand the ramifications of such a decision. It concludes with examples of specifications and tests to help a good throttling valve meet application performance objectives. A broad view of nonlinearities is taken to include anything that changes the valve’s response.
The knowledge presented gives guidance, including examples of specifications and tests in Table A.1, to improve loop performance and shouldn’t be taken as requirements. The goal is to make suppliers and users aware of the impact of valve response on loop performance, so better decisions are made about selecting throttling control valves. The need for tight shutoff can be met by a separate on-off valve coordinated with the throttling valve.
Valve specifications must emphasize achieving the best response metrics, including response time, deadtime, resolution, lost motion, travel gain, flow gain and installed rangeability. Response metrics must be based on the change in effective flow coefficient reflecting the actual movement of the internal closure member. Due to lost motion in positioner readback, actuator shaft to stem, and stem to internal closure member (e.g., ball or disk) connections, readbacks of valve position aren’t necessarily representative of actual closure member positions. Consequently, bench tests need a travel indicator attached to the actual closure member. Since what we’re really interested in is the change in effective flow coefficient, and because process temperature and pressure can affect resolution and lost motion, response tests done with a precise low-noise flow measurement in a pilot or actual plant or flow lab can provide the most representative response metrics.
The time to 86% of the final valve response (T86) for a step change in signal is critical for many loops. This response time often increases with actuator size and step size due to slewing rate. The response time greatly increases for small step sizes for many pneumatic positioner and actuator designs, particularly as signals reverse direction. The deadtime part of the response time increases for these positioner designs and systems with significant resolution limits and lost motion often aggravated by higher friction. Constant speed actuators (e.g., electric and electro-hydraulic) can result in a fast T86 for small steps and a slower T86 for larger steps. The deadtime part of the response time is most detrimental, especially in terms of the peak error for a load disturbance because a control loop can’t begin a correction until the valve starts to respond.
The response time is critical for compressor surge control and most pressure control loops. The large T86 response time for small signal reversals can cause a limit cycle when the longer T86 response time significantly slows the overall control loop step response time. In general, the T86 valve step response time should be less than 10% of the desired closed-loop time constant for self-regulating processes or arrest time for integrating processes.
Using a volume booster on the positioner output, with booster bypass opened just enough to stop position hunting by enabling the positioner to see part of the actuator volume that’s much larger than the booster volume, can make valve response faster without causing oscillations. Using volume boosters , instead of positioners mistakenly advocated for fast processes, can cause seriously unsafe instabilities.
For pneumatically actuated valves, the valve response deadtime for a step change in signal is a combination of pre-stroke deadtime and the deadtime due to positioner sensitivity limits interacting with friction induced dead band. The pre-stroke deadtime depends on actuator volume and fill-and-exhaust rates. It’s applicable when moving from an end-point cutoff.
Such deadtime can be estimated by the fill and exhaust factors for an actuator type and volume that’s divided by the corresponding () flow coefficients exemplified. During mid-travel reversals, deadband induced from positioner sensitivity and friction is inversely related to step size, and can increase dramatically for small signal changes. Higher friction forces require a larger change in actuator pressure to reverse direction and more deadtime. In general, the valve deadtime should be less than 10% of the total loop deadtime.
There are additional sources of deadtime due to gradual changes rather than step changes in controller output. The gradual change can be approximated as a ramp, and the additional deadtime can be estimated as the lost motion and resolution divided by the average ramp rate in the controller output. For a reversal in direction of controller output, the additional deadtime occurs for the deadband, which is the sum of resolution and lost motion. For steps continuing in the same direction, the additional deadtime is the result of resolution.
For pneumatically actuated valves, the stair-step response due to resolution is often the result of the difference between static and dynamic friction of piston seals, stem packing and valve seat or seal components, which can be worse due to wear and corrosion. Movement doesn’t start until the force exceeds the static friction. The movement of the internal closure member (e.g., plug, disk or ball) jumps and doesn’t stop because the dynamic (sliding) friction is less than the static friction. This leads to a stair-step response. Clearance between the gear teeth on piston-actuator, rack-and-pinion connections worsens resolution caused by differences in static and dynamic friction. The hole pattern of a “drilled hole valve cage” can cause resolution issues of the flow coefficient and the process response. A non-zero resolution causes a limit-cycle if there’s one or more integrators anywhere (e.g., PID, positioner, process). The limit-cycle amplitude for a self-regulating process is the open-loop gain multiplied by the resolution. The open-loop gain is the product of the valve-travel gain, valve-flow gain, process gain and measurement gain. Steep installed flow characteristics, oversized valves and sensitive processes such as pH, as well as narrow measurement spans, can result in extremely large amplitudes in the limit-cycle. Low-friction packing and minimizing rubbing of seats and seals, particularly near-closed positions, can greatly improve resolution.
Lost motion is the magnitude of the percent offset between the percent-position and percent-signal input after a reversal of input signal minus the initial offset. Lost motion can be estimated as the deadband minus the resolution. Major sources of lost motion are friction, backlash and shaft windup. Lost motion from friction is proportional to friction forces and inversely proportional to I/P and/or positioner gain. Backlash is often due to play in linkages seen in piston link-arm and Scotch-yoke actuators, and in pinned or keylock shaft-to-stem and stem-to-ball or disc connections for rotary valves. Lost motion is also caused by shaft windup in rotary valves, when the piston or diaphragm actuator shaft twists before it moves and increases with friction in packing and seat or seal. A reversal in signal requires a reversal in twist, causing lost motion. Lost motion causes a limit cycle if there are two or more integrators anywhere in the control loop. Lost motion in rotary valves can be greatly minimized by converting linear actuation to rotary motion using a zero-clearance drivetrain (e.g., lever arms with rod end bearings), clamped actuator to valve stem connections including clamped splined connections, large stem diameters, and zero-clearance stem to flow element connections (e.g., stems integrally cast with ball or disk or plug).
Valve travel gain is the final change in closure member position divided by the step change in signal, which are both expressed in percent of full scale. If the positioner has a characterization of the input signal, then the change in output of the characterizer is used as the change in input signal. Valve travel gain is particularly affected by resolution and lost motion for signal changes slightly larger than the resolution or lost motion since the change in closure member position is reduced. Using integral action in the positioner may improve travel gain but the ability to help the PID reject fast load disturbances is reduced by the need to decrease the positioner gain and set an integral dead band to reduce limit-cycles from resolution. The positioner can be thought of as a secondary loop, where fast immediate response to demands of primary loop is most important. Errors in the primary loop from offsets in the secondary loop can be quickly eliminated by feedback control correction of positioner signals.
The flow gain (product of travel gain and valve flow gain) contribution to the open-loop gain is the final change in flow in engineering units divided by the step change in percent signal. It’s affected by travel gain, input characterization in the positioner and installed flow characteristic. The result is often a severe nonlinearity especially for small signal changes due to travel gain nonlinearity, and for a low valve pressure drop to system pressure drop ratio due to installed flow characteristic nonlinearity.
The flow gain nonlinearity can be reduced by a more precise valve (better resolution and less lost motion), minimal excess capacity, and a more linear installed flow characteristic. Given a precise and properly sized valve and a well-known and constant installed flow characteristic, signal characterization can greatly reduce the flow gain nonlinearity. The PID gain can then generally be increased since it’s no longer set to deal with the steepest slope (highest gain) of the installed flow characteristic. The higher PID gain can decrease the deadtime from resolution and lost motion by increasing the rate of change of the PID output signal. The increased change in signal on the flatter portions of the installed flow characteristic also helps reduce the dead time from resolution and lost motion by the magnification of the change in signal by the signal characterizer. Furthermore, identifying open-loop gain depends less on step size due to less local changes in flow gain making tuning more accurate.
For installations with a low valve pressure drop to system pressure drop ratio (e.g., < 0.1), inherent flow characteristics develop severely distorted installed flow characteristics. The distortion results in linear inherent flow characteristics approaching a quick-opening flow characteristic with a large flow gain amplifying resolution and lost motion effects, and 50% of maximum flow reached below 20% valve position. The distortion results in an equal percentage inherent flow characteristic with nearly zero flow gain below 5% valve position. There’s a severe loss of linearity for a linear inherent flow characteristic and severe loss of installed rangeability for both characteristics.
A more useful term than “inherent rangeability” is installed rangeability, which is the maximum controllable flow divided by the minimum controllable flow. The minimum controllable flow is the deadband that-s the corresponding flow on the installed flow characteristic near the closed position. For example, if the deadband is 0.4%, the minimum controllable flow would be the flow from the installed flow characteristic at 0.4% valve position. The resulting installed rangeability raises awareness about the consequences of trying to select valves with large capacity, tighter shutoffs and lower prices that appear to use less energy. Rangeability is greatly improved in valves that are more precise and optimally sized with the valve-to-system pressure drop ratio greater than 0.25 for an equal percentage inherent flow characteristic, and a valve-to-system pressure drop ratio greater than 0.5 for a linear inherent flow characteristic.
