Loop type tuning cheat sheet
A quick reference for what each class of loop normally wants, and what usually goes wrong with it. It gives shapes and directions rather than numbers — anything that hands you a constant without having measured your loop is guessing.
| Loop type | Usual structure | Character | Check first |
|---|---|---|---|
| Flow | PI, no derivative | Flow loops are fast and noisy. Dead time is a fraction of a second to a few seconds, and the measurement carries turbulent noise that derivative action amplifies straight into the valve. | Rate should be zero. Then check reset against dead time. |
| Pressure | PI, no derivative | Gas pressure loops behave like fast self-regulating loops; liquid pressure is faster still. Both are noisy enough that derivative rarely helps. | Which kind you have — step the output in manual and time it. |
| Level | P or PI, tuned loose | Level is integrating, not self-regulating. It needs averaging tuning far more often than tight tuning, and it responds too slowly for iterative bump testing. | Whether the vessel is there to absorb variation. Most are. |
| Temperature | PID, derivative on PV | Temperature loops are slow and lag-dominated, which is the one common case where derivative action genuinely earns its place. | The thermowell, before you add derivative. |
| Speed | PI, no derivative | Pump and fan speed loops are dominated by the drive's own ramp rates. The process is usually faster than the drive is allowed to move. | Drive accel and decel times, and the minimum speed setting. |
| Analytical | PI, gain scheduled | pH and composition loops carry long sample-system dead time and often a strongly nonlinear process gain. Dead time, not tuning, usually sets the limit. | Mixing, sample path length, and reagent valve sizing. |
No constants here, deliberately. Nobody can give you a correct gain without knowing your process gain and dead time. What transfers between plants is the shape — which structure suits a loop type, and what usually goes wrong with it.
How to use this
Find your loop type, check the structure column against what is actually configured, and work through the common problems before touching any constants. Most loops that get re-tuned repeatedly have a structural problem in that list.
Questions that come up
Why does this not give tuning numbers?
Because nobody can give you a correct constant without knowing your process gain and dead time, and any source that offers one is guessing. What is genuinely transferable is the shape — which structure suits a loop type, which direction to move, and what usually goes wrong — and that is what this covers.
Related
Last reviewed 2026-08-01.