PIDSnap

Cascade loop hunting

Cascade hunting is a fight between two controllers that were supposed to be in a hierarchy. The master sets the slave's setpoint; the slave moves the valve. When both are too fast, or the slave is not actually faster than the master, they chase each other and the process variable never sits.

The instinct is to retune whichever faceplate is on the screen. That is how the fight started. Split the pair: put the master in manual and see whether the inner loop still hunts. The period of the cycle usually tells you which one is generating it before you touch a constant.

Put the master in manual

If the hunt stops, the master was driving it — too much gain or reset on a slave that was already doing its job. If the hunt continues, the slave is unhealthy: stiction, a noisy inner PV, or constants that would hunt even as a standalone loop.

A fast hunt (seconds) is almost always the slave. A slow hunt (minutes, matching the outer lag) is the master, or the two beating because the inner loop was never several times faster.

Do not retune both

Tune the slave first, with the master in manual, the way you would tune any loop of that type. Only then put the master in auto. Moving both sets of constants in the same hour produces a picture that cannot be attributed to either loop.

What causes it, most likely first

Both loops retuned at once

most common tuning can fix this

Master and slave were tightened together. Neither is stable in the presence of the other.

How to confirm it
Put the master in manual. If the inner loop sits and the outer measurement drifts slowly, the structure is fine and the master needs to be slower.
What to do
Restore the slave to a working tune first. Then loosen the master — especially reset — until the outer loop no longer chases the inner setpoint around.

Slave is not actually faster

common tuning cannot fix this

The inner measurement has similar lag to the outer one, so the cascade adds a second PID without adding a faster actuator path.

How to confirm it
Bump the slave in manual and time it. If it is not several times faster than the master, the cascade is the wrong structure.
What to do
Find a genuinely faster inner loop (flow, not another temperature) or drop the cascade. Tuning will not create a time-scale separation.

Stiction or saturation on the inner valve

common tuning cannot fix this

The slave cannot hold its setpoint because the valve is mechanical. The master sees a wandering inner PV and hunts trying to correct it.

How to confirm it
The slave hunts in manual-master as well, with a constant-amplitude cycle or an output on a limit.
What to do
Fix the inner valve or the constraint. Do not tighten the master.

Questions that come up

Should I add derivative to the master to calm it down?

No. Derivative on a slow outer loop that is already chasing a noisy inner setpoint usually makes the fight worse. Slow the master. Make the slave healthy.

Related

Last reviewed 2026-09-10.

Run a free PIDSnap check on this loop

Two photos, thirty seconds, no signup. The pre-flight check finds the structural problems that no amount of tuning will fix — before you touch the controller.

Start a free check