Bump test procedure

A bump test is the foundation of every loop diagnosis: put a known change in, and watch what comes back. Without a deliberate disturbance there is nothing to analyse.

This is the procedure PIDSnap uses, written so it can be followed from a phone at the console. The printable version is available by email.

Get the printable version

A one-page PDF you can print and take to the console. We will send it to your email and will not pass your address to anyone.

Before you touch anything

Confirm you have authorisation from whoever owns the process.
Confirm the loop is not safety-related, interlock-related, or feeding a critical downstream constraint. If it is, stop — diagnosis only.
Check the process is steady. A bump test during an upset tells you about the upset, not the loop.
Know how to put it back. Write the current values down before you change anything.

The test

With the loop in manual and the process steady, step the output by a few percent and hold it. Record the time the step was made.

Watch for two things: how long before the measurement starts to move at all, and how long before it stops moving. The first is the dead time. The second, less the first, tells you the response time.

If the process will not tolerate a manual step, do the same in automatic with a small setpoint change of a few percent of range. You lose the clean separation of dead time but you still get the response shape.

What to record

A trend showing PV, SP and output on the same time axis, starting before the change and running until the loop has settled.
The size of the step and the direction.
The time to first response and the time to settle.
Anything else that changed during the window — a load change, another loop moving, an operator action.

What the response tells you

A clean settling response with one small overshoot is a healthy loop. A decaying oscillation is a tuning problem. A constant-amplitude cycle is mechanical. A response that never arrives is a saturated output, a heavy filter, or a rate limit.

That is the whole diagnostic tree, and it is why a photograph of the trend is enough for PIDSnap to reach a verdict.

Trend of a healthy PID loop: after a setpoint step the process variable overshoots slightly and settles within two peaks. One small overshoot, a second much smaller peak, then flat. The output moves decisively and then stops moving. This is what every stopping criterion in a PIDSnap session is describing.

Questions that come up

How big should the step be?

Big enough to see clearly above the noise, small enough not to upset the process. A few percent of range is the usual compromise. If the response is indistinguishable from noise, the step was too small and the test tells you nothing.

Do I need to do it in manual?

Manual gives a cleaner measurement, because you see the process response without the controller reacting to it. But a small setpoint change in automatic is safer on a live plant and still shows the response shape, which is what most diagnoses turn on.

Symptoms

Glossary

Last reviewed 2026-08-01.