Won't transfer to AUTO
A controller that will not go to AUTO without kicking the valve is not a tuning problem. It is a tracking problem. The integral term, or the setpoint, does not match the current output when you transfer, so the first thing the PID does in automatic is step the process.
Until that is fixed, you cannot run the manual test cleanly, you cannot put a cascade master in auto without upsetting the slave, and you will keep getting blamed for bumps that happened at the mode change rather than at the constants.
What bumpless transfer is supposed to do
On the way to AUTO the controller should initialise its internal state so the output does not move. Setpoint tracks PV, or the integral is back-calculated from the current output, depending on the block. If either of those is off, or the setpoint is sitting somewhere other than the PV, the first scan in AUTO is a step.
Setpoint first
Look at SP versus PV at the moment of transfer. A setpoint that was left at the last AUTO value, while the operator has been running in manual at a different PV, produces an immediate proportional kick. That is not windup and it is not aggressive reset. It is a setpoint that was not lined up.
What causes it, most likely first
Setpoint not tracking PV in manual
most common tuning cannot fix this
SP stayed where it was. AUTO closes a large error in one scan.
- How to confirm it
- Read SP and PV just before the transfer. They differ by more than noise.
- What to do
- Enable setpoint tracking in manual, or line SP up to PV by hand before transferring. Then see whether the jump remains.
Integral not tracking the actual output
common tuning cannot fix this
The PID's internal output does not match the valve position — a tracking input left disconnected, a cascade slave not initialised, or anti-reset feedback pointing at the wrong signal.
- How to confirm it
- SP equals PV, and the output still steps. The block's working output and the I/O output disagree.
- What to do
- Fix tracking / BKCAL. Do not lower gain to hide the step.
Derivative on error at the instant of transfer
less common tuning can fix this
A remaining SP–PV difference, even small, produces a derivative kick on the first scan if rate acts on error.
- How to confirm it
- The kick vanishes with derivative off or with derivative on PV.
- What to do
- Put derivative on PV. Line up SP. Then transfer.
Questions that come up
The jump is small. Can I live with it?
On a flow loop, sometimes. On a heater or a compressor recycle, no. A small step at transfer is also a sign that tracking is wrong, which will not stay small on the next transfer from a different manual output.
Related
Free tools
Glossary
Last reviewed 2026-09-10.