PID engine reference 1.60 (Power User)

Note for onboarding: This chapter is for advanced users. You can skip it during initial Brautomat setup and return later if needed. Start first with AutoTune step by step and Brew-day quick guide.

Scope

This reference covers runtime behavior introduced in 1.60: Compute() sequencing, L/R AutoTune behavior, and tuning-factor style control.

Core engine idea

Control since 1.60 is not a single bare PID loop. It is a multi-stage heating strategy:

• boil region with fixed output (thresTemp/thresOut),

• heat-up phase for fast approach,

• enzyme limiter in enzyme window,

• ramping/coasting for controlled approach,

• phase-aware PID behavior with anti-windup.

Goal: fast when far from target, controlled when close.

Terms (quick reference)

• Setpoint/measured/output: target temperature, sensor value, and resulting output power (0..100%).

• P/I/D: P reacts directly, I removes residual error, D damps response dynamics.

• L and R: plant delay and slope parameters used by L/R AutoTune.

• Enzyme limiter: dynamic output cap inside enzyme temperature window.

• Ramping/coasting: shaped setpoint approach and glide phase to reduce overshoot.

• sa/psa: PID sample time and output power window.

• thresTemp/thresOut: boil transition point and fixed boil output.

Practical Compute() sequence

Compute() runs in a fixed order:

1. mode check (AUTOMATIC)

2. boil handling (thresTemp/thresOut)

3. heat-up outside enzyme window

4. enzyme limiter (effective outMax reduction)

5. ramping/coasting guidance

6. PID math (phase-dependent scaling)

7. anti-windup and output limits

Enzyme limiter, ramping, and coasting are active guidance mechanisms, not cosmetic layers.

AutoTune 1.60: robust L/R workflow

AutoTune is state-driven:

1. commit-protected start (WAIT_HEAT_COMMIT) after real heat commit

2. L via trend criterion (not one-shot threshold)

3. R via linear fit with plausibility checks

4. derived values: Kp/Ki/Kd, sa, psa, thresOut, heat-up windows

PID derivation from L/R

RecalculatePIDFromLR(...) derives control values from plant model inputs:

• system class from R,

• actuator path (IDS vs relay/webhook),

• volume influence,

• tuning factor as style parameter.

Continuous model values are mapped to discrete sampling time (sa) so controller behavior matches real runtime timing.

Tuning factor behavior

Tuning factor does not change measured physics (L/R), it changes control style:

• lower: more conservative, earlier braking

• higher: more direct, later braking

It affects multiple layers simultaneously:

• P/I/D effective scaling

• heat-up/approach/coasting transitions

• enzyme limiter bounds

• ramping dynamics

In practice, tuning factor is usually the first and most effective fine-tuning lever.

Practical brew-day order

1. run AutoTune with realistic volume and circulation

2. verify plausibility of L, R, sa, psa, thresOut

3. adjust tuning factor first

4. adjust P/I/D only in small isolated steps

See also

• AutoTune process (Power User)

• Brew-day quick guide

• FSM process flow 1.60 (Power User)

• What's New 1.60.x