Skip to content

Commissioning — Full Explanation

The wizard screens carry only what you need to run each step. This page is the long version: what each step actually measures, how the resulting numbers are derived, and which protections are live while it runs.

For the practical walkthrough — what you press, in what order — see Automatic System Characterization. The final step has its own page: Stress Test.

Protections during characterization

The steps that drive the field directly with no feedback control — Field curve, Current Control Autotuning, Verify Current Control, Voltage Control Autotuning, and Keep-Alive Floor & Field Decay — run with over-voltage protection only. The current-based protections are paused because they would fight the measurement; the fast over-voltage hard shutdown is never disabled.

Every other step — Prep, Disturbances, Fault Threshold Autotuning, and the Stress Test — runs under full normal protection. Each wizard screen states which case applies at the bottom.

Step 1 — Prep

Commissioning runs semi-automated tests on your alternator and uses the results to set configuration settings. Budget about 15 minutes. Prep checks the preconditions and snapshots every setting the wizard might touch — the state-of-charge guidance and what Abort and revert all restores are under Before you start.

Other charging sources (solar, shore charger, DC-DC, wind, hydro) and large on/off loads (inverter, windlass, watermaker, electric cooking) must be off. The tests measure how your battery bank responds to the alternator alone; another active charger or a cycling load distorts the measurements, and a tune fitted to distorted data can misbehave later when that source is off.

Step 2 — Field curve

An open-loop field-drive ramp across the full range maps drive → amps, finds the saturation knee — where more drive stops producing more current — and proposes the next step's wave floor (baseline current) and step size. No feedback control; raw drive only.

Step 3 — Current Control Autotuning

An open-loop sine sweep on field drive identifies the plant — time constant τ, gain, and dead time — and proposes the current-loop gains and filter time constants.

This characterizes the alternator's own field → current behaviour, so it runs on alternator current, the same signal the inner loop regulates in every mode. Wave floor and step size carry over from the Field curve step's proposal (the Plant Delay tab).

The sweep frequency range is fixed at 0.5–20 Hz, widening to 0.3–30 Hz on an automatic retry.

Engine speed is never gated. Run near the Field curve's speed, because the carried-over sizing and the fitted gains both track speed — the wizard only flags a result more than ±500 RPM off, or one where speed wandered more than 500 RPM mid-sweep.

Step 4 — Verify Current Control

A closed-loop check: it drives the target current as a sine from slow to fast and measures overshoot (peak gain must stay at or below 1.15, or the loop rings toward over-voltage) and speed (bandwidth), on alternator current.

It sweeps past the speed the loop can follow on purpose — the response going small and ragged at the top end is how the bandwidth is found, not a bad reading. Eyeball the raw sweep on the Current Control plot before accepting.

Step 5 — Disturbances

This step records the low-frequency disturbance at each engine speed into a map. You can stop as soon as the worst-ripple speed is pinned down and no unswept gap could hide a bigger peak — there is no need to cover every speed.

During the sweep the wizard holds the alternator at a fixed test current (25% of the largest current your RPM/Amps table allows at or below 2000 RPM) and captures the measured filtered ripple — the same IExcessTau-averaged signal each over-current detector trips on — per RPM bin, for both the alternator and battery detectors. A value only commits when two readings at that speed agree, so a throttle transient cannot poison the table.

The optional current check then commands three current levels at the worst-ripple RPM and fits ripple = a0 + a1·I per detector. The same windows also capture how fast that ripple moves the smoothed battery voltage (V/s, through the voltage-loop damper's own filter) and fit it the same way.

This produces the measured projections only — it never sets a threshold. Review them against your settings in the next step, on the Protections ripple plots, and on the Tuning deadband plot.

Step 6 — Fault Threshold Autotuning

This step sets the over-current trip line from the ripple measured in Step 5.

The trip line is Slope · current + CV base, floored and capped. Slope is set to the measured ripple slope, and CV base to ripple-at-idle plus the Safety Margin, so the line runs parallel to the ripple, that margin above it. The current-limited (CC) line runs the same slope, the CC offset above the voltage-holding (CV) line. Any current where the ripple would cross the line is shaded red.

These are the exact Group 3 settings on Setup ▸ Alternator ▸ Protections — editing them here writes them live.

Why lead-acid and AGM get a much larger margin

On a lead-acid or AGM bank the recommended margin is large — 30 A — because a brief current spike does these batteries no harm. Their wear accumulates over minutes to years, not milliseconds, and there is no lithium-style BMS cutoff. Setting the trip high lets a momentary over-current from an engine-speed blip ride through, so the current loop settles on its own instead of cutting the field and dropping the charge for a fraction of a second. The battery-voltage protections remain the fast defense against an actual over-voltage. The ceiling is raised to match, so the larger margin is not clipped back down.

On lithium, 5 A suits most installations.

Voltage-Rise Tolerance (the D-term deadband)

The second section sets the voltage-loop damper's deadband line the same way: slope = the measured rise-rate slope, base = the measured intercept plus its own Safety Margin (V/s), clamped by its floor and ceiling. Those are the Voltage-Rise Tolerance settings on Setup ▸ Alternator ▸ Tuning ▸ Voltage. If Step 5 was skipped there is no fit, and both must be set by hand.

The damper (the D term) ignores voltage movement slower than its deadband, so ordinary belt and diode ripple cannot work the field. Commissioning has already measured this installation's own ripple rise rate — the Step 5 current check — and set that threshold automatically, as a line that climbs with output current, since ripple shakes the voltage faster at higher current.

The Safety Margin on that screen is not the threshold itself; it is the safety factor riding on top of it — how far above the measured ripple the whole line sits. Larger means fewer nuisance engagements but a later reaction to a genuine fast rise; the over-voltage protections remain the fast defense.

The fit is valid only for the D-term voltage filter it was measured with. Change that filter and re-run this step.

Step 7 — Voltage Control Autotuning

This step measures how stiff the battery is — how many millivolts it moves per amp — at the voltage loop's own (~0.6 s) reaction timescale, so the loop's stability margin comes out the same on any battery. A multi-second reading is mostly slow diffusion "soak" and swings 2–3× between lithium, AGM, and flooded banks; the fast reading does not.

The sequence: settle at a baseline current, size a safe test step (automatically reduced if the bank nears its over-voltage ceiling), practice-hold at the step through the current loop to learn its field level, come back down to the baseline to learn that one, then fire four abrupt field pulses, reading the settled voltage change 550–650 ms after each edge against the settled current change. The result is the median over the eight edges, so a disturbed edge is thrown out rather than biasing the answer — at least three must survive.

The two field levels are learned back to back, one settle apart, on purpose. The pulses replay them as fixed drive levels, so both have to describe the same alternator at the same moment. Learning them minutes apart lets engine speed move between them, and the replayed pair then steps the output barely at all.

Engine speed does not have to be perfectly steady. Adding the test step loads the engine and pulls it down a little, and that sag is part of what gets learned, so the pulses reproduce it rather than fight it. Speed drift is measured and reported alongside the result — compared at matching points, so the test's own sag is not counted as drift — but it never blocks a run. What does stop a run is the alternator having no room to move: if the two learned levels come out less than 2 A apart, there is no step to measure and the wizard says so instead of spending 18 seconds pulsing.

Measurement is taken at the battery shunt, or at the alternator sensor when no shunt is fitted (with loads held steady the two are equal). On anything but lithium the run refuses outright if the step would drive the bank into gassing, which reads the resistance low — draw it down with some loads first.

Gains follow as Kp = α ÷ stiffness, Ki = ρ × Kp, Kd = Td × Kp, computed in 12 V-equivalent terms (the stiffness is scaled by 12 ÷ system voltage first) and each clamped to a safe range. The voltage-loop gain source switches to Auto on Apply.

Keep all other battery loads constant during the test.

Step 8 — Keep-Alive Floor & Field Decay

Two measurements per held speed.

The onset point. Field ramps briefly until output current just begins. Onset follows a 1/RPM law, so three points fit the whole Keep-Alive column, parked a margin below. Automatic Keep-Alive Learning then maintains them over alternator life, but only ever lowers a floor — it never raises one back up. Above your highest captured RPM the floor is forced to zero, so the field can always shut fully off at speed.

The field drain. The current loop ramps to the commissioned test level, the drive is frozen for 5 s to get a settled baseline, and the field then steps to the exact floor a real over-voltage cut drives to (the flat Min Field %) while the decay is recorded on the 20 kHz current channel — calibrated against the precision sensor during the same run. If that channel is unavailable the precision sensor carries the fit instead, and the result is marked coarse.

The stored number per speed is the measured time from field-off command until output has decayed to within 10% of the residual floor — that is, 90% of the way down from its pre-cut level to that floor — read directly off the trace, with no model. The fitted time constant (τ = L/R) is reported as a cross-check.

Drain time varies with engine speed, so the three points are fitted with a straight line, shifted up to sit at or above every measured point. The over-voltage response reads that line at the engine speed latched when the episode began, held constant beyond the tested range and never extrapolated.

Both here and during the test, speed is taken before the cut, because the cut scrambles the tach signal for a few seconds; the tach gates are suspended during each run for the same reason. Field at the cut floor cannot over-volt, so the drain runs are inherently safe.

Step 9 — Stress Test

The final exam, with its own page: Stress Test. It charges at idle (at least 10 seconds) until the voltage settles, parks the bus at a constant-voltage target a set headroom below that level (Target Headroom Below Idle, on the Stress Test tuning tab), then has you snap the throttle to deliberately fire the over-voltage protection, and grades the recovery — clamp count, any hard current cut, recovery time, post-recovery stability. The settle rules, arming logic, and grading thresholds are all on the Stress Test page. It writes no settings, so skipping it never blocks the COMMISSIONED badge; it is a reference check you can run any time.

A few mechanics recorded only here: settling onto the lower target is given up to 90 seconds, and a voltage holding steady near the target — even a whisker short of it — when that wait runs out still arms the test, graded marginal (the stability limit is adjustable on the same tab). The recovery watch gets its own full window from the trip, so taking your time before snapping can never inflate the recovery time and a late snap can never starve it. A snap under about 300 RPM above the armed speed does not register as a snap at all, and the run ends without a grade. There is no Finish button — the test ends itself once the voltage has settled back, and Cancel only aborts the run ungraded.

Cross-references