Stress Test

The last step of Automatic System Characterization is a live-fire exam: it deliberately makes the over-voltage protection fire and grades how the whole tuned system — control loops and protections together — rides out the worst transient your vessel can produce, a hard throttle snap.

The trick that makes this safe and repeatable is the target. The test charges normally at idle until the battery voltage settles, then holds the bus at a constant-voltage target 0.10 V below that settled level — a target the alternator can reach on a battery at any state of charge. When you snap the throttle, the alternator briefly pushes the bus above that reachable target, the protection fires exactly as it would in real use, and the recovery is measured. Every protection stays fully active the whole time.

What you do

1. Bring the engine to idle and press Start stress test.
2. Wait while the regulator settles the battery and parks the bus at its target — it announces each stage. When the screen shows ARMED, snap the throttle: accelerate as briskly as you can to about half of maximum RPM, then come straight back to idle. A rapid blip, not a hold.
3. Hands off. The test watches the recovery and ends itself a few seconds after the voltage has settled back at the target — you never press anything to finish it.

A protection trip is expected — that is what is being graded. A gentle snap that never trips is also a pass (the loop absorbed it); the result just notes that a harder snap would prove more, if your engine allows one.

What happens behind the scenes

1. Settle at idle. The regulator charges normally until the filtered battery voltage holds within 0.10 V over 3 seconds. It must also stop creeping upward — charging keeps nudging the bus up long after a 3-second window looks flat, and a target fixed against a still-rising average reads low, leaving the current limit rather than the voltage loop in command. So the voltage must additionally hold within 0.10 V of a drifting reference for 10 seconds, waited at most 30 seconds. If it can't settle within 45 seconds, the run proceeds anyway with the swing reported as a marginal stability grade (as long as it stays within the fail band); only a larger swing is graded a fail on basic stability — the bus never held still enough to fix a target.
2. Park at a reachable target. The target is set 0.10 V below the settled idle voltage and the regulator enters constant-voltage hold at it, waiting for the same stability rule again — this time also requiring the bus to arrive within half the target headroom, so the approach transient is over before the armed watch begins.
3. ARMED. From here the test counts every protection event, tracks the voltage peak and the dip after the cut, and watches engine speed to measure how hard the snap was.
4. Auto-end. Once the voltage has climbed back to the target from below and held the stability window, the run grades itself and shows the verdict. If no snap is ever detected the run times out and aborts; if the voltage never re-settles, that is recorded as a failed recovery.

All voltage figures here are for a 12 V system — they scale automatically with the system voltage class.

What a recovery looks like

A snap produces a characteristic shape, and knowing it makes the grades easy to read:

The voltage spikes above the target, the protection cuts the field, the voltage dips well below the target (the field is gone and the loads are still there), then crawls back up as the control loop rebuilds the field. The spike is the smallest part of the story — the protection caught it. The time and the risk live in the dip and the climb back, which is why the grades focus there.

How it's graded

Category	What is measured	Grade
Pre-test stability	The battery voltage held 0.10 V / 3 s so a target could be set	Clean = pass; a swing within the fail band = marginal; beyond it = fail (the test can't arm)
Protection events	Every over-voltage protection firing counts — rapid re-clamping counts as several, because that chatter is itself a fault	1 ideal, 0–1 pass, 2 marginal, 3+ fail
Hard current cut	A field cut fired by the battery current sensor (INA228) — a different, harsher protection than the voltage clamp	None = pass; any = fail
Recovery time	First protection event to the voltage re-settled at the target	≤ 12 s pass, 12–20 s marginal, > 20 s fail
Post-recovery stability	The 0.15 V / 5 s re-settle achieved after the event	Clean settle = pass; settled but still swinging within the fail band (after a further 5 s of grace) = marginal; never re-settled = fail
Overshoot	Peak voltage above the target	Reported for context — never fails on its own
Recovery dip	How far below target the voltage fell after the cut	Reported for context — never fails on its own

The overall verdict is the worst grade among the gated rows. Overshoot and the dip never gate because on their own they aren't the harm: a tall spike the protection caught once and recovered from cleanly is the system working. Repeated clamping, a current-sensor cut, or a recovery that drags on is the system not working, and those are what fail.

Example outcomes

Blue is battery voltage, the grey dashed line the constant-voltage target, the amber dashed line the protection trip level, and each red dot one protection event. The horizontal axis is seconds after the snap.

Clean single event — PASS

What good looks like: the protection fires once, the voltage dips and climbs back within about 10 seconds, then holds the stability window. One event is the ideal — the protection did its job exactly once and the loop cleaned up.

Rode it out, no trip — PASS

The voltage bumps above the target but never reaches the trip level — the control loop absorbed the whole transient and there is nothing to recover from. This is a pass. If the snap was gentle (a slow-revving engine), the result adds a note recommending a harder snap if the engine allows one; if it can't rev faster, this stands — there is nothing more to prove.

Double-dip — MARGINAL

The protection fires, the voltage climbs partway back, then trips a second time before it ever re-settles. It does recover, but slowly. Two events and a 12–20 second recovery both land in the middle tier: not ideal, not a failure — worth a corrective action and a re-run.

Repeated cycling — FAIL

Spike, dip, partial climb, spike again — the loop keeps overshooting into the protection and never re-settles. Three or more events is a fail on its own, and so is a recovery that never completes. This is the signature of a voltage loop tuned too hot for the installation, and exactly what the corrective actions below address.

Reading the result

The verdict card lists each gated category with its grade, plus a context line: how hard the throttle snap actually was (RPM rise rate and peak), the target used, and the ungated overshoot and dip figures. The last result is kept and shown when you come back to this step later.

If it fails

A poor result comes with one-press corrective actions matched to the failure signature:

• Soften voltage loop 20% — reduces the voltage-loop gains so the loop overshoots less into the protection. Offered when there were repeated protection events or a hard current cut.
• Increase damping 25% — strengthens the damper (derivative term) that leans against fast voltage rise, catching the spike earlier.
• Enable assisted recovery — turns on the assisted post-protection recovery, which rebuilds the field faster after a cut. Offered when recovery was slow and the feature is off.

Each button writes the live setting immediately; press Re-run the test to confirm the fix. Keep it to two or three snaps per session — throttle fatigue is real, and later snaps tend to be gentler than the first, which makes back-to-back results hard to compare.

Relation to the other steps

The verdict grades the tuned system, so re-running any tuning step upstream — the Field curve, either current-loop step, or Voltage Control Autotuning — invalidates a previous Stress Test result, and the wizard automatically re-selects this step for the next pass.