Why isn’t it charging?

8 September 2026

A 170 amp machine is making 40 amps — why?

This is a class of unanswered question when looking at some of today's more complex lithium installations.

Correct and broken look the same

Modern regulators are interpreting a dozen+ limits at any given time — many internal, and increasingly, also arriving from other devices on a smart network.  The lowest of the bunch (or the physical limit of something in the system) defines the true output.

The variety of limits makes it impossible to answer our question with a simple ammeter display or current vs. time plot. Forty amps might be a temperature derate doing its job, belt slippage, degraded brushes, a battery that says it's full, or even a programming bug. Same reading, five possible causes and the appropriate response to each is different.

XREG-010 continuously highlights limiting factors

The app/web interface banner has been updated to better share information. Select a condition from the menu and observe the response in the interface:

Tapping any data card in the banner opens a detail popup with more than the headline: what is limiting, the active limit value, and where that gets set. The last one saves a lot of time in a complex interface.

Try it out...

As of September 2026, there are fifteen different reasons XREG-010 charging can be reduced, and twenty-six reasons it can stop altogether. Each is clickable to display the same popup as will be shown in the user interface:

Special mention for one case — the one that generates the most wasted time for owners, technicians, and OEM tech support.  If a BMS or gateway such as a Victron Cerbo GX is publishing charge limits on the network, and the regulator is set to follow them, then the number in charge is not the regulator’s. Changing settings on the regulator in that condition is fruitless.  XREG-010 saves time by highlighting who is sending the active limit.

Physical flaws

The banner described so far does not identify physical issues; for that we need another extra. A slipping belt, a corroded crimp, or an alternator losing diodes could all show up as field high, yet low output.  So enters an additional principle: comparison of performance now against the same alternator’s own past.

Short and long term trends

During operation, when engine speed, field drive, bus voltage and alternator temperature all hold steady for long enough, the regulator logs what the alternator produced at this operating point. Over time this builds a record book — the best, average, and minimum output this machine has ever managed under each set of conditions.

A single value, Charging System Health, is today’s output as a percentage of the "best ever" for similar conditions. Both sides of the comparison are the same statistic, a sustained multi-second average, so a momentary dip doesn't read as lost health. MEASURED means a recorded best exists at this operating point. ESTIMATED means it was interpolated from trustworthy neighbors. When the surrounding data is too one-sided to grade fairly, no number shows up.  The system is tuned to balance sufficient data entry with outlier elimination — set by the factory but user-adjustable. A detail page on that tuning is coming soon.

That percentage rides in the same banner beside the rest of the readings, and an additional plot is available in the Live Data → Diag tab:

Sample data. Forty-five minutes exported from the bench regulator.  Sixty-one graded points, mean 100%, tenth percentile 99%. The plot shows the last thirty minutes; the empty stretch on the left is a period it declined to grade (system still learning/unsteady conditions).

The system above measures by minute. Those points are summarized hourly and fed into the long term health trend:

A healthy charging system holds a roughly flat line near 100. The darker line is each hour’s low 10% mark — a tenth of that hour’s readings fell below it.

Waveform diagnostics

There is a third layer in development as well. The regulator samples alternator output current at 20 kHz — fast enough to see the rectifier ripple, pulse by pulse. The development system works similarly to the one previously described in that it compares "now" to reference behavior from the past.  In this case it's a visual flipbook, where the operator can scroll through waveforms from each 1000 rpm band, and compare them side by side to "reference". This is diagnostic only and currently plays no part in regulation.

In the background, a rectifier fault classifier analyzes two-second windows of the raw stream on the regulator’s extra CPU core, so the control loop never waits on it. It returns one of four verdicts — quiet, healthy, trending, fault — and on a fault it separates a diode-class pattern from a phase-class one by the way the crest train repeats. It then files the waveform into the flipbook, so the bad page sits next to the good one at the same engine speed, with a counter that survives reboots.

So far this has only been tested on modeled faults. Real fleet waveforms delivered over the internet will in the future inform this system, and lead to more accurate warning and diagnostics.  Ultimately, better fleet reliability.