WEATHER ↗A–Z
The real-data investigation

Inside the Vanilla Files.

Three dramatic battery events and the careful analysis of what 888 real hourly observations can—and cannot—tell us.

AI-generated illustrative scene
AI illustration · NIX learning series

A month's worth of battery history, rather than a single flattering screenshot

I prefer a proper set of readings to a glowing 100% battery icon. An icon tells you almost nothing about the previous night, how the battery recovered, or whether it has spent long stretches without a full-charge synchronisation. Vanilla's uploaded file is therefore an unusually useful companion to a SmartShunt guide.

The export contains 912 timestamped rows, of which 888 have numeric values for both Charge (Ah) and SOC (%). The valid readings run from 6 October 2026 at 04:00 to 12 November 2026 at 03:00, one per hour, without a gap between the first and last valid observations.

The data contains only three columns: date/time, Charge (Ah), and SOC (%). There is no logged battery voltage, SmartSolar yield, temperature, source identification or appliance status. That is a limitation, not a reason to ignore the very useful story the two measurements can tell.

20–21 October: the deep dip and fast recovery

At 00:00 on 20 October, Vanilla's SOC was 73.2%. It fell during the early hours and at 09:00 was 51.6%. Later the same day it climbed to 97.1% at 14:00.

The next morning was particularly revealing. At 09:00 on 21 October, SOC reached the lowest valid reading in the entire file: 32.1%. By 13:00 it read 98.0%.

That is a dramatic change over a short period. It would be wrong to attribute it automatically to the roof panel, just as it would be wrong to regard a one-hour signed charge figure as a definitive live solar amperage. The logged Charge (Ah) value at 09:00 on 21 October was +15.92Ah even while SOC had fallen since the previous hour. This apparent mismatch is a reminder that timestamps, sampling convention and the logger's interpretation matter; you cannot assume every column is an instantaneous, perfectly aligned measurement.

The useful question to take away is: what charging source was active, when did it begin, and did the monitor genuinely recognise full charge? A future export with MPPT, Orion and hookup activity would answer more.

5 November: from 37% to a useful reserve

At 08:00 and 09:00 on 5 November, the battery reported 37.0% SOC. At 10:00 it read 43.6%, at 11:00 61.0%, at 12:00 72.6%, and at 14:00 83.8%.

This is the clearest charging-recovery story for an everyday reader: below 40% in the morning, then well above 80% by the afternoon. The source of the replenishment remains unknown from the file; the result at the monitor is clear.

The evening entries were remarkably stable by comparison: 82.0% at 18:00, 81.6% at 21:00 and 82.5% at 23:00. That variation is small compared with the earlier recharge, but it is still not enough to determine exactly which appliances were running.

11 November: a return to 100%

On 11 November at 09:00, the shunt reported 63.2%. The next two readings were 79.6% at 10:00 and 94.2% at 11:00. At 12:00, SOC reached 100.0%.

That is a striking example of why a full-charge synchronisation matters. A displayed 100% can be legitimate after full-charge detection, but the SOC export by itself cannot prove the configured charged-voltage, tail-current and detection-time conditions were met. Nor can it establish the battery's long-term health.

By 23:00, the battery was back down to 81.7%, indicating that net use or system behaviour reduced the estimated charge over the following hours.

What happens throughout the clock day?

The average signed Charge (Ah) reading is more positive in late morning than in the night hours; in the cleaned dataset, the hour-of-day mean reaches its highest value around 11:00. That makes sense as a possible charging window, but it is not proof of solar alone. Driving times, mains charging habits and how the logger timestamps values could produce or alter the pattern.

Also, means can hide dramatic days. A few high positive charge values can pull the average upwards while most periods show relatively small net changes. For a new analysis I would always look at median, the range and individual dates as well.

Why there are no 'watts generated' claims in these charts

Energy in Wh and power in W require voltage and current or an equivalent valid direct measurement. This export does not contain voltage readings, instantaneous current in amperes or measured PV production in watts. We can plot the supplied signed Charge (Ah) column as labelled, but we cannot responsibly transform it into verified solar watts or total generated kWh.

Likewise, the data is not a battery-health certificate: cell balance, actual usable capacity and diagnostic BMS information would require other observations or tests.

A better NIX data collection plan for the next trip

Record the following alongside the SmartShunt export: daily SmartSolar yield, charging status of the Orion, whether EHU was connected, battery temperature if available, Fogstar BMS alarms and the starting/ending SOC around overnight stops. If possible, log the fridge's duty cycle or measured energy consumption separately.

With these inputs you could produce a genuinely strong campervan energy case study: what did a winter night cost, how much charge returned the next day, and how much came from each source?

The real prize: confidence, not just numbers

After enough nights away, a good battery dashboard should tell you something practical: when a 105Ah lithium system copes easily, when weather or appliance use begins to matter and when it is time to change plans. That's the real reason this NIX series exists.

Explore the source: Open the interactive 888-row dashboard, try the Power Lab or read the SmartShunt flagship guide.