WEATHER ↗A–Z
← ALL STORIES
Read the text version

The Great Cold-Night Showdown — Solar, Battery & Heat: Field Report | NIX Heat Lab

❄️ The November cold-night experiment • Edition 01

The great
cold-night
showdown.

One campervan. One living room. Three tiny sensors. And a frosty Staffordshire night that revealed two very different stories about staying warm.

13–14 NOV 2026📍 Staffordshire, England◷ 9 hours, 9 minutes📈 550 matching readings per sensor
SCROLL TO INVESTIGATE
The moment of truth · 06:09 Saturday

One night. Three temperatures.

All readings below come from three separate SwitchBot exports, matched to the same minute.

❄
Outside · coldest recorded
1.0°C

A damp, near-freezing Staffordshire dawn.

▣
Vanilla · campervan interior
22.6°C

Webasto set to 20°C. Comfortably warm throughout the overnight comparison.

⌂
The living room · home
16.3°C

After an evening at 21.7°C, the room had gradually cooled overnight.

Edition 01 · Expanded engineering field report · 12–16 minute read

What follows is the more forensic version of the story: less “it was warm”, more why it was warm, what the controls were doing, what a falling temperature curve means and what the data can never tell us on its own. Every figure labelled measured comes from the exported SwitchBot readings; every hypothesis is clearly identified.

01 / The story

Not just a test. A confidence check.

After frustrating problems with the diesel heater in my campervan, Vanilla, I wanted something more convincing than "it seems to be working". So I left the Webasto set to 20°C and let the sensors do the talking.

At the same time, I had readings from the living room at home and another sensor outside. That gave me a rather wonderful experiment: three places experiencing the same cold night, but responding to it in completely different ways.

“At 9pm, the living room was warmer than Vanilla. By dawn, Vanilla was a full 6.3°C warmer.”

That dramatic reversal isn't a verdict on which is better insulated. It is a story about heating strategy: the Webasto was being asked to hold warmth, while the home system was operating around a lower overnight setting.

AI-illustrated winter home and warm living room, with a fictional NIX character
AI illustration • not an actual photograph of the property
Vanilla from 9pm to 06:09
+2.5°C
20.1°C → 22.6°C

Despite outside cooling steadily, Vanilla finished warmer than it started.

Living room from 9pm to 06:09
−5.4°C
21.7°C → 16.3°C

The room cooled most quickly early in the night, then much more slowly towards morning.

02 / The temperature battle

Don't take my word for it.
Watch the lines.

This interactive chart is drawn from the actual CSV sensor readings, not AI-generated artwork. Move your finger or mouse over it, or hit Replay.

Three temperatures, one timeline

Temperatures in °C · 1-minute original samples · exact timestamps

OvernightFull 23 hours
VanillaLiving roomOutside
▶ Replay the cold night
● Saturday 06:09 · Slide to inspect any minute
Vanilla22.6°C
Living room16.3°C
Outside1.0°C
Vanilla–room gap+6.3°C

Live inspection means the chart responds to your interaction; these are historical readings, not a connection to the sensors. Full-day chart displays the exact common overlap from 06:57 Friday to 06:09 Saturday.

03 / The night, decoded

Five moments that
tell the story.

The ramp-up
20:15 Friday

Vanilla: 15.6°C

The campervan was below the 20°C setting, just before its rapid evening rise.

Target reached
20:49 Friday

Vanilla: 20.1°C

Up 4.5°C in 34 minutes. That's roughly 7.9°C per hour across that interval.

The crossover
21:41 Friday

Vanilla takes the lead

The van's sensor first equalled or exceeded the living-room sensor.

Overnight peak
03:47 Saturday

Vanilla: 23.0°C

A 3°C overshoot above the selected Webasto setting, worth investigating.

The finish
06:09 Saturday

22.6° · 16.3° · 1.0°

Vanilla, living room and outside. Three numbers, one striking contrast.

04 / Home versus campervan

The house was warm.
Then the schedule changed.

The living room peaked at 24.3°C at 14:55 Friday, then entered a long cooling phase. Here's how the overnight decline unfolded.

Living-room cooling by three-hour blocks
21:00–00:00
−3.3°C
00:00–03:00
−1.4°C
03:00–06:00
−0.9°C

Each bar shows the absolute temperature drop, not energy consumed. The corresponding average cooling rates were about 1.10, 0.47 and 0.30°C/hour.

A room doesn't cool at a constant rate. As the indoor–outdoor temperature difference narrows, heat loss often slows too, assuming everything else is unchanged.

Your previously reported Tado schedule was 20.5°C daytime / 17°C night-time, with the thermostat in the hall rather than the living room. The overnight profile is broadly consistent with a night setback, but the CSV doesn't record actual Tado calls for heat.

That distinction matters: the room's 16.1°C minimum at 05:58 doesn't automatically mean a failed thermostat. The hall and living room may simply have different temperatures.

⚠️ Not an insulation contest: Vanilla is smaller and was actively being heated towards 20°C; the house followed a different control strategy. We cannot rank insulation or running cost from temperature alone.
05 / Sensor forensics

Not one reading. 550 of them.

A screenshot tells you how warm a room was. A time series tells you how the heating system behaved. Here's the properly nerdy version of the comparison, distilled so you don't need a statistics degree.

🌡️Measured · Vanilla

21.94°C

Mean temperature from 9pm to 6:09am. Median: 22.1°C. Lowest: 20.1°C.

🏡Measured · Living room

18.00°C

Mean overnight temperature. Median: 17.6°C. The room cooled gradually from its evening peak.

❄️Measured · Outdoors

2.27°C

Average outside during the comparison. Minimum: 1.0°C, exactly at the final sample.

The details hiding in the averages

62.4%

of Vanilla's recorded minutes were at 22°C or warmer — 343 of 550 samples. A 20°C controller setting is not the same as the temperature at every point in a cabin.

180 min

of living-room readings were below 17°C. The hall-mounted Tado thermostat and the living-room SwitchBot were not necessarily reading the same air.

Visualising the temperature advantage

When did Vanilla pull ahead?

This chart shows Vanilla minus living room. Above zero, the campervan was warmer. Below zero, the living room was warmer. The line is calculated from the actual matching timestamps, not an AI illustration.

-2°+0°+2°+4°+6°+8°21:0000:0003:0006:0006:09By morning: +6.3°C

Hourly comparison points, plus the final 06:09 reading. Vanilla caught up at 21:41. The largest observed gap was +6.5°C at 05:58. The finish was +6.3°C.

Temperature stability

Vanilla's overnight standard deviation was 0.64°C; the living room's was 1.49°C. That measures the spread of the recorded readings.

It does not prove the Webasto regulates more accurately. A room intentionally cooling towards 17°C will naturally have a larger spread than a heated van hovering near 22°C.

What “550 samples” really means

There are 550 one-minute observations per sensor across 549 elapsed minutes, from 21:00 to 06:09 inclusive. Adjacent observations are correlated: 550 samples do not equal 550 independent heating experiments.

SwitchBot exports report temperatures in tenths of a degree, so a line that's perfectly flat for several minutes may simply be hidden variation smaller than the display resolution.

Open the hour-by-hour source table
TimeVanilla °CRoom °COutside °CVan − room
Fri 21:0020.121.73.4-1.6
Fri 22:0021.120.43.2+0.7
Fri 23:0021.619.22.8+2.4
Sat 00:0022.118.42.6+3.7
Sat 01:0022.317.82.5+4.5
Sat 02:0022.617.42.2+5.2
Sat 03:0021.817.02.0+4.8
Sat 04:0021.216.71.6+4.5
Sat 05:0022.416.51.4+5.9
Sat 06:0022.616.11.1+6.5
Sat 06:0922.616.31.0+6.3

Every displayed cell comes from the original CSV reading at the listed minute. More granular data remains inside the chart and downloadable package.

The striking part isn't that the van was 6.3°C warmer at the end. It's that the temperature advantage built gradually, as the campervan stayed heated and the house followed a completely different overnight strategy.
06 / Thermal engineering, minus the headache

Where does all the heat actually go?

Heat doesn't simply vanish. Warm air and warm surfaces release energy to colder surroundings. A heater's job is to replace that energy fast enough to hold the temperature you want.

🔥Heat input
Q̇

Webasto, radiators, DREO or Vortex add energy to a space.

→
🏡Warm space
C

Air, seats, furniture, walls and cabinets store thermal energy.

→
🌬️Heat loss
UA

Windows, roof, floor, ventilation and leaks transfer heat away.

The basic energy-balance model
C × dT/dt = Q̇heater − UA × (Tinside − Toutside)

Translation: the rate at which temperature changes depends on heat added, minus heat escaping, divided by how much thermal mass must be warmed. C represents effective heat capacity; UA is the overall heat-loss coefficient; dT/dt means the temperature change per unit time.

This is a simplified one-zone model. Real rooms and vans have multiple surfaces, air currents, solar gains and ventilation paths, so it is a teaching tool rather than a calibrated model of Vanilla.

Why the room cooled faster at first

Between 21:00 and midnight it fell 3.3°C, versus 1.4°C during the following three hours. When the room is much warmer than outside, the temperature difference drives greater heat loss. As that difference narrows, the loss often slows.

But radiators, thermostatic valves and hidden heat gains could also influence this curve. It cannot be used to isolate the building's insulation without heater on/off data.

Why a small van can warm quickly

Vanilla rose from 15.6°C at 20:15 to 20.1°C at 20:49: 4.5°C in 34 minutes, or 7.94°C/hour across that interval.

There's less enclosed air than in a house, though seats, cabinets and the vehicle structure still absorb heat. A fast sensor response is encouraging, but it does not reveal the heater's instantaneous 2 kW output or the heat stored in the van's structure.

The “degree-hour” trick

We can measure how much warmer each space remained than outside over time. Imagine adding up every minute of its inside-minus-outside difference. That's a temperature-exposure quantity in degree-hours, not energy consumption.

Vanilla
179.95 °C·h
Living room
143.89 °C·h

Trapezoidal integration across the exact 549 elapsed minutes, using the original one-minute measurements. These numbers describe the indoor–outdoor temperature gap accumulated over time. They do not reveal fuel, electricity, U-values, efficiency or thermal capacity.

Why we cannot award an “insulation winner” Even if two spaces had identical degree-hours, the energy required would differ depending on their size, windows, air leakage, insulation and operating controls. Without the relevant UA and heat-input data, this is an apples-to-oranges comparison — albeit a deliciously nerdy one.
Extra geekery: what's the theoretical effect of a 1°C temperature change?
For an unchanged room at the same outdoor temperature, the simplified conductive heat-loss term is proportional to the indoor–outdoor difference. At 2°C outside, maintaining 20°C gives a difference of 18°C; maintaining 19°C gives 17°C. That's about a 5.6% lower instantaneous heat-loss driving difference in this simplified example. It is not a prediction that the total household fuel bill drops 5.6%, because boiler cycles, occupancy, ventilation and reheating all matter.
07 / A little science, without the lecture

The interesting bit
isn't just temperature.

💧

The moisture story

Relative humidity looks ordinary in both spaces. But the actual quantity of water vapour tells a richer story.

Overnight averageLiving roomVanilla
Relative humidity39.5%44.8%
Absolute humidity6.09 g/m³8.66 g/m³
Dew point4.0°C9.4°C

At a dew point around 9.4°C, condensation could form on a surface colder than that, such as glazing or bare metal. This is a risk indicator, not proof that condensation occurred.

⚙️

Why 20°C can mean 22°C

Vanilla's Webasto was selected at 20°C; its SwitchBot sensor averaged 21.94°C between 21:00 and 06:09.

Possible contributors include different sensor locations, stratified air, delayed control response and residual heat. The recording shows temperature variations, but not the burner state, fuel-pump pulses or actual fuel use.

A successful warm night is reassuring after the repairs, but it does not certify the installation. The fuel-pump mounting/bracket issue still warrants completion and inspection.
AI illustrative concept of Vanilla in a frosty field with lights glowing inside
AI concept scene · Vanilla against the cold
AI illustrated cozy campervan sleeping area with a blanket and dog
AI concept scene · a cosy camper interior; not a documentary image
✦ Open the full AI editorial storyboard (decorative artwork, not source data)AI-created visual editorial board about a three-way cold-night heating test

Artwork may stylise vehicles, rooms, timelines and graphs. Only the interactive charts and written measurements in this article should be treated as the sensor-data record.

08 / Inside the Webasto brain

Set to 20°C. Reads 22°C.
Fault, feature or sensor location?

Now for the interesting control-systems bit. Your diesel heater has its own temperature feedback loop. The SwitchBot is a separate witness watching from another point in the van.

🎛️Reported setting

20°C

The requested Webasto controller setting, as you confirmed.

📡SwitchBot measured mean

21.94°C

The independent sensor's overnight average, not the heater's own feedback reading.

↗️Difference in readings

+1.94°C

Average observed offset from the selected setting. Not necessarily thermostat calibration error.

How the heater actually regulates

01 · Sense

The heater measures intake air via its internal sensor or a compatible external room sensor.

→
02 · Modulate

It varies its fan speed and metering-pump fuel delivery as heating demand changes.

→
03 · Pause / restart

At minimum output it can stop combustion during a control pause, then restart when its sensor cools.

This is described in the Webasto Air Top 2000 STC workshop manual, section 3.3–3.5; it explains a possible mechanism, not what definitely happened inside Vanilla. Read the workshop manual ↗

Event A · 02:20–02:30

−1.5°C

Vanilla went from 22.6°C down to 21.1°C over 10 minutes. Outside held at 2.2°C at both ends. This looks like a local cooling episode, but the fuel pump and fan weren't logged.

Possible explanations: a control pause, airflow change or a sensor exposed to briefly cooler air.

Event B · 03:47–04:02

−1.8°C

From the overnight peak of 23.0°C to 21.2°C in 15 minutes, while outside dropped from 1.8°C to 1.6°C.

This is another candidate for a control-state change, but a graph of air temperature alone cannot tell us whether the burner stopped.

A useful clue from the manufacturer's documentation Webasto says intake temperature may be measured locally or via an external sensor; control changes also take time to take effect. If your SwitchBot sits in a warmer air layer than the heater's sensing point, both could be “right” while reading different temperatures. That hypothesis needs placement details and simultaneous readings to confirm.

What about the previous fuel-system troubles?

After the repair, this run gives reassuring evidence that the cabin received sustained heat. But the exported data contain no combustion signal, no pump pulse log, no error-code record and no exhaust or battery-voltage trace. It would be premature to call the fuel system permanently reliable on that basis alone.

What the readings support

Rapid recovery, high cabin temperature and repeated warming after cooling dips are consistent with an effective heat source across the test.

What the engineer still needs to finish

The reported cable-tie pump mounting should be checked and replaced with the ordered appropriate vibration-damped mount. Webasto's installation documentation specifically requires a vibration-damping mounting and a suitable pump orientation. Have the professional verify this and the fuel connections.

Practical overnight safety Use a suitable working carbon monoxide alarm and keep the combustion air/exhaust system unobstructed. Because the fuel-pump fixing is reported as unfinished, get the installer to complete and verify it before treating the system as fully repaired. Don't work on the heater's fuel or exhaust installation yourself.
09 / The fuel-and-battery question

Warmth is one thing.
What did it cost?

This is where most heating comparisons go wrong. Temperature is a result. Energy consumption is an input. Without a heater-specific measured input, we can describe performance but not true running cost. We now have total battery-history data, analysed in the next chapter.

Webasto Air Top 2000 STC D

Manufacturer technical ranges for the diesel version:

SpecificationRange
Nominal heat output0.9–2.0 kW
Fuel while firing0.12–0.24 L/hour
Rated electric running draw14–29 W

The ranges are published operating specifications under specified conditions, not logged readings from Vanilla. Manufacturer installation manual, technical data ↗

Living-room heating inputs

The Tado controls your oil-fired central heating with its thermostat in the hall. The DREO and Vortex provide supplementary electrical heat in the living room.

A Meross export showing kWh by timestamp would tell us what each electrical heater consumed. To assess the oil boiler separately, we would need burner runtime, fuel metering or another defensible oil-use estimate.

Without these sources we can say “warm” or “cool”, but not “cheaper per night”.

Interactive thought experiment · not a measured result

What if the Webasto fired for only part of the night?

Move the slider to choose an assumed firing fraction for the 9 hours 9 minutes between 21:00 and 06:09. The calculation multiplies assumed firing hours by Webasto's published 0.12–0.24 L/h fuel range. It deliberately doesn't claim to reconstruct the real burner cycles.

Assumed proportion of the night firing: 50%
Assumed firing time
4.58 h
Illustrative fuel range
0.55–1.10 L
At 100% firing
1.10–2.20 L

At a notional 100% continuous firing fraction, 9.15 h × 0.12–0.24 L/h gives 1.10–2.20 L. Actual operation may involve modulation, pauses, restart loads and installation-specific conditions. This slider is for understanding sensitivity, not for billing or proving efficiency.

And the battery? We now have its history.

At the stated 14–29 W rated electric running draw, nine hours and nine minutes at those draw levels equates to roughly 0.13–0.27 kWh — an illustrative running-power calculation only. Startup/glow-plug demand, varying fan speed, control pauses and other van loads can change the actual measured battery energy use.

The measurement plan that would settle it

01
Meross electrical logs

Export minute/interval energy and power for DREO and Vortex. Ensure both monitors use the same clock and the same test night.

02
Victron SmartShunt

Use synchronized battery current/energy records for Vanilla, then account for the fridge, lighting and other simultaneous loads. A net battery trace alone cannot isolate the Webasto; see the new time-aligned battery evidence below.

03
Webasto operating evidence

A proper heater runtime/diagnostic record or trustworthy fuel measurement is needed to turn the diesel range into a real overnight consumption figure.

04
Oil boiler heating demand

Log Tado calls for heat and, where possible, oil burner runtime or metered fuel use. The hall thermostat's temperature must be kept distinct from the living-room sensor.

Best scientific conclusion today Vanilla clearly delivered better overnight warmth in this specific test. Whether that required less fuel or more efficient heating than the house remains unknown. This distinction makes the report more trustworthy, not less interesting.
09A / The fourth trace · new evidence

The van was warm.
But what did the battery do?

This is the missing subplot. We've already watched the air temperatures. Now, a new 960-row battery-history export lets us follow Vanilla's electrical reserve through the same cold night. It spans 6 October to 14 November, with usable readings through 07:00 on 14 November. The data do not include an appliance-by-appliance power meter — a critical distinction.

The overnight electrical detective story

The cabin warmed and stayed comfortable. Meanwhile, its 105 Ah LiFePO₄ leisure battery's reported state of charge moved down by 14.7 percentage points between 9pm and 6am.

That is not inherently alarming; the real question is which devices used those amp-hours. The Webasto, the compressor fridge, the control electronics and any other active circuits all share the battery budget. The file records their net effect, not separate costs.

21:00 → 06:00
84.8→70.1%
−14.7 percentage points
Measured SOC values · nine elapsed hours
Battery SOC at 20:0086.3%Start of evening watch
Battery SOC at 06:0070.1%Still substantial reserve
Battery SOC at 07:0067.0%Most recent valid hour
Approx. nominal capacity equivalent15.4 Ah14.7% × 105 Ah · estimate, not a shunt total

Trace 04: Battery state of charge

The green curve shows the reported SOC at each hour. It never climbs overnight. This is an actual data trace, not a rendering invented by the AI artwork.

Vanilla lithium battery state of charge on 13 to 14 NovemberHourly state-of-charge readings declined from 86.3 percent at 8pm to 84.8 percent at 9pm, 70.1 percent at 6am, and 67 percent at 7am. Direct measurements from the uploaded battery history CSV.60%65%70%75%80%85%90%20:0022:0000:0002:0004:0006:0007:0084.8%70.1%67.0%

Source: Solar History - 2026-10-06.csv. Hourly data. The source does not specify the SOC algorithm, calibration error or precise sampling method.

Charge balance, hour by hour

The file's Charge (Ah) field is negative at every hourly reading between 8pm and 7am, consistent with net discharge. The larger negative bar at 7am deserves a separate equipment check.

Hourly charge in amp-hoursAll twelve readings from 8pm to 7am are negative. The final 7am hourly value is minus 5.07 amp-hours, larger in magnitude than earlier hours.0-1-2-3-4-520:00-2.4822:0000:00-1.8402:0004:0006:00-3.1607:00-5.07

The bars depict the CSV's signed hourly values. The export doesn't explain how the hours are assigned to timestamps, so these bars are not automatically a time-matched Webasto consumption trace.

Turn percentages into a rough energy budget

How much electricity is 14.7%?

Given the nominal 105 Ah leisure battery, a 14.7-point SOC change corresponds to roughly 15.4 Ah of nominal capacity. At an assumed 12.8 V nominal battery voltage, that is about 0.20 kWh across nine hours — an average net battery draw of approximately 22 W.

0.147 × 105 Ah × 12.8 V ≈ 198 Wh

A teaching estimate, not a measured energy total. Battery SOC depends on shunt configuration, capacity assumptions and synchronisation; real voltage varies. Other loads are included. No fair comparison with mains-electricity tariffs or diesel cost follows from this number alone.

The previous night as a clue

That draw wasn't the usual baseline

On the immediately preceding comparable 21:00–06:00 window, the SOC went from 99.8% to 96.8%: a fall of just 3.0 points. This night, the fall was 14.7 points.

That is a meaningful change in net battery behaviour, but it's not a controlled comparison. We don't have verified heater on/off times, solar input, alternator or hook-up status, fridge cycles, or an inventory of switched-on devices for the two nights.

A before/after difference is a lead for investigation — not proof that the Webasto consumed the difference.

A potentially interesting morning event

What happened around 7am?

The CSV records −3.16 Ah at 6am and −5.07 Ah at 7am, compared with overnight hours more commonly around −1 to −2 Ah.

That may reflect an extra load, a change in heater operation, or the file's reporting intervals. There is no evidence here to identify the cause, and it would be poor diagnostics to label it a glow-plug event without a time-aligned heater or current log.

Best follow-up: inspect a finer-resolution SmartShunt current trace and note any switches or heater state changes around 06:00–07:00.

The nerd corner · what the data genuinely mean

Four things a battery-history CSV can — and cannot — tell us

01 · Charge isn't the same as solar harvest

The filename says “Solar History,” but its three columns are timestamp, Charge (Ah) and SOC (%). There is no dedicated PV current, panel watts or charger-source column. Positive daytime readings cannot be credited specifically to solar.

02 · Battery discharge is not Webasto discharge

Even a precise shunt measures the total net current through the battery. Fridge, lighting and standby electronics can all contribute. To isolate Webasto electrical use, we'd need a controlled on/off comparison or a dedicated heater current channel.

03 · Battery amp-hours and battery percentage won't match perfectly

For example, the nine hourly “Charge (Ah)” entries stamped 22:00 through 06:00 sum to −16.72 Ah, while the nominal-capacity equivalent of the 21:00-to-06:00 SOC change is about 15.4 Ah. These are not necessarily measurements of exactly the same interval. Adding logged hourly values gives one figure; converting SOC change using the nominal 105 Ah gives another. They needn't agree because of timestamp boundaries, SOC estimates, capacity calibration and the export's unknown aggregation method. We should not quietly make them agree.

04 · Heating effectiveness and energy efficiency are separate questions

We know Vanilla held above 20°C overnight while outside reached 1°C. We don't know how many litres the Webasto burned, its precise firing duty cycle, or how much power was used by the heater alone. Comfort is demonstrated; efficiency is still an open measurement.

Open the original 12 hourly battery readings · 20:00–07:00
Local timestampSOCCharge field
Fri 20:0086.3%-2.48 Ah
Fri 21:0084.8%-1.43 Ah
Fri 22:0083.3%-1.86 Ah
Fri 23:0081.8%-1.43 Ah
Sat 00:0080.1%-1.84 Ah
Sat 01:0078.5%-2.26 Ah
Sat 02:0077.0%-1.41 Ah
Sat 03:0076.0%-1.34 Ah
Sat 04:0073.2%-1.33 Ah
Sat 05:0072.5%-2.09 Ah
Sat 06:0070.1%-3.16 Ah
Sat 07:0067.0%-5.07 Ah

The label “Charge (Ah)” is preserved exactly as supplied. The export's timestamp convention and the inclusion/exclusion of partial hours are undocumented.

Updated NIX verdict The Webasto proved it could keep Vanilla warm near freezing. The new electrical evidence shows the battery still reported 70.1% SOC at 6am, but also identifies a more substantial net overnight draw than on the previous night. This is the start of a useful energy budget, not yet a measured Webasto electricity bill.
10 / The lab notebook

How trustworthy
is the evidence?

A good experiment records its limits as carefully as its wins. Here's the audit trail: what was measured, what was reported, and what we would need to claim more.

Measured directly in the CSVs

High-confidence observations

Time-stamped air temperature, relative humidity, calculated dew point and absolute humidity for Outside, Vanilla and Living Room. Data were aligned to the same minute.

Overnight interval: 21:00 Fri 13 November to 06:09 Sat 14 November 2026. 550 matching timestamp rows per sensor; 549 elapsed minutes. The common exported overlap spans 06:57 Friday to 06:09 Saturday, 1,393 rows.

Reported settings and conditions

Context, not sensor measurements

Webasto set to 20°C. Previously reported Tado target of 20.5°C daytime and 17°C overnight, with Tado in the hall, plus DREO and Vortex supplementation. The fuel-pump bracket was reportedly still pending installation.

These details come from your description of the systems, not from the temperature files. The exact Tado program in force that night, heater cycles, window state and pop-top position weren't recorded.

The evidence ladder

ClaimEvidence levelWhy
Vanilla stayed at ≥20.1°C after 9pmMeasuredAll 550 interior readings
Living room cooled from 21.7°C to 16.3°CMeasuredMatching timestamp readings
Webasto was regulating via pausesPossibleTemperature dips fit the behaviour but don't log burner state
Webasto's thermostat is miscalibratedUnknownSensor placement and actual feedback reading unknown
Vanilla has better insulation than the houseNot establishedDifferent areas, heat input and heating strategies
Vanilla costs less to heat overnightNot establishedNo comparable metered energy use

Five upgrades for NIX Heat Lab 02

Keep conditions repeatable

Mark the SwitchBot positions and leave them unchanged. Record roof raised or lowered, any cracked windows/vents, and whether anyone was in Vanilla. Keep the next test duration and setting consistent.

Log the controls

Photograph or note the exact Webasto setting and controller, Tado setpoint and schedule start times. Record heater noise, pump clicks, fault codes and any deliberate manual changes.

Catch the missing energy

Align Meross power history, SmartShunt current and voltage data, and any Webasto diagnostic/fuel-runtime information. Add the oil boiler's runtime if feasible.

Ask the right comfort question

Not simply “which space was warmest?” — ask whether the living room felt comfortable at waking time, whether Vanilla's 22–23°C is hotter than desired, and what energy each change required.

What about measurement uncertainty and sensor placement?
The exported temperatures are displayed to 0.1°C, but 0.1°C display steps are not a guarantee of ±0.1°C absolute accuracy. A sensor close to a heater outlet, a window, the floor, a door seal or a warmer air layer may differ from another position. As a result, the 1.94°C SwitchBot-versus-Webasto-setting difference cannot be interpreted as a calibration error. Better future tests would compare multiple colocated sensors briefly, then return them to consistent measuring positions.
Why the humidity averages are useful but not a condensation diagnosis
Relative humidity depends strongly on air temperature. Dew point describes when air reaches saturation if it cools without changing moisture content. Condensation requires an actual surface at or below the local dew point, and surface temperature was not measured here. The 9.4°C mean dew point in Vanilla is therefore a threshold to consider, not evidence of water droplets or damp damage.

Sources and data provenance

  1. Primary evidence: Original exported SwitchBot CSVs — Outside, Vanilla and Living Room, 13–14 November 2026. The downloadable ZIP contains the unmodified files plus the timestamp-aligned temperature table.
  2. Webasto Air Top 2000 STC workshop manual — heater intake temperature sensing, heat modulation and control pause behaviour (section 3).
  3. Webasto Air Top 2000 STC installation manual — diesel heater specifications and vibration-damping fuel pump mounting (sections 8.5 and 16).
  4. Energy Saving Trust: heating controls — background on room thermostat setpoints and timed setbacks.

No external research was used to alter the raw measurements. Technical documents support descriptions of how the equipment is designed to work; they do not establish what actually happened during this individual test. AI-generated visuals throughout the feature are concept illustrations, not photographs or quantitative evidence.

11 / The verdict

So, who actually
won the night?

✅
Vanilla's heating capacity

The cabin stayed between 20.1°C and 23.0°C for all 550 compared minutes, as the outdoor temperature fell.

✅
Living-room pattern explained

A controlled evening-to-morning decline is plausible with the previously reported Tado night setback.

🔎
Temperature regulation needs a closer look

Vanilla ran about 2°C above its selected 20°C setting on average.

⚡
Energy efficiency is still an open question

To compare costs, we need measured electricity use, fuel consumption or heater operating logs — not temperature alone.

🏆 NIX HEAT LAB · COLD NIGHT 01

Vanilla wins for overnight warmth.
Not for proven efficiency.

From a repair-confidence perspective, this was an encouraging night. The Webasto demonstrably kept the van comfortable even as the outside sensor reached 1.0°C. But there is more to learn about its thermostat response, the pump mounting and the true energy cost.

Meanwhile, the house did something very different: it cooled as the night went on. Improving when and where that heat is delivered may make the next morning feel much nicer without simply turning everything up.

— NIX Heat Lab
09B / The fifth trace · the sun joins the investigation

What did Vanilla's roof panels
actually contribute?

We knew how the cabin felt, we knew what the outside temperature did, and we could follow the leisure battery's reported state of charge. A second file called Solar-History.csv finally tells us what the Victron solar charger harvested on each of the previous 30 complete days. The two “solar history” exports are not the same type of record: one is hourly battery SOC, and this one is the MPPT's daily photovoltaic history.

FRIDAY · 13 NOV120 Wh

Total PV energy converted by the MPPT that day

30 FULL DAYS3.85 kWh

15 October–13 November, 128.3 Wh per day on average

LAST SEVEN DAYS95.7 Wh/d

670 Wh in total, 25.4% lower than the 30-day mean

FRIDAY PEAK POWER52 W

26% of the panels' nominal 200 W rating at that instant

A month of measured autumn sunlight

Solar generation, day by day

Each bar is one complete day's recorded MPPT solar yield, not an estimate of sunlight and not a battery-charge percentage. Hover over the bars for daily readings. The final orange bar is Friday 13 November; the pale-blue bars show the latest seven complete days.

Measured solar energy each full day from 15 October to 13 November 2026Thirty-day daily harvest ranges from 40 to 250 watt-hours. The last seven days average 95.7 watt-hours compared with the thirty-day average of 128.3 watt-hours. November 13 harvest was 120 watt-hours.05010015020025015 Oct: 180 Wh harvested; peak 102 W15 Oct16 Oct: 250 Wh harvested; peak 110 W17 Oct: 230 Wh harvested; peak 112 W18 Oct: 240 Wh harvested; peak 87 W19 Oct: 230 Wh harvested; peak 120 W20 Oct: 50 Wh harvested; peak 24 W20 Oct21 Oct: 130 Wh harvested; peak 43 W22 Oct: 170 Wh harvested; peak 97 W23 Oct: 70 Wh harvested; peak 104 W24 Oct: 60 Wh harvested; peak 16 W25 Oct: 180 Wh harvested; peak 89 W25 Oct26 Oct: 100 Wh harvested; peak 48 W27 Oct: 140 Wh harvested; peak 82 W28 Oct: 150 Wh harvested; peak 35 W29 Oct: 140 Wh harvested; peak 65 W30 Oct: 130 Wh harvested; peak 62 W30 Oct31 Oct: 120 Wh harvested; peak 47 W1 Nov: 120 Wh harvested; peak 71 W2 Nov: 150 Wh harvested; peak 74 W3 Nov: 100 Wh harvested; peak 62 W4 Nov: 140 Wh harvested; peak 72 W4 Nov5 Nov: 40 Wh harvested; peak 13 W6 Nov: 60 Wh harvested; peak 32 W7 Nov: 120 Wh harvested; peak 37 W8 Nov: 110 Wh harvested; peak 50 W9 Nov: 60 Wh harvested; peak 21 W9 Nov10 Nov: 80 Wh harvested; peak 28 W11 Nov: 90 Wh harvested; peak 53 W12 Nov: 90 Wh harvested; peak 52 W13 Nov: 120 Wh harvested; peak 52 W13 Nov30-day mean · 128 WhDAILY SOLAR YIELD • WhFULL DAYS ONLY · partial 14 November excluded

On a phone, swipe the chart sideways to examine every daily reading.

Earlier days Last 7 days Friday's test day 30-day average
The big comparison — and its limitation. The battery's fall from 84.8% to 70.1% between 21:00 and 06:00 corresponds to a nominal 198 Wh equivalent if we assume 105 Ah and 12.8 V. Friday's solar harvest of 120 Wh is roughly 61% of that numerical equivalent. But the solar energy was harvested during Friday's daylight, while the battery draw covers Friday night into Saturday. The MPPT output may have fed electrical loads directly as well as the battery; neither dataset isolates Webasto consumption. This is a useful scale comparison, not a same-period energy balance or a diagnosis of an energy deficit.
The power ledger

Why 120 Wh doesn't mean “120 Wh stored”

☀️ Solar array120 Wh13 November · measured MPPT yield
→
⚡ MPPT outputSharedBattery charging + active DC loads
↔
🔋 Battery−14.7 pts21:00–06:00 SOC change

The 120 Wh is energy converted by the solar charger. It does not tell us how much remained in the leisure battery after contemporaneous loads. On a parked van, the fridge, standby electronics or other connected equipment may consume some as it arrives.

Useful mental arithmetic: 198 ÷ 120 ≈ 1.65. So the overnight nominal battery change is comparable in size to 1.65 days of Friday's PV harvest, if that energy were all available to refill the battery. In real use, loads and charging losses mean the recovery takes longer.

Charge-controller detective work

What did the Victron charger report?

Friday PV peak52 W
PV peak voltage35.94 V
Battery voltage range13.04–13.56 V
Time labelled Bulk539 minutes
Absorption / Float0 / 0 minutes
Recorded error historyAll zero in this 31-row export

Bulk throughout the recorded charging period, with no Absorption or Float minutes, suggests the MPPT did not progress to its later charge stages that day. That does not establish the battery's absolute SOC, because the MPPT sees its own charging process, and alternator/DC-DC charging may also exist.

A very important zero

“Consumption: 0 Wh” does not mean Vanilla used no electricity

The history includes a field named Consumption (Wh), showing 0 Wh on Friday. This is not a whole-van energy meter. Victron's MPPT load figures monitor loads connected through the charge controller's load output; loads wired elsewhere can be invisible to it. The separate SmartShunt is the more suitable source for measuring net battery current. Victron's own explanation ↗

In other words, the zeros in this column cannot contradict the overnight fall in battery SOC. Both datasets can be perfectly consistent.

A subtle Saturday-morning trap

14 November's “0 Wh” isn't a failed day

Today's row was exported in the morning, before Saturday's full daylight period had occurred. At the moment of export it recorded 0 Wh yield, a maximum PV power of just 2 W, and 31 minutes in Bulk. We must exclude this incomplete day from monthly and weekly averages. A daily 0 Wh figure early in the morning is not evidence the panels stopped working.

Across the preceding 30 complete days, the daily solar yield ranged from 40 Wh (5 November) to 250 Wh (16 October). Even in the same month, cloud, low sun, vehicle orientation and shading can radically change harvest.

Nerd corner · how to read the MPPT log correctly

Three useful comparisons — and three traps

01 · Power ≠ energy

52 W is the strongest instantaneous solar reading that Friday; 120 Wh is the accumulated daily energy. They answer different questions. A 200 W panel is a nameplate test rating, not a winter output promise.

02 · Five days reached Float

Only 5 of 30 complete days recorded any time in Float, and 6 of 30 recorded Absorption. This suggests limited solar-charging opportunity in the period, not proof that the battery was never charged by another source.

03 · No MPPT error codes

The four error-history fields were zero on every exported row. Encouraging for the charge controller; not a test of the Webasto, the entire solar installation, the panels' cleanliness or the pump mount.

04 · A solar peak isn't a diagnosis

Friday's 52 W peak is 26% of the 200 W nominal panel rating, but the file doesn't provide irradiance, panel temperature, cloud cover, shading, tilt or charger-limiting status. We cannot call that an underperforming array from this record alone.

05 · Shunt versus charger

The SmartShunt helps explain what changed in the battery reserve. The MPPT history reveals solar input. Neither can identify the heater's exact fuel burn or individual appliance wattage without a dedicated channel.

06 · The next proper experiment

Record Solar MPPT power and SmartShunt net current at the same time as the Webasto's operating state, fridge cycles and cabin temperature. Then we can separate daylight replenishment from overnight demand.

Definitions checked against Victron's official SmartSolar monitoring guide. Reported figures are preserved from the export; the 198 Wh overnight comparison uses nominal assumptions and should not be treated as a watt-hour meter reading.

Open the last 14 complete days of solar-controller readings
DateYieldPeak PVBattery voltage rangeAbsorption / Float (min)
Sat 31 Oct120 Wh47 W13.10–14.19 V0 / 0
Sun 1 Nov120 Wh71 W13.10–14.22 V128 / 0
Mon 2 Nov150 Wh74 W13.08–13.81 V0 / 0
Tue 3 Nov100 Wh62 W12.92–13.38 V0 / 0
Wed 4 Nov140 Wh72 W12.89–13.59 V0 / 0
Thu 5 Nov40 Wh13 W12.91–13.76 V0 / 0
Fri 6 Nov60 Wh32 W13.16–13.79 V0 / 0
Sat 7 Nov120 Wh37 W13.12–14.06 V0 / 0
Sun 8 Nov110 Wh50 W13.10–14.11 V0 / 0
Mon 9 Nov60 Wh21 W13.12–14.18 V0 / 0
Tue 10 Nov80 Wh28 W13.08–13.87 V0 / 0
Wed 11 Nov90 Wh53 W13.09–14.21 V41 / 52
Thu 12 Nov90 Wh52 W13.12–14.22 V30 / 317
Fri 13 Nov120 Wh52 W13.04–13.56 V0 / 0

Original export retained in the accompanying ZIP. These are Victron MPPT daily summaries, not time-matched hourly records.

Updated NIX verdict: Vanilla's Webasto delivered convincing comfort, while its leisure battery lost approximately 14.7 percentage points during the night. Friday's solar array contributed 120 Wh during the preceding day — useful, but insufficient by itself to quantify battery recovery, heater consumption or true energy efficiency. We now have two sides of the energy story, but not a complete energy balance.
12 / The next experiment

Same three sensors.
One smarter test.

Vanilla, round two

Repeat the test with the Webasto at 20°C, with the SwitchBot in the same position. Note heater starts/stops, any faults, the pop-top position and battery behaviour. Have the outstanding pump mounting issue resolved.

Aim: thermostat behaviour + reliability

Living room, round two

Try a controlled night at an 18°C Tado setting, keeping the rest of the schedule stable. Compare morning comfort and DREO/Vortex energy use against the 17°C baseline once Meross exports are available.

Aim: comfort versus kWh

The real prize isn't the highest reading on a thermometer. It's a campervan you trust and a home that feels beautifully warm, without using more energy than necessary.

Methodology: all three original CSV exports were aligned by timestamp. The formal comparison is 21:00 Fri 13 Nov through 06:09 Sat 14 Nov 2026 inclusive: 550 one-minute values per sensor. Other daytime numbers are taken from their own source files. Setting information comes from your reported configuration, not from the sensor files. All AI illustrations are artwork rather than historical photographs.
◉
NIX HEAT LABCuriosity meets a proper set of readings.

Independent personal experiment, Staffordshire, 13–14 November 2026. Temperature, relative humidity, dew point and absolute humidity from exported SwitchBot CSV files. Not laboratory-grade calorimetry or heater certification.

Back to the top ↑

Story and charts built from original data.
Original visual concept art generated with AI.
🖨️ Print this field report

↑ Article design image