Sunlight Power Box

The Sunlight Power Box is a low-cost solar energy storage system built from lithium-ion cells salvaged from discarded laptop battery packs. A 3S4P pack of twelve screened 18650 cells stores about 111 Wh at 11.1 V nominal, charges from a 100 W solar panel, and runs LED lighting, phone charging, and a small communications router. It is designed for households and small facilities in off-grid or unreliably-supplied areas, and to be built and repaired by non-specialists using hand tools and a multimeter.

The design deliberately avoids soldering, welding, and permanent assembly. Cells are held in a screw-terminal holder so that a failed cell can be swapped by unscrewing one plate, which means a village can maintain the system without outside help.

This page documents the design, the cell screening method, and the modelling behind it. It accompanies a picture-based build manual and a free offline simulator, both linked below.

This is important to state up front. The quantitative figures below (screening yield, balancing convergence, cost of storage, carbon saving, solar balance) are model-based results, generated from Monte Carlo and parameter models using values drawn from published literature. They have not been validated against a long-term measured dataset.

A physical prototype using 3S/4S BMS boards, a B6 balance charger, and an 11.1 V pack was built and charge/discharge tested. That prototype confirmed the system configuration is physically workable; it is not the source of the numbers below.

Anyone deploying this should treat the modelled figures as design guidance and measure their own cells.

Appropriate if you can obtain discarded laptop battery packs, have a multimeter, and are willing to test every cell individually. It is not a beginner project: salvaged lithium cells can vent or catch fire if mishandled, and the screening step is not optional.

The system can run LED lighting through the evening, charge phones daily, and power a small internet router. It cannot run a fridge, iron, TV, pump, or anything on mains AC. Output is 12 V DC and 5 V USB only.

Read this section fully before sourcing any cells. The build manual expands each of these into a picture page.

• An adult must be present when opening packs, connecting wiring, and during the first charge.
• Never use water on a lithium fire. Keep a bucket of dry sand within reach at all times. If a cell smokes: move away, cover with sand, move it outdoors with a long tool.
• Never bridge + and −. Remove rings and watches. Never carry cells loose in a pocket with keys or coins.
• Never solder directly onto a cell. Sustained heat damages the separator and can cause a failure days later. This design uses a screw holder specifically to avoid it.
• Never build without both the BMS and the fuse. See below for why this is quantitatively, not just nominally, essential.
• Heat means stop. Touch-check with the back of the hand. Anything hot means switch off and investigate before continuing.
• Eye protection, gloves, and closed shoes, every session.
• Keep small children away from cells and from the finished box.

Discard without further testing any cell that is dented, crushed, corroded, leaking, swollen, burnt, damp, or that smells of solvent. Discard any cell reading below 2.5 V; do not attempt to revive deeply discharged cells.

A damaged cell can appear to work and fail weeks later. There are always more discarded packs. Rejected cells go in a metal container with sand and then to an e-waste facility, never to general waste and never to fire.

Only an adult should open a pack, using a plastic opening tool along the case seam. Never push a screwdriver into the body of a pack — puncturing a cell causes immediate fire. Inside are six or more cylindrical cells joined by thin nickel strips. Bend the strips off gently with pliers, one cell at a time, taping the positive terminal of each cell as it comes free and numbering it.

Cells with different histories drift apart in voltage as the pack cycles. Modelling three cells starting with a 0.35 V spread over 40 cycles:

Balancing mode	Voltage spread after 40 cycles	Outcome
Active	0.12 mV	Converged
Passive	19.2 mV	Converged, more slowly
None	590 mV	Diverged; highest cell exceeded 4.2 V, reaching overcharge and thermal-runaway conditions

Without balancing the spread grows without limit and the strongest cell is driven past its safe ceiling. In a pack of mixed-history salvaged cells the BMS is the safety system, not an accessory. (Modelled result.)

Screening quality determines both safety and pack lifetime, and it is the step most hobby builds skip.

1. Voltage check. Multimeter on DC volts. Below 2.5 V, discard. 2.5 V or above, continue.
2. Capacity test. Run the cell through a 4-bay tester-charger on its test cycle, which charges, discharges, and reports capacity in mAh. This takes several hours; run it on a stone or metal surface, never on bedding or cloth.
3. Accept at 2200 mAh or above. Write the measured capacity on the cell in marker.
4. You need twelve accepted cells, plus two or three spares.

Internal resistance rises as a cell degrades and catches problems that a capacity test alone can miss. Where an IR meter is available, grade on both axes:

Grade	Internal resistance	Capacity	Use
A	under 50 mΩ	2200 mAh or more	Main pack
B	under 80 mΩ	1800 mAh or more	Auxiliary or low-load use
Discard	above thresholds	below thresholds	Recycle; do not reuse

Monte Carlo modelling of packs at 70% mean state of health gives roughly 86% of cells reusable: 48.9% Grade A, 37.5% Grade B, 13.7% discarded. For a sample of 20 packs (120 cells) this gives 63 Grade A cells at a mean of 2542 mAh, comfortably more than the twelve needed for a 3S4P pack.

Yield falls sharply as pack health drops, which is why screening rather than sourcing is the limiting process. (Modelled result; consistent with published observations that 60–90% of cells in retired packs remain serviceable.)

Sort accepted cells into three parallel groups of four so that the group capacity totals are within about 100 mAh of each other. Deal them out like cards from a list sorted by capacity, then swap between groups to even the totals. Before assembly, confirm every cell sits within a narrow voltage band (for example 3.6–3.8 V); joining cells at different states of charge causes large equalising currents.

Parameter	Value	Note
Configuration	3S4P (extensible to 3S8P)	3 series × 4 parallel
Nominal voltage	11.1 V	3 × 3.7 V
Full charge	12.6 V	4.2 V per cell
Cut-off	9.0 V	3.0 V per cell
Capacity	~10 Ah / 111 Wh	12 accepted cells
Charging	MPPT controller, lithium mode	Lead-acid 14.4 V profile must never be used

Salvaged: 5–10 discarded laptop battery packs, yielding the 12 cells needed plus spares. Repair shops, offices, schools, and e-waste collection points will often give these away.

Purchased: 3S BMS, 3×4 screw-terminal cell holder, MPPT solar charge controller with a lithium mode, 100 W solar panel, 20 A fuse and holder, main on/off switch, 5 V USB output plate, 12 V LED lighting, heavy red and black wire, a ventilated plastic enclosure.

Shared tools: multimeter, 4-bay cell tester-charger, screwdrivers, plastic opening tool, insulating tape. A single multimeter shared between several households is enough, and it is the tool that keeps everyone safe.

Safety equipment: eye protection, gloves, a bucket of dry sand, a fire blanket.

A full kit including tools and personal protective equipment came to roughly 1.52 million KRW (about US$1,130). The recurring per-unit cost is far lower, since tools, meter and PPE are one-time shared purchases and the cells themselves are free.

This is the single most common way to make the system dangerous. The pack is full at 12.6 V. Controllers configured for lead-acid batteries push 14.4 V, which overcharges lithium cells and can cause fire.

Before connecting anything, confirm the controller is set to lithium / Li-ion / 3S mode and that it terminates at 12.6 V. If it cannot be set, or you cannot confirm it, do not use that controller. There is no exception to this.

Connection order: battery to the controller's BATTERY terminals first, then the panel to the PANEL terminals (do this in the evening or with the panel covered, since a panel in sunlight is live), then loads to the LOAD terminals last.

Three independent layers of protection:

1. BMS — automatic, continuous. All four balance leads must be connected in order from the negative end. Before inserting the balance plug, meter the leads against lead 0: they must read approximately 3.7 V, 7.4 V, 11.1 V in ascending steps. Wrong readings mean wrong wiring; inserting the plug anyway destroys the BMS immediately.
2. Fuse — 20 A on the positive line, next to the pack. Keep spares. A fuse that blows has done its job; find the cause before replacing it.
3. Main switch — the household emergency stop. Everyone in the building should know where it is. Switch off first, ask questions after.

Once the BMS is fitted, its two output leads are the pack's only connection point. Nothing connects to the bare pack.

Modelling a 100 W panel at 5 peak sun hours with a 0.75 loss factor for MPPT, wiring, and charge/discharge inefficiency:

• Daily generation: 375 Wh
• Daily consumption (lighting, phone charging, router): 150 Wh
• Surplus: +225 Wh per day
• Autonomy from full charge with no sun: about 1.5 days

The surplus is what makes the system survive cloudy days and gives headroom as the pack ages. (Modelled result.)

Levelized cost of storage over 20 years at 300 cycles per year:

System	LCOS (USD/kWh)
Recycled 18650 pack	0.222
New lithium-ion	0.288
Lead-acid with desulfator	0.300

The recycled pack is cheapest despite a shorter cycle life, because its initial cost is so much lower. The trade-off is more frequent replacement, which means screening strictly (favouring Grade A cells) is the main lever on lifetime cost.

On manufacturing carbon, for 48 cells (0.444 kWh): approximately 35.5 kg CO₂ for new cells against 3.6 kg for recycled, a saving of about 32 kg or 90%. The real-world benefit depends on the charging source and is maximised when charging from solar, as here. (Both modelled, using published emissions factors.)

Weekly: touch-check for warmth, look for insects or water ingress, meter the pack (should read between 9.0 V and 12.6 V).

Monthly: clean the panel, tug-test every wire, check holder screws are tight, inspect cells through the ventilation holes for swelling or leakage.

When a cell fails: switch off, unscrew the holder plate, remove the cell to the sand container, fit a tested spare. Nothing is glued or welded, so the system is repairable indefinitely. This is the point of the design: a system a community can repair is a system a community actually owns.

Placement: box indoors, in shade, on a high shelf, ventilated, away from children, flooding, and cooking fires. Panel outdoors in full sun, angled toward midday, secured against wind. Heat is what shortens battery life more than anything else.

• All quantitative results are modelled, not measured over time. Screening distributions, cycle life, and thermal behaviour in particular need calibration against real salvaged cells.
• The hardware prototype confirmed feasibility of the configuration; long-term cycle testing has not been done.
• Recycled cells have a shorter cycle life than new cells, so replacement is more frequent.
• The system is DC-only and cannot support resistive or motor loads.
• No field pilot has yet been run. Deployment data from a real installation is the obvious next step.

A picture-based build manual covers the whole process with minimal text, intended for readers who may not share a language with the designer.

A free browser-based simulator lets students and builders practise cell screening, pack design, BMS balancing, and safety procedures before handling real cells. It runs offline with no installation and supports English, Korean, Bahasa Indonesia, and Filipino:

• Simulator: https://solja6654.github.io/Sunlight-Power-Box/
• Source code: https://github.com/Solja6654/Sunlight-Power-Box

Practising on the simulator first is strongly recommended for school groups.

Designed and documented by Hajoon Kim. Contributions, corrections, and especially measured data from anyone who builds this are welcome on the discussion page.