These are samples from the 40 A discharge test above. The percentages refer to remaining measured test capacity. They are included so you can inspect the data, not use it as a universal fuel gauge.

The 100% sample is near the start of discharge after charging; it is not a rested full-charge voltage. The 0% sample is near the test endpoint, not a recommended everyday shutdown setting. Values are rounded from the underlying samples.

For a monitor, 100% is a reference point established by its full-charge detection rules or a manual reset. For a capacity test, it is the starting point under a specified charge procedure. For a complete battery, displayed SOC can also include manufacturer-defined reserves.

Those definitions need not be identical. A display reaching 100% is not proof that every cell is balanced, and a battery showing 95% is not automatically missing 5% of its rated capacity. Look at the measurement method, actual charge completion and usable energy.

For a Victron SmartShunt, use the current manual and your battery’s charge procedure. These controls have separate jobs: [3] [4]

• Battery capacity: the Ah basis used to calculate percentage.
• Charged voltage: the voltage condition for recognising full charge. It is not the charger’s voltage command.
• Tail current: how low the current must fall while the voltage condition is met.
• Charged detection time: how long those conditions must persist together.
• Charge efficiency: compensation for the difference between charge put in and recovered.
• Peukert exponent: correction for discharge-rate effects; Victron suggests 1.05 for lithium unless the supplier specifies otherwise.

Victron’s lithium guidance suggests 99% charge efficiency. Tail current must also suit a charger that terminates at its own current threshold. Those are device-specific starting points, not universal settings for every BMS.

Do not copy a lead-acid charged-voltage table into an LFP system without checking it. Select a full-detection condition that your completed LFP charge cycle can reliably meet, without allowing an ordinary partial solar charge to reset the monitor.

Suppose a 100 Ah monitor has a 4% tail-current setting. Its current condition is below 4 A. If the charged-detection time is three minutes, the required voltage and current conditions must remain satisfied together for that period.

A passing cloud can also reduce solar current below 4 A. That does not mean the battery is full. This is why the voltage threshold and detection time matter alongside tail current. These numbers explain the logic; they are not a universal configuration prescription.

Calibration corrects the gauge; balancing addresses differences between cells. Resetting SOC to 100% does not move charge between cells. Changing a capacity setting does not fix a weak cell. A balancer also needs enough current and time to do its job.

Small cell-voltage differences in the flat middle do not prove equal SOC. Watch how the cells behave towards charge completion and under load. A cell that repeatedly reaches a limit before the others warrants checks of connections, measurements, balance and capacity. Do not solve a high-cell trip by raising its protection limit.

No. The 12°C cutoff in our long-life guide belongs to an optional conservative profile for repeated cold operation. It is not a universal manufacturer prohibition.

Actual limits depend on the battery. For example, Victron specifies charging its Lithium Battery Smart range between +5°C and +50°C. A different cell’s permitted temperature/current map does not override those completed-battery limits. [6]

Near or below freezing, use the battery’s explicit low-temperature procedure and protection. A heated battery must warm its cells before charging them outside their allowed range. Air temperature is not necessarily cell temperature.

Keep the distinction clear: manufacturer operating limits are boundaries; an extra longevity margin is a choice.

Temperature, time, charge/discharge conditions and the battery’s design all contribute. LFP ages while sitting as well as while cycling. Published research on LFP degradation shows that temperature changes the ageing mechanisms; there is no honest universal conversion from one cold charge to a fixed number of lost cycles. [7]

For everyday use: avoid unnecessary heat, repeated protection trips and leaving a discharged battery with parasitic loads attached. Use the capacity you need. If you choose a lower daily charge target, keep a deliberate plan for the full-charge events your battery requires.

For a cycle-life claim, ask for temperature, depth of discharge, rate and end-of-life capacity. “Thousands of cycles” without those conditions is incomplete information.

Usable energy ≈ nominal V × actual Ah × (starting SOC − reserve SOC) / 100.
Battery draw ≈ AC watts ÷ inverter efficiency + extra DC watts.
Runtime ≈ usable energy ÷ battery draw.

Efficiency is a decimal in the formula: 90% = 0.90. The calculator estimates energy; it does not verify that the battery can supply the requested power. A BMS or inverter may shut down before the chosen SOC reserve is reached.

For a better estimate, log real watt-hours used over a representative day and compare them with usable battery energy. Direct measurement captures load cycling that a simple constant-load calculation misses.

An SOC estimator tries to infer an internal state from measurements. Voltage feedback becomes weak where the open-circuit voltage (OCV)–SOC curve is flat: a small error in voltage, temperature correction or model parameters can imply a large error in SOC.

This is an observability problem, not simply a need for a voltmeter with more decimal places. The measured curve above illustrates the practical difficulty, although a dynamic curve includes effects that an OCV model treats separately.

Accurate estimation therefore combines current integration, a capacity model and useful reference events. More advanced estimators can also use temperature-dependent models and filtering. The objective is a plausible state estimate, not an exact percentage derived from a single voltage sample. [9]

Under discharge, a simple model is terminal voltage ≈ OCV − I × R − polarisation. Charging changes the direction of those current-related terms. Remove the load and some voltage recovers quickly; other processes relax more slowly.

Hysteresis adds dependence on charge/discharge history. It is not the same thing as the immediate I × R drop, and “rested” is not one universal number of minutes. The rest protocol, temperature and required accuracy must be defined before comparing OCV measurements.

Our charge/discharge plot does not isolate these mechanisms. It shows their combined practical effect during the recorded tests. Analog Devices’ LFP fuel-gauging note explains how a flat voltage curve and hysteresis complicate estimation. [10]

For discharge, ignoring secondary corrections:

SOC change (percentage points) = −100 × ∫ I dt / C.

Use hours for time and Ah for capacity C, with discharge current positive. Charging adds charge, with an appropriate efficiency correction. Current offset, missed loads, a wrong capacity basis and unobserved self-discharge can each accumulate into error.

A full-charge reference corrects accumulated estimation error. It does not establish the battery’s usable capacity on its own; capacity measurement needs a defined charge/discharge test.

Pack voltage is the sum of cell voltages. A sum does not tell you how those values are distributed. As an arithmetic example, a four-cell pack at 13.2 V could consist of four cells at 3.30 V, or one at 3.00 V and three at 3.40 V.

The total is identical; the cell-level situation is not. That is why a BMS monitors individual series groups and why pack SOC should be considered alongside highest/lowest cell voltage, temperature and active limits.

Entering “13.2 V” and receiving “67%” would look helpful. Without the cell model, current, temperature, history and a validated model, the precision would be unjustified.

Instead, the chart lets you inspect an identified experiment, and the runtime calculator makes its assumptions visible. When a measurement cannot answer the question reliably, understanding that limitation is useful information.

The source is Off-Grid Garage’s EVE MB31 test using a ZKE-Tech EBC-A40L. The published method charges with a 40 A limit to 3.65 V, ending near 15.3 A, then discharges at 40 A to 2.5 V. The CSV headers report 329.70 Ah charged and 330.60 Ah discharged. Temperature was not stated.

The prepared dataset integrates current against time and samples at 1% intervals. Charging is expressed as charge put in relative to the charge-test total; discharging as charge remaining relative to the discharge-test total. They are independent normalisations, not an independent measurement of absolute electrochemical SOC.

Lines connect those samples without smoothing. Four-, eight- and sixteen-cell views multiply the cell voltage by series count. No cell variation, cable drop or pack imbalance is simulated. The endpoints are test boundaries, not normal operating recommendations.

Download the exact plotted dataset · Original shared measurement files