LiFePO4 battery bank sizing comes down to one line: daily watt-hours × days of autonomy ÷ 0.8 depth of discharge ÷ inverter efficiency. A homelab averaging 60 W needs about 4 to 5 kWh of nameplate LiFePO4 for two days without sun, which is one 48 V 100 Ah rack battery or four 12 V 100 Ah batteries.
Everything after that is choosing the inputs honestly. The load figure is the one people fudge, autonomy is the one that drives cost, and bank voltage decides how much copper you buy. If the rest of the system is still a sketch, the full design order is in sizing an off-grid solar system for always-on homelab loads. This piece stays on the battery.
What goes into the sizing formula?
Five inputs and an optional sixth: average load, hours per day (24, because it is a homelab), conversion efficiency, days of autonomy, allowed depth of discharge, and an ageing margin. Only depth of discharge comes off a datasheet. The rest are measurements or decisions, and a bad measurement costs more than any chemistry choice.
Daily energy. Meter the lab at the wall for a week, and make sure that week includes a backup window and a ZFS scrub. Average watts × 24 is your daily watt-hours. Power supply nameplates are ceilings the hardware never touches at idle, so sizing from labels buys battery you will never discharge.
Conversion losses. If the lab runs through an inverter, divide by its efficiency at your actual load, not the headline figure. Use 0.9 as a placeholder, then read the datasheet for idle draw, because it dominates at homelab-sized loads: an inverter idling at 20 W adds 480 Wh a day, a third of the example below. Gear that runs from a DC brick can often be fed from the battery through a DC-DC converter instead, skipping the inverter entirely. Check each device’s input voltage range before getting clever.
Days of autonomy. This is the expensive input. IEEE 1562, the recommended practice for sizing stand-alone PV systems, calls for 5 to 7 days for non-critical loads in sunny areas and 7 to 14 days for critical loads or low-sun areas, designed against the month with the least sun and the highest load, according to SunWize’s summary of the standard. That guidance is written for stand-alone sites where the array is the only charger. A homelab with a grid connection, a generator, or an owner willing to shut down the media server for a wet weekend can live on one to two days.
Depth of discharge. Victron’s comparison of lithium and AGM puts the practical depth of discharge at 50% for AGM and 80% for lithium iron phosphate, and concludes that an AGM bank needs about twice the amp-hour rating of a lithium one. Use 0.8.
Ageing margin. Rated cycle life ends at a capacity threshold, conventionally 80% of original. That is also the end point Sandia National Laboratories used in its multi-year cycling study of commercial LFP, NCA and NMC cells (Preger et al., 2020). If you want two days of autonomy in year ten rather than year one, divide by 0.8 again.
Worked example: a mini-PC, a NAS and a switch
Say a week of metering shows an average 60 W for an N100 mini-PC, a four-bay NAS, a 2.5 GbE switch and the router, all fed through a small inverter.
| Step | Arithmetic | Result |
|---|---|---|
| Daily energy at the load | 60 W × 24 h | 1,440 Wh |
| Off the battery (90% inverter) | 1,440 ÷ 0.9 | 1,600 Wh |
| Two days of autonomy | 1,600 × 2 | 3,200 Wh |
| 80% depth of discharge | 3,200 ÷ 0.8 | 4,000 Wh nameplate |
| End-of-life margin | 4,000 ÷ 0.8 | 5,000 Wh nameplate |
To get amp-hours, divide by nominal bank voltage. LiFePO4 cells are 3.2 V nominal, so a “12 V” battery is four cells at 12.8 V, “24 V” is 25.6 V, and a “48 V” rack battery is 16 cells at 51.2 V.
| Nominal bank | Ah for 4,000 Wh | Ah for 5,000 Wh | Off-the-shelf fit |
|---|---|---|---|
| 12.8 V | 313 Ah | 391 Ah | Four 100 Ah batteries in parallel |
| 25.6 V | 156 Ah | 195 Ah | One 200 Ah battery |
| 51.2 V | 78 Ah | 98 Ah | One 100 Ah rack battery |
All three fits land on 5.12 kWh, a common module size. EG4’s LifePower4 48V v2, for example, is a 51.2 V, 100 Ah, 5.12 kWh unit built from 16 cells in series and rated for more than 6,000 cycles at 80% depth of discharge, according to EG4’s product page. One of those covers this lab with the ageing margin included.
Now move the dials. Three days instead of two takes the end-of-life figure to 7,500 Wh. Running the same loads on DC and dropping the inverter brings the two-day figure down to 4,500 Wh. Every 10 W shaved off the continuous load removes roughly 830 Wh from the bank, which is why retiring the old dual-socket server humming in the corner is almost always cheaper than buying the battery to keep it alive.
Which bank voltage should a homelab use?
Under a few hundred watts of continuous load with no large inverter, 12.8 V or 25.6 V is fine and keeps DC-direct options open for mini-PCs and network gear. Once a kilowatt-class inverter, a bigger array or loads beyond the lab enter the picture, go 48 V. The reason is current: the same 1 kW is about 83 A at 12 V and about 21 A at 48 V, and cable gauge, fusing and voltage drop all scale with it.
Charge current scales the same way, which matters when choosing a controller; the trade-offs are in MPPT vs PWM charge controllers for homelab solar. The 48 V rack ecosystem also grows cleanly: EG4 lists up to 64 LifePower4 units in parallel. You will not need 64. You might need two.
Check the second constraint while you are here: BMS continuous current. The LifePower4 above uses a 100 A BMS, roughly 5 kW at 51.2 V, and inverter surge plus charge current has to fit inside that, shared across parallel units. For a 60 W lab this never bites. For a bank that also feeds a well pump or 240 V loads through a split-phase inverter, it does.
Does depth of discharge matter for LiFePO4?
Yes, but far less than for lead-acid. LiFePO4 handles deep cycling well enough that 80% is a sensible planning figure where AGM would get 50%. The cycle count on a datasheet, though, is a lab result under specific conditions, not a promise.
Sandia’s study is the useful corrective. Even with every cell operated inside its manufacturer’s specifications, varying discharge rate, depth of discharge and temperature changed the time to reach 80% capacity by thousands of hours and cycles within each chemistry (Preger et al.). Your shed in August counts as one of those conditions.
The stronger argument for planning at 80% is operational. The bottom of the bank is where the BMS low-voltage cutoff lives, and hitting it drops every machine at once, mid-write. A 20% reserve gives a shutdown script polling the battery monitor or inverter time to stop the NAS cleanly, well above the cutoff. Use a shunt-based monitor for that signal: LiFePO4’s flat voltage curve makes resting voltage a poor state-of-charge gauge.
Where the bank lives changes its size
Cold is the hidden sizing input. Battle Born’s guidance is that LiFePO4 should not be charged below about 25°F unless the battery has internal heating, although it can discharge down to about -4°F, and available capacity drops temporarily in the cold (Battle Born Batteries). The BMS enforces that by blocking charge current.
The consequence: a bank in an unheated garage can sit under a clear January sky and refuse charge until the enclosure warms, so effective winter autonomy is shorter than the spreadsheet says. Options, cheapest first: move the bank indoors near the rack, which is already a space heater; buy self-heating batteries; or add a day of capacity. Cold-blocked charging looks like a fault from outside, and the diagnostic order is in LiFePO4 solar battery not charging?.
Heat is the opposite problem: the same page puts the upper operating limit around 135°F and notes that sustained heat accelerates wear, so keep the bank out of an unventilated closet beside a warm NAS.
Mistakes that wreck the number
- Forgetting inverter idle draw. On a 60 W load, 20 W of idle draw needs about a third more battery.
- Buying autonomy instead of a fallback. Fourteen days makes sense for a remote telecom site. A Plex server is not one, and two days of battery plus a generator or grid transfer costs far less than five days of battery.
- Sizing the bank without the array. A bank the panels cannot refill in the worst month only delays the outage. Run the numbers in solar panels for a homelab: how many do you need? before ordering.
FAQ
How many 100 Ah LiFePO4 batteries do I need?
Divide your nameplate watt-hour target by 1,280, the energy in one 12.8 V 100 Ah battery, and round up. The 60 W homelab example above needs 4,000 to 5,000 Wh, which is four batteries in parallel either way. At 48 V, one 51.2 V 100 Ah rack battery stores the same 5,120 Wh in one box.
How many days of autonomy should a LiFePO4 bank have?
One to two days is enough for a homelab with grid power, a generator, or loads you are willing to shut down. IEEE 1562 recommends 5 to 14 days, but that guidance targets stand-alone PV sites with no other charging source. Each extra day adds a full day of energy, divided by 0.8.
Can I add more LiFePO4 batteries to my bank later?
You can, but new batteries will not share load evenly with aged ones, and the bank tends to behave like its weakest member. Rack systems designed to parallel identical modules over a shared communication bus cope better than mixed 12 V drop-ins. Sizing once, with an ageing margin, is usually cheaper than expanding.
How long does a LiFePO4 battery bank last?
A quality bank should outlast most homelab hardware: EG4 rates its 48 V rack battery for more than 6,000 cycles at 80% depth of discharge, over 16 years at one cycle a day. Temperature and charge settings shorten that, and Sandia found operating conditions alone, all within spec, moved time to 80% capacity by thousands of cycles.
Related across the network
- UPS Sizing for a Mini PC Homelab: Runtime and VA Math — minilabhq.com
- Homelab Watt Budget: Planning a Build Around Idle Power — minilabhq.com
