What is the application of cell balancing in lithium batteries?
Cell Balancing
I've been doing battery pack teardown and BMS design for almost 12 years. Every time someone asks me "what exactly does balancing do?", I just show them two photos: one is a 2-year-old 36 V e-bike pack with no balancing (half the cells at 1.8 V, the rest at 4.18 V), the other is the same batch with a 100 mA passive balancer (all cells still within 12 mV after three years). That's the whole point in one picture.
In real packs, cells are never born equal. Take 100 brand-new Samsung 50E cells from the same tray: the capacity spread is usually ±30 mAh, self-discharge rate can differ by 2–3 times, and internal resistance can vary 3–5 mΩ. After 200 cycles the weak ones fall behind even more. Without balancing, the pack capacity is dictated by the weakest cell on charge and the strongest cell on discharge. You lose 20–40 % usable capacity in the first year alone.
There are only two ways people actually balance in production:
1. Passive (bleed) balancing
This is what 95 % of packs under 20 kWh use.
Real chips you'll see on boards:
TI BQ769x2 / BQ769x3 / BQ78z100
Analog Devices LTC6804 / LTC6811 / LTC6813
Maxim MAX14920 / MAX14921
NXP MC33771 / MC33772
Typical bleed current: 80–150 mA (BQ76952 can push 220 mA if you parallel two pins).
Resistor value in mass production: almost always 27 Ω or 33 Ω 2512 1 W package.
Trigger settings I actually use:
Start bleeding at 15–25 mV above pack average during CV, stop when delta drops below 6–8 mV.
If you try to chase 2 mV you'll just cook the board forever.
Heat is the real problem. A 16s scooter pack with 100 mA bleed can dump 7–8 W straight onto the BMS PCB. That's why those cheap 20-dollar e-bike boards warp and crack after one summer in the sun.
2. Active balancing
Once you go above ~30 kWh almost everyone switches to active.
Real-world examples I've torn down:
Tesla Model S/X (2012–2021): small DC-DC bricks on the HVJB, ~1 A per module
BMW i3 (all years): inductive transformer module, 24-to-1 ratio, about 2 A
BYD Blade packs: flying-capacitor boards based on Infineon TLF35568 reference design
Chinese stationary storage monsters: isolated DC-DC modules moving 15–20 A across the whole 700 V string
Off-the-shelf boards people actually buy:
1–2 A green AliExpress flying-capacitor boards (the ones with 8–16 little relays)
Heltec / Orion BMS Gen4 / REC Q / Dilithium: 5–8 A with decent cooling
Trigger strategy in actual cars
Most OEMs only balance in the top 10–15 % of SoC because the voltage curve is too flat below 3.45 V (NMC) or 3.40 V (LFP).
Tesla waits until >90 % SoC and only runs balancing for the last 30–60 min of charge.
My 2017 BMW i3 logs show the inductive balancer only wakes up when delta > 30 mV and pack current < 5 A.
What happens when you cheap out and skip balancing
I have a drawer full of 2016–2018 low-speed EV packs from mainland China that saved $0.80 by deleting the bleed resistors. After 18 months the weak cells go negative every discharge, copper dissolves, dendrites grow, separator melts, fire. I've watched 72 V 20 Ah packs drop to 8–9 Ah usable in 14 months. Slap a $12 active balancer on them and they come back to 18–19 Ah. Lesson learned the hard way.
Bottom line
Under 15 kWh → passive is fine and will stay forever (scooters, power tools, drones)
400 V+ or >50 kWh → active or you'll be swapping modules every year
Every single proper EV sold today has balancing. The only variable is whether it's a 33 Ω resistor or a $300 active board.
That's cell balancing. Nothing magical. Just stopping the good cells from being dragged down by the one lazy brother that decided to age faster.
