Raise the AXP228 charge current with a udev rule
On the original CM4 image the kernel left the charger at 300 mA while running; a one-line udev rule sets constant_charge_current to 2.4 A.
What it takes
- Config change
Difficulty 1 of 3 — easy.
Details
supercom32's thread started as 'why does it not seem to charge' and ended with a fix. yatli explained that the kernel failed to set the AXP228's constant charge current on CM4 cores, so a running console charged at the lowest setting, 300 mA — switched off, the PMIC falls back to its own default of around 2 A, with no LED to show it. The rule, in /etc/udev/rules.d/99-devterm-charging.rules:
KERNEL=="axp20x-battery", ATTR{constant_charge_current_max}="2500000", ATTR{constant_charge_current}="2400000"
Values are in microamps. The thread dates from 2023 and a proper fix was expected in a device-tree overlay, so read /sys/class/power_supply/axp20x-battery/constant_charge_current on your image before adding it.
Links
Open at Forum threadRelated mods
With no real suspend on the CM4, grandrew's script fakes it: busy processes throttled to 1–2 % CPU, backlight off, powersave governor — about 13 h on a charge.
Michael_Wolf's Lua monitor for clock, power draw, temperature and battery, plus the benchmark showing the backlight is what actually drains a uConsole.
Reference thread for anyone touching the battery board: J1 and J2 are identical, the two 18650s are simply paralleled, and the AXP228 is a single-cell charger.
Any genuine branded 3,000–3,500 mAh 18650 works, flat-top or button-top; listings claiming much more than 3,500 mAh are fakes.
The 21700 version of HG's NVMe battery board fits the stock backplate and roughly doubles runtime: 4 h 31 min of 1080p playback against 2 h 27 min on 18650s.
paragonnov's Debian package remaps the power key to a real screen-off sleep: backlight and LCD off, keyboard unpowered, CPU pinned to its minimum clock.