Power architecture

Power flows both ways over the module pogo pins (user-confirmed):

• Plugged in (USB): USB → half → docked module (base charges module).
• Off USB: docked module → half (the big module pack powers the half).

The 50 mAh base cells (FH301217) are hot-swap buffers — they keep a half alive while you swap modules, they are not the runtime source.

FCC teardown sheet — top-middle cell reads FH301217 3.7V 50mAh 0.185Wh (the buffer); bottom row shows an opened module with the big blue pack and the copper Qi coil (FCC internal photos).

Consequences (all verified live)

• Discharge tests: unplugged + docked drains the big module packs slowly. To sag the base cells, undock the modules overnight.
• “True cold boot” (the only real MCU reset besides ee/10ce): USB out + modules undocked + on/off switches. Flipping the switch with USB plugged in is NOT a reset (proven in the boot-trap era).
• Module battery % is host-computed, not on the wire: de/100b payload [00, HI, LO, 00] = module-rail voltage in millivolts (big-endian). NayaFlow maps it to % with calibration ≈ 4222 mV→100 %, 3910 mV→67 %, 3709 mV→45 % (~9.3 mV/% in that region). Samples: full Touch 0x1064=4196 (app showed 4222), discharged Track 0x0E82=3714 (app 3709) — within charging drift.
• Base voltage candidate: fe/1006 bytes[1..2] big-endian mV (dumps read ~4088 left / ~4087 right; screenshots 4091/4098 later — drift direction consistent with USB charging).
• Empty-dock quirk: with no module, right reports de/100b zeros while left floats at ~0x1060 (charger rail unloaded). Use de/1001 byte1 as the true presence flag, never de/100b.

Rails at a glance (healthy board, USB)

Source	Value
Module rail (de/100b)	~4200 mV full, ~3700 mV low
Base rail (fe/1006)	~4090 mV, rises on USB charge
Module FW / base FW	0.2.3.3 / 0.3.41.0
Base cells	50 mAh buffers only
Module packs	~600–1000 mAh (the real batteries)