TPS65186 stuck in partial power state

[tkreyche]

TPS65186 stuck in partial power state after extended operation, causing silent display failure

Hardware: Inkplate 5 V2
Library: InkplateLibrary

Symptom: After 1–2 hours of continuous operation, the display stops updating silently. display.display() and partialUpdate() return without rendering. The ESP32 continues running normally.

Diagnosis via readPowerGood(): The TPS65186 PGSTAT register is stuck at 0xA0 (VDDH + VPOS asserted, VPP/VNEG/VCOM/VN not asserted) instead of the expected 0x00 (panel off) or 0xFA (all rails good). This stuck state persists indefinitely — it does not self-recover.

Root cause: Suspected I2C glitch during einkOff() leaves some TPS65186 power rails partially asserted while the software panelState variable is set to 0 (off). Subsequent einkOn() calls attempt to power up from this partial state; the TPS65186 does not complete its startup sequence, readPowerGood() never reaches 0xFA within the 250ms timeout, and einkOn() returns 0. Both display.display() and partialUpdate() then silently bail out with no visual change or error. The TPS65186 WAKEUP behavior makes readPowerGood() return 0xFF during normal sleep, which any application doing health checks needs to account for.

Workaround: Manually power-cycling the device recovers it. Soft recovery (einkOff() + 1s delay + einkOn()) also recovers it.

Suggested fix: einkOn() (and/or the display/partialUpdate functions) should detect and report failure rather than returning silently, to allow application-level recovery. Ideally, einkOff() should verify via readPowerGood() that rails actually powered down before returning.