Small update: the current consumption analysis is paying off: i found and fixed two issues in my mod, where the date entry screen after boot and the menues would consume 2.5mA instead of 250uA (10x) because of one of the chips not sleeping.
Also, in stock, there is always a 10ms busy wait before any inter chip uart communication (which means higher battery consumption), to ensure the other chip wakes up and receives the message. This is not enough wait in 1% of the cases so I first rose it to 25ms so no messages were lost (which resulted in a ble command not responding).
But in reality, even 5ms is enough 90% of the time, so i made both chips respond to the other’s wake request. This means the wait time is not anymore a fixed amount, resulting in battery savings of ~15uA. This is more important in my mod than in stock because there is a lot more inter chip communication between the chips (to e.g feed the bthome advertisements). The net result should be that the mod version has slightly better battery life than stock despite having many more features.
By the way, I will publish a complete table of how many culombs of energy each activity costs (
each advert, each button event sent to home assistant, each state update from the stm8 to the ble chip, each ada and cal cycle, etc.) I find this fascinating.