Update on 14.10.2022:

1. In order to calculate the estimate battery life time, we need to know the current consumption for the sensor node at different stages.  In this case it's mainly include:

A. All the devices sleep/suspend (around 1.5uA to 2.5uA)

B. EPD refresh

C. BME680 measure T&H&P

D. BME680 measure AQI(gas)

E. STM32WL send data to LoraWan(TTN) and open receive windows(RX1/RX2)

make a rough calculation, the system can running around 2 months, with EPD refresh once every minute, AQI update every 5 mins,  and commit date to TTN every 15 mins. And it seems not bad.

But when I build another one in order to test the real battery lifetime, the Node stopped working after several hours, even it equipmed with 2 new CR2450 batteries. then check the battery voltage, it's only around 2.6V, the nominal value is 3V, and new battery voltage is 3.3v, it's dropped to fast.  Then reconnect the 2.6V batteries with the node, it works again. After analysis, found it's caused by the peak current when the RF working together with EPD refresh, in some conditions the BME680's heater also workings in parallel. At this moment, the peak current is more than 100ma, which is far beyond the poor battery's capacity. There is no detail specification for the peak current of CR2450 battery, Energizer said it's 9mA(https://data.energizer.com/pdfs/cr2450.pdf),  Murata said it's new battery CR2450R: "The maximum pulse discharge current*1 has been doubled to 50mA compared to that of Standard. "(https://www.murata.com/en-global/products/productdetail?partno=CR2450R), which means the standard CR2450 can provide 25mA peak current.

This current waveform captures the EPD refresh and Lora send, it's already over 40mA, and the battery doesn't survive when the AQI sensor's heaters (plus ~16mA 1.6s) work in parallel..

I want to order Murata's CR2450R battery, but found the delivery time is quite long due to they are "dangerous goods". Then using supercapacitor is also may a solution, even it may has several uA leakage current , I found out out 2PCs 0.1F/5.5V supercapacitors which purchased 2 years ago, but found they are defected, they don't have any capacity anymore.

I nearly want to give up, then checked the STM32WL55 datasheet again,  found there is an application note to reduce the RF power. After optimization, the peak current for RF reduced from 33mA to 20.5mA, the SNR seems not influnced by the optimization.

Before RF Optimization:

After Optimization:

encouraged by this, I creat a new Look Up table for the EPD partial refresh, which reduce the partial refresh time from 1.8s to 1.2s, the payback is stronger afterimages. but it doesn't matter, after 10 cycles(equal 10 mins) faster refresh execute one time slower partial refresh. And make a full dispay refresh every 2 hours.

Then add status machine to avoid EPD/RF/AQI_Gas heater working in parallel, in this way we can limit the peak current within 25mA, which means ideally 12.5mA peak current for each battery. 

To check what's the real battery lifetime, a new battery is installed in the Node at running from 12.Oct evening, let's see how long it can last(seems the battery voltage strongly influnced by the temperature).

2. Some interesting tests

    2.1 Put the Node into the refrigerator freezer(-18deg), the EPD can't working under this low temperature but it can recovery when the temperature recovers.

    2.2 The cracked EPD can still keep the last image

 2.3 Two Sensor Nodes has similar measurement result

2.4 The PCB antenna has similar SNR as the spring antena when the housing is not mounted, but when the housing is mounted, the PCB antenna is around 1.5db weaker then the spring antena. This may caused by the PCB antenna is partialy covered by the battery, which has metal housing.

You can find the PCB gerber file and FW for the MCU, having fun.