No wakeup after complete discharge of the batteries

[pino_otto]

The small UPS board seems to work better than the TP4056 with regard to the clean shutdown issue of the ESP.
When the battery has not enough power to keep the ESP working, the UPS board provides ~0V, so the ESP is shutdown well, without going into a zombie status (due to low voltage).
However I still have to solve the problem of the clean boot of the ESP. I am going to try to make a twilight circuit with a photoresistor (LDR) and a relay, that will close the power line to ESP, only when there is enough sun light, so that the solar panel can provide enough power to boot the ESP well.

[chemmex]

Don't you consider implementing deepsleep mode?

[pino_otto]

I have already implemented deep sleep and it's working well, when the ESP is properly powered.

My system is powered by a solar panel. The problem is how to make the ESP boot well, when the battery is discharged (if it happens at night) and then the sun is coming back. If I connect directly the solar panel to the system ((mini-ups or battery charger) + wemos/esp), the power provided by the solar panel increases very slowly in the morning and it has not enough current to boot the ESP. The ESP goes in a zombie status and it cannot boot again, even when there is a big sun light later.

So I am trying to build a twilight circuit that will give power to the system only when there is enough sun, that means enough current to boot the ESP well.
This is a reference for a simple twilight circuit: https://www.electronics-tutorials.ws/wp ... o-io20.gif

[pino_otto]

Maybe I have to buy a bigger solar panel.

[chemmex]

You should decide on the strategy to achieve your goal. In my opinion, you should first calculate the energy your MCU and peripherals can consume in active and sleeping time. Then choose appropriate solar panel and battery. Then try to optimize your circuit for unnecessary power losses. Below are my thoughts on the subject:

1. Use a switching regulator for powering the ESP, they are more efficient in the whole voltage range of the battery than any type of LDO. It should sit between the battery and 3.3 Vin of the ESP. Do not use 5V boost regulators, they will feed the Wemos' LDO rather than your circuit. As I recommended on a previous page, TPS63020 is a good choice for that
2. Try to minimize the consumption in deepsleep mode. The ESP itself draws ~20uA, while your peripherals can consume much more. You can also switch the load with FETs. Also take a look at voltage dividers and quiescent power of the regulator.
3. If you have an undervoltage detector or protection circuit, set the hysteresis large enough (0.3V or more) or you will be stuck in a boot loop. With 18650type LiPo battery, you can go as low as 2.9-2.8 V, and recover at 3.2-3.3V. My real setup has rising threshold set at 3.2V, this is enough to wake up and power a PMS7003 sensor with a fan during almost a minute then go to sleep again. With 10 cm2 5V panel with almost no sun in winter
4. Optimize the active time, you may use static IP setup and minimize the time needed to read sensors and publish the results
5. Last but not least, your charging circuit should be efficient, too. In low sun time your panels' available energy is usually 10 times less than in full sun, and it is not that easy to extract. Basically, your battery's charging ability should match the panel's output power while the sunlight and battery state of charge changes, that's what the ideal solar charging circuit is. The TP4056 is rather a dumb linear charger, but you have a chance to adjust a typical charging current to the maximum power of the panel (in maximum sunlight) by choosing a right shunt resistor. Andreas Speiss has a good video explaining that hack. If you want to go further, there are MPPT-type controllers which can charge even in low light conditions

[pino_otto]

Chemmex, thank you very much for the many suggestions.

The twilight circuit seems to work. I will keep it monitoring to see if it is stable in the long run.