random thoughts of an engineer

 Till now I only drawn parts of schematic and talked about a mysterious control logic. Now I'll write a bit more how this logic is actually implemented. Possibilities "Typical" solution Output section can be made using one of many commercially available current mode SMPS ICs available. Input section is much more problematic. ICs usually only describe stability criteria and math for standard application and none of them have a stabilization of the input voltage as a typical application. Dedicated MPPT ICs are a few and far between and all of them are notoriously expensive. Also, there's problem of protecting a battery from over charge / over discharge - probably another IC, as well as some delay to turn on LED predefined time after the sunset. "Discrete" solution A few EEs much more skilled than me designed LED drive circuits using transistors only, One of examples is: https://www.romanblack.com/smps/a05.htm . Some of those circuits can be adapted to perform

Intro  I decided to make the input power path simpler than in the output section. Instead of a complex semi-synchronous rectification I opted for a simpler Schottky diode rectifier. But apart from that everything else got more complex. final circuit The current polarity problem When the sun is up the solar panel's output voltage is somewhere between 6V and 12V, which is higher than li-ion voltage. But when sunsets solar panel behaves like LED and sinks current from battery even with voltages as low as 3V. This means that either way the MOSFET is connected at some point in time, the body diode will conduct. To get around this fact that I used 2 MOSFETs in typical back-to-back configuration to fix the body diode "issue". It doubles Rds(on) and gate charge doubling switching losses and conduction losses. Not perfect but acceptable. I could drive both MOSFETs independently to reduce switching losses a bit (as one would be switched rarely) but then I'd need 2 level shifter