Andreas Spiess has a new video on converting the typical inexpensive TP4056 charger/protection module into a solar charger with power path, also known as load sharing. I have a couple questions about his design, and hope someone here who has solar panels could test it out.
First, the mosfet he uses is not logic level, which is a poor choice for switching voltages in the 3V-5V range. An NDP6020P (TO-220) would be better.
Second, I've run a test (see picture) using a logic level mosfet with an adjustable power supply standing in for the solar panels. And for a range of input voltages below 4V (passing cloud), the battery is powering the load through the mosfet's body diode, either fully or partially. Of course that's using a low impedance power supply and a fully charged battery, so I don't know what happens with solar panels and a partially discharged battery. I think there's even some chance the system would oscillate with a higher impedance solar source. Hopefully someone will be able to test that in real daylight.
With a fully-charged (4.2V) 18650, as the input voltage dropped, the output voltage got as low as 3.73V when the body diode was carrying the load, before rising back to 4.2V as the mosfet finally turned on fully. If the battery was 3.7V instead, the output might not support even the best 3.3V LDO when clouds are passing.
So I wonder if this is really a good solution. One thing that might help some would be to add another Schottky across the mosfet - in parallel with the body diode. That would provide a lower voltage drop than the body diode, and move any diode-related dissipation out of the mosfet. When I tested that, the lowest output voltage was 3.99V, so about a quarter volt better.