I am trying to devise the simplest/cheapest possible circuit for using a small solar cell to maintain the charge on a single-cell lithium battery. The reason for low cost is that I need to build fairly large number of devices employing this circuit, and the solar cell and battery alone will probably add up to a significant fraction of the total unit cost even in the best case.
The solar cell in question is nominally 6V and produces only ~5 mA of current in full sunlight. It's that very low current output that makes things difficult. I don't want to use a larger cell if I can avoid it, due to both cost and space.
A key requirement is also that there be minimal current leakage when the battery is not fully charged, as the goal is run a very low power circuit indefinitely off the battery with only intermittent sunshine to keep the battery topped off.
My amateur's thinking is to send the solar cell current directly through the battery but with a limiter to prevent the battery from ever seeing more than 4.2V at its terminals, which is the max voltage at full charge. That requirement rules out a simple Zener diode shunt, because the knee is too soft: the battery voltage would have to drop far below 4.2V before the current effectively shuts off.
One off-the-shelf circuit I've tried is this one: 1. But it consumes all of the current from the solar cell and passes nothing to the battery, even after removing the on-board LEDs. It's clearly not intended for something that can only source 5 mA of charging current.
So what I'm hoping for is a circuit that behaves somewhat like a Zener diode and will shunt >5 mA near 4.2V but have minimal leakage at voltages much below, say, 4.0V. And of course it must not consume more than a milliamp or so of the scarce available current from the solar cell when the battery is being charged.
I am not an electrical engineer, so I don't even know whether such a circuit is achievable, but I'm imagining something involving the combination of a Zener diode and a low-power comparator or similar to convert the soft knee to a sharp threshold. Forgive my ignorance if this makes no sense.
EDIT: Some comments ask for additional details:
The device in question will operate outdoors and is tentatively expected to draw ~0.1 mA on average (it will be turned of completely most of the time and wake up 4—6 times per hour).
The proposed battery is an 18650 nominally rated at 2600 mAh. This is therefore sufficient to run the above device for at least a year without charging.
The solar charger must therefore be capable of replacing the ~0.1 mA used by the device averaged over the entire year. It is expected that the bulk of the charging will occur during the sunnier summer season and that a charging deficit will prevail during the winter.
The solar cell I'm trying out is nominally 6V (approximately 6.3V actual for I < 2 mA or so). I don't have a datasheet for it, but I've made my own measurements of I vs. V for less-than-ideal sunshine. When connected to my battery (or any other moderately conductive load), it behaves more or less like a sunlight-dependent current source of up to 5 mA. This is a factor of 50 above the expected average yearly demand. If its output averaged over a year nevertheless proves inadequate, I'll go to a larger-area cell, but I'd prefer to squeeze as much efficiency as possible out of the smaller, cheaper cell before giving up on it.
Expected temperature range most of the time is –10°C to +30°C, though it may very occasionally get much colder (–20°C or even —25°C).
Location is in the U.S., initially in Madison, WI, at 43°N, 89°W, but the device is intended for use at other locations in the future. I do have the ability (with moderate effort) to evaluate available solar energy as a function of location and date, but again, it's the yearly average that matters giving the battery capacity.
Yes, I can program a microcontroller, but I really don't want to put a second one on the device, and the first one will be powered off most of the time.