Thank you so much for the replies!
I played with charging and discharging a lithium cell I had laying around, to get an intuitive feel for how it works. After doing that, what
amyk said made a lot more sense. For whatever reason it didn't click in my head that the battery has a voltage, and that the current flow is dictated by the difference between the power supply voltage and the battery voltage. Everything makes a lot more sense now. I suppose my confusion was derived from seeing all the charge graphs which actually show the charge voltage, not the battery voltage. In fact, I don't think I ever found a charge graph that graphs the battery voltage... Trap for young players I suppose.
Lithium based rechargeable batteries can be damaged (over time) by slow charging currents.
Out of curiosity, do you have a reference that goes into more detail? I tried googling for "trickle charge dendrite" and variations but couldn't find anything that said that dendrite growth was related to slow charge. One website actually discusses slow charging and doesn't mention any ill effects (
http://www.powerstream.com/li.htm), but that talks about 0.18C. I imagine charging under 0.05C or 0.01C could be bad, because it makes cut-off more difficult, but that's really just about overcharging and causing plating.
New questions:* With respect to battery balancing, I found this post which argues for bottom balancing instead of top balancing:
http://www.myelifenow.com/2012/10/lifepo4-charging-method-dont-ruin-your.html He suggests bottom balancing each cell to 2.75V, and then setting the charge/discharge profile to 3.0V min, 3.5V max. No active balancing circuit needed. He also argues that top balancing is bad (at least for LiFePO4s). Any thoughts on that? I haven't found any technical documentation that suggests one way or the other.
* Are there advantages/disadvantages to an MCU controlled buck type lithium charger, versus something more analog?
* Most systems I've seen put the cells into packs, and then hook all the packs up, and charge the whole kit using one charge controller. Is there any reason why we can't have one tiny charge controller per cell instead? I vaguely imagine a system where you use an MPPT buck and/or boost to generate a constant voltage rail from the solar panels (say 48V). Then each cell has an MCU controlled buck that drops that rail to charge its cell, combined with a boost to supply voltage back to the rail when in discharge mode. Then the rail can be hooked up to the mains inverter. Possibly a master controller to switch the cells between charge and discharge based on system load and limit their charge based on system supply. To me that makes it so each cell is individually balanced, and cells can be added and removed from the system very easily (for replacement, or just growing the system slowly over time). Also seems like you could take better advantage of economies of scale by buying lots of small components, rather than a few very large components.