A few more updates* to Rev^2, but this is working well, now. At first it made sense to snap this in next to my project boards, but I do not want those large enclosures floating around in the yard, so while stressing about getting this solar buck charger to work (the way I wanted), the reality is it is going to be integrated into my controller board. With that said, I am so glad I decided to do this stand alone, it has been a challenge, but seeing it work is awesome.
I will try to sum this up, but there is a lot going on, so see the LT3652 datasheet* also. I did not quite go with the LT datasheet for the battery temperature compensation after finding some SLA data** from PowerStream and Panasonic. The room temperature float voltage is 6.83V, and it takes any DC power source above 14.5V and bellow 32V, but will pull any power it can from a constrained source at 14.3V. That means when the rain clouds show up 20 mA may be the max from PV (34mA after converted down to 6.8V) and the battery will have to make up the difference (these MCU's only work if there is a power glut). The power up sequence I ended up with prevents starting the load (Vout) until the battery has some reserves, although the value (6.5V) used is not really optimized. This allows a clean power up event (at 6.5V) even if the PV is providing a constrained (morning) power flow. And if the sun does not keep up with the load it will disconnect (at 6V). The buck converter would then immediately raise the PWR voltage (unless the sun is blocked) and connect the battery which starts to charge up. This slow connect and disconnect process is acceptable for an MCU or any CMOS device. An MCU can also track Vout and sleep or cause a disconnect at the time of its choice rather than wait for the 6V sensor to decide.
This is the everycircuit simulation* of the power sequence control, I know it is impossible to follow, and that is my complaint. This could really be a useful tool to show how things work, but since I can't add my own box of things that should be in a sub-circuit (TLVH431) and lock the layout paths it is almost, but not quite, worthless. It is still fun to play with. The device parasitics have also been a problem, e.g. the Zener between a BJT base and emitter and the body diode found in discrete MOSFET, which is a problem in all the SPICE simulators on my list (ICAP, LTSPICE, PSPICE).
- EveryCircuit - Solar Power Sequence Control (needs Chrome to run)