Not so grand plans
April 6, 2009
I decided, that the next circuit I build, will do one thing well, and I’ll worry about the rest later.My brother and I have been over the issue of “what do you want in a bicycle light”, and what we both basically want (and both currently compromise on) is something that is bright as all get out, silent, requires no fiddling to keep it running, and also stays on for a bit when stopped.
Crudely, a “dynamo plus a battery plus LEDs”, but the problem is that you can overcharge your batteries, and the battery control chips that I can find seem to assume conditions that do not hold on a bicycle, like civilized temperatures that don’t change much. So, instead, I plan only to either “do no harm”, which means that the standlight eventually starts to dim and the batteries need charging (that might take months), or perhaps to be far less ambitious in my charging plans — stop at 80%, instead of aiming for 100%. It looks like voltage changes suffice for Sanyo Eneloop, if I stop early (and I want to use Eneloop because they have low self-discharge).
It turns out that 4 lithium batteries would probably be just right. Compact, light, and if you stay below a critical voltage you won’t overcharge them. 3 volts per cell is the minimum output, 4.1 volts per cell is the maximum charge, call it 4 to be safe, and the operating voltage will normally be in the 3.5-3.7 range. Times 4 gives a minimum of 12, usually 14, which is plenty to run the lights at a medium level. I still need to scope out the temperature effects, but wonder of wonders, the ATTiny chips have an on-board temperature sensor.
I also decided that the previous circuit was just too spread out; real estate costs. It should be possible to stand things up, perhaps with some tape and/or glue to help lash them together and cut down on vibration. I also simplified the 3-part ramp to the BuckPuck control, and settled on a 2-part circuit instead. The shunt is not quite as sharp, but is still better than what I am using now.
The trickier part (for me, now merely a hobbyist EE) was to figure out a circuit that could be powered from the battery itself, that would allow current to flow in either direction when switched on, and that could be switched with TTL voltages. Given that, a microprocessor can sit and watch the wheels go round, or not, and watch the system voltage, and when things look slow, it can turn on the batteries. Optionally, it can also turn off the lights, so that they can be blinked or varied while stopped, and it can get limited inputs (inconveniently, using analog-to-digital conversion) on its reset pin.
A bicycle light system does get to play one trick that is not available to (say) a cell phone or a personal computer; it can turn the lights on or off for a few milliseconds while the batteries are charging, and see how the battery voltage responds to the change in the load, and nobody’s going to notice. This might be a good indicator, I don’t know yet. It is, of course, a Simple Matter of Programming.
The software does NOT get to control the shunt that clamps the system voltage below part-frying levels.
The other tricky part is not knowing exactly how much noise there will be in this system.
The BuckPuck is a literal black box, but is presumed to contain a switching power supply. The capacitors feeding it are big. The microprocessor is alleged to need noise reduction, so I took the recommended precautions and followed them as well as I could (so, the ground fed to the voltage dividers for the analog signals, is at the other end of the circuit from the ground feeding the processor, and I decorated the processor with capacitors. Is this enough?