Well, it's a success, and as suspected, doesn't need nearly the juice that was first assumed, except for during bulk/absorb with high ambient temp. Before, they didn't stand a chance. Now, it does lose ground, and the temp does rise some, but indeed can pull the temp right back down to where it needs to be for the next round. It gets close on "solar" alone (as of this moment, the sun is still being simulated).
When the temp is still above the set threshold, it then goes into a trickle mode, pulling a small amount of power from the batteries themselves to meet the demand. I know someone reading this is facepalming right about now... Peltier? On battery? Yep. It's not as bad as one might think...
Soon I'll put up real graphs using real solar for charge and cooling, and even tho the ambient temps are beginning to drop back down again as fall approaches, it'll still be clear that it can and does keep up with minimal stress on the batteries. If for some reason I can't get a nice hot day and uncomfortable evening captured, I do have some artificial sun graphs captured that illustrate it readily. I'd throw them in right here but I'd have to dig a bit for the best candidate. In due time, friends.
For this post, I'm going to concentrate on the nightmare that came of developing the controller for the Peltier unit, and hopefully ditch some gotchas for anyone else that wants to take this on as a viable solution to battery temperature management. It's worth the effort in the end... but ermahgerd... LOL
First go round involved what I thought would be a relatively simple discrete component design using a single transistor as the amplifying element. It "worked" but was very flaky. It turned out that the transistor was sensitive enough to the ambient room temp that it would self adjust the target temperature as things began to warm up in the morning, effectively causing it to have to "wake up"...
Here you can see that the cooling wasn't even beginning until well into the day (days 15 and 16 capture the issue). This just wasn't any good. It still had a lot of work to do long after charging was complete and the "sun had gone down". I fought with this design for quite some time looking for a way to compensate for the temperature stability, to no avail. I suppose there's a way, but I was more determined to make it fundamentally stable rather than hacking in extra bits to make it right.
The only way this was going to happen was to get the "transistor" temperature compensated from the gitgo, and the only practical way to do that was to use an IC op amp that had all that built in.
Things got messy. Quick. To complicate matters, I was trying to use two different buck converters to run this thing; one for solar, the other for running on battery. But, it was something the LM324 could handle, right? There's 4 amps on the chip, no problem.
Breadboarding is always a joy, and this was no exception... Sorry, I can't type that with a straight face...
So, some SPICE modeling (electronics simulation software), and basic ideas all working in theory, but as anyone knows that has played with sims, they have to be taken with a grain of salt. This one didn't do too bad actually, given the complexity; real world wasn't too far off from what the screen said for a change. But there inevitably were more gotchas.
One of the first simulations for this controller
The biggest problem off the bat with this design was the dual buck concept. In theory they just seamlessly take over from one to the other as availability/demand allows/dictates, and all is well, right? Well for whatever reason, there was a hysteresis-like effect that took place whenever the gain control that brought the two into alignment got close to the threshold. All of a sudden, the output would swing full tilt, and the pots had to be spun back a good ways to once again get the output to change, at which point everything had been driven out of whack and the output went straight to zero. Lather, rinse, repeat... there wasn't any getting around it.
I finally gave in and removed the low power (battery) buck converter and decided to try and drive the whole show with the high power converter instead. This has pros and cons over the dual design, but stability was pretty high up on my list; the lower output when running on battery was of less concern (and as it turns out, not a big deal at all - more on that when I get to the "real" graphs in another update). The simplicity factor was also a real plus... not as much to keep track of.
The controller after graduating from bread to proto. Much neater, no?
... and what it looks like on paper.
I'm publishing that schematic under "use it however you want" licensing... I don't care for lawyer lingo. Have at it.
I'll say one thing about it... it's a touchy SOB that's difficult to get calibrated, and I wore out one of the cermet (trimmer) pots on the breadboard trying to do exactly that. All of the pots on the proto were installed from new stock - and good thing too - they've seen their share of tinkering, even though I halfway had a process down to tame the thing.
You'll notice by the schematic that it also does heat. It's particularly effective at this of course, since Peltier has horrific efficiency (I love it when I can play both sides of the fence and be right on both counts! LOL).
The trick to getting it to work is patience, finesse, OCD, and knowing how to get it back under control when it starts swinging wildly from too much gain
Here's a teaser showing off some of its ability... It doesn't auto-changeover, so there are dips and peaks here and there, but the tight regulation is obvious when it has something to work with. It has been turned off completely for about the last 24 hours (no charging occurred during this time either). You can see where just very recently power was reapplied). In the 4th plot (year), the reason this whole thing came about is apparent... and actually, this plot only displays the *averages* - the peak temps were much higher, well over 40C when I began to realize I had a serious problem on my hands.
As presently set, it will hold the temp within +/- 0.2C of 25C as long as it has somewhere to go with power input control. I'll get into that relationship more in the posts to come as well.
More as it comes...
Steve