Odd behaviour: Solar charging voltage crash mid-day

I’ve built a small solar-powered weather station that connects to my home system using a Jeenode. I’m using three AA NiCds that will, at full charge, deliver around 4v. As this is a Jeenode, I’m running my ATMEGA328 at 3.3v. I’m sending three pieces of data to Xively: Battery/Solar Voltage, Temperature and Humidity. I was running it indoors for a couple of months, just to make sure it worked, then I placed it in my freezer to test its behaviour in low temperatures, and it seemed to work as expected.

I’ve placed it on the side of my house, above a south facing roof section, where it now gets very warm in the early afternoons: above 48C these past couple of days. While indoors in test mode, the battery level would top out around 4.4v depending on how much sun it would get, and drop to about 3.7v just before sunup. But now, the level crashes from about 4.4v to 3.8v around 1pm, and then slowly recovers to 3.9v later in the day, then dropping at night to 3.6v.

I’m confident that the solar panel is delivering enough juice to the NiCd’s, but I’m curious about the apparent voltage crash: I don’t know if

1/ the level is really going down (I don’t want to have to sit on a hot section of roof with my voltmeter to confirm this)
2/ if there is something in the circuit design or
3/ if the components themselves are contributing to the problem.

I’ve shared a 1-day graph of Temperature and Battery Voltage along with a schematic of the charging circuit:


What type of voltage regulator (3.3V) do you use? PS: OK, I've seen the pdf - that LDO reg shall work from 3.5V up reliably.

It's an MCP1702-33.

The 1Meg/1Meg divider is not optimal, even you have there a capacitor - my experience with such values is the ADC readings were not precise (unless when averaging them).


Also I would add few larger cpacitors (10-100uF) around the voltage regulator..

What about a sw bug where you switch something on which causes a larger current draw?

I’m pretty much discounting a sw bug, unless there is something in the code that changes based on the temperature. The processor is effectively off for 98% of the time. The code is pretty simple and I don’t think there’s much chance of something in software causing a runaway current draw.

However, I did some research on the effects of temperature on rechargeable batteries: I chose NiCd’s partly because they were lying around unused and partly because I wanted batteries that work in sub-zero temperatures. NiMh batteries don’t deliver much of a charge below 0C, but NiCd’s are effective down to -20C. But I didn’t pay too much attention to high temps; here in Halifax Canada we’re on the ocean and our climate is quite moderate compared to the middle of the continent. Apparently rechargeables will not charge at any temperatures higher than 40C, and in a sealed box I’m pretty sure that by 1pm that’s what the temperature is.

We don’t normally get very many days of high heat like this, so I’m going to leave things be for now and revisit this later. I’m not concerned that the unit isn’t charging at that time, because our days are quite long during the summer. 8)

When Solar panels get hot,the output voltage drops. Could that be the problem.

either that or you have a shadow from a tree... or part of the house.

A weather station (well the temperature/humidity measuring part) shouldn't be in direct sun in the first place? Google "Stevenson screen" - you want to measure air temperature in the shade. I'd bury the battery pack 6 inches down and then the temperature variation will be negligible!