I figured I'd measure my ESP8266 test system, so I scrounged a 1Ω resistor and estimated the average current at 70mA, i.e. 230mW, even when just sitting on a timer. This is more than the solar panel at peak, and no amount of MPPT will help us with this, so luckily the ESP8266 has a deep sleep mode which various people have estimated < 50μA, or 165μW. Downside is until my packages get here I'm stuck with an ESP-01 which will require some precision soldering to enable deep sleep...

In the meantime I took a closer look at my logs. Occasionally analysis is no match for experimentation but I might as well do some predicting, as well as confirm previous hypotheses.

• There are times the solar panel voltage drops below the regulator input / supercap when the supercap still has a few 100mV of capacity and was below its earlier peak, which means energy is just getting wasted. Effectively at those times the panel is reporting open circuit voltage which is good for a daylight metric but not much else. All the while the capacitor is discharging through the regulator.
• This would be much less of a problem with a AA cell as there is a less direct relationship between voltage and charge.
• I really need to hook up a load and log that for a while, because really all I have proved is I can log some data and that 22F will keep an unloaded 3v3 supply up overnight which isn't that useful by itself

A combined MPPT / boost circuit would help with the wasted sunlight: varying the switching to track peak power of the panel, efficiently regulating to 2.4V to 95% utilise the capactor, followed by a step up to 3V3... but thats starting to sound a little over complicated.

Of course, I could revert to a AA rechargable, but that would be less interesting to me at this point. Rechargeables need maintenance: eventually they need replacing and likely to die at an inconvenient time. They also do cost a lot less.

So back to the back of the very rough envelope: lets assume a rather bad efficiency of 25% and 8 hours daylight, then 2.5V @ 0.02A x 6= 300mWH, and if we use 60s/hour for reading sensors and transmitting and assume 200mA in that small work period then the approximate power required is (.983 * 165μW + .017 * 0.2 * 3.3) x 24 or 273 mWH so we have 10% margin even at those extremes.

I should have done that calculation more explicitly before, luckily it works out OK. But regardless, to date this has been useful as an exercise for getting my lab back into shape and getting my mind back into the swing of working with electronics.