A solar supercap powered Moteino (15Farad charged by BQ25504)

[perky]

You'll need a diode, preferably ideal. Once the FET has turned on in your circuit it is like a switch that remains on until the voltage drops below the threshold, but that means when the panel output drops it will just pull the voltage on the supercap down with it.

Mark.

[WhiteHare]

I added a diode to the design, and it seems to have addressed the leakage issue.  Attached is the updated schematic.

[WhiteHare]

You'll need a diode, preferably ideal. Once the FET has turned on in your circuit it is like a switch that remains on until the voltage drops below the threshold, but that means when the panel output drops it will just pull the voltage on the supercap down with it.

Mark.

The diode I used was this Schottky:  https://www.digikey.com/product-detail/en/toshiba-semiconductor-and-storage/CUS520,H3F/CUS520H3FCT-ND/5114381
It allegedly drops 280mv at a current of 10ma.  I agree it would be interesting to try your ideal diode in its place.

[perky]

Yeah, if your panel outputs a voltage that's high enough under low light conditions you'll probably get away with a normal schottky diode. It would be interesting to see how the ideal one behaves though.

Mark.

[WhiteHare]

Yesterday I also prototyped and tested a different circuit (see attached).  It doesn't need a diode because it instead uses TI's "ultra low leakage" load switch.   It doesn't seem to incur the voltage drop of the "dumb" diode-based design once the voltage on the supercap creeps up to the 1 or 2v range.  I guess because of that, at the margin under overcast light conditions it does seem to outperform the "dumb" diode-based design.

[WhiteHare]

I added a diode to the design, and it seems to have addressed the leakage issue.  Attached is the updated schematic.

After additional testing the new circuit with the dumb diode, I'm realizing I now need to wire the NCP301 input to the supercap instead of the solar panel.  Otherwise, it won't reach the intended voltage.  So, attached is the updated schematic.

[ChemE]

Your last file uploaded seems borked.

[WhiteHare]

Your last file uploaded seems borked.

Oops.  Sorry.  Attached is what I meant to attach.  However, I then discovered that it doesn't charge terminate!  Not even sure why.  Back to square one I'm afraid.  Maybe I'll be forced to use a smart diode as the workaround.

[perky]

The problem as I see it is that the output of the NCP301 remains off (i.e. high impedance) until the votage gets to about 0.25V, after which it is on, then goes off when it reaches the threshold (see figure 10, the little hump at the beginning). However, the very condition that is off turns off the charging FET. So startup is a problem, if the solar cap ever goes below 0.25V, or starts from 0V, it'll never turn on again. I can't see how you fix that without significant more complexity.

I think you might need to abandon this approach and instead use a 2.7V LDO to clamp the voltage from the solar panel and then follow with an ideal diode or use a 3.0V regulator followed by a schottky diode.

Mark.

[WhiteHare]

I think you might need to abandon this approach and instead use a 2.7V LDO to clamp the voltage from the solar panel and then follow with an ideal diode or use a 3.0V regulator followed by a schottky diode.

Out of those two options, which would be better?

[perky]

I would say the 2.7V with ideal diode is marginally better as the schottky diode version wastes power across the diode drop, and it would work down to 2.7V output rather than 3V. Also the actual output voltage is a little bit better controlled at very low currents. If your panel is powerful enough though it's much simpler with a schottky.

Mark.

[WhiteHare]

Of the two, which one do you believe would have the lowest reverse leakage?

I think for now I will shelve my p-fet design, as it seems to be going sideways.  Meanwhile, the load switch design continues to work, so that's what I'm using for now.  I changed the 10K pullup resistor for a 1M resistor.  It still works,  and doing that may have slightly reduced the current drain off the capacitor at night.  Not sure.  I also ordered some zener and Schottky diodes, so I'll give those a try once they arrive.  I plan to try Perky's ideal diode as well once the breakout board (posted above) arrives from OshPark, in about another week or so.

[WhiteHare]

Because at the moment I'm using the load switch design, this morning I made what I hope will be a slight improvement.  I used an MCP6S22 opamp to hopefully decrease the sensing load on the supercap.  See attached schematic.  I prototyped it this morning, and I am running it now.  So far, so good.  I'll have a better idea after tonight whether it's an improvement or not over the previous load switch design, and early indications are that it is.  The objective is to minimize any unnecessary current leakages or drains from the supercap.

[Edit: The inspiration for using the MCP6S22 came from the following blog, where the author thought it was a good tool for measuring capacitor leakage:  http://www.robotroom.com/Capacitor-Self-Discharge-1.html]

[WhiteHare]

Scratch that.  The results are in, and it turned out to be much leakier than the earlier load switch circuit.  I'm reverting.

That concludes the grand search.  I'm generally happy with the load switch circuit.  It works reasonably well, and compared to a BQ25504 it's less expensive and easier to assemble.  When the additional parts that  I've ordered arrive I'll try both a regular diode circuit and the ideal diode circuit.  If they turn out to perform better, then I'll likely go that route.  If not, then--at least for me--this is it.  The load switch circuit is my backstop.

[perky]

Schottky diodes with low forward voltages can have pretty bad reverse leakage (relatively speaking), so the ideal diode which is about 1uA should be better especially over temperature. If you can afford more voltage drop then you could use a silicon diode instead, but remember the actual forward voltage drop you get can vary between very low and high currents as per the diode's curve. That's a pain when you trickle charge but take gulps of current now and then.

Mark.