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

[WhiteHare]

From what I've seen so far, Schottky diodes that are rated above 200ma current do seem to allow surprisingly high reverse currents.  So, I've focused on 200ma and below.  I found this one:  https://www.digikey.com/product-detail/en/nexperia-usa-inc/1PS79SB30,115/1727-4782-1-ND/2531260
which I have on order.  It claims to have a reverse current of 500na at 25v, and so I'm assuming (?) that it will have an even lower reverse current at 5.5v and below.  At least, that's how I hope it works.     Anyhow, I should receive it on Friday, so we'll soon see.

Meanwhile, I changed the load switch circuit to use a 10 MegaOhm resistor, and it still works.  So, that's my new baseline.

[WhiteHare]

From what I've seen so far, Schottky diodes that are rated above 200ma current do seem to allow surprisingly high reverse currents.  So, I've focused on 200ma and below.  I found this one:  https://www.digikey.com/product-detail/en/nexperia-usa-inc/1PS79SB30,115/1727-4782-1-ND/2531260
which I have on order.  It claims to have a reverse current of 500na at 25v, and so I'm assuming (?) that it will have an even lower reverse current at 5.5v and below.  At least, that's how I hope it works.     Anyhow, I should receive it on Friday, so we'll soon see.

It arrived a day early, and... good news!  I tested the diode with Dave Jones's uCurrent Gold.  At 6v, reverse leakage is about 33na.  At 3v it's about 26na.  And at 1v, it's about 22na.  So, the performance seems quite good indeed!

[perky]

Good work! Nice to know there are decent schottky's out there.

Mark.

[ChemE]

Interesting.  I couldn't help but linearize those currents and plot them against voltage and the R-squared was 0.9945 which is damn near unity.  So to find all the reverse leakages just use: current (in nA) = 17.24exp(0.1335*Voltage).  Don't know jack about EE but I'm betting dollars to donuts that the theory that describes this has an e in it!

[WhiteHare]

Presently I'm trying the schottky diode in series, after an LDO.  In some sense,it works, because if you give it enough time, that arrangement will achieve the voltage target on the capacitor.  However, what I'm observing is that starting at about 0.4v below the charge target, the rate of charge (i.e. the measured current going into the capacitor) drops rather dramatically--this while being powered from a bench power supply, so the amount of available current simply isn't an issue.  This makes sense, because the amount of voltage drop across the diode is a function of the current through the diode.  The bigger the current, the bigger the voltage drop.

So, although it "works," I'm less than sanguine about the non-optimal rate of charge.  If the "smart diode" would avoid that, it would help justify its choice.

[WhiteHare]

On the other hand, the LDO plus diode approach is cheap and seems to be bulletproof.

I used a Maxim LDO that output 2.706v, and with the schottky diode (above) in series with it, the mini 5v solar panel charged a supercap to 2.526v (unloaded measurement).  Therefore, leaving a bit of safety margin so that I don't inadvertently charge a supercap beyond its 2.7v rating, what I actually need is probably a 2.85v  LDO, such as:  http://www.digikey.com/scripts/DkSearch/dksus.dll?Detail&itemSeq=224811086&uq=636271637818000871

Then the supercap should charge to about 2.65-2.68v.  Well, at least on sunny days like today.  I haven't yet tried it on a heavily overcast day.

[perky]

Yep, diodes are annoying at very low currents because you're then on that curvy bit of the forward voltage curve. That can sometimes be a couple hundred of millivolts less than when supplying a large current.

It'll be interesting to hear the results of the ideal diode experiment (assuming you still are planning to do one)

Mark.

[WhiteHare]

It'll be interesting to hear the results of the ideal diode experiment (assuming you still are planning to do one)

Yup, I have the parts, so I'll run the experiment after the breakout board arrives from Osh Park.

Eventually, I think I'd like to somehow run the Moteino from TWO supercaps: a small one, and a big one.  The idea being: quickly charge the small one first, so the Moteino can get up and running right away, even if all it does is transmit, "I'm alive" and give a steady stream of wireless reports on what the rising charge voltage is on the bigger supercap.  That would be more satisfying than setting it up only to wait a very long time just to hear that that the voltage on the big supercap finally crossed the 1.8v threshhold, because in the interim you're left wondering just what the heck, if anything, is going on with it.     

 

[WhiteHare]

The attached schematic is, so far, the best of the solar charge terminations I've tried.

Here's the notion of how it works: In the absence of any applied voltages from the solar panel, the 10K pull-down resistor on the gate to the P-channel Mosfet ensures that the p-fet remains "ON".  So, any current developped by the solar panel then passes through the p-fet and through the diode (1PS79SB30), which was discussed earlier above and which ensures that any captured charge on the supercap doesn't flow backward and discharge through the solar panel at night (or, really, whenever the voltage produced by the solar panel is less than what's stored on the supercap).

This, then, continues until the voltage on the supercap rises to the point where further charging should be terminated.  The NCP301 voltage detector is powered the voltage on the supercap and monitors the voltage.  When the monitored voltage rises to the point where further charging needs to be terminated, the NCP301 voltage detector triggers, going HIGH on its output.  That in turn turns on a load switch (which is a small circuit unto itself), which delivers the voltage produced by the solar panel to the Gate pin on the P-FET, which has the effect of turning "OFF" the current flowing through the P-FET.  Eventually, the voltage on the supercap falls, which turns off the load switch, which in-effect turns "ON" the P-FET, allowing the current to flow once again through the P-FET until the supercap reaches its voltage termination point.  The cycle repeats like this indefinitely.

I tried out the circuit this morning, and it charges at warp speed compared to the other circuits I've tried to date.  I think that's because it doesn't throttle the panel voltage, as, for instance, the LDO design does.  It can probably be further optimized, but I'm already happy with it.   

[perky]

Well that fixes the annoying problem of the NCP301's output being high impedance up to about 0.25V, that is not enough to turn on the load switch to turn off the pass FET. After that it behaves in the way you want. Excellent work!

Mark.

BTW I would consider using a NCP300 totem pole driver here and remove the 100k pull-up. If the solar input voltage is low and the supercap is below the threshold you'll sink 10uA through that pull-up. Now you've got a voltage controlled load switch you can use the totem pole output instead

[WhiteHare]

BTW I would consider using a NCP300 totem pole driver here and remove the 100k pull-up. If the solar input voltage is low and the supercap is below the threshold you'll sink 10uA through that pull-up. Now you've got a voltage controlled load switch you can use the totem pole output instead

Excellent suggestion!  I'll be sure to give that a try.

[perky]

You might also want to add a 100k pull-down on the load switch gate. It's belt and braces really, I was concerned that on power up with no voltage on the supercap the output of the NCP300 will be high impedance until it reaches 0.25V or so, so the gate of the load switch might be able to float. It'll probably get clamped by the substrate diode in the NCP300 so won't reach more than 0.55V or so but that might just be high enough to turn it on, adding the pull-down will guarantee the load switch is off (simple mod, just cut the track from the existing 100k resistor from the supercap and wire to GND). This resistor will only sink current when you're above the threshold so won't load anything.

Mark.

[WhiteHare]

Excellent suggestion!  I'll be sure to give that a try.

Good news: I just now tried it, and it works!  I didn't have an NCP300, so I substituted this:  https://www.digikey.com/product-detail/en/sii-semiconductor-corporation/S-1000C26-M5T1U/1662-1010-1-ND/6601150 It arrived just yesterday and seems notionally similar.  The modified circuit works, and this time without the pull-up resistor!

I must say I still don't grasp the gist of what the difference is between the NCP300 and the NCP301.  However, whatever the difference is, the salient difference is that a push-pull totem pole (whatever the heck that is) doesn't seem to need a pull-up resistor.

Thanks for the suggestion!

[perky]

If you look at page 2 of the spec you linked to you'll see the difference between the two types. The CMOS output version has two FETs, one for driving high and one for driving low, while the open drain version only has one for driving low. Thie open drain version requires a pull-up so that it sinks current while driving low. The CMOS version can drive high or low by switching on the relevant FET leaving the other one off. That is what is called a totem pole output in electronics terms. So you can see that the CMOS version can drive low or high without needing a pull-up.

BTW I think you might need to protect against the power-up condition between 0V and 0.95V accidentally turning on the load switch (see figure 13, the shaded bit at the bottom). If that happens the main FET is turned off and it will stop charging and remain in that state. So it might be an idea to put a silicon diode or an LED from the comparitor's output to the load switch gate and a 100k pull-down to ground on it just to make sure the load switch remains off during that initial period.

Mark.

[WhiteHare]

Here's some documentation I didn't have time to post yesterday.

Attached is a photo of the prototype I made yesterday.  On the left is the push-pull totem pole chip, mounted as a module.  In the middle is the voltage detector, also mounted as a module.  On the right is the load switch--which is actually a circuit for which I had previously made a breakout board--that's also mounted as a module.  Not shown is the diode, which is soldered on the back.

Also attached is a PDF of the schematic for the TPS22860 Load Switch Breakout Board which was used in the prototype.