Solar Energy Harvesting for wireless motes

  • Hero Member

    This new thread is a continuation of an earlier discussion started here:

    Something interesting: if you want to minimize the voltage drop across your solar cell's blocking diode, pick a schottky that's rated for high current (even if you're sending low current through it):

  • Hero Member

    I added an alternative LDO (tolerant of input voltages up to 30v) to the BOM for:
    so that up to 4 of the keychain solar cells can be linked in series and yet still managed as a single input.

  • Hello! 🙂

    The topic of powering MySensors nodes with solar energy interests me and I have been looking for solutions lately to do so.

    I have found an awesome energy harvesting chip with a lot of features to create hassle-free solar-powered devices. It's called the AEM10941 by e-peas.

    What is really great with this chip is that it takes a solar cell as input, charges a battery / supercapacitor and outputs two stable selectable voltages. So it does energy harvesting & boost / buck converter.
    Moreover, if the sun isn't shining for a long period of time, you can also add a "backup battery" that will be used when your rechargeable medium runs out of juice.

    Jasper Sikken built a nice board during the Hackaday "energy harvesting" competition with this chip: As you can see, it requires very few components, which is great!

    I plan to use this chip for my outdoor weather station that I am building with MySensors (I have a BME280 which measures temperature, humidity & pressure). I'll very probably build my own board with 2.54mm-spaced pins to manipulate it more easily! 🙂

  • Hero Member

    @encrypt Sounds promising! Please do let us know what you think of it after you've had a chance to try it.

  • Hardware Contributor

    @Encrypt they look interesting but where do you buy those chips and at what price ?

  • There is a "Where to buy" button at the top of the page I linked.

    As you can see on, the page, they can be bought worldwide on the Fujitsu webshop for €4 per unit:

    You can also buy the "ready to use" board built and sold by Jasper Sikken on Tindie:

    I've just found now that he has posted the schematics on GitHub, that's great:

  • Hardware Contributor

    @encrypt said in Solar Energy Harvesting for wireless motes:

    There is a "Where to buy" button at the top of the page I linked.

    As you can see on, the page, they can be bought worldwide on the Fujitsu webshop for €4 per unit:

    Yes thank you, I have seen that but I hoped there would be another source for the bare chip. I have to enter all my personal information to create an account and have an idea of the shipping price, which is annoying. Can you or anyone else with already an account check how much the shipping is for a few units for "non EU" destination ? Thank you !

  • Hello!

    Sorry for the delay, it seems I haven't been notified...

    As far as I'm concerned, I have no account on that website.
    I could create one, even if I'm planning to use the chip on the future...

    I'll come back to you if I do so 🙂

  • Hero Member

    Here's my current thinking on what to do next:

    It's a roll-my-own LDO, which has two advantages: 1. I can pick whatever maximum voltage I want for the solar input (I'm thinking a max of 40v would cover everything), and 2. It gives me access to the charge finished signal (similar to a power_good signal), which will set things up for the next stage, which is dumping surplus charge into a much larger capacitor. i.e. that next stage after this will give priority to quickly charging a small capacitor (say 100uF) to immediately power-up the MCU, and then only afterward to charge the really big supercap.

    What do you guys think?

  • Hero Member

    I forgot to mention: another advantage is that it can start charging the capacitor at lower solar voltages than what a pre-made LDO (at least the ones that can withstand 40v) can. AFAIK, the pre-made 40v LDO's don't pass current until the voltages are 2v+, or thereabouts. In theory, this one could start charging at around 0.4v to 0.8v (depending on how many diodes I end up needing to guarantee a full-shutoff at the PMOS).

  • Hero Member

    I have doubts the earlier version would have charged the capacitor when the harvested solar was of weak voltage. However, this new version should do that:
    It somewhat depends on the behavior of the voltage detector during the power-up phase, so I won't know for sure without building it and then testing it. Also, it does not address how to cleanly disconnect the charger so that the MCU can take open circuit voltage measurements of the solar cell.

  • Hero Member

    OK, hopefully this new version has addressed all the issues, including the clean disconnect for the voltage measurement:

    I guess if no one has any comments, then like the little red hen I'll be going this alone.

  • Personally, I'd just use the AEM10941, ah ah.

    It has MPPT module, starts charging the storage medium when your solar cell reached 50mV... It's hard to beat that 😛

    That being said, it's made to work with small solar cells, so if you plan to use "big" panels (via V > 5V or I > 110 mA), then building your own circuit will be better.

    I can't say much about your design though, I haven't enough knowledge on that matter 😉

  • Hero Member

    @encrypt You make a persuasive case. The $4 chip pricing seems not unreasonable. Maybe I should instead put my effort into figuring out how to reliably solder those kinds of tiny chips.

  • You can also go for the TI Bq25570 which has support to solar Panels, Thermal and Piezo Electric Generators and available to buy everywhere !

  • Hero Member

    Here's a simplified version:

    How it works:
    This particular NCP301 voltage detector goes high when the voltage reaches 2.7v (most voltage detectors only go high when they fall below a target voltage). Thus, that should trigger the NFET to open, which should provide a positive bias to the base of the PNP transistor, which I'm hoping will be enough to completely turn-off the current flow through the PNP transistor. If I'm lucky, fewer than 3 diodes will be needed. The resistor values may need tweeking. For the NCP301, the typical quiescent current is 500na, which is lower than any of the hysteresis chips I checked.


    1. This should allow charging of the storage capacitor at any voltages above the diode and PNP voltage drops, which should be low compared to any pre-made LDO (at least all the ones I'm aware of).

    2. It will allow input voltages up to 40v from the solar cells, which is higher than what the pre-made LDO's allow (again, at least all the ones that I'm aware of). Thus in a very dimly lit environment, a string of solar cells could be put in series to counteract the dimness and yet still charge the capacitor.
      As always, comments are welcome.

  • Hero Member

    OK, I think this one's the winner:

    It uses only 4 parts and is compatible with any input voltage: Thus you can stack as many or as few solar cells in series as you want to, and the circuit should work the same regardless.

    How it works: current from the solar cell/panel flows through the diode to charge a capacitor (either surface mounted to the PCB or attached to the PCB using the provided through-holes). When the voltage reaches 2.7v, the voltage detector goes high, burning off 50ma of current through the 56 ohm resistor until the voltage drops below its hysteresis point. As long as the solar cell/panel's current does not exceed 50ma, this design should work. If you need to handle an input current of greater than 50ma, then simply modify the circuit to instead connect the voltage detector output to an appropriately sized mosfet for that current, and then use that mosfet to dissipate the surplus current through a suitable resistor to ground.

    In my case I'm be choosing a diode with a maximum of 100na reverse current leakage, but you can choose whatever diode you want to fit your particular trade-offs.

    I presume that by choosing a different voltage detector you could just as easily charge a battery instead of a supercap, if that's what you wanted to do.

  • Hero Member

    I just now sent the files to fabrication. If it tests out as expected, then I'll post the gerber files.

  • Hero Member

    Just for fun I added an LED that will flash each time the capacitor discharges a little to stay within its maximum 2.7v. Although brief, it indicates that solar harvesting is working and that the capacitor is fully charged.


  • Hero Member

    BTW, I found a voltage detector that consumes just 150na, so I'll probably switch to using that because it will be important for the nextgen version which prioritizes the charging of a bootstrap cap before dumping solar charge into a much larger supercap.

  • Hero Member

    It finally dawned on me that a very solid minimalist circuit can be accomplished using just two diodes:

    The trick to making it work is selecting a diode D2 that has a forward voltage drop of 2.7v. For instance, CMF05(TE12L,Q,M) is such a diode, and on Digikey it costs a mere 40 cents:

    The result is a circuit that's not only inexpensive but can withstand any voltage that might be applied to it and, in realistic terms, any charge current that it's likely to encounter as well. And by picking diode D1 to have a low forward current (and for that, any common diode will do), it will charge quickly as well. So, better, faster, cheaper. Usually you only get to pick two of those. 🙂

  • Hardware Contributor

    @neverdie what will prevent your supercap from discharging through D2 ?
    When current flowing through the diode gets low, the forward voltage gets lower too so your supercap will be drained.

  • Hero Member

    @nca78 Thanks for pointing that out. I don't have in my possession the diode with the 2.7v forward voltage drop, so I ran some tests on a red LED insteaad. According to my multimeter, the red LED has a forward voltage drop of 1.8v. I hooked it into a uCurrent Gold to measure current and then decreased the voltage below 1.8v to see how the LED current reacted. You are right. Voltage had dropped all the way to 1.4v before I could no longer see any detectable current on the micro amp scale. Then, switching to the nano-amp scale, it wasn't until I had reduced the voltage to one volt that I could no longer discern any current on the nano amp scale. I had thought the current would cut-off much sooner than that, but I was wrong. Thank you once again.

    Back to the drawing board!

  • Hero Member

    @nca78 What if a 2.7v zener diode, reverse biased, were used instead? Would it have essentially the same problem? Scratch that. Probably not, except in limited cases, and even those might require hand selected zeners.

  • Hero Member

    OK, for an extra 40 cents, this new circuit should work perfectly without more than 150na of current drain:

    I suspect 150na is less than the self discharge rate for any supercap.

    For larger solar panels, expunging surplus capacitor charge through an nfet would be the prudent way to go, but this circuit should work for the tiny solar panels that I'm currently focused on, which for sure are producing less than a milliamp of current. The circuit retains the ability to withstand any input voltage, provided that the input current is guaranteed to be less than 10ma, which is the absolute maximum provided by the datasheet.

    It's a good time to be alive. Not long ago these ultra low current drain parts didn't even exist--at least not at $0.82 for single unit quantities. 🙂

    That said, there are a number of different ways to attack this problem. Maybe an even cheaper way exists that can withstand any input voltage (say, up to 50 or 60v DC worst case)? That's really the only complicating factor. If one assumes less than 12v input voltage, or even less than 20v input voltage, then I can see at least some other possibilities that would work just as well, if not better.

  • Banned

    This post is deleted!

  • Banned

    Looks interesting. How effective are these devices? What type of solar panel is compatible with this device?

  • Hero Member

    @marcosbe We're using solar cells scavenged from solar keychains. They put out less than 10ua in current under indoor lighting conditions.

    I tested the version 11 circuit (above), and it works. I subsequently added an LED to the discharge path so that it blinks during the discharge. This way you can confirm visually when the cap is charged. With just a 100uF cap instead of a supercap, it blinks fairly often even with just relatively weak indoor lighting.

  • @neverdie An interesting experiment, but can you clarify that by indoor lighting conditions you mean that harvesting is active only when lights are on, or does daylight continue to produce at say a lower rate?

  • Hero Member

    @zboblamont These particular solar cells seem to perform about the same regardless of whether the weak indoor light is sourced from indoor LED lights or from outdoor light that has leaked into the room from behind closed shades.

    That's a lot different than most solar cells, which may output no current at all from even bright indoor LED lighting but which will outperform these cells from sunlight that has leaked into the room from behind closed curtains. That's one of the two things that makes these scavanged cells interesting. The other thing is that they have a much higher voltage, even under weak lighting. Go figure. I don't know why they are that way, so I just accept it as empirical fact.

    FWIW, for testing purposes, the solar cell industry has arbitrarily defined weak indoor lighting as 200 lux, which in my book is actually fairly bright. I would estimate that these cells still perform at below 200 lux. For instance, I can get LED blinks off my charge circuit just from pointing the cell at my TV from across the room in an otherwise completely dark house late at night. Maybe later I'll dig out my lux meter and take some measurements.

  • @neverdie Thanks for that clarification, intriguing possibilities indeed, who'd have thunk...
    Not concerned over what what light levels generate which level of power rather than the principle of continuity, i.e. day/artificial light agnostic.
    It would indeed be ironic in the leap from tungsten to LED for illumination, that alongside lower energy consumption, the possibility of recycling part of that reduced energy coincides.

  • Hero Member

    @zboblamont I made the circuit able to withstand any input voltage because I wanted to stack of bunch of solar cells in series and thereby, if possible, manage to charge even using just moonlight. Well, after experimenting with that concept last night, it looks as though solar cells just aren't additive in that way, at least not under ultra low lighting conditions like moonlight-only.

  • @neverdie Ah, Texans - "don't the stars look purty t'night?".... "Yeah, but the moon is only producin' 2mV honey..." Romance is not dead but buried...😅
    Just kidding, brilliant stuff... 😉

  • Hero Member

    @zboblamont I still have hopes if that if I can just get even 1ua from moonlight, even at a mere 20mv, that I'll be able to collect it and eventually harvest it using:

    That may yet still work.... I think maybe the reason why stringing a bunch of cells in series doesn't work is that at very low light levels, the leakage currents become dominant. Except, if that's so, why does it still show a high voltage?

    Fortunately, I don't have an actual need to harvest only moonlight. It's just that if I could, then my gear would be good enough to run pretty much anywhere, other than in complete blackness.

    Of course, if anyone has ideas on how to harvest moonlight, then please do post.

  • @neverdie "I still have hopes if that if I can just get even 1ua from moonlight..."
    Sounds like a job for a parabolic concentator increasing luminance on the panel, albeit at the risk of frying at the merest hint of sunlight..

  • Hero Member

    @zboblamont said in Solar Energy Harvesting for wireless motes:

    @neverdie "I still have hopes if that if I can just get even 1ua from moonlight..."
    Sounds like a job for a parabolic concentator increasing luminance on the panel, albeit at the risk of frying at the merest hint of sunlight..

    It would be ungainly, but that could work. For instance, ultra-low-lux security cameras have a little flag that rapidly pops up (faster than you can blink) to cover an imager and protect it against overexposure in those kinds of scenarios.

  • Hero Member

    @zboblamont I have the Buck Energy Harvester working now:

    However, fresh out of the box it turns out not to be compatible with the keychain fob solar cells. Why? What's not even mentioned In the LTC3388 datasheet'S executive summary introduction page is that the active quiescent current is 150-250ua, which is enough to completely overwhelm the keychain fob solar cell output of around 10ua. I thought perhaps it would instead accumulate a charge and later do a forced conversion in a burst mode, but it doesn't appear to work that way. Such a mode could probably be configured, but I don't see anything like it discussed in the datasheet, so it would require some design and prototyping. More to the point, though, is that in such a scenario, I no longer see that there's anything special about the LTC3388 versus any other buck converter. i.e. I might as well dump the LTC3388, which is fairly expensive, and replace it with a much less expensive, highly efficient buck converter that's also highly efficient running in a burst mode.

  • Hero Member

    I notice that TI has a relatively new buck converter chip, the LM5164. In comparing it to other buck converters, what stands out about it is that it has high efficiency, even at a load current of just a milliamp:


    So, at least that much is impressive. However, its active current is 600-880ua, which is through the roof compared to the LTC3388, which is what led me to search for an alternative buck converter in the first place. The only way I can see to make use of it would be in very short bursts, which it fortunately seems designed to manage.

    So, for that reason, for now I'll stick with the LTC3388 and leave the perfecting of the buck converter as a future optimization...

  • Hero Member

    So, this smoking gun more or less settles it the matter regarding the LTC3388:

    Because of low efficiency at the microamp range, neither would appear to be a good match for solar cells extracted from the key fobs. Nonetheless, it would be a good match for higher current solar cells.

  • Hero Member

    I hadn't been aware of it until now, but apparently there do exist batteries with a very low 0.7% annual self-discharge rate, which translates into still having 70% of their original capacity a full 40 years later! They're fairly pricey, but do they seem like a much simpler method for getting past the cold-start problem on most energy harvesters. Presumably nearly all motes would be obsolete after 40 years anyway, and probably even long before that.

  • Hero Member

    Looks as though the EM8500 solar energy harvester can still function even if fed a current of just 1ua, for a total of 1uw power, and perhaps even less:

    If I'm not mistaken, that's lower power than any other chip! And if it's price really is $2.30 for quantity 1, then that would make it one of the least expensive energy harvester chips as well.

    Unfortunately, neither digikey nor mouser carry it.

  • Hero Member

    Intended for TEGs instead of photovoltaic, the EM8900 is nonetheless impressive. According to its datasheet, it can both cold start and operate with an input voltage as low as 5mv. That's far and away better than the LTC3108, which before now I had thought was the world leader with its 20mv minimum operating voltage.

  • Mod

  • Hero Member

    @mfalkvidd Yup. It was when looking there for the EM-8500 that I first noticed the EM-8900 (just 79 cents, quantity one!).

  • cough cough , so much dust here in the archives... 😉

    @NeverDie SPV1050 is what you are looking for, only Maximum Power Point Tracking IC from all IC's you mentioned/tested in both your topics. And trust me, it makes a whole world of a difference. ST even have online component calculator for matching it to solar panel and battery used. My usage scenario was outdoor rechargeable coin cell battery and calculator solar cell. Only drawback: you have to determine maximum power point of your solar cell.

  • Hero Member

    @Sasquatch Cool! How did you wire yours up? Did you use an eval board of some kind, or did you do something custom? I actually have one of these: Sheets/DFRobot PDFs/DFR0579_Web.pdf
    but I hadn't gotten around to testing it. So, thanks for the feedback! Now I have a reason to try it sooner rather than later.

    Looks as though the setting of the "MPPT" may be similar to how the LTC3105 does it:
    It's not really MPPT, though, so Linear Technology calls it MPPC instead.

  • I made pcb for it it's on the roof of my flat, will dig trough my archives to see if i have kicad project for it.

Log in to reply

Suggested Topics

  • 87
  • 7
  • 6
  • 7
  • 9
  • 1