Computing stuff tied to the physical world

Re-thinking solar options

In AVR, Hardware, Musings on May 28, 2012 at 00:01

So will it ever be possible to run a JeeNode or JeeNode Micro off solar power?

Well, that depends on many things, really. First of all, it’s good to keep in mind that all the low-power techniques being refined right now also apply to battery consumption. If a 3x AA pack ends up running 5 or even 10 years without replacement, then one could ask whether far more elaborate schemes to try and get that supercap or mini-lithium cell to work are really worth the effort.

One fairly practical option would be a single rechargeable EneLoop AA battery, plus a really low-power boost circuit (perhaps I need to revisit this one again). The idea would be to just focus on ultra-low power consumption, and move the task of charging to a more central place. After all, once the solar panels on the roof of JeeLabs get installed (probably this summer), I might as well plug the charger into AC mains here and recharge those EneLoop batteries that way!

Another consideration is durability: if supercaps only last a few months before their capacity starts to drop, then what’s the point? Likewise, the 3.4 mAh Lithium cell I’ve been playing with is rated at “1000 cycles, draining no more than 10% of the capacity”. With luck, that would be about three years before the unit needs to be replaced. But again – if some sort of periodic replacement is involved anyway, then why even bother generating energy at the remote node?

I’m not giving up yet. My KS300 weather station (868 MHz OOK, FS20′ish protocol) has been running for over 3 years now, I’ve never replaced the 3x AA batteries it came with – here’s the last readout, a few hours ago:

     :41   KS300 ookRelay2 humi             77
     :41   KS300 ookRelay2 rain             469
     :41   KS300 ookRelay2 rnow             0
     :41   KS300 ookRelay2 temp             18.2
     :41   KS300 ookRelay2 wind             0

And the original radioBlip node is also running just fine after 631 days:

    1:32   RF12-868.5.3 radioBlip age       631
    1:32   RF12-868.5.3 radioBlip ping      852330

Even the JeeNode Micro running on a CR2023 coin cell is still going strong after 4 months:

    1:42   RF12-868.5.18 radioBlip age      139
    1:42   RF12-868.5.18 radioBlip ping     188449

So ultra-low power is definitely doable, even with an Arduino-compatible design.

No worries – I’ll keep pushing this in various directions, even if just for the heck of it…

  1. I don’t know if it is truly practical, but it is surely interesting. I spent a long time watching this series: http://www.youtube.com/watch?v=JwjmzKz8DqY&list=PL05DCC80A8E4EAB09&index=16&feature=plpp_video

    He is facing some interesting challenges as well. There are energy harvesting chips out there — I don’t know if those would help?

    • Thanks for the pointer. It’s not quite clear to me at what point he is right now – then again, I didn’t go through each of the videos…

  2. I would like to have a single Eneloop (or other large capacity NiMH) AAA battery in a garden solar light running the JeeNode Micro. With the LED under control of the JeeNode, plus one other sensor (temp, PIR etc) My nodes are 90m from the nearest power and I want/need around a dozen.

    I think at this level it should be quite feasible given the work you have done to date. For me the battery is required to keep it running for maybe a week at the most given heavy overcast days.

    • Yes, one AAA or AA is perhaps the best candidate, even though a boost regulator introduces losses (especially at these low power levels). A constant 40 µA current on a 800 mAh AAA is still 20,000 hours, i.e. over two years. With an Eneloop claiming to hold 85% of its charge over 2 years, this would be slightly less.

      With a PIR sensor added, that’s 50 mA more, i.e. perhaps 150 µA on the input side of the boost regulator. So then total consumption would be say 200 µA – still enough for a year on a 2000 mAh AA cell.

      These considerations are exactly why I think that we don’t really have to go all out on solar. Maybe even just top up an Eneloop with a trivial solar setup – I’ll chase the path in a future post.

  3. @Tim, certainly looks feasible. I’d be inclined to use LiPo chemistry though – any energy conversion steps are lossy, so better to do these upstream of the storage when there is more likely to be excess than downstream where conservation is the priority.

    Choosing a panel with open circuit voltage greater than 4.2+0.4v allows a simple charge circuit with a precision shunt regulator at ~4.2v to protect the battery, and the output side can drive the JN micro directly (with maybe a diode drop to the RFM12 logic supply to get closer to spec). The LED can be PWM to get the target average forward current.

    Many variations possible e.g. max energy transfer control from panel to storage, but perhaps not worth the complexity. IMHO oversizing the harvesting side and then minimising the extraction rate through hardware design and then software decisions as storage depletes is perhaps the best way to bridge several poor capture days.

  4. I have been wondering how practical it would be to use a current transformer clamped around a mains cable to harvest enough power to run a jeenode. I guess it would need an energy harvesting chip but…. has anyone got thoughts on this or perhaps has even tried it?

  5. I have a setup with a small 6V, 150mA panel (about 10cm x 10cm) directly charging 4 2100mAh Ni-MH cells (HEMA variety for those in Holland). This is used to power a JeeNode with a RFID reader attached to it. Although I do some low power stuff to minimise consumption, just the reader spends about 30 to 40mA when active (it too has a low power mode), which I need to use often, to check for cards and read them. I do this once a second. The panel, although it is all a bit of brute force, has managed to keep the unit powered for about a year now, including through the rough winter we had this year.

    Just as an additional note, if I redid it, I would get one of those low power PIR sensors to wake up the node when a card approaches. The whole thing could then probably run, like the room sketch, for a long time from the unaided AA cells.

  6. thank you for bringing this subject – - we are planning to make autonomous pieces of art making noise powered on solar – - we are looking for the solar-panel-rechargeable combination to have enough Amps to feed a small audio amplifier – - yes we have to do a lot of research :-)

  7. Very nice to follow these articles.

    It got me wondering about two things…

    1. are there similar projects aiming at having equipment running for millenia? :-)

    2. instead of doing something at a fixed interval would it perhaps be a better option to make the lowest power consuming mode duration dependent on the available power and/or estimated power at the time of transmission? Perhaps it’s even more energy efficient to bring the Jeenode back to life whenever a certain amount of power is available as determined by some external electronics.

    cheers, Eric (Utrecht, Netherlands)

    • A project that aims to have a computer running for millennia exists – and they might even succeed :) Look at http://longnow.org/clock/ It is a mechanical computer aimed at keeping track of time for ten thousand years. Very geeky, very cool.

  8. @ericvrp – stretching to millennia brings in other tough problems, not least component predicted lifetimes bite after a few decades.

    Your theme of varying the timing of consumption by skipping timeslots for example makes sense. This can be done quite well without additional ‘electronics’ by the Jeenode estimating power remaining and modifying its own sleep pattern. Keeping accurate time is key to a responsive 2way dialogue in these circumstances.

  9. varying the timing of consumption by skipping timeslots

    Check tomorrow’s post :)

  10. Within what constraints are you looking for a working solar-powered setup? Above, amvv mentions one setup that works, and I have a slight smaller setup that is also successful. I guess the question is just how small/inexpensive can the panel and battery go?

    I have a house power meter pulse counter sending a count wirelessly back to my PC several times a minute, and it’s been running for a month or so on a small 0.5W panel and a Li-po battery. The Jeenode sleeps between pulses, and it’s woken by an interrupt triggered by an IR photodiode on a pin with the pull-up resistor enabled.

    Specifically, the parts I used are:

    It all fits nicely in this waterproof enclosure.

  11. For my own experiments I’ve been trying a (nominal) 6V solar panel – with a resistor and 5.1V zener across it – feeding through a schottky diode to give a 0.7 V drop to a cheap LiPo – that then feeds a low drop 3.3V regulator.

    So far nothing bad has happened and the cells seem happy enough over the 2-3 months I have been running them so far.

    (The 5.1V zener seems to start conducting at about 5V – and I rarely see 6V from the solar panel – so the voltage at the bettery seems to sit at 4.2V quite happily).

    But as per everyone I shall wait (a long time) and see…

  12. I think you should decide how small the solar panel should be. This setup is working out very good for me so far; http://forum.jeelabs.net/node/147

    It isn’t expensive (including LiPo) and might include a nice housing to cram in the a JeeNode. But it might not be small enough for everyone. Would be nice to compare the performance of the different types of cells in indoor lighting conditions.

Comments are closed.