I Am A Battery: Harvesting Heat Energy

If you get tired of charging batteries, you might be interested in [Hackarobot’s] energy harvesting demo. He uses a peltier device (although he’s really using it as a thermocouple which it is). A 1 farad super capacitor stores energy and an LTC3108 ultra low voltage converter with a 1:100 ratio transformer handles the conversion to a useful voltage.

The truth is, the amount of energy harvested is probably pretty small–he didn’t really characterize it other than to light an LED. In addition, we wondered if a proper thermocouple would work better (some old Russian radios used thermocouples either in fireplaces or kerosene lamps to avoid requiring batteries). Although a Peltier device and a thermocouple both use the Seebeck effect, they are optimized for different purposes. Thermocouples generate voltage from a temperature differential and Peltier modules generate temperature differentials from voltage.

However, as [Hackarobot] points out, the same technique might be useful with other alternate power sources like solar cells or other small generators. The module used has selectable output voltages ranging from 2.35V to 5V.

We’ve looked at other electric harvesting devices including a camera and a magic ring that also uses an LTC3108. You probably aren’t going to replace a hefty battery with something like this anytime soon, but for a low power project, it could be just the ticket.

16 thoughts on “I Am A Battery: Harvesting Heat Energy

  1. Problem is that the russian radios used a different thermocouple, with a temperature difference of some 400°C…while they still do make high temperature thermocouples in ceramic a package similar to the one used here, they are stupidly expensive…

    1. The “Carrier-powered” option looks like it could be interesting. I might do a little more research into the sort of ranges they were able to obtain. I always enjoy a good “Tesla would be proud” style project.

        1. A lambda sensor is a fuel cell (O2 in exhaust vs O2 in air), not a thermocouple…
          Also, only a zirconia narrow band sensor will produce power (less then 1V), other need some input power to work…

    1. I tried for the longest time to believe a peltier could be useful for converting heat into electricity but they simply suck. The best options are stirling generators and it takes a rather large one to make useful power from a thermal difference. No wonder steam is still used. It makes far more sense, but is a bit dangerous.

  2. I find everything about this story absolutely fascinating, except the cost of the LTC3108 device. There’s gotta be a cheaper way to construct the boost circuit for no-light wearable computing?

    1. Unless there has been huge advancements in the realm of world breaking physics, peltier powered microcontrollers from body heat isn’t happening. There are basic ultra low power chips from TI but they really are just good for transmitting short radio data. LTC chips are so expensive because energy harvesting is a niche market. There really aren’t a lot of options. Check TI stuff.

  3. “Although a Peltier device and a thermocouple both use the Seebeck effect, they are optimized for different purposes. Thermocouples generate voltage from a temperature differential and Peltier modules generate temperature differentials from voltage.”

    Just… no. This is wrong. While they’re both based on the seebeck effect, one uses a pair of metals and the other is a semiconductor. And a peltier device is bi-directional: it is capable of producing electrical power from a thermal flux, which arises from a temperature differential; it’s not just for producing a thermal flux from a current.

    You can even buy off-the-shelf products – http://www.biolitestove.com/ – that work on this principle.

    1. I don’t disagree with what you’ve said but I don’t think what I said was wrong either. I’m quite aware of the composition of both devices. In fact, both devices will work as the other (although thermocouples make very poor Peltier devices because–if I recall correctly–because of the high conductivity of the metals). It was the use of semiconductors that allowed the devices to be practical (I used them in the 80s to cool semiconductor devices under test). But that’s what I mean by optimized. Either device generating electricity is exhibiting the Seebeck effect. Either device generating temperature difference is exhibiting the Peltier effect. https://en.wikipedia.org/wiki/Thermoelectric_effect

    1. It does. Don’t ask me how, but LTC3108/9 in my tests started up at a hair over 20mV (slightly more for the bidirectional -9 variant), even with a very slowly rising supply (0~20mV over the course of a minute or more). I have not tried it over hours/days however.

  4. There is thermally regenerative ammonia-based battery (TRAB, doi:10.1039/C4EE02824D) [Cu|Cu(NH3)4++||Cu++|Cu]. Cu(NH3)4++ is generated from copper and ammonia in anode when discharge. For charging, heat is used to distill the ammonia from electrolyte Cu(NH3)4++ for using in cathode that will be anode.

  5. Cool! (Er, hot?)
    I’ve also been playing with the LTC310x series and thermal energy harvesting (Eagle files at https://github.com/drmn4ea/mosquino.hardware/tree/master/pwr-pelt). The chip self-starts and generates usable energy (1.8 ~ 3.3V) from ~20mV in my informal tests, corresponding to about a 2degC gradient over a no-name 40x40mm Peltier device (128 junctions?) and ~7degC from a 25x25mm one.

    As someone mentioned above, the hard part is the cold side – as soon as you start heating the hot side, the heat very quickly transfers to the cold side, meaning you need to supply a rapidly-replenishing source of heat to maintain the gradient, and reliably dispose of it on the other side. For example, heat from your hand will probably not keep it alive indefinitely, even with a big heatsink on the other side (or probably cold water, ice, snow…). If you have a good heat source and a good cold source though, it works well for small amounts of consistent power.

    If your setup can maintain larger gradients (~350+mV, or >>10degC over a large TEG), the TI BQ25504/25570 type parts work well – they have a much larger minimum startup voltage, but can also handle considerably higher input currents.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.