I did my PhD on energy harvesting (specially focussing on hostile environments with high temperature or high radiation) around 15 years ago and harvesting from stray EM radiation was the holy grail for room temperature stuff where vibrations or heat gradients couldn’t be found.
If you’re willing to sacrifice always on connectivity and have a node report in on an infrequent basis then I always figured EM harvesting would be the way to go for most applications since even a tiny amount of energy can build up over time to become a useful amount.
I knew I’d gone deep into this world when I started thinking that micro watts was a large amount of power!
About 10 years ago I did a research/what-if project, involving energy-harvesting tracker tags .. essentially a smaller/lighter version of the AirTag.
It used energy-harvesting for the sensor packages, essentially charging caps if things got wet/moved/etc., but would boot up to do the sensor-data processing and communication (BLE/ANT) .. my lead on the project brought his dusty old home phone in, to give us all a stable power source, but we didn't really need it much - lots of 800mhz-900mhz in the city.
Was fun to realize, even back then, that we are literally surrounded by energy and don't really have an energy problem. We have an energy management problem, which we have yet to localize at the individual, and are instead spreading the problem across our society.
But, in between builds on this project, I would often daydream of a scenario where computing devices ran simply on background radiation, and there was no longer any need for cables, really, in order to compute.
Its fun to note we're getting closer and closer to that point. Okay, I could build a solar powered computer and also feel such satisfaction, but .. an all-in-one wafer that simply ran on cosmic noise?
This was basically my feeling when doing my PhD - there is energy everywhere and if we make systems power efficient enough there is no real reason we can’t make significantly great use of that energy.
That being said, it’s not the energy per se that we can tap into, it’s the gradient created as it dissipates/moves that we tap into. Heat is great, but not if it’s uniform, solar is great, but not if it harvesting it casts an area in shadow that needs it, vibration is great so long as it doesn’t damp and further load the system it’s being taken from.
" doesn’t damp and further load the system it’s being taken from."
Don't all of these Harvesting systems load a system? For example one of the transmitters somewhere. Yes it is microwatts and no transmitter will notice a single unit. However if we start to talk large quantity of sensors, who is really paying the price? Otherwise we are admitting that Free Energy is a real thing.
You and others may find the recently released Nexperia NBM5100 and NBM7100 of interest. While aimed at batteries it functionally translates high impedance power sources to lower impedance to charge up caps for IoT usage.
I mean yeah obviously thermodynamics aren't being violated, I think the base idea is that the background noise you're harvesting from was going to be lost as inverse square law losses anyway. Large amounts of sensors would ultimately be as a whole coupled to antennae broadcasting on those frequencies, damping them, and causing the gain circuits to automatically increase output power.
Oh for sure, it’s relative to the system being harvested from and the amount of energy being extracted though.
If it’s a train cart/carriage then not really a problem as there is so much mass and vibration to go around, if it’s the surface of a small industrial machine that needs to be reasonably balanced only producing a few hundred milli g then more of a concern.
It would need a lot of these devices for the odd mW to really make a difference. It adds up, but it's in the ballpark of a router's status LEDs. These costs are already accounted for as part of the device's operation. Critical systems might start to better direct their transmissions and to reduce side bands instead of saturating the area with energy.
Specifically for moving object like air tags, watches “automatic” movements are already a “solution” that would be easy enough to implement, and probably even provide up to a few joules (!!) per day.
(If you want to start a business off this idea feel free to contact me ;) half joking but half serious)
I have zero interest in this idea, business-wise - it is already a crowded space, Apple are mere new-comers to this - as the more you see these sensor nets grow, the more you realize we are making a big, big psycho-social 'smog' around the ways they can be used ..
So yeah, I'd rather make musical instruments. If anyone wants to talk about that idea, I'm all in.
With you on this, haha. Big tech has heavily crowded the wearables space, with questionable intentions and uses of personal health data.
Additionally, when doing research around kinetic / thermal energy harvesting technology for a wearable design project (gesture recognizing smart ring) for school, we found modules to have far too low a volumetric energy density (W/mm^3) to be useful, where size matters more than anything else.
Musical instrument design is a well saturated space as well…
Big Tech may have crowded the wearables space, but Garmin is still quite a player in that field and the description "Big Tech" seems almost as bad a fit for them as "Olathe tinkerer collective" would be.
Have you ever thought of a commercial location tracking solution based off this tech? It’s another heavily, heavily crowded industry, but it has oddly high margins, and a basically-infinite battery life would be a huge selling point
As a user, I would emphasis the importance of analogical input and feedback to musical instruments. The more analogical input means the more freedom to express nuances. Some but few electronic instruments include easy to use analogical input, with sliders, gyroscope or even the voice directly for example.
For example, the Dualo touch uses a slider and a gyroscope for analogical input:
I wish there was a small and affordable guitar pedal that had shimmer and bitcrush, and only that, with two potmeters and a passthrough button. I think the combination of those two effects is underused. Could it be done only with analog circuits?
.. can both do what you want, and a ton more to boot - I know you specifically asked for simple and cheap - well, these devices are cheap and .. amazing ..
By it's very definition, _bit_crush is going to be digital. Shimmer is a variation on reverb and might be possible using analog only but would be an interesting circuit.
I think there is space in the IoT market for a micro-watt platform that harvests energy from the environment. Always on connectivity is silly if it sits idle most of the time (think monitoring systems that take a temperature and moisture readings every 15 minutes), especially if it can communicate via a low power/low bandwidth channel.
I build a system for an agriculture client like this with each node only having a very small solar panel (6in^2) and some super capacitors that kept it reporting data throughout the night. We even used performance metrics from the voltage being output by the solar panel to measure how cloudy it was. These devices reported via LoRa to a base station setup that was connected to a PoE system so we just ran a single Cat-6 STP line to a pole in the middle of the property.
I think the other aspect of “always on” is that the operator isn’t going to pick up the sensor years down the line. They’re just going to leave it dead and buried, especially if the sensors have to live inside the earth There’s ongoing work on harvesting electricity out of soil
AirTags get about a year of battery life from a 2033 in moderate/typical use cases, according to Apple. That’s 0.675 Wh (Panasonic 2032) / 1 year = 77 microwatts. Or about 7 joules/ day. That’s a lot to guarantee for production. Watch movements are also at least an order of magnitude more expensive than batteries and have a service life.
Yes, you need a reservoir to store the charge (energy) which in my case was the limiting factor to deployment in high temperature environments as capacitors get very leaky as temperature increases.
Harvesters don’t need their own power supply and ideally the system can come back to life from totally empty (meaning it doesn’t need to be deployed with any charge level)
I was doing my thing back before the packaged harvester management chips were available and I made a PIC based microcontroller system that could monitor the charge level in a capacitor fed from a piezoelectric vibration energy harvester and decide when to wake up and transmit a data reading based on the capacitors charge level.
The trick was getting the PIC’s sleep current draw low enough that the capacitor was charging even at milli g level vibration levels - from memory you can get a PIC’s sleep current well down in the tens of nano amp range.
PICs have a built in voltage reference that is stable (a zener or schotkey diode from memory of 0.6V?) and a way to read what the Vsupply (basically the capacitor) is compared to it. So the PIC would wake up every 30 seconds, check Vsupply to the built in reference and, when high enough fully wake up, take and transmit a reading and go back to sleep.
From memory I got it running at 50mg transmitting every few minutes.
I got a nice journal publication in IEEE for that work.
The linked paper made sense once I realized mg is not milligrams. You sound like a former academic like me. All those papers/citations were so important (publish or perish), but now you just have to get things done, not so much writing.
Thanks, it didn’t report back progress of charging, it was connected to a sensor (light sensor in this case but could be anything really) and would transmit that back to a base station.
The main use of energy harvesting is for powering wireless sensor nodes, so this was a simple implementation of that.
How did you deal with the 'black start' scenario where you need to bring up the harvester after it is first assembled? Did it require a jump start with an external energy source or did it auto start? And did you use Silicon or something more exotic?
Black start wasn't problem as the current draw from the PIC waking up to check vref was so tiny it didn't stop the system charging (so long as there was a base line minimum amount of vibration).
I forget the minimum amount but so long as there was more than say 20 mG (milli-gravity) of vibration acceleration then it could harvest from empty.
That's super cool. Lots of these systems never get past the blackstart issue and I think that pre-charging the system is cheating because if the system then ever goes for a long time without enough energy input it will die for real.
So really neat to see your system did not suffer from that.
Thanks, to answer your other question (that I somehow missed). This project was with totally off the shelf components - it basically formed the back bone to my literature review as current state of the art.
However, the research group I was in was focussed on using silicon carbide devices for hostile environment sensors (think high temp gas sensors primarily) and looking at how to make FETs from SiC that could work in high temp/high radiation environments etc.
This is why my thesis and papers primarily focussed on testing energy harvesting devices at temperature - batteries won't work at 300degC and running power cables up and into a volcano fumarole is a tad awkward.
I had this feeling though that if I didn't get a room temperature, off the shelf components, system working I would have missed a key part of the learning - I just got lucky that the system was novel enough for a paper publication.
I'd love to see an open source energy harvester, especially if it is based on off the shelf components. Low power sensors without a battery would be a major plus for the environment.
Seems like a potential nightmare when your sensor network's operation is contingent on one badly tuned power circuit in a crappy Chinese LED floodlight in the janitors closet being perpetually left on.
It is not just ambient electrical noise that contains usable energy. Others have demonstrated practical sensor energy harvesting from ambient WiFi and ambient broadcast TV signals. I'd expect that the Sony device can probably use those too, so if you've got WiFi in your building you should be fine.
If there is ambient radio, you can also use that for extremely low power data transmission from your sensors. The trick there is to not actually transmit any of your own energy. You instead either absorb or reflect the ambient radio. This is called backscatter communications.
Here is a video showing a demonstration of both energy harvesting from ambient radio and backscatter data transmission [1].
If you use WiFi signals as your ambient radio for backscatter, the research group in that video has also demonstrated that you can make your backscatter data stream match standard WiFi protocols so that you can use ordinary WiFi equipment for the receiver.
The above was fairly short range, but even so there are a lot of interesting applications. For instance you might want to have moisture sensors inside a wall to detect water leaks early. If your sensors are powered by ambient EM energy (noise or radio), you don't have to limit the sensors to places where you can easily reach them later to change batteries. They are fine if they are behind brick or drywall with no openings.
Seems like an interesting way to identify machines that are breaking down.
If these devices can be built cheaply, just place them all over a factory, chirping their ID whenever they have enough power to do so. You map the IDs to locations and get a real-time map of where there's EM noise spitting out. When the map starts changing from it's usual patterns, then you identify the source of that change.
Reminds me of a neat video I saw once on the inner workings of a sheet metal stamping factory (for car parts). They used a directional microphone array pointed at the inspected machine and compared audio waves and overlayed “hot spots” on a display to see how the machine was working and where it was malfunctioning.
Don't worry, by the time you work out the source of your data outage, Sony will have a wireless power module to sell you as well. It will be wildly inefficient however, so you'll want to only run it when there is actually a data outage. By the time you work that out, there will be orchestration services available that notice the outage and respond by temporarily flipping on the wireless power module. A sensor on one side of the room driving a machine on the other side, and somehow we've got it dependent on a cloud service on the other side of the planet. The future is looking bright.
Question is whether this could harvest enough energy to store it in a small battery to continuously provide power. But really sounds frightening if you change the setup of a few appliances and then suddenly half of your autonomously powered devices don't get enough power anymore because the noise is not high enough anymore.
could be very useful in-doors where solar is not possible and wiring is not practical. A lot of sensors do not have to be on all the time - occasional reading is more than enough.
Those are claimed numbers and are pretty similar what Atmosic needs as ambient input to their rf harvesting chips. Field strength and antenna aperture/efficiency/size left as an exercise.
Okay, this is pretty cool. Effectively lowering the EM noise floor in their vicinity by rectifying the energy is a neat trick. The catch has always been that the level of energy was such that the conversion circuitry for ambient noise was unable to process enough to both power itself and provide a bit extra (efficiency losses effectively prevented any net energy conversion). That they managed to do that is what I find so impressive.
The second thing is the low frequency stuff. Its taught in most EE programs that you can use an antenna that is tuned to a frequency to transfer energy wirelessly from xmitter to receiver (see NFC or RFID) but antennas that tune sub-MHz frequencies are typically quite long in order to have some level efficiency. That they do this by exploiting geometry of various conducive elements and get enough energy is super impressive too.
If you followed the old BEAM stuff[1] that Mark Tilden promoted, you could see some ideas about two-phase devices that live in a "collect" phase charging up an energy reservoir and then an "execute" phase where they dump that energy doing their thing. BEAM was focused on robotics and bug like behaviors but there is no reason you couldn't have a sensor that measures temperature and humidity and transmits that to a receiver somewhere. Or a passive "game" camera that, once it has collected enough energy, takes a snapshot when it detects motion and sends that along.
I'll be interested to see this stuff get commercialized and designed in. It does have shades of "Smart Dust"[2] though which is pretty ripe for abuse.
Yes, BEAM! It's been years since I thought of that.
What a great point of entry into robotics and hardware design it was for a whole generation of engineers and roboticists. It provided a very low cost (it was encouraged that the parts for these robots came from old broken electronics. They were also super simple designs which were explained in great detail, very well. This combination made for a very good education in the foundations of robotics, engineering, mechatronics, hardware and system design, etc.
BEAM, the Parallax Basic Stamp, the Tandy 1000RLX, and the RB5X robot and the summer classes I took with the makers the RB5X all had a huge impact on my ability to architect, design and build robots professionally.
Just to get a raw idea, can this accumulate enough power to gather a gps position and send it over wifi / 4g nb-iot signal ? I have only very basic knowledge so not enough to make an estimation of the kind of things this can power
While possible it is unlikely. It would take a considerable amount of time to accumulate that amount of energy needed from these sources. These systems are generally used for transmitters that transmit for a couple of seconds at most and may be off for the rest of the day. Enough energy would have to be accumulated to run the GPS long enough to get a valid fix etc.
The nice thing about RF harvesting is you can have a beacon that keeps an area charged. Atmosic powered remotes with some TVs do this by having a beacon in the TV providing power.
I did some work in this area about a decade ago. The issues I ran into involved converting the energy into something that can be stored. If semiconductors are used for detection/rectification, the band-gap voltage must be overcome before any energy is recovered. Specialized devices with very low band-gap voltages are available, but losses are still significant.
I suppose "noise" requires a more precise definition here – I assume they don't mean a thermal background. Or otherwise I'm left wondering how this squares with the Second Law of Thermodynamics: You start with high entropy (thermal noise) and end up with low entropy (a charged battery)?
Yes, exactly, you are talking about a coherent bass tone (sine/cosine) for which the window happens to be resonant. This energy can be extracted of course.
Yes, but EMI noise is (Its called frequency). Which is the mechanism by which the the Sony module in question operates.
It's weird for yoi to bring up thermal "noise". What you are referring to is thermal decay, entropy. Yes, you can harness this too, but it's a different mechanism and not really pertinent here.
This is Sony, so their electrical engineers clearly know what they are doing.
But at the same time I can't understand how such a thing works using pure white noise. Feynman famously described (and disproved) a similar energy-harvesting device [0] except one that uses mechanical noise and a mechanical rectifier instead of the electrical noise and electrical rectifier of Sony. He showed that such a device will eventually stop working due to the second law of thermodynamics. Why doesn't that happen in the Sony chip?
Because RF energy harvesting depends on an external energy source - it’s the same reason a crystal radio or solar calculator can keep working as long as it’s illuminated.
So, to capture this better, would you then design your factory with acoustics in mind (more so than now), too? Like areas where soundwaves would bounce to resonate and concentrate to hit sensors?
This is really cool, but also seems rather indexed on physical cards. I have to imagine as we go towards more tech centric physical checkouts, we’re probably leaning more into device-based methods (phone, watch, etc). I imagine we will see less physical cards moving forward due to devices being able to fill this niche in a more secure way.
Agreed - that's the commercialisation approach they've gone with, but it's not the only one. I think access cards have legs still because corporations won't necessarily want to enroll people into BYOD situations (what happens if my phone battery dies, etc), nor issue everyone phones, and cards are just simple to administrate. But yeah I can imagine the opportunity isn't infinite.
Wow. Scientists at Stanford figured this out to a demonstrable idea over ten years ago I believe. I am looking for the article, ambient power for IoT devices.
I thought it was illegal to use an antenna to harvest RF energy. But, I guess not in these situations at least. All I can find from the FCC on this is https://www.fcc.gov/media/over-air-reception-devices-rule Can anyone comment on how these devices are legal?
The novelty seems to be the form-factor. It's energy harvesting made simpler for low-power applications. (A small microprocessor for example, LED's, sensors, etc.)
The ones I've read about before harvested from radio such as WiFi or broadcast TV. With this one they are talking about harvesting from EM noise from things that aren't intentional transmitters. Is that new?
overhead power lines aren’t intentional transmitters but some form of “could i power a device from the RF leakage?” gets asked on EE stack exchange at least once a year since its inception.
it’s hard to point at anything in the press release and confidently say “that’s new”. i think what’s happening is that IoT/sensing is a big enough market that companies can finally invest in engineering and commercializing some of the things that have been researched over the last several decades. which imo is great to see.
I assume this is already being used for spying on people, since the military has the incentive and resources to stay a few years ahead of what is publicly known.
If you’re willing to sacrifice always on connectivity and have a node report in on an infrequent basis then I always figured EM harvesting would be the way to go for most applications since even a tiny amount of energy can build up over time to become a useful amount.
I knew I’d gone deep into this world when I started thinking that micro watts was a large amount of power!