Before Lunar New Year, I had ordered two 3000 F, 2.7 V supercapacitors from China for about $4 each. I don’t actually remember why, but they arrived (unexpectedly) just before the holiday.
Supercapacitors (often called ultracapacitors) fill a niche somewhere between rechargeable lithium cells and ordinary capacitors. Ordinary capacitors have a low energy density, but a high power density: they can store and release energy very quickly. Lithium cells store a lot of energy, but charge and discharge at a comparatively low rate. By weight, supercapacitors store on the order of ten times less energy than lithium cells, and can deliver something like ten times lower power than capacitors.
Overall they’re an odd technology. Despite enthusiastic news coverage, they are a poor replacement for batteries or capacitors, but their long lifespan and moderate energy and power density make them suitable for some neat applications in their own right. Notably, they’re used in energy harvesting, regenerative braking, to extend the life of or replace automotive lead-acid batteries, and to retain data in some types of memory. You’re not likely to power your laptop with supercapacitors.
Anyway, I had a week-long holiday, and two large capacitors of dubious origin. Sometimes we live in the best of all possible worlds.
Each capacitor is allegedly capable of storing about eleven kilojoules, and delivering that energy at a frankly alarming rate. Exactly what rate is unclear, as I don’t have any documentation for the internal resistance, thermal tolerance, or even the polarity.
Further complicating matters is that all my suppliers are closed for the next two weeks or so. Whatever I was going to build, the first challenge was that it had to be built from the parts on hand. Second, most hospitals are running at reduced capacity right now – so even more than usual, it had to be reasonably safe.
For that and other reasons, directed energy weapons were out of the question. (Use science for peace please!) I did really need a spot welder though, and there was no shortage of examples of people who had built their own from supercapacitors, typically in the 500 F range.
However, I’m using six times that capacity, and I didn’t feel the safety precautions used in these builds (most often none) scaled particularly well. My concerns were in two main areas: a safe and efficient charging method, and preventing unintentional discharge.
The latter was simple to fix. As the supercapacitors were roughly the shape of a large battery, I put a collar around one electrode as a spacer. The tape is not structural, it’s just to hold the plastic in place for a photo and while inserting the capacitor.
Then I made a large version of an AA battery holder out of a weatherproof plastic box and some 0.7 mm thick copper plates I had left over from other projects. Later on, I’ll drill holes in these plates and attach thick copper wires with bolts. I may also pack the empty space inside the enclosure with something fireproof, like packed sand.
The capacitor fit firmly in place and I wasn’t able to dislodge it by shaking the case vigorously. The next challenge was to determine capacitor polarity so I could clearly mark it on the case. Out of habit, I pulled out a multimeter and measured the voltage across the terminals, without realizing that probably wouldn’t work.
Unfortunately, it did work. Someone has decided to ship the supercapacitor containing significant charge (about one volt). They were pretty well packed when I received them, but I still can’t imagine any reason they would ship them bearing charge. If any of our readers can think of one, please let us know in the comments!
In any case, now that I could store and connect things to the capacitor more or less safely, it was time to build a charger. Certainly, there are many examples on YouTube of people manually connecting and disconnecting a lithium cell and current limiting resistor to the battery while also holding a multimeter in place, or connecting a large variable-voltage power supply with current limiting function. There are also voltage limiting and cell balancing supercapacitor protection boards that are a very reasonable option, but my suppliers don’t carry them and are closed for the next 2 weeks.
Directly using a battery was out of the question. A current limiting resistor wastes a lot of power, and human error can lead to overcharging the capacitor or polarity errors, reducing its lifespan (and possibly mine as well). The fail state of large capacitors seems to vary between ‘large firecracker’ and ‘small hand grenade’, both of which I prefer less. A high current benchtop power supply wouldn’t be great either, because it requires a wall outlet, and I was hoping to make this a mobile unit.
I had a few DC – DC buck converters lying around that were rated to 5 A of current. These are fairly efficient, around 85%, and can output down to 0.8 V with a DC input between around 5-32 V. My intended power supplies for this are a bunch of lithium batteries at different voltages, a 40 W 18 V solar panel, and possibly other things like old laptop power supplies.
I tore apart one of the broken DC – DC converters to see what chip it used. It was an XL4005E1 (PDF). The datasheet said that it has a ‘maximum duty cycle of 100%’ (hilarious for reasons we’ll discuss later) and had an active high enable pin. This seemed like a good way to limit the output current of the power supply to within a safe current using pulse width modulation.
While a 555 timer would have been sufficient, I used an ESP8266 because I plan to expand functionality later. The capacitor charge rate slows as it charges, so the ideal PWM duty cycle depends on the present charge level. Later on, I want to use the onboard analog to digital converter on the ESP8266 to optimize the charge rate, display charge state using an LED or screen, and turn off the circuit automatically to save power when charged. For now, the only goal was to find a reasonable duty cycle to charge the supercapacitor to over 2 V for testing. To this end I set the output to 2.5 V and connected the PWM output from the ESP8266 to the XL4005E1 enable pin. The PWM output was determined by the following simple program on the ESP8266:
pwm.setup(1, 1000, 650) pwm.start(1)
Unfortunately, this didn’t work at all. The XL4005E1 always stayed on, and thankfully it shut down before any combustion occurred. Tying the enable pin to ground or VCC also had no effect – I suspect a counterfeit chip. Ironically, the datasheet was still exactly correct to say it has a ‘maximum 100% duty cycle’, it just turns out this is also the minimum duty cycle!
I looked around for something that could switch enough current to be a useful alternative. I found a stack of LED dimmers that someone had asked me to buy for them, and never picked up. They were super cheap, and I’m happy I kept them now. Inside I recognized a 9 V voltage regulator, a 555 timer, and an IRF530N power MOSFET (PDF) with heatsink.
I couldn’t help but think that if I just removed the 9 V voltage regulator, I could use the dimmer directly to control the output current of the power supply. However, that would mean no automated control later, and a dial without any particularly useful markings.
I desoldered the MOSFET, and used the 3.3 V output to drive the gate. The datasheet suggested this probably wouldn’t work, and frankly I shouldn’t have bothered. What did eventually work was connecting the input of the DC-DC regulator — it will always be 7.4-18V for my use cases — to the MOSFET gate via a 2N2222 transistor. The base of the transistor is connected to the ESP8266 PWM output. The PWM output allows the transistor to switch the DC-DC regulator input, driving the power MOSFET effectively.
I had found earlier that the DC-DC converter shut down if you draw more than 8 A. Through trial and error, I found a 40% duty cycle worked acceptably well while safely accounting for vendor optimism. The downside is that with a fixed duty cycle, the charge rate drops to nearly nothing when the capacitor hits about 2.15 V. My goal for testing was to charge it to 2 V, so this was acceptable for now. For convenience, I attached a small LCD voltmeter to monitor charge, and called it good enough for now.
Admittedly, after a few unplanned revisions to the circuit this became a bit of a kludge of parts. If I had to redo this, I might keep the power MOSFET controlled by a BJT, but replace the ESP8266 with an 8-bit Atmel MCU. It would be smaller, more power efficient, a little cheaper, and I used to love programming those chips in ASM. Another big advantage of that approach is I can use pulse sequencing to add multiple power supplies in parallel, whereas that might not be practical on the ESP8266. Normally this would be unsafe – among other things one power supply failing means things get out of hand rather quickly for the others. However, if each power supply is connected to the supercapacitor via its own power MOSFET, and a microcontroller manages them so that only one is on at a time, it should be fine and offer a faster charge rate at an acceptable cost.
The next steps are to build electrodes and a high-current switch, then try welding different materials. I also want to optimize the charging circuit. I think this has been enough for one day, so stay tuned for sparks coming up soon.