Introduction: Super Capacitor Powered Raspberry Pi Laptop
Depending on general interest towards this project, I can add more steps, etc if that helps simplify any confusing components.
I have always been intrigued with the newer capacitor technology popping up over the years and thought it would be fun to try implementing them as a battery-of-sorts for fun. There were a lot of quirky problems I came across working on this as they are not designed with this application in mind, but wanted to share what I have found out and tested.
This is more to highlight the difficulties of charging, and pulling power from a bank of super capacitors in a mobile application (though with how heavy it is, its not all that mobile...).
Without the great tutorials below, this would not have come to fruition:
- www.instructables.com/id/Lets-learn-about-Super-Ca... - In-depth information on Supercapacitors
www.instructables.com/id/How-to-Make-Super... - Tutorial to build a charging and discharging circuit
I will attempt to dig up more that I used if I can find/remember them.
- If you have any tutorials you think are relevant, let me know so I can throw it in here.
The main reasons I wanted to try this are:
- Charging to full within SECONDS (high amperage involved limits this system to minutes... safely).
- Hundreds of thousands of charge cycles without degradation (over a million under the right conditions).
- A very niche technology that could find its way into the mainstream battery industry possibly.
- Environmental operating conditions. Temperatures of +60C to -60C for the capacitors used here.
- Charging efficiency is >95% (batteries are on average <85%)
- I find them interesting?
Now for the ever necessary warning when working with electricity... Even though there is very little chance of injury working with low voltages of ~5V, the incredible amount of amperage that super capacitors can output will cause burns and instantly fry components.The first article mentioned provides an excellent explanation and safely steps. Unlike batteries, fully shorting the terminals does not risk an explosion (though it can shorten the super capacitor's life depending on the wire gauge). Real problems can arise when over-volting (charging past the marked max voltage) where the the super capacitors will fizzle, 'pop' and die in a smoky mess. Extreme cases can be where the seal pops quite loudly.
As an example of how much power can be released, I dropped a 16 gauge copper wire across the fully charged bank at 5V (accidentally of course) and was slightly blinded by the wire exploding in a white and green flash as it burned. In under a second that 5cm piece of wire was GONE. Hundreds of amps travelling across that wire in less than a second.
I settled on a laptop as a platform as I had a Raspberry Pi lying around, an aluminum suitcase, a kiosk keyboard and a 3D printer to prototype on. Originally the idea was to build this laptop just so it could run for 10-20 minutes at minimal effort. With the room I had extra in the suitcase, it was too tempting to try and push more out of this project by cramming in more super capacitors.
Currently, the amount of usable power is under that of a SINGLE 3.7V 2Ah lithium ion battery. Only approximately 7Wh of power. Not astounding, but with a charge time of less than 15 minutes from empty, its interesting at least.
Unfortunately, only just about 75% the stored power in the capacitors can be pulled out with this system... A much more efficient system could definitely be implemented to pull power at lower voltages around 1V or less. I just didn't want to spend any more money on this as well as, under 2V in the capacitors leaves only about 2Wh of power available out of a total of 11Wh total.
Using a low power 0.7-5V to 5V converter (~75-85% efficiency) I was able to charge my 11Wh cellphone battery from 3% to 65% using the capacitor bank (though phones are extremely inefficient in charging, where 60-80% of input power is actually stored).
For parts used in this project, there are probably better parts to use than I had on hand. But here they are:
- 6x super capacitors (2.5V, 2300 Farad - from a car regenerative braking system. Can be found on Ebay,etc.)
- 1x Raspberry Pi 3
- 1x 5V powered display (I am using a 5.5" AMOLED display with HDMI controller board)
- 2x ATTiny85 micro-controllers (I will include the programing)
- 2x 0.7V-5V to constant 5V 500mA DC-DC converters
- 4x 1.9V-5V to constant 5V 1A DC-DC converters
- 1x suitcase
- 3x 6A PWM capable mosfets
- 2x 10A Schottky diodes
- 10x Aluminum T-slot frame (with joints etc. depends on what you want to use to hold things in place)
- kiosk keyboard
- 20W 5V solar panel
- USB to micro USB cables
- HDMI cable
- Assortment of basic electrical components and prototyping boards.
- many 3D printed parts (I will include the .stl files)
These parts can easily be interchanged for more appropriate/efficient parts, but this is what I had on hand. Also, dimension constraints will change with what components are chosen.
If you have any feedback on the design, don't hesitate to leave a comment!
Step 1: Power Characteristics
To give an idea of what to expect power-wise when using capacitors for something they were definitely not designed for:
When the capacitor bank voltage drops too low (1.9V), the ATTinys have been programmed to not power on any system components. This is just to ensure the components are not drawing any power when they cannot run consistently at lower voltages.
This system runs using DC-DC converters at voltage levels of 4.5V to 1.9V from the capacitor bank.
Input charging voltage can be from 5V to 5.5V (no higher than 5A at 5.5V). Adapters of 5V 10A or higher will damage the mosfet and will burn it out at half PWM charging rate.
With the charging characteristics of the capacitors, a logarithmic/exponential charging rate would be best, as it gets harder to push the power in the closer you get to full charge... but I could never get the math function working with floating type variables on the ATTiny for some reason. Something for me to look at later...
At full processing power, approximate run time is 1 hour. On idle, 2 hours.
Using LowRa transceiver cuts life by another ~15%. Using external laser mouse cuts life by another ~10%.
Lower capacitor bank voltage = less efficiency converting to 5V to power components. About 75% at 2V capacitor charge, where a lot of power is lost as heat in the converters.
While plugged in, the laptop can run indefinitely using a 5.3V 8A adapter. Using a 2A adapter, the system requires full charge before powering on for unlimited use. The ATTiny PWM charging rate is only 6.2% of power input when the capacitor bank is 1.5V or less climbing linearly to 100% charging rate at full charge.
This system takes a longer time to charge using a lower amperage adapter. Charge time from 2V to 4.5V with nothing running off the capacitor bank:
- 5.2V 8A adapter is 10-20 minutes (usually around 13 minutes).
- 5.1V 2A adapter is 1-2 hours. Because the diodes drop the voltage by about 0.6V some adapters at exactly 5V will never fully charge this system. This is ok though, as the adapter will not be negatively affected.
- 20W solar panel in full sunlight is 0.5-2 hours. (lots of of variance during testing).
There is the inherent problem of using capacitors where they do not hold their charge very long the closer you are to the max voltage.
Over the first 24 hours, the capacitor bank self discharges from 4.5V to 4.3V on average. Then over the next 72 hours will slowly drop to a fairly constant 4.1V. The ATTinys coupled with a small self discharge will drop the voltage at 0.05-0.1V per day after the first 96 hours (exponentially slower as the voltage drops closer to zero). When at 1.5V and lower the capacitor bank voltage drops at around 0.001-0.01V per day depending on temperature.
With all of this taken into consideration, a conservative approximate would be a discharge to 0.7V in ~100 days. I left this sitting for 30 days and was still left with just over 3.5V.
This system can run indefinitely in direct sunlight.
* * * TO NOTE: * * Critical voltage of this system is 0.7V where the DC-DC converters powering the ATTinys will fail. Luckily, the mosfet controlling charge rate will pull itself ~2% high when power is connected at this voltage or lower, allowing slow charging. I still haven't figured out WHY this happens, but its a lucky bonus.
I had to fully charge and discharge the capacitor bank ~15 times before they balanced chemically and held a decent charge. When I first hooked them up, I was extremely frustrated with the amount of stored charge, but it gets much better over the first 15 full charge cycles.
Step 2: Pi Power Controller
In order to turn the Pi on and off I had to implement a power controller with 4 DC-DC converters and a mosfet.
Sadly the Pi draws about 100mA even when off, so I had to add the mosfet to completely cut power to it. With the power controller in play, only ~2mA are wasted at full charge(~0.5mA at low charge).
Essentially the controller does the following:
- Regulates the voltage level below 2.5V in the capacitors to avoid overvolting while charging.
- Four DC-DC (1A max each, 4A total) pulls directly from the capacitors from 4.5V to 1.9V for a constant 5.1V.
- At the press of a button, the mosfet allows power to flow to the Pi. Another press cuts power.
- The ATTiny watches the voltage level of the capacitor bank. If too low, the mosfet cannot be turned on.
The silver button, when depressed indicates the power remaining in the capacitor bank. 10 blinks at 4.5V and 1 at 2.2V. The solar panel can charge to the full 5V and blinks 12 times at that level.
The capacitor voltage is regulated with the green disc 2.5V regulators that bleed off any excess power. This is important because the solar panel passively charges the capacitors through a 10A diode directly up to 5.2V which would over-charge them.
The DC-DC converters are capable of providing up to 1A each and are variable constant voltage output. Using the blue potentiometer on the top, the voltage can be set to any level you require. I set them to 5.2V each which drops about 0.1V across the mosfet. One will be the tiniest bit higher voltage output than the others and will get moderately hot, but the others will handle power spikes from the Pi. All 4 converters can handle power spikes up to 4A at full capacitor charge, or 2A at low charge.
The converters draw ~2mA quiescent current at full charge.
Attached is the Arduino sketch I am using to get this done with the ATTiny (Plenty of notes added). The button is attached to an interrupt to pull the ATTiny out of sleep and power the Pi. If the power is too low, the power LED is blinked 3 times and the ATTiny is put back in sleep.
If the button is pressed a second time, the Pi power is shut off and the ATTiny put back to sleep until next button press. This uses a few hundred nano amps in sleep mode. The ATTiny is running off of a 500mA DC DC converter that can provide a constant 5V from a voltage swing of 5V-0.7V.
The power housing was designed on TinkerCAD (as are all the other 3D prints) and printed.
For the circuit, see the crudely drawn schematic.
Step 3: Charging System
The Charge Controller consists of three parts:
- The controller circuit driven by an ATTiny
- The mosfets and diodes (and fan for cooling)
- I am using a 5.2V 8A wall charger to power the laptop
The controller circuit wakes up every 8 seconds to check for a connection to ground on the charging port. If the charging cable is connected, the fan starts up and the charging process begins.
As the capacitor bank gets closer and closer to full charge, the PWM signal controlling the mosfet is increased linearly to 100% ON at 4.5V. Once the target voltage is reached, the PWM signal is turned off (4.5V). Then wait until the defined lower limit is reached to begin charging again (4.3V).
Because the diodes drop the charging voltage from 5.2V down to ~4.6V, theoretically I could leave the charger running 24/7 with the voltage capping out around 4.6-4.7V. Time of charging to discharging when at or near full is about <1 minute charging and 5 minutes discharging.
When the charging cable is disconnected, the ATTiny goes to sleep again.
The mosfets are from Ebay. They can be driven by a 5V PWM signal and can handle up to 5A each. This is on the positive line using three 10A schottky diodes to prevent back-flow to the wall charger. Double check the diode orientation BEFORE connecting to the wall charger. If oriented incorrectly to allow the power to flow from the capacitors to the wall charger, the charger will get very hot and probably melt when plugged into the laptop.
The 5V fan is driven by the wall charger and cools the other components as they get very hot below halfway charged.
Charging using a 5.2V 8A charger only takes a few minutes, where as a 5V 2A charger takes over an hour.
The PWM signal to the mosfet only allows 6% of power through at 1.5V or less climbing linearly to 100% at full charge of 4.5V. This is because capacitors act as a dead short at lower voltages, but become exponentially harder to charge the closer you get to equalization.
The 20W solar panel drives a small 5.6V 3.5A USB charger circuit. This feeds directly through a 10A diode to the capacitor bank. The 2.5V regulators keep the capacitors from over-charging. Its best to not leave the system in the sun for extended periods of time as the regulators and charger circuit can get quite hot.
See attached Arduino Sketch, another badly drawn circuit diagram and .STL files for the 3D printed parts.
To explain how the circuit is wired together, the charge controller has one line to test for input voltage from the charger and one line to the pwm pins on the mosfet modules.
The mosfet modules are grounded to the capacitor bank negative side.
This circuit will not turn off without the fan being connected from the negative side of the capacitors to the high side of the charger input. Because the high side is behind the diodes and the mosfets, very little power will be wasted as the resistance is over 40k resistance. The fan pulls the high side low while the charger is not connected, but does not take enough of the current to bring it low while the charger is plugged in.
Step 4: Capacitor Bank + Additional 3D Prints Used
The capacitors used are 6x 2.5V @ 2300F supercapacitors. They have been arranged in 2 sets in series of 3 in parallel. This comes to a bank of 5V @ 3450F. If ALL energy could be pulled from the capacitors, they can provide ~11Wh of power or that of a 3.7V 2.5Ah Li-ion battery.
Link to datasheet: https://www.chemi-con.co.jp/e/catalog/pdf/dl-e/dl-sepa-e/dl-dle-e-170401.pdf
The equations I used to calculate the capacitance and subsequently the available watt hours:
(C1*C2) / (C1+C2) = Ctotal
2.5V 6900F + 2.5V 6900F
(6900*6900) / (6900+6900) = 3450F @ 5V
Using 4.5V to 1.9V of available potential at 3450F capacitors
((C * (Vmax^2)) / 2) - ((C * (Vmin^2)) / 2) = Joules Total
((3450 * (4.5^2)) / 2) - ((3450 * (1.9^2)) / 2) = 28704J
Joules / 3600 seconds = Watt hours
28704 / 3600 = 7.97 Wh (theoretical maximum available power)
This bank is very large. at 5cm tall x 36cm long x 16cm wide. It is quite heavy when including the aluminum frame I used... About 5Kg or 11lbs, not including the suitcase and all other peripherals.
I hooked the capacitor terminals up using 50A terminal connectors soldered together with 12 gauge copper wire. This avoids a resisting bottleneck at the terminals.
Using an aluminum T-bar frame, the laptop is incredibly sturdy (though also VERY heavy). All components are held in place using this frame. Takes up minimal space within the laptop without having to drill holes everywhere in the case.
Many 3D printed pieces were used in this project:
- Capacitor bank holders full
- Capacitor bank holder bracers
- Capacitor holders bottom
- Separator between positive and negative capacitor terminals
- Raspberry Pi holder plate
- Top covers for around keyboard and capacitors (for aesthetics only)
- AMOLED screen holder and cover
- AMOLED controller board holder
- HDMI and USB wire guides to display controller from Pi
- Button and LED plate top access for power control
- others will added as I print them
Step 5: Conclusion
So as this was just a hobby project, I believe it proved that supercapacitors can be used for powering a laptop, but probably shouldn't for size constraints. The power density for the capacitors used in this project is more than 20x less dense than Li-ion batteries. Also, the weight is absurd.
That being said, this could have different uses than a conventional laptop. For example, I use this laptop mostly from solar charging. It can be used out in the woods without worrying too much about charging and discharging the 'battery' repeatedly, multiple times per day. I have slightly modified the system since initial build to incorporate a 5v 4A outlet on one side of the case to power lighting and charge phones when out checking sensors in the woods. The weight is still a shoulder killer though...
Because the charging cycle is so quick, never have to worry about running out of power. I can plug it in for 20 minutes (or less depending on current level) anywhere and be good to go for over an hour of intensive use.
One drawback of this design is it looks very suspicious to a passerby... I would not take this on public transit. At least not use it near a crowd. I have been told by a few friends that I should have made it look a little less 'threatening'.
But all in all, I had fun building this project, and have learned quite a bit on how to apply supercapacitor technology to other projects in the future. Also, fitting everything in the suitcase was a 3D puzzle that was not overly frustrating, even quite an interesting challenge.
If you have any questions, let me know!