I am looking to build a UPS circuit using supercapacitors.

I have seen a number of circuits posted online, and a lot of them look very complex (some using lipos, some using NiMH etc).

Please note, I do not know electronics and the extent of my knowledge is pretty much being able to follow a circuit and soldering it (as long as it's not too complex).

I found the following circuit and was wondering if it would work for the Raspberry Pi and other 5v based SBCs.


Could I potentially use 300f supercaps with this circuit?

The web site says that it supports 5v at 2.5A. How do I calculate how long I could sustain 5v at 1.0 - 2.5A with this circuit (or using larger caps e.g. 300f supercaps)?

Can I modify the circuit to trigger the GPIO pin on the SBC when it switches to backup mode (i.e. DC power in fails)? The GPIO Pins on most SBCs are 3.3v.

My goal is to build a circuit that could keep a board running for a couple of minutes at the most (assuming a current draw of around 1 - 2A on backup power). Around 2-3 minutes would be great, or around 1 minute minimum.

I also need the circuit to be able to provide around 2.5A when powered from the wall wart (and whilst the battery is charging).

The other problem I am having is when I download the LTspice version of the circuit, it shows the supercaps as 5m. When I change it to 50f, the simulation does not seem to work correctly.


  • 1
    \$\begingroup\$ Try juice4halt.com to see a complete unit with schematics and software downloads. \$\endgroup\$ Commented Mar 18, 2015 at 4:52

1 Answer 1


In theory you can substitute any super caps on the input so long as the original source can handle the inrush current. I would certainly add some inrush current protection before this.

Simulating DC-DC converters is extremely computationally expensive, which is likely why the provided SPICE model used 5mF caps. In theory there should be no negative effects (other than what I mentioned about inrush current, which would be a problem even with the 50F caps).

Also, keep in mind that capacitors in series have reduced capacitance, so the net input capacitance of the example circuit with 2 50F caps is only 25F. You need these 2 because most super caps can't handle 5V by themselves.

Suppose you wanted to provide 5V@2A continuous output for 1 minute, and that the converter is 100% efficient. That means you will be drawing 10W constant power from the capacitors, or a total of 600J of energy over 1 minute.

The voltage across the input capacitors can go from 5V to 0.5V. That means the total available energy is:

\begin{align} E = \frac{1}{2} C (5^2 - 0.5^2) \end{align} Substituting in the 600J and solving for C, C = 48.5F. So having 150F (2 300F caps in series) is sufficient for keeping this circuit alive for at least 1 minute.

Now suppose the converter is 50% efficient. All you need to do is double the capacitance required, which means the 300F caps you wanted to use are still sufficient for the job.

This circuit is capable of supplying the 2.5A along with charging the caps so long as your wall wart is capable of providing the current for both (i.e. it must have >2.5A current rating).

One very easy way to sense if the wall wart is on is to add some one-way device and a sense pin upstream. For example, you could use a Schottky diode:


simulate this circuit – Schematic created using CircuitLab

Keep in mind, though, that you'll have to take into account the voltage drop across the diode (typically ~.15-.5V for a Schottky diode).

This circuit will not work as-is with pretty much every modern rechargeable battery chemistry available (i.e. you can't just replace the super caps with a battery pack).

  • \$\begingroup\$ Thanks a lot for the response. I figured the circuit won't work for Lipo or NiMH as is (Lipo circuits seem pretty complex due to the overvoltage / undervoltage issues). \$\endgroup\$
    – Driv
    Commented Mar 17, 2015 at 19:17
  • \$\begingroup\$ I will give it a try with some 350f supercapacitors to see how it works. The data sheet for the LTC3122 does mention that is has inrush protection. I assumed it was already baked into the circuit. In your sense circuit above, what does C1 do? Will minor fluctuations in the input power from the wall wart (e.g. flickering lights, dip in power) cause the GPIO to trigger? \$\endgroup\$
    – Driv
    Commented Mar 17, 2015 at 19:28
  • \$\begingroup\$ "typically ~.25V for a Schottky diode" no, typical voltage drop in schottkys is 0.5V. With the more expensive silicon ones dropping around .2-.3V. \$\endgroup\$
    – KyranF
    Commented Mar 17, 2015 at 20:14
  • \$\begingroup\$ @user69907 The inrush protection is for the output, it doesn't protect the inrush current to charge the supercaps. \$\endgroup\$ Commented Mar 17, 2015 at 20:16
  • \$\begingroup\$ Thanks for the updates. I am ordering the parts and will test out the circuit. \$\endgroup\$
    – Driv
    Commented Mar 18, 2015 at 1:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.