Sometimes we have power outages that last about 100 ms that are long enough to discharge capacitors inside router, when that happens we need to wait about 10-15 minutes for DNS, which is pretty annoying. What I'd like to do is have (cheap) UPS only for the router that will withstand 0.5 s power outages. Router specs (I will test them with a multimeter to be sure about amperage) are 12V 1.5A. I think my options are:

  • SLA battery with a charger and DC boost converter, 4V 4.5Ah (smallest one I can get). I searched for 4V SLA chargers but I can't find any, I would need to make one from standard 12V switching power supply via timing circuit and step down converter. I would need to make further research about charging SLAs.

  • Capacitors / supercaps. Adding electrolytic caps in parallel between DC barrel jack. Don't know the viability of this solution. Would need to research how to calculate appropriate capacitance.

The solution with capacitors would be much simpler.

Edit: after doing some more research I came up with this setup:

2 × 18650 batteries connected to 2S PCM circuit 2S 5A PCM that go through boost buck converter to give 12V (also with 2S battery indicator). Not sure how to make it only work when there is no power so that the wall wart is still connected. Only minus is that after few discharges/long use time I would need to recharge the batteries, but that won't be a problem.

  • \$\begingroup\$ Why do you want to go as far as big SLA batteries? I think looking at a standard 18650 battery solution, with the needed DC/DC convertor and charge controller could be easier. \$\endgroup\$
    – Joren Vaes
    Aug 6 '17 at 16:01
  • \$\begingroup\$ I'm not sure about this, but a 12V lead acid can't be overcharged by the power brick. A few meters of wire should have enough resistance to limit the charging current. A bad car battery would probably not have any issues with being connected in parallel, should be reasonably cheap, and could probably handle 0.5 hours. But they are not suitable for indoors use! If overcharged or charged at too a high current they will produce hydrogen gas. They may leak sulfuric acid. They are very dangerous when shorted. \$\endgroup\$
    – Oskar Skog
    Aug 6 '17 at 16:17
  • 1
    \$\begingroup\$ electronics.stackexchange.com/questions/19837/… \$\endgroup\$
    – Oskar Skog
    Aug 6 '17 at 17:25
  • \$\begingroup\$ A few supercapacitors in series to withstand the voltage. Connect in paralell with the power supply. Perhaps a series resitor to limit the inrush or diode-resitor. Boom! Done. \$\endgroup\$
    – winny
    Aug 6 '17 at 17:27
  • 1
    \$\begingroup\$ No need for UPS. What's the DC voltage it accept? \$\endgroup\$
    – Jason Han
    Aug 7 '17 at 11:42

Calculating the needed capacitance is easy for a simple parallel cap: 1 microfarad will supply 1A of current for 1 microsecond while dropping the voltage by 1 volt. 1A is a typical steady-state current for a wireless routers, and I suppose most of these can survive an input voltage of 11V, so you roughly need as many farads as the seconds you want your router to last. Maybe your router can withstand a drop to 10 or 8 volts, in which case you can get away with a half or a quarter of that value, respectively.

What's more tricky is to make such caps work as intended. First, most power supplies won't like it when you cold-start them with a huge cap connected to the output. You could quickly kill the internal resettable fuses or rectifier diodes if you do this. This can be solved with a pre-charge circuit, such as this one (look for a Schottky diode with low voltage drop):


simulate this circuit – Schematic created using CircuitLab

Second, many supercaps look more like batteries than capacitors when you use them. Pay attention to their current rating and internal resistance to make sure they can actually supply 1A.

For SMPS, you can probably keeps the caps even smaller by tweaking the hot (high-voltage) DC bus instead of the cold (low-voltage) one. Most SMPS can run with anything between 80 and 230VAC, so (assuming your mains voltage is 230V) the caps can give away almost 2/3 of their charge while keeping the SMPS running. This fact together with their higher voltage (leading to higher stored energy) reduces the required capacitance by about 2000 for an SMPS with a decent efficiency. Photo flash caps are great for this, since they are relatively inexpensive and rated at 400V. The obvious disadvantage is the need to properly test and isolate your hack, because getting shocked by a 400V cap is often deadly.

Batteries are certainly an option, but you'll need to implement some sort of charger (depending on the chemistry you chose) and a boot converter to raise your battery voltage to 12V. You will also need to replace the battery once in a while (compare to good caps which will last for 20 years or more).

  • 2
    \$\begingroup\$ Trying to upvote one more time after you added the pre-charge circuit, but you can only upvote once. :( \$\endgroup\$
    – pipe
    Aug 7 '17 at 11:52
  • \$\begingroup\$ @DmitryGrigoryev I have few 1n4148 on hand but these are rated for 300mA. Unless you don't need full current capability. \$\endgroup\$
    – Oxmaster
    Aug 7 '17 at 17:28
  • 1
    \$\begingroup\$ @Oxmaster 1n4148 is a signal diode with small max current and a huge forward voltage drop. You need something like sr240. \$\endgroup\$ Aug 7 '17 at 19:59

Chances are your problem is gone if you just get a power supply that is overrated. Like a 3A supply brick from an external hard-disk.

For a UPS the sub-second power outages are problematic, those would require an online UPS. An Online UPS has an AC->DC->AC topology instead of a bypass relay.

Capacitors or overrated supply can help you survive these sub-second glitches, but why not go full UPS instead?
You can get a DC UPS from Mean Well like the DRC-60A, or Phoenix QUINT-UPS.

  • \$\begingroup\$ Even 0.1s at 12W is going to require the PSU to store 1.2J internally. That's what I'd expect from a 12V 10A or 20A brick. But this is a viable approach, especially if the OP overestimated the length of the power glitches. \$\endgroup\$ Aug 7 '17 at 12:15
  • \$\begingroup\$ @DmitryGrigoryev I don't have stopwatch with me when power goes out :(. Is there any easy way that I could calculate time that router can withstand without any support? Maybe if I use voltage divider and detect with arduino when pin goes LOW? I have relays that could switch off AC power, that's an idea. \$\endgroup\$
    – Oxmaster
    Aug 7 '17 at 17:48
  • \$\begingroup\$ @Oxmaster: You could try my cheeky answer to electronics.stackexchange.com/questions/322127/…. In your case you would have a contact at the top and at the bottom. Your circuit would be interrupted while the 'contact' falls. \$\endgroup\$
    – Transistor
    Aug 7 '17 at 17:55

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