Battery backup options

[akis]

I have a device which works at +/-18V and also contains an MCU and some other peripherals that work at 5V. The draw at 5V is around 130-170mA.

Sometimes one or both of the supplies fail and then the MCU simply stops and there is no warning that the device has stopped working. In other cases the MCU monitors the device for good operation and will beep loudly if there are issues.

I was thinking of adding a small 5V battery (rechargeable, haven't found one yet, it's hypothetical) and stick it/hack it inside the device, so that if the 18V supplies fail, the MCU will still be powered and will beep to attract my attention.

It does not need to last hours of course, but a few minutes of beeps will be enough.

So I have sketched the schematic below, could you please tell me if it looks OK?

The schematic shows the current setup with the addition of R2 to current-limit-charge the battery and D2 to bypass the R2 and provide voltage to the LOAD when power fails (switch S1 on the far left). The 33R resistor is in place to heat up (a lot) and spare the 7805 which would otherwise cook.

[ZeroStatic]

Possably a supercap may fill your need for backup of the controller system only, otherwise I would recommend a pack of 4 NiMH cells to provide the backup if you want a rechargeable solution. If recharging is not required a set of 4 x AA cells with a diode will work and is cheap and easy.

Brenden

[fcb]

A supercap will have to be pretty huge to give you 5V @ 170mA for a few minutes.

I'd use a LM317T regulator setup in constant current/constant voltage mode (I think National used to have an app note that explained how to do it - it was only perhaps 3 resistors as well as the 317T) to trickle 4xNiCd/NiMh cells.  You'll have to figure out how to feed the battery voltage onto the processor rail - probably via a diode and LDO? How is the 5V regulated in the existing design??

[akis]

I have done some research.

The NiMH is cheap £4.80 for the 4.8V/2550mAh pack. But it needs to be charged at C/10 (or less) to be allowed to over charge without damage. Still the charging has to stop at some point and that would require some components. Then, if power is lost, the battery has to be disconnected after say 2-3 mins of alarms buzzing, else it would go on for ever even when you do not want it, and would drain the battery when the device is off. More components required then,

The .47F supercap would provide 2T cycles of discharge to go from 99.33% charge to 95%, into a 30R load, in almost 3 minutes. According to my calculations that is. 3 minutes is long enough to sound the alarm buzzer. And we do not care if the supercap gets drained completely, it is not a NiMH battery, and it cannot overcharge!

However this super cap would be like a super-short and will probably cause all sorts of harm. So it would need to be charged through some limiting resistor/LM317 in CC, but for discharge it would need a diode, which would then drop 0.4V, so the MCU would get 4.5V or thereabouts, not 5V.

This is what I have come up with.

[Red Squirrel]

I like lead acid for stuff like that since you can just float it and not worry about overcharging and they're cheap and relatively safe (no fireworks display if something goes wrong).  I'd look into a 6 volt battery and voltage regulator for the MCU.

6.75v PSU (to keep battery on float) -> battery -> 5v regulator -> MCU

MCU could also monitor battery voltage, when it hits around 5v it turns off. (5v relay or something)  If PSU fails MCU does not even notice and keeps running.