Battery backup: capacitor leakage

[sjb]

I want to charge a capacitor from a buck/boost ic, then run a PIC from this capacitor. The PIC will often be in a 'low power' mode.

"Why not simply run the buck/boost ic continually?" Typical 90% efficiency, so I think less enery is used if I charge a reservoir capacitor to ~= 3v3, and re-charge when voltage falls to ~=2v9.

I would like to simply make a Excel chart showing 'e^-t/rc' of various capacitors and choose one that is a suitable performance/price tradeoff. However, it seems that capacitor leakage is not usually specified simply as a parallel resistance.

Does anyone have a specific capacitor recommendation ?

[phonoplug1]

Use a TPS61025 instead? 96% efficiency. I don't think the last 4% will make much difference to overall battery life - the main factor will be the amount of time the system is not in 'sleep mode' and the amount of current consumed at those times.

Still, if you are determined to charge up a capacitor and do it that way, you should avoid electrolytics - they are leaky. But then your choice may be limited by the capacitance you need. Why not charge up a 'gold cap' to run the PIC when its in sleep, and when not in sleep turn the switcher on to power directly.

[sjb]

Hi Phonoplug,

The backgound is this: Assume 2*AAA alkaline cells. When they are low, the total voltage is barely enough to run the PIC & other parts (external flash etc). When they are healthy, total voltage is sufficient. So to get maximum battery life I put in a boost circuit.

I'm not set on the capacitor idea, simply wondering about pros & cons.

"I don't think the last 4% will make much difference". That's a good point, but the situation is complicated by the fact that when the PIC is idle, it does not use much current and then the efficiency of the buck/boost is low.

TPS61025. Device Quiescent Current: 25 μA (Typ)
PIC24FJ256GB1: 100 nA standby current

Interestingly, I came across the similar TPS61220 (and friends) only yesterday:
5.5 μA Quiescent Current

[phonoplug1]

How long does it need to run from these batteries? With a decent set of AAA batteries (Procell) that have a capacity of 1175mAh (remember thats based on a significantly higher loading than 25uA), a 25uA loading should be powered for 47000 hours, or nearly 5 and a half years!

Thats why I say the significant factor to the battery life will be the amount of current used when not in sleep, and/or the proportion of time its not in sleep mode. Otherwise if 5 and a half years isn't long enough I'm not sure any battery will be any good after 5 and a half years, loaded or not!

I think if there was much advantage in charging a capacitor as you originally suggested, then there would be chips around that work like that. There are many switcher chips that have a 'discontinuous mode' to make them more efficient at light loads. This does increase output ripple at light loads though in most cases (like when a micro is in sleep anyway) thats not a problem. In discontinuous mode the switcher effectively goes into a low power sleep mode between bursts of switching.

[sjb]

Thanks for the update. When you say 'discontinuous mode' I see that (for example) The MCP1640 DS has this to say "The MCP1640/C devices use an automatic switchover from PWM to PFM mode for light load conditions to maximize efficiency over a wide range of output current". Also, "Low Device Quiescent Current: 19 μA, typical PFM Mode".

Basically I'm wondering about taking advantage of the 100nA sleep current.

One thing I see is that some switchers allow you to short input to output: the PIC in question runs as low as 2v0, and so I now think that may be a better/easier solution that the capacitor.

On the (MCP1640C/D) chips, that is called 'Input to Output Bypass Option' (I would need to 'manually' select this mode, presumably just before entering sleep).

[thanx]

Thanks for the update. When you say 'discontinuous mode' I see that (for example) The MCP1640 DS has this to say "The MCP1640/C devices use an automatic switchover from PWM to PFM mode for light load conditions to maximize efficiency over a wide range of output current". Also, "Low Device Quiescent Current: 19 μA, typical PFM Mode".

[thanx]

I want to charge a capacitor from a buck/boost ic, then run a PIC from this capacitor. The PIC will often be in a 'low power' mode.