Low quiescent current supercapacitor power management.

[xani]

Hi,

I want to build a project, tiny sensor station that is powered by solar power and radios its result to nearby base station (just nrf24l01, nothing fancy). Basically micro will sleep for most of the time and occasionally (when there is enough power left in cap), send a burst of data with sensor info.

So I need about 3.3V 20-30mA peak, with few uA on standby and tolerance to low/high temperatures. From what I've calculated *basically* I need enough storage to get it thru the night which at the vey worst case is around 12 hours

So it seems to me that slapping a supercap on it is not a bad choice as I don't want to really bother with changing batteries or want to include charging circuit

From what I've calculated having 1F cap discharging from 5.1V  to 3.4V (I just want to slap a zener in to control the voltage, but chip with dedicated cap circuit could just go 5.4V) would allow for around 40uA of average current, which seems doable as micro in standby only takes around 1uA with 32khz timer running and I dont plan on running radio at night (or batch measurements and run at reduced rate).

Question is... how ?

I've looked into dedicated chips like LTC3105 but they seem to be overkill on price (but separate outputs and power good would be handy), and a bit big on quiescent current, also no step-up from supercap so it still wastes quite a bit of its capactity.

From what I've calculated just using LDO MCP1711 would be not only MUCH cheaper (well I doubt I will make more than 10 of those but still) but savings on Iq alone would be worth it. But that's still whole lot of wasted capacity.

So I've started looking into DC/DC converters. I need 3.3 from 5.1 to(preferably) around 1V (but it doesn't need to start from that) so I need SEPIC/buck-boost  converter (forgive me if I fucked up something, never really builded a DC/DC converter into anything yet)

LT1615 looked interesting at first sight but 20 uA Iq and "costs more than micro I used" not so much

I've also looked into microchip ones but all of the low Iq seem to be of boost type. Digikey/mouser dont have Iq on parametric search so I dont really know where to look for next

[JS]

As you've seen low power SMPS at low voltages aren't that efficent, as you need some power to run the thing itself.

You could try with a μC which runs from a wide range, like 5V to under 3V, and get away without regulator, or a low Iq LDO, from 5.5 to 3.5 you average 4.5V, to go to 3.3V you are getting 73% efficency which is not that bad at this low powers, you won't be getting 95% out of a SMPS here, more like 80% if you are very lucky with some added complexity and cost, so if you can get away with the LDO or no regulation at all, maybe just a zenner to limit the maximum voltage and avoid overcharging and you get the "battery life" you need you probably should stay with that.

Once you get to a few tens of mA thinga might change but you are not using that for too much time so I don't think it will be useful to have a step down to use when transmitting and then LDO to keep powered while sleep.

JS

[xani]

In theory both CPU and radio can run from ~2V, I wanted 3.3V so it could power variety of sensors too.

SMPS idea was less about getting max efficiency of conversion itself (as that doesn't matter that much compared to idle current) but more about getting the most out of cap capacity (as going to 1V would basically double it)

But you've made me think about something else, I could just run micro at 2V and make it activate a step up (or just... act as one, it does have adc after all) when cap is below 3.3V to power the peripherals when it is measuring stuff

[thm_w]

1F cap is about $1.10, changing to a 2F cap is around $2, from china (or put two in parallel). From digikey its more like $2.50 vs $5.00 or so.
If you can spare the additional space that would be the simple solution.

LTC3105 is crazy expensive, there are probably cheaper options but to make something for $1 would be a challenge.

Another option might be 3 AAA nimh batteries, a lot of the cheap solar LED lights get away with attaching the solar panel directly to the cells with a diode. Slightly more expensive with much more capacity, and lower lifespan.

[xani]

Actually something like MCP16251/2 is only like $0.6 + maybe a $0.2-0.3 in passive components so it is pretty comparable cost-wise

[JS]

In theory both CPU and radio can run from ~2V, I wanted 3.3V so it could power variety of sensors too.

SMPS idea was less about getting max efficiency of conversion itself (as that doesn't matter that much compared to idle current) but more about getting the most out of cap capacity (as going to 1V would basically double it)

That's exactly what chews your efficency in low power, in high power you go over 95 no problems. In low power having the thing running might waste more energy than the load.
[/quote]
But you've made me think about something else, I could just run micro at 2V and make it activate a step up (or just... act as one, it does have adc after all) when cap is below 3.3V to power the peripherals when it is measuring stuff
[/quote]
Depending on the power requeriments of the sensor you could even make a charge pump with the micro.

JS