# Efficient power supply for embedded project

I'm working on wireless sensor network project consisting of sensors based on ATmega328 and MRF24J40MA. The sensor is expected to stay asleep most of the time, waking up on regular intervals to collect sensor data and sporadically sending collected data over 802.15.4.

I'm not sure on how to power the circuitry for longer life on batteries. I consider two options:

Use LVD1117V33 to scale down 4xAA batteries to 3.3V. I guess this means batteries operational voltage range of 6V..3.3V.

Use NCP1402 to scale up 2xAA or 1xAA to 3.3V. I expect the battery to be drained nearly to its minimum until the circuity fails.

The second option sounds more promising, but don't I overlook something? Are there any options?

• The first part is a linear LDO regulator with a dropout of about 1V, which means that you need at least 4.3V input voltage to keep regulation. It also means that (because of the voltage drop over the regulator), you will waste a considerable amount of energy through heat dissipation on the regulator. – us2012 Jul 2 '13 at 15:11

Your ATmega328 works down to 1.8V, the MRF down to 2.4V. A very simple solution would be to use two alkaline (or one LiSoCl2!) batteries without any conversion. The MRF will work down to 1.2V per battery. I don't know how much juice is left at that level, I could not find a graph in a few minutes googling.

Another simple option would be to use 3 alkaline batteries, and feed the MRF via a linear 2.7V regulator that is switched on only when needed (either a P-fet with a 3-leg regulator, or a regulator that has an off pin). If you parallel a few ATmega328 output pins you can probably even feed a 3-leg regulator directly. An MCP1702-2.7 has 0.7V dropout (?), so the batteries would be useful down to ( 2.7 + 0.7 ) / 3 = 1.13 V. If you want to be really clever: at 1.13V per battery you could let the ATmega328 feed the battery directly to the MRF, bypassing the regulator.

• Forgot to mention that I also need regulated 3.3V for my ADC sensor. – Farcaller Jul 2 '13 at 18:18
• If true that changes the playing field a lot, but why would you need 3.3V? – Wouter van Ooijen Jul 2 '13 at 18:20
• Disregard "ADC" in the comment above, I don't know why it came to my mind :-) I actually meant "temperature sensor", I'm using DHT22, which operating voltage is 3.3-6V. – Farcaller Jul 2 '13 at 18:38
• So in both suggestions I made you could feed the DHT directly from the batteries, or maybe from an ATMega pin so you could power it down. – Wouter van Ooijen Jul 2 '13 at 18:44
• is voltage regulator a more effective option? I.e. if it regulates voltage by stepping down, it sounds like less effective comparing to step up. – Farcaller Jul 2 '13 at 22:02

The NCP1402 looks a decent choice for one AA battery but with two batteries in series you might find the output voltage (3.3V) rises a bit above 3.3V on light loads. Figure 4 doesn't indicate a voltage higher than 2.5V can be used to generate 3.3V whereas fig 58 implies that for larger load currents you may be ok with an input voltage of 3V.

Figure 57 also implies (due to the graph curve) that slightly less than 3V will be OK for 3.3V output on no-load.

The suggestions so far imply you have two options:

1) Use a linear regulator, dissipating Vin-Vout as heat

2) Bypass the regulator by dropping 100% of unregulated Vin through device

A third option would be to use a switching regulator. Boost converters produce higher voltage at the output than the input and, therefore, require a higher current at the input than the output which none of the batteries described above should have an issue supplying. I didn't pull the datasheet for your specific component, but there is probably a drop-in (package- and pin-compatible) switch-mode equivalent for it requiring few or no external components.

Depending on your sensor refresh rate, a rather tiny solar cell (if sunlight is available to your device) could power the boost converter and run the device for "free" while simultaneously charging a super-capacitor to power the system through the night until morning. Two supercaps in parallel would approximately double your run-time without input power. I suspect that a couple of supercaps could potentially run your setup for many days between power applications, but you could also just substitute an appropriate rechargeable cell if you want to guarantee power for longer periods. The boost converter will (slowly) float the cell back to full all day long as long as there is enough sunlight.

These aren't spec'd especially for your application but to give you some ideas:

These are extremely tiny: https://www.sparkfun.com/products/9962

Since I didn't sum your power requirements, this is possibly too extremely tiny to even parallel cost-effectively for your purpose; in that case, select one slightly less tiny. :)

DIP8 step-up/down: https://www.sparkfun.com/products/317

1F, 2.5V supercap: https://www.sparkfun.com/products/10068

Two in series would make a 0.5F, 5V backup.

10F, 2.5V supercap: https://www.sparkfun.com/products/746

Linear also makes a thousand different families of power-management chips. Many are SMPS's at heart, but come equipped with a variety of handy features like output-enable, reverse-polarity protection, brown-out/low-power detect, redundant battery fail-over and/or backup power supply, load balancing, balanced battery charge and/or discharge, over-current protection, etc, etc... You should be able to find the perfect chip very quickly using their parametric search.