ARM with a dual switched regulator

[brucebertrand]

I'm looking at running an ARM micro and a few other devices on a board with an LTC3558, which has both buck and buck-boost regulated outputs.  The buck-boost appears to be the four-switch type (non-inverting).

Everything should run happily with a single 3.3V supply.  Even though the ARM will work down to 1.8V, at least one device on the board has a dropout voltage of just under 3.3V.  I'm planning on using a single LiPo cell to power the device, and would like to eek out as much runtime from it as possible.   The entire project will draw around 100mA max, but with the possibility of frequent sleep periods on the micro and other devices, during which I might see something like 5 to 15mA total draw.  I'm looking at LiPo cells in the neighborhood of 1000mAh. 

Considering the voltage requirements, I'm assuming that I'll need to use the buck-boost regulated output of the 3558.

My question is this:  Is there any reason I should even think about using the buck regulated output?

• Is there really a significant improvement in efficiency with buck topology vs buck-boost?
• Should I just run the micro on the buck and have it sit at 1.8 while everything else is at 3.3 on the buck-boost? (FWIW, I'm pretty new to this stuff, so this might be a big can-o-worms that I'm not quite ready for...)
• Should I come up with a single 3.3V rail topology that switches from the buck to the buck-boost at a threshold voltage?

Thanks in advance!

[nctnico]

Just do the math when using a switcher versus a linear regulator! A switcher uses a few mA at least so it is likely not the best choice for powering a circuit which needs less than 100mA.

[mark03]

I'm not sure there is much point in buck-boost from single-cell lithium ion, as there is not much recoverable energy from the cell below 3.2V (which your "3.3V" rail could drop to without the system failing).

But as nctnico said, you just need to do the math.  I would amend what he said, though, in that you can get switchers made specifically for low-power operation, maintaining reasonable efficiency down into the 1-10 mA region, with quiescent current far below a few mA.  Just go to TI's web site, drill down to buck regulators with integrated switch, and have a look at the available Iq given your other requirements for packaging, input/output voltages, etc.

Another factor, somewhat counter-intuitive, is that micros with a large supply range usually draw less current at, say, 1.8V, than they do at 3.3V.  This means that even if you obtain the micro's rail from a linear regulator ("LDO" in industry-speak), you still use less power overall.  Any efficiencies gained from a switch-mode regulator are over and above that.