Please check my reading of this boost converter datasheet

[danderson]

I'm designing a battery-powered board that alternates between relatively high load (15mA), and a very low power sleep mode (0.9uA). The circuit needs to run at 3.3V, so I've been researching boost converters for the power supply. To stretch battery life, I'm looking for a converter that has very low current draw when unloaded.

I think I've found a converter that would work for my needs, and is also well stocked and relatively cheap. However, my datasheet reading skills are still in their infancy, so I'd like a sanity check on my reading and reasoning.

The datasheet in question: http://www.torex.co.jp/english/products/dcdc_converters/data/XC9140.pdf . I would be using the XC9140A331: 3.3V out, load disconnect in shutdown, no UVLO. In circuit, it would have Vbat = 1.5V (1 alkaline/lithium AA).

To figure the supply's current consumption when my board is at full power (15mA load), I took the middle typical efficiency figure from the table on page 4, which is 85% efficiency for Vbat=1.8V and Iout=30mA. I cross-referenced this with the top-left graph on page 18, which shows substantially similar efficiency for 15mA @ 1.8V. From that, and assuming that the efficiency is comparable at 1.5V input, I conclude that the supply will draw a total of 17.65mA from the battery.

To figure the sleep mode current draw (0.9uA load), I assume that the supply will be quiescent most of the time, as the output cap will hold ~3.3V for quite a while at such low draw. So, the supply's current draw is just its quiescent current, listed as Iq in the table on page 4. Looking up the value for E2 on page 8 says that it's 10uA maximum for the 3.3V version of the device. Therefore, I conclude that the average current draw from the battery will be ~11uA, with 90% of that draw going to supply losses and the remaining 10% going to the MCU.

Unfortunately I can't cross-reference this with the efficiency graphs, because they stop at 0.01mA Iout, an order of magnitude higher than my sleep mode draw. That said, the top-left graph on page 18 does show the efficiency dropping dramatically, as you'd expect when the load current approaches the supply loss currents. Extending the curves to 0.001mA suggests that 10% efficiency is plausible.

As I said above, two major assumptions I'm making are that the efficiency is substantially similar at 1.8V and 1.5V Vbat, and that the supply's quiescent current dominates current draw when load drops to a microamp.

Does my reasoning and reading of the datasheet make sense? Have I made a huge mistake somewhere that would significantly change the current draw estimates? Is there some well known low-quiescent-current supply option that I should know about, or is ~10uA losses about as good as it gets?

Thanks in advance!

[T3sl4co1l]

Sounds right.

Efficiency pretty strongly tails off at low currents, suggesting Iq limited operation there.  If Iq were subtracted out, one would have the switching efficiency, which probably remains fairly stable.  But yes, overall efficiency stinks down there, as it should.

You could very possibly add your own micropower comparator to switch the controller between run (Iq 10uA max) and sleep (0.9uA) modes... but you'll still spend a few uA doing it, and it won't respond well to rapid transients (those micropower comparators are s-l-o-o-o-o-o-w!).  For which you may have to increase capacitance or something, and it ends up being a worse idea overall (area, cost, simplicity, stability..).

Tim

[danderson]

Sounds right.

Efficiency pretty strongly tails off at low currents, suggesting Iq limited operation there.  If Iq were subtracted out, one would have the switching efficiency, which probably remains fairly stable.  But yes, overall efficiency stinks down there, as it should.

Thanks for the sanity check. That's my reading also, switching efficiency remains substantially the same, but the constant overhead of the circuit starts to dominate.

You could very possibly add your own micropower comparator to switch the controller between run (Iq 10uA max) and sleep (0.9uA) modes... but you'll still spend a few uA doing it, and it won't respond well to rapid transients (those micropower comparators are s-l-o-o-o-o-o-w!).  For which you may have to increase capacitance or something, and it ends up being a worse idea overall (area, cost, simplicity, stability..).

That seems like an interesting thing to play with for a v2, but assuming the datasheet isn't lying, I can meet my battery life requirements with 11uA worst case sleep current. Interesting thought though, I'll keep it in mind for future improvements.

Thanks!