9

When browsing for buck-boost converters I've looked at many datasheets and I've made an observation that I'd like to question.

I have a low-power application and energy consumption is one of my primary concerns. The converters specify efficiency curves, but no manufacturer plots the curve into the very light loads I'm interested in. Most curves start at 1 or 10 mA. I'm curious about the 1-10 μA range, what to expect there?

ICs that don't have some sort of power saving mode all drop very quickly from reasonable 80-90% eta towards 10% at low loads (tens of mA). ICs that do have power saving (everyone calls it a different name... pulse skipping, power saving, green mode, etc.) stay in the desirable range throughout the plots. But what about the μA range. What happens there? Why does no one plot it?

Here are two examples (with power saving)

Graphs

ocrdu
  • 8,705
  • 21
  • 30
  • 42
Alexander Ohm
  • 341
  • 1
  • 8
  • 4
    You should find ones designed for sub-mA loads if you want to see graphs that include sub-mA efficiency numbers. – Hearth Jan 03 '23 at 14:47
  • @Hearth The thing is that my IoT device, like many others, also requires ~1A, but very infrequently. The MAX77827 (right picture) is designed for just such a use case, still I find the datasheet lacking... – Alexander Ohm Jan 03 '23 at 14:58
  • 8
    why would one plot a µA range? That makes no sense, that's in the order of self-discharge of the output capacitors. "Efficiency" is basically 0 where quiescent current >> output curent, so that plot would be pretty senseless – what empties your battery isn't bad efficience, it's the current draw. Efficiency is only relevant under significant load. and even then, all you would need efficiency for is converting the power used to the current drawn to calculate the battery life. – Marcus Müller Jan 03 '23 at 15:01
  • @MarcusMüller Example: the the quiescent current of said MAX77827 is listed as 6-14uA in skip mode, and hundreds of times higher in PWM mode, even if not switching. – DKNguyen Jan 03 '23 at 15:10
  • 1
    @DKNguyen that's between pulses in pulse skipping mode. The typical average quiescent current is likely to be around 75 uA at very light loads. You can infer that from the graph shown in the question. Of course it may be a tad less. – Andy aka Jan 03 '23 at 15:15
  • @Andyaka Thanks! Yes in skip mode it lists Iq as 6 uA. My load is about the same. So if the IC spends a lot of time between pulses in skip mode then efficiency DOES become relevant. But you are correct that it doesn't matter much if Iq is higher, and perhaps it is. Could you explain how you infer 75 uA from the graph? – Alexander Ohm Jan 03 '23 at 15:24
  • @OhmInpw I based it on 90% efficiency at 1 mA load current from an input source of 4.3 volts. 10% is wasted in the device hence, 10% of the output power of 1 mA and 3.3 volts is 10% of 3.3 mW or 0.33 mW. If 0.33 mW is taken from a 4.3 volt supply, the current has to be about 75 uA. OK it's actually 76.7 uA. – Andy aka Jan 03 '23 at 15:30
  • @Andyaka Thanks, that's perfectly clear. But that means you defined your load based on the graph. Light load by my definition (well, my use case) is several magnitudes lower than what's plotted. So then your reasoning doesn't really hold up, right? – Alexander Ohm Jan 03 '23 at 15:35
  • I calculated the losses in the chip for a 1 mA load but, for lighter loads, these losses will not increase and probably, won't decrease much but, in all fairness, the losses could halve to maybe 0.165 mA with zero output current. That would imply a quiescent current of maybe 39 uA. – Andy aka Jan 03 '23 at 15:46
  • @Andyaka Adding 40 uA constantly would decrease my product's lifespan by about 25% so the difference is quite a big deal for me. I'd prefer to find a more robust method of estimation, but truly thank you for contributing. Could you explain the reason why you don't think it would decrease more at minuscule loads? Experience (from what)? A hunch? A guess? – Alexander Ohm Jan 03 '23 at 16:10
  • 1
    Experience in electronics really. You should look for one that does specify low current working and is unambiguous about quiescent current draw. – Andy aka Jan 03 '23 at 16:22
  • Is the candidate buck boost powering only the radio or also the micro? Basically, looking for other options, like disabling it when not in use, but I need to know what voltages and currents you need. – bobflux Jan 03 '23 at 17:09
  • If ultra-low load current efficiency is not shown, it simply is worse :) Look further for parts where this use case is advertised and documented. – datenheim Jan 03 '23 at 20:24
  • You can implement your own super-skip mode by usig a converter at optimum efficiency to charge a suitably large capacitor of suitably low leakage and then shutting it down completely. Depending on the desired accuracy of the output voltage you might follow this with a ultra low differential LDO - but ideally the converter alone will be adequate. A 100 uF capacitor with zero leakage (!) will drop 0.1 V/s at 10uA. – Russell McMahon Jan 06 '23 at 23:46

1 Answers1

13

Why is buck-boost efficiency not specified for ultra light loads (uA)?

Buck devices first

You need to dig a tad deeper and find devices like this one: -

enter image description here

A closer look at the efficiency curve on the front page: -

enter image description here

At 1 μA load current, efficiency is still circa 70%. I'm sure there are others of course like this: -

enter image description here

I used this search link to find them:

https://www.analog.com/en/parametricsearch/11463#/p5573=min|3.6&p5574=5|max&p5347=min|3.3&p5357=3.3|max&p5349=0.1|1.5&qsfv=vinmin|3.6_vinmax|5_vout|3.3_iout|100m&p5362=Buck

Buck-boost devices

The LTC3130 looks useful. It has a power efficiency of 70% at 10 μA load current (LH graph): -

enter image description here

The LTC3129 also looks good: -

enter image description here

Andy aka
  • 434,556
  • 28
  • 351
  • 777
  • 1
    This part only does buck though. It sounds like he is searching for one that also does boost. – Linkyyy Jan 03 '23 at 18:01
  • 1
    @Linkyyy it's a valid point and I'll have a trawl round for examples of buck-boost. – Andy aka Jan 03 '23 at 19:29
  • 2
    I'll consider this an accepted answer with the conclusion that there are components that specify what I asked for, I just didn't look hard enough. Thank you! – Alexander Ohm Jan 04 '23 at 12:41