Are linear voltage regulators more efficient than switching regulators at very light loads?

Question

I'm planning to make a battery application with a PIR sensor. The sensor consumes 70 μA and will be supplied with 4-5 V. I have two Li-ion batteries in series which gives around 8-9 V in total.

I have 2 options:

1. Supply directly from the batteries (there is a battery protection module in between) and let the sensor's linear regulator obtain 3.3 V;

2. Put a buck converter in between and take the voltage down to 4-5 V before sending it to the sensor.

I was planning to use an LM2576 as a buck converter but I realized it draws 80 μA even in off mode so I'm sure it draws even more than that for its operation.

Let's assume 80 μA input for the IC itself and 100% efficiency from the converter to be optimistic. Input voltage will be 9 V and input current will be 70 μA * 3.3/9 = 26 μA so wasted power is {input power} - {necessary power} = 9*(80+26) μA - 3.3*70 μA = 723 μW. I don't know if the power calculation is relevant or not, but the necessary current is more than doubled with a buck converter anyway. On the other hand, a linear regulator will only waste (9-3.3) * 70 μA = 400 μW.

So is it true that linear voltage regulators can be more efficient than switch mode ones?

Answer 1

So is it true that linear voltage regulators can be more efficient than switch mode ones?

If you choose an inefficient-enough switcher and an efficient-enough linear regulator, sure.

I was planning to use LM2576 for buck converter but I realized it draws 80 microamps...

See above. The reason there's a bazzilion switching regulators on the market is roughly the same reason that there's a bazzilion different op-amps on the market: it's a hard problem to solve with one generic solution, so there needs to be lots of options.

That's an old part, designed at a time when the notion of selling a switching regulator than any circuit designer could just plop onto a board and gain success was revolutionary. For most consumers of that part, the quiescent current was the least of their concerns.

So -- find a different part, or go with your linear regulator scheme.

I did a bit of digging, and it looks like really low-quiescent current switchers are out there, but not common. It also looks like the selection is significantly better if you're looking at parts that only go up to 5V or so of input voltage -- so you may need to trim your system design to the parts available, and go with a single LiPo cell and either a buck/boost converter, or just run the thing down to 3.3V on the battery and call it "discharged". This will reduce your effective cell capacity, but you'll have longer cell life.

Answer 2

That comes down to which regulator you can engineer to have lower quiescent current.

Both need some kind of reference voltage generation.

The switching circuit needs some kind of comparator. The linear regulator needs some kind of op amp.

I would argue that a hysteretic switching regulator would have the lowest quiescent current, but of course wouldn't be quite so precise.

An actual switcher would use a pretty fast oscillator which usually needs more quiescent current than a slow opamp for a linear regulator. You might even skip the opamp if the output impedance of the reference circuit is low enough, which might happen for ultra-leight loads.

Answer 3

In this context I would prefer to use the Lithium cells connected parallel and use a capacitive charge pump IC with low operating current (S-882140AMH, LTC1522, LM2775) to lift the cell voltage a little bit with high conversion efficiency). 30-40 µA operating current and efficiency > 82 % should be possible.

Because the cells are parallel, you have twice the runtime compared to the series circuit at the same operating (or GND) current. And you don't need a cell balancing circuit.