Most efficient LIiPo based power supply for ESP32?

[AloyseTech]

Hi,

I'm working on an ESP32-PICO-D4 project. I need to power the device with a LiPo battery (~2000mAh). The device will mostly be in deep sleep (5-10uA typical), with periodic wake-up every 10min to 6 hours, for 5-15seconds, at 120mA average (400mA peak). The ESP32-PICO minimum supply voltage is 2.7V but it might be too low to ensure no brown-out occurs at WiFi peak power consumption.
Here are the few option I have in mind :

• Use a very efficient LDO set at 2.8V or 3V, able to supply 400mA (TPS7A26 for example)
• Use a buck/boost converter to supply steady 3.3V : TPS63030, TPS63001, LTC3130(very expensive)
• Use a buck converter to supply 2.8-3V : TPS62742

What do you recommend? Do you know any other good existing part for the job?

[Siwastaja]

1) Calculate the average duty cycle and average current
For example,
average wake up period = 30min = 1800s
active period = 10s @ 120mA
-> average current = (1790s * 10uA + 10s * 120mA)/1800s = 680 uA

2) Now compare the efficiency of the LDO vs. switch mode supply: look at the consumption at 680 uA load.

It's very likely the LDO will be the winner, although some small, modern switchers with good light load modes could work out if the assumptions change (for example, more frequent wake-up).

If you are going to need, say, 3.0V output, you are going to drop only 0.65V average, and the linear solution offers max efficiency of about 80%, which is achievable in practice, too, by choosing a very low quiescent current part. It's almost impossible to beat that.

For extreme optimization, one could use a hybrid LDO + switcher solution, but probably not worth the complexity.

Peak dissipation of 1.2V*120mA = 144mW for < 15 sec, or 1.2V*400mA = 480mW for some milliseconds max I guess, should be thermally trivial to handle as well, for any LDO package.

Do take a close look at the dropout voltage at your max peak (400mA) load, remember that in order to fully utilize your 2000mAh cell, you'd like to want to take it down to about 3.0V. If your absolute minimum at load is already more than 2.7V, you may need to sacrifice a bit. But clearly LDO's with dropouts well over 300-400mV start hurting your battery life by not allowing you to run the battery empty, so aim for ~200mV dropouts. Your large battery capacity is helping here: 400mA is only 0.2C, so there is no need to discharge to very low voltages; stopping somewhere around 3.1V, maybe even 3.2V, already gets you close to 0% charge left.

[AloyseTech]

The TPS63030 buck/boost has more than 80% efficiency at 3.6V input from 200µA to 400mA output. Most of the time the periodicity will be 2h or more with 15seconds on time, so the average current will be around 200µA.

LDO's with dropouts well over 300-400mV start hurting your battery life by not allowing you to run the battery empty, so aim for ~200mV dropouts..

The TPS7A26 is not a good candidate then, since the max dropout is around 450mV at 400mA... I liked it for the low quiescent current.

The TPS62742 buck has a quiescent current of less than 1µA and an efficiency of >90% starting at 10uA load. Maybe this could be a good candidate if I set the output voltage to 3V... It should be better than the LDO solution with the average efficiency of 80%. It is not clear what is the true continuous current I can expect from the buck at VBAT = 3V. It seems the current is limited to 100mA if the headroom is less than 0.7V...

[Siwastaja]

The TPS62742 buck has a quiescent current of less than 1µA and an efficiency of >90% starting at 10uA load.

This is really impressive, thank you for pointing it out.

Look out for ripple voltage, though, as the control is hysteretic, depending on voltage ripple at the output cap.