nRF52832 efficient power management solution

[dingari]

I'm working on a design that incorporates an nRF52832 along with some sensors and drivers.

Up to this I've been sacrificing power consumption in prototypes by using LDOs since they're cheap, small and easy to work with. But now I'm starting to look into a more efficient power management setup.

The device is small and runs off a single LiPo (3.0 - 4.2V range). Using just the nRF52's internal DCDC converter alone is not sufficient. And since I have to step the supply voltage down for the nRF52 anyway, and the board has sensors and drivers operating on some milliamps, I'm looking for an efficient external buck converter to supply power to all the chips.

I'm thinking of stepping the battery voltage down to 2.5V or 1.8V (depending on chip availability for one of my sensors). The device is never "off" in the sense that the voltage regulator is never shut down. Instead all the chips go into sleep/off mode. The total sleep current (apart from the regulator quiescent current) is about 10uA, so I'll need a pretty efficient regulator.

I'm mainly looking at QFN/DFN/WFN chips that I can easily assemble by hand. Some chips I've been looking at:

• Texas Instruments LM3671
• Texas Instruments TPS6224x
• Texas Instruments TPS62740

The TPS62740 is looking extremely good since it claims to get 90% efficiency at 10uA, while others are typically around 20-30%. However, it is a little bit more pricy, and a little bigger. Are there some chips that are frequently used with nRF52?

Then there is another thing. With a supply voltage of 2.5V (or even 1.8V), how much gain is there by using the nRF52's internal DCDC converter?

Edit: The TPS62740 also offers the benefit of an optional load switch that can be controlled by the nRF52. That way the sleep current would just be 0.3uA (the system OFF current of the nRF52) + regulator quiescent current.

Edit 2: I have white LEDs with a forward voltage of 2.8-3V, and a haptic motor that requires 2.5-3.3V. So if I want to lower the main power rail below 3.0V, I'll run into trouble with the LEDs, and if I go further down, my motor won't work.

What I'm contemplating right now:

• Use TPS62740 to provide the main voltage rail for the nRF52 (1.8V), and use the LOAD output to power my sensor (this way I'll be able to shut it down completely from the micro).

• Use another buck converter such as TPS62230 to generate ~3V for the LEDs and the haptic motor.

This way, in "off" mode, only the VOUT of the TPS62740 is active (LOAD disconnected) and the nRF52 is sleeping (total Iq less than 1uA), and the secondary (~3V) regulator can be shut down from the micro.

This is a bit more convoluted and takes up more board space, but I'm looking to maximize the battery life. But the question is, is it worth it?

[dingari]

Hi @AmbystomaLabs, thanks for your input. Really appreciated. I'm working with a LiPo whose cut-off voltage is 3.0V, and I plan to integrate standard protection circuitry on my board (Texas BQ297xx or equivalent). The other chips on the board use 1.8V logic levels so that shouldn't be a problem, and any LED's I'll configure as active low. I'll have a look through ST's site for some regulators.

[dingari]

I've posted an update in the original post, any thoughts?

[AmbystomaLabs]

Even if you are using 1.8V for all your logic you still need to populate the onboard dc/dc of the nRF. It will reduce the current consumption by 1.3/1.8 or 72% vs. the internal LDO which doesn't reduce current consumption at all.

The load switch on the 62740 does provide a convenient way to handle disconnecting your unused circuitry, however you could easily implement this with a p channel fet in a 0.6x0.8mm package.

Make sure you can control your haptic motor with low side switching. Since gpio vdd is 1.8v you can't high side switch unless you level shift.

[AmbystomaLabs]

I just wanted to clarify the logic bit, since vdd is 1.8 you cannot try to talk directly to 3v logic. The 3v logic will have internal pull ups to create 3v high. 3v on the nRF gpio will just turn on the high side fet of the nRF and drive the nRF and all your other circuitry with 3v from the logic. The only way for you to talk to your haptic is via the gate of an N channel fet that does low side switching of the power on the haptic.

[dingari]

Yes I understand level shifters and why and when they are needed. In my application the nRF52 is talking to two I2C slaves (a MARG sensor and a haptic driver), each operating at 1.8V slave logic level (regulated internally by each chip). So no level shifting should be necessary. The main reason I'm eager to go with the TPS62740 is the high efficiency at low currents. And another thing: My understanding is that the nRF52 GPIOs can be configured as proper open-drain/collector, so writing 0 closes, and 1 leaves the pin open. That should allow supplying LEDs with higher supply voltage than the nRF52, without having voltage drop (LED_VDD - nRF52_VDD) across the LED in an open/high state (writing 1 to the pin).