You can just power the pro micro with 5V, it won't complain. Actually, that's what happens when you connect it to USB.

I'd still need to convert the 3.3V for the NRF24 module and I have a FTDI that is switchable in voltages. But damn, the Arduino IDE doesn't know what voltage is on the Arduino, that selection is only for crystal speed.

The pro micro has a build-in voltage regulator, it outputs 3.3V, should be enough for the NRF24?

Personally I would just connect the lipo directly to the RAW pin, and then power everything from the VCC pin.

It's the pro mini but I guess it stays valid. I'm gonna remove the LEDs on everything to keep the power consumption down/low.

I'd use low-power LDO like MCP1702 directly from Li-Po. Actually, MCP1702 is my favourite part for this, as its quiescent is 2uA, so for most of the applications I can let circuit powered all the time, just kicking MCU down to low power mode. Just like here https://github.com/jaromir-sukuba/mini_alt/blob/master/design/mvm1a.pdf in project #Mini Altimeter

MCP1702 comes in TO-92 package too - for breadboarding - and you can choose from a few output voltages, where 3,3V and 3,0V version is most suitable for this one. If you don't specifically require 3,3V, go for 3,0V - you'll get few more percents of accumulator capacity, enabling to discharge Li-Po deeper. 

Even 3V3 output is not a huge problem, as under 3,5V the Li-Po capacity falls down really fast and there is not much juice left under 3,5V anyway where the MCP1702-33 still works with load up to few tens of miliamps. I can probably find better spec'd parts, but this one is always in my junkbox.

Thanks, that must be the part I didn't get from @K.C. Lee comment. How could I forget the endless rants from Dave Jones about the batterizer... so the LiPo is almost done anyways when around 3.5V, which would also be the point where a proper regulator drops out. I also wasn't aware of that such low dropout voltages would be possible.

On the MCP1700, the dropout is 50mV if you keep the load under ~70mA.  Lower currents gets you even lower dropouts.  So you can actually get down to 3.35V at 70mA while still maintaining regulation.  I like this part because it goes up to 250mA while dropout is 178mV typ.  :)

The 200mV was just an example.

BTW if you look at the curve, the average battery voltage is 3.8V.  So on average you are getting 3.3V/3.8V = 86.8% efficiency for the LDO.

Yes, MCP1700 looks even better than MCP1702.

If possible I would run all directly from the lipo voltage an would just add some lipo under voltage protection.

If you need 3.3V I woul use some switching regulator (like LM3671 or MCP1603)

You don't need a buck boost converter for 3.3V if you are running off a LiPO.  You might be gaining about 10% or so more capacity near the end assuming the boost didn't come with a price.  Also bare in mind that the battery cycles can be extended if you don't deep discharge these Li-ion.

Instead you should look at the min operating voltage of your parts to see if they could run below 3.3V. Real nRF24 parts operates down to 1.9V.  If you lower the clock speed of the Atmega to 10MHz, the datasheet say it can run down to 2.7V.

Ditch the 1117 regulator.  It is not designed for battery operation - huge quiescent current, large dropout.  I am using MCP1700 series LDO for my battery operated projects.  For buck converter, I use AAT2120.  0.4mm pitch DFN buggers - probably not for you.  Depending on your load - if it is sleeping all the time and wake up with low current 20-30mA burst , it is not worth using a switcher.

so to recap in own words: go down with voltage to get out of dropout range of the (better) converter?

My general concern/wish is to have a stable 3.3V out of a lipo (3.7 to 4.2), not how I should power this project in particular. 

But thank you for your input :)

Not sure what you mean.

In general a buck-boost converter is 5-10% lower efficiency vs a buck converter as there are more parts that the current has to go through.

Using a LDO would get your regulated voltage.  e.g. with 200mV drop out will get you regulated output between 3.5V and up.  At below 3.5V, the output is essentially the input minus the drop out. IRL you have to read the datasheet to find out *exactly* what your part would do under the load current, voltage, temperature.

Also unless you need exactly 3.3V, there is nothing to gain by fixating on a voltage.  Ratiometric ADC conversionan get you the gamepad pots readings regardless of the actual supply voltage.  

i.e. use of supply voltage as ADC reference to read divider powered off the same rail, the supply term is cancelled in the math.  Now if you have to read off an external sensor with absolute voltages, then it is another story.