I've spent almost 12 hours trying to get two NRF24's "using an UNO & a MEGA-2560" talking to each other when I stumbled upon the power issue topic considering the Mega boards. Here's a screen shot showing the 3.3v power lines on the UNO "Blue trace" and the Mega "Yellow trace":
and the MEGA
Apparently these RF modules are quite sensitive to how clean your voltage line is and a stable supply waveform seems to be essential for proper functioning.
I'll mention the various methods I tried to overcome this issue and their impact in case it helps anyone:
-Tried adding decoupling capacitors starting from as low as 100nf up to 10uf
---> Minimal improvement of the waveform shape but it didn't work.
-Tried powering the MEGA from an external power source
---> No effect on the waveform so it didn't work either.
-Used an external bench supply to power up the RF module on the MEGA side
--->Bob's your uncle, works like charm.
-Used an arduino Due 3.3v rail
---> Worked as well.
So in conclusion:
-It was the crappy Mega's 3.3v line that was problematic as you folks mentioned earlier. I have an original Mega that I bought off arduino's site.
-The Due's 3.3v line isn't that clean & seems to be border-line"I'll post the waveform when I've time".
-Overall the UNO's 3.3v waveform is the cleanest & most stable one.
This issue gave me all sorts of funny & unexpected output on the terminal while running the examples of the RF24 library.
I hope the 3.3v rail can be improved in future revisions on both the Mega & the Due as this can be a serious limiting factor for sensors & modules working off the 3.3v supply.