Good idea. I've had a look, and I think it's the LED's I'm driving. It's actually a PCB with 4x1watt LED's on; RGBY. It uses buck drivers, so lots of inductors etc. The LED pcb has RGBY GND and V pinouts. So I draw 9v off the LED board, and use a 3.3v lin reg to get the power for the atmega, nRF24L01+ and flash chip. I can then use the atmega to PWM the RGBY pins. It all works great if I reset the board manually!
It's quite frustrating that I'm so close to having it work. I wish I had an oscilloscope so I could better analyse what's going on.
I might try putting some more caps on the power supply... any other ideas?
Cheers!