I have in the past successfully used the AT86RF230 radio chip. In that case I copied the layout quite accurately from the Atmel application note AVR2005 (http://www.atmel.com/Images/doc8092.pdf). This resulted in good enough performance of about 50+ meters line of sight range with a small chip antenna.
Now I am prototyping with Atmega128RFA1 and closest to a reference design I have found is the application note AVR2044 (http://www.atmel.com/Images/doc8339.pdf). I would like to design a system with similar level of performance (about 50 meters range is enough).
However I am a bit confused about which way to go. In the AVR2005 a star-like ground separation/splitting scheme is used and it's importance is clearly explained. Also, I have seen it work in my own designs. But in the AVR2044 the ground plane is not split at all, it's completely solid over the whole PCB. This application note also talks about separating the digital and analog ground currents, but I see no evidence of it happening in the layout. Digital and analog grounds get happily mixed everywhere. Even one digital ground pin (55) is directly connected to the paddle under the AVR which is internally connected to analog ground. Also, the crystal layout looks quite vulnerable in contrast to the AVR2005.
Do anyone of you have any comments or experience? What kind of range can one expect between two RCB128RFA1 boards? Do you know of any other reference designs for Atmega128RFA1 that we could discuss here?