AVR Freaks

> Any idea's what might be limiting the range?

Apart from the obvious things (like a broken PCB / bad solder), here
are some ideas about what to check:

• are all channels poor, or just a single one?
• any interferers around?
• what's the RSSI or ED value of the received signal? Keep in mind that RSSI reflects the current actual energy on the channel (updated each
  microsecond), while ED reflects the median RSSI value from the last
  ED measurement (either explicit, or implied at the start of a received
  frame); two neighbouring boards should yield something around -40 dBm,
  assuming the Tx side is set to 0 dBm
  
• how's your chip antenna attached? there's probably going to be a balun then -- on my board, I could benefit from the symmetrical antenna design
  of the reference board, and the symmetric RF pins
  

p.s.: My tiny230 boards are working across a distance of about 10...20 m
within a building, i.e. some (light) walls between

> The best I could get was about RSSI = 12 when the
> boards were about 5mm apart...

Just to make sure: this is still measured while the frame is still "on air",
e.g. at the RX_START IRQ, is it?

A value of 12 would be (-91 + 3 * 12) -55 dBm, which is pretty weak for
such a tight setup. Something must be wrong within your RF path.

Any chance you could use a different board as a 3rd party reference?

Also, maybe you can unsolder the coupling Cs for a test, and use a
simple folded dipole antenna for a comparison? As a DC path is
permissible between RFN and RFP, you could solder it directly onto
the pads of the coupling Cs. With a lambda of 13 cm, the length of
the dipole should be about 6 cm, just simple 0.5 mm copper wire
folded as a dipole loop.

> There are but 3 things after the AT86RF230b...
> The 22pf c0g caps, the balun, the antenna ...

Not too much. While chip antennas aren't too brilliant in their gain
(which is usually rather a loss instead), they should for sure get
somewhat better results.

> As it stands, with the folded dipole wire antenna, I can get between
> 6 to 8 RSSI at 8 meters distance...

...which translates into -73 ... -67 dBm. That looks good to me.
Each 6 dB means a theoretically doubled link distance (without any
interference, of course), so this should max out at about 100 m.

> So, the important question is... is it allright to use the dipole?

If you can get it mechanically stable, yes, the dipole is certainly a
good solution. After all, if you get it into resonance, it'll yield
you a sure 2.15 dBi gain.

What your design is lacking though is any kind of filtering against
radiating harmonics. While there's a theortically good suppression of
the first harmonic in a symmetrical design, the radiation of the
second harmonic (3 * f) could be above the regulatory level.

I think the Atmel appnote uses a dual Pi lowpass filter (which I
also designed into my tiny230), but you might as well get away by
simply introducing a couple of small resistors (about 5 Ω each)
instead. In theory, this causes a higher loss than the full Pi
lowpass filter, but I didn't notice much difference.

> Also, if the dipole is allright for long term use, What do you
> suggest for making it a bit more durable...

> Can I glue it between two pieces of paper to make it rigid?

As long as the paper doesn't get wet (or the glue), that's probably
OK.

You can also use thicker wire. In theory, this will lead to slightly
smaller dimensions of the antenna, but as long as you don't check the
resonance anyway, that doesn't matter much. (If you could tune it for
resonance though, it will yield you optimal radiation, as well as act
as an additional supression of non-harmonic spurious radiation.
Unfortunately, it doesn't help for harmonics though as the antenna is
also resonant for them.)

Maybe you could use the PCB antenna from the Atmel appnote then?

> Also... I have some 0.7mm and 1mm magnet/trafo wire available...
> Would that be a better choice than the 0.5mm copper wire I scavenged
> from a LAN cable?

Probably, the only question is how you are going to fix it to the PCB
then without breaking it due to the mechanical forces applied.

> Just an stupid question, since I'm not an RF guy... Is the chip
> antenna located just above a ground plane? What say the antenna
> datasheet respect to ground planes? IIRC, the ones I read about
> (three years ago!) didn't like ground planes too close.

I agree.