JeeNode frequencies and signal quality

A question arose on the HackerSpace Brisbane mailing list about the supported frequencies of the JeeNodes. I was under the impression, in the vacuum of contrary information, that the units I had would work at least to some extent in all the supported frequencies (433, 868 and 915 MHz), although it is difficult to make a single radio that can do that. I was under this impression because I had tested such with the RF12demo on the units when I first unpacked them.

It turns out that because I bought them at the actual JeeLabs Shop, in Europe, I got the 868 MHz versions, rather than the 915 MHz version sold at Modern Device. This might be problematic.

Anyway, I did a number of tests to determine the suitability of each frequency with the modules I have.

Testing more thoroughly with the RF12demo again, I determined the following radio capabilities:

• changing both to any of the 3 frequencies allows communication both ways (including ack)
• changing one to 868 and the other 433 allows data 868 -> 433 but not the other way (and no ack reply)
• sending from 915 cannot be received by either 868 or 433

This indicates all frequencies work, or at least some kind of shared delusion of using the same frequency.

Then I tested the relative signal quality of each frequency. I wrote a sketch that spams my car about every 50ms (it has an echo function for pinging and tests just like this) and reads the echo reply. I ran that on the USB JeeLink attached to my laptop used the highly scientific approach of walking around the house with it. I had it print out a dot for every send, and an R for every receive, so the ideal output looks like this:
.R.R.R.R.R.R.R.R.R.R.R.R.R.R.R.R.R.R

... and so on.

• 915 was ok, in the same room it got all the packets, and I could get through two or three walls and still receive a significant number of packets.
• 868 was awesome. I went all over the house and hardly dropped a packet. In fact I had to get the laptop halfway into the fridge before I started losing more packets than I sent! ...R...R.R.R.R....R..R... 
• 433 was disappointing. It lost packets within the room and lost significant amounts just outside the door!

Note that this was all with unoptimised antennas. Default length of about 90mm. According to this talk page, 868 should be at 82mm (pretty close), and 433 at 165mm, which might explain its poor performance. I don't think 1/8 wave works very well.