Tom,
We have the same problems at busy sites, though most of the links are on 40mhz channels and using 20mhz bandwith, so 10mhz on either side of "free" space or 20mhz between the signals. We have started experimenting with shields on nanobridges and nanodishes and see a speed increase from 10mbit to 30mbit. Havent looked into filters or cavities yet.
On the very big sites we are starting to use 24ghz equipment like the airfiber, as you can see on the "keyser" picture. However, these are still very expensive and cut in the budget. We also see that on longer links (10km+) we are having issues with rainfade on 24ghz, where the link drops entirely when a decent rain passes, and then everything gets rerouted over 5ghz links with more hops which is also not optimal.
There are also problems with ubnt equipment on hamsites, as they arent shielded properly by default and it interferes with HAM rf. We try to use mini-pci boards on the routerboards there, which are in a shielded copper box inside a watertight aluminium enclosure and then use 5ghz panels or grids.
I'm sorry that the camsite got passworded, it seems there was a lot of load and the sysops decided to put a password on it. I was out of the country until today without access the any datanetwork so I didn't know until today.
73s Robbie ON4SAX
We have started deploying a similar network here in western Washington (see hamwan.org). One of the problems we have encountered is crosstalk/IMD between our 5 GHz modems. We typically have 5-6 modems on a tower (three sectors and 2-3 PtP dishes). We're using 10 MHz guard band between channels. With all the modems running, we see significant increase in noise floor across the band. Throughput is degraded. By turning off all but one modem, we get the throughput back.
We've employed spatial diversity where possible, but never have enough space to completely escape the noise. We're experimenting on other fronts--testing with cavities, isolators, and antenna shielding (for improved front-to-back ratio).
I notice you have many 5 GHz antennas in very close proximity on some towers and I'm curious what that does to your noise floor. How are you mitigating it? It looks like you mix GPS-sync gear with non-sync gear, so that cannot be the whole solution.
Tom KD7LXL