I mentioned the
Tassajera WeatherElement station in the WeatherElement area back in April, and again a couple of weeks ago re: an
Inversion Layers article.
It occurred to me today that I should post here in the Remote Weather Monitoring board too, with some details about the Ethernet back haul.
The link consists of 2 Nova Engineering NovaRoam 900 wireless routers, Yagi beam antennas, and a band pass filter in the antenna line at the mountain top end. The routers were a little hard to learn to configure because they are, well, routers. I was used to simple 802.11B bridges, clients, and access points before so delving into routers and router tables was a whole 'nuther realm.
Each unit has to have it router table configured to know where to send data based on IP addresses. Each unit has a different private IP address on the RF and wired ports, and those are also on different networks! So, my home network might be 192.168.0.xxx, the RF network 192.168.200.xxx, and the Ethernet jack on the mountain top unit could be a network of 10.1.0.xxx.
I used to use that one as a repeater to supply Internet from my house to a friend out in the country. Then his wired port and computers would have a different network again, like 10.1.2.xxx.
Oh, yeah, each unit has a serial port on it for local configuration. That can also have yet another IP address/network, and routing can be between serial ports, too.
It too a lot of local testing before I deployed to make sure that everything was routed properly. I labeled each router with all its config data to keep them all straight.
At the mountain top, the router and WeatherElement data hub are on the same static network, and the hub sends data to the server once a minute. There is a lot of interference up there, probably in-band. The NovaRoams are direct sequence spread spectrum, and any 900 MHz frequency hopping spread spectrum nearby stomps on them pretty badly. Using the filter didn't really help at all.
Continuous pings from home to the site show about 45% packet loss. The Data Hub buffers data when the path is interfered, so we still get back filled data and about 95-98% of theoretical data downloads.
If I were really concerned about it, I could try it on other channels. However, remotely changing the remote channel over the link is a cr@p-shoot. If the new channel is worse and I lose access, it's a long drive to fix it.

The path from our house to the weather station is about 9 miles. The path from there to where my friend used to live was 23 miles. Not bad for 1 watt at 384K or so speeds.