On Tuesday we started assembling the mesh nodes for the pilot network.
About 22 people attended the basic training course on Monday, and 18 today for the more heavy duty work where we actually started building the Dili Mesh Potatoes and deploying our first mesh nodes. Building the nodes involved taking MPs out of their indoor cases and mounting them in weatherproof boxes. Also 15m phone and DC power cables need to be fabricated.
We were impressed by the enthusiasm of the Timorese guys. I didn’t have to do much at all. They have really taken ownership. Lots of enthusiastic sawing, drilling, and soldering going on in a room that was close to 40C and 100% humidity. No one seemed to care much. Hacking, Timorese style…..
So I could just sit there and work on some little tools to help with the deployment. One problem that worried me was the 40 networks we detected when we ran “iwlist atho scan” on the Supernode. However it’s hard to understand pages and pages of iwlist output so I spent a pleasant morning hacking some scripts to visualise the Wifi environment.
Here is a sample output, each cross represents a network:
I am also working on some scripts that report signal strength of adjacent nodes. This can help explain low Batman scores and poor packet loss results with ping. For example if the signal strength is OK, but you have poor packet loss, interference from other networks might be the problem.
At 4pm we headed off to install the first node (MP-10) at the Blind Society, about 300m away from Fongtil HQ. I watched in amazement as Ase (Ah-say), one of the Fongtil guys, kicked off his shoes and scrambled up a large tree to manipulate a 15m mast into position. He can climb a tree faster than I can fall out of one. All in steady rain. Thirty minutes later the Potato was in position and we had dial tone on our first Mesh Potato.
You can get an idea of the local tree height in these photos. You can just see the MP01 poking through the top:
I sat inside with another battery powered MP connected to my laptop via Ethernet. That way we didn’t need an Ethernet cable to the MP up the pole, my indoor node just meshed with it. Pretty cool feature of mesh networking.
However there were problems with lots of packet loss on pings between MP-10 and the Supernode. Signal strength looks OK at -70dBm but MP-10 could detect 25 other networks. So I am guessing interference. However it still worked well enough such that I could scp files across from the Supernode. We couldn’t make any phones calls as the Supernode (mounted on a 20m mast at Fongtil) actually has problems seeing MPs inside the Fongtil building. This may also be a sign of interference, or just the nature of the directional Nanostation 2 antenna. More research required.
The mechanical side of installing a node for a real world site can be a lot of work. This is something I hadn’t realised until now. For a link of several 100m you need a good clear line of site, which means you need to get the node up high. If you have 15m trees or a two story building between you and the next node, that means a 15m mast at each site. Plenty of 25m trees and two story building around here.
I found a similar problem with my test mesh in suburban Adelaide. Flat terrain, plenty of trees and two story buildings. So in Kilkenny I also require tall masts for decent links.
After one or two of the 12V DC power cables failed I discovered some of the soldering on the power cables needed a little touching up. The Fongtil guys had used some solder they had laying around. This seemed to have a really high melting point (maybe it was lead free?) and was making joints that didn’t really stick.
I had brought some solder from Australia which flowed much more freely. I also showed the guys the importance of really getting the joint on the DC connectors hot before attempting to solder.
Here are the before and after results (sorry about the soft focus effect on the after photo, but you get the basic idea):