Date: Sun, 2 Oct 2005 05:45:37 -0500 (CDT) Subject: electronics box almost done again X-UID: 94 Content-Type: IMAGE/JPEG; name="img1850.jpg" Content-Type: IMAGE/JPEG; name="img1851.jpg" Content-Type: IMAGE/JPEG; name="img1852.jpg" The wirewrapped discrete components were soldered to help keep the wires from coming loose. It's not a perfect job but the design itself is inherently flawed in this sense. A perfect job would use a printed circuit board (trivia: the printed circuit board is another American invention). I put everything back together inside the electronics box and powered the digital side from the terminal block. It works which is reassuring. I'm not 100% sure of the wiring as I did catch one mistake a few days ago. That night, there was a bad lightning storm in the Dallas metro area. It took out power where I live for about three hours. Fortunately - the primary computer is now a laptop with two fresh battery packs for at least 6 hours of runtime; I have lots of batteries and battery powered lighting sources; I live next to a hospital so relocated to the emergency room to keep working (hospitals almost always have power). No one hassled me there. I got all of the wiring written down in my notebook. Then when I came back and checked the wiring with a flashlight, I noticed one microcontroller pin was wrong. The surface mount keystone RJ45 jack (white box by the intake fan) is for expansion. If I add a "weapons module" of some kind with pyrotechnics like rockets, flares, or smoke generators, then it should be completely removable from the system. That's for legal and safety reasons. Adding the keystone only took an hour. It was done by hand too, quickly and without any power tools. Contrast that with the two internally mounted keystones that took many hours to get right. It's an object lesson on how to design for low cost. Labor involved in assembly dominates technology manufacturing and assembly. The electronics box is stuffed really tight. It feels like some metal bodied digital cameras. The internals bulge out against the metal shell so it feels like a brick. Powering the fan, it does push air through the box. But there must be a lot of resistance due to all of the wiring in the way. My original thought for the 900 MHz Maxtream radio module was to connect a getty to it. So it would be a login console. The problem with this is that there's no way to have secure communications out of the box. I haven't seen a "secure serial port" device abstraction. So the comms architecture will use the tap and tunneling network driver over the 900 MHz radio. It will appear as a network interface. Then using the bonding driver (CISCO terminology for this is "etherchannel", even though I know it is marketing speak, it does sound cool) to connect the WiFi radio link and the 900 MHz link in an active-backup mode. When the robot is within range of the base station, comms will go over WiFi. As it moves out of range, comms shifts to the lower speed but much longer range (20 miles with high gain antennas and line of sight) Maxstream radios. This way, the software doesn't have to be aware of the comms hardware available. In practice, I expect that a lot of tuning will be required. What happens if the robot is obstructed for short periods by buildings? There may be lots of networking havoc as the bonded network interface fails over and then back. Anyway, the primary command interface will be through ssh. So all traffic will be encrypted. For now, I think that a command line interface should be the focus with graphical tools for visualization of position and robot state only. I read that Mayor Newsom wants to have free ubiquitous WiFi in San Francisco in a matter of months, not years. Google has offered to do this for free with a speed of 300 Kbps. I've thought for a long time now that hardware and infrastructure drives technological development. Ubiquitous wireless digital comms could allow for interesting field robotics products. It would allow for decoupling the robot body from computation and sensing. A lot of stuff becomes practical when technology is woven into the civic infrastructure. E.g. wireless voice communication isn't practical without a huge network of cellular towers, washing machines aren't practical without a huge network of electrical power generation and distribution...robotics isn't practical without a huge network of wireless communications?