Date: Sun, 27 Feb 2005 03:41:28 -0600 (CST) Subject: latest mockup X-UID: 27 Content-Type: IMAGE/JPEG; NAME="img1441.jpg" Content-Type: IMAGE/JPEG; NAME="img1442.jpg" Content-Type: IMAGE/JPEG; NAME="img1443.jpg" I have two inoperative torchiere lamps, one of which caught on fire last year. They have been disassembled to save space. A tube from one of these lamps makes a good frame component. Some of the major components are in the photo of the mockup. The two webcams will be mounted and possibly rebuilt (disassembled and IR filters removed) into an aluminum bar for stereo vision. The USB card is necessary for the webcams to interface to the Geode mainboard (about 5.5" square). The large black brick with yellow tabs is one of the drill battery packs. It supplies 3.5 Ah at 18 volts. I will be using two, one for each motor. The red cordless screwdriver is the cheapest 3.6 volt one I could find. The low voltage is advantageous as it means I can use six AA batteries and two transistors without need of an H-bridge for forward and reverse. It has another advantage in that the case is square in cross section, not ergonomically rounded (which looks cool but is difficult to mount) like the more expensive tools. The screwdriver will be converted into a linear actuator along with the 10 Kohm (it's really 12 Kohm) slide potentiometer. This will be used for steering. At the top of the photo is an OEM GPS receiver. It operates off a 5 volt serial interface. In the center is the 900 MHz radio module. It outputs 100 mW (the latest version can radiate 1 watt) and should have far superior range compared with WiFi. The fiberglass spars will have to be shortened by about an inch. They're just a tad bit too long. I figure that drive and system components will be bolted to the main tube. I'm thinking that if the electronics where mounted on a board suspended on cords, then it would be effectively shock and vibration isolated. Carmack wrote that elastomeric (rubber) isolators really don't work very well. A guy at XCor (made engines for Rutan's space ship) confirmed this. He said that wire rope isolators do work. A few years ago, I was really into making kites for a while. So this is more like kite construction with rigid spars held in compression by cords. Anyway, the significance is that GPS receivers lose accuracy when shaken. I figure that the lock is lost on some of the satellites. But in general, the electronics won't be happy if shaken. So keeping it stable is good. I feel kind of bad because I seem to be unable to design anything. Everything evolves as it is being built. I've noticed this phenomenon at my jobs. The best work usually happens when there is no design, just hacking in a time sensitive crisis. The Borg Appliance was made the same way. There was never any fixed design for it. Mistakes became "features" or "industrial design elements". Pat's observation that I will be unhappy with this robot is true. This design isn't "tight". I am happy with the Borg Appliance because it is a very tight design. Everything fits together without much clearance. It feels dense and organic. This design is the opposite. It feels bolted together and loose. What I hope is that as everything is assembled and it grows, it will become dense. So I want to get to the point that it becomes difficult to fit everything into the system. Basically, I don't like having underconstrained problems as I become nervous. Highly constrained problems are more like math problems in school, predictable. The fine tweezers are necessary when working with steel. Together with a 10x loupe and good light, I now stop and try to remove metal fragments from my hands when I first feel them. This mostly works. The mistake is to keep working. That will drive them deeper into flesh. On Friday, I spent about a half hour in the restroom with an Xacto knife and loupe removing the larger fragments. Altex is an interesting place. The workers there are the purest Silicon Valley type supernerds I've seen anywhere. Fry's doesn't have this culture. Altex doesn't have everything. But they do have stuff you are unlikely to find anywhere else except for mail order. I noticed at $45 multimeter with a serial interface. That's pretty neat. There were also a half dozen $70 industrial power supplies. I was tempted. These have four channels on terminal blocks with tunable output from 5 to 24 volts and up to 20 amps each channel. I just wish they would go down to 3 volts. But I may end up getting one of these supplies. Altex is not the place to go for computer shopping. The selection of cases is limited. What they have is too expensive. They just don't have the volume of Fry's. They do have a lot of DC power supplies at reasonable prices. They also have miscellaneous components either obtained as surplus or pulled from equipment. But they are not a one stop shop for electronics. I didn't see any microcontrollers. I also don't remember any clock crystals. I remember a pancake stepper motor with very high resolution. They had a lot of stuff for technicians and operations types who have to run network cabling and build rack mounted assemblies. Also, the Altex employees are probably more knowledgeable than elsewhere. If Fry's is like Home Depot, then Altex is like Elliott's Hardware. There was one younger guy there who looked like one of the nerdy guys in the movie "Demon Seed" (plot summary - scientist builds strong AI that is wired into home, it takes over, traps his wife, somehow bioengineers itself into an egg and zygote hosted by the wife, and then is born as a baby with super intelligence). I don't remember if the Altex guy wore horn rimmed glasses. But even if here was not, you'd still think he was. I didn't mention it before. But I tried wrapping some 500 pound test nylon cord through and around the axles and wheels. The wheels have integral ball bearings. This is great for the front wheels. But the rear axles are live under power of the motors. The wheels should be rigidly attached to the axles. I didn't want to mess with the bearings in the wheels so had a quick and dirty scheme to wrap nylon cord around the wheel spokes and axles. Surprisingly, this works pretty well. I had thought it wouldn't work and was planning on making a mounting plate with studs as for an automobile wheel. What this shows is that crazy ideas sometimes do work. Looking in hindsight at this project (which is still at least one year away from any kind of functional demonstration), the choice to build everything only makes sense if the next robot is going to be a modified full-size vehicle. A lot of the experiences gained with this project are about improvising solutions from COTS (common off the shelf) components and systems integration. If I had taken a large scale RC toy, all of the fabrication, drive system, and control electronics problems would be taken care of. There would be very little analogue electronics. Most everything would be software. As this project is, there's more basic problems. But I think that these problems will arise when modifying a vehicle built for people to use. Does anyone know how much warehouse space costs to lease? It's just a crazy thought. I know how much a home costs. But I have no idea how much commercial space costs.