Date: Sun, 28 Aug 2005 05:25:06 -0500 (CDT) Subject: analog electronics is rocket science X-UID: 84 Content-Type: IMAGE/JPEG; NAME="img1828.jpg" The picture gives some idea of what the operational vehicle will look like. The battery tray I machined earlier looks ugly to me. I prefer a more organic look (the postmodern gothic style of machine design in Japanese art appeals to me - not that I'm able to do that with this robot). That tray also adds significant weight. I'll probably use a combination of lexan, urethane foam, zip and velcro ties for mounting everything. The exception might be the lasers and webcams. For them, an aluminum rectangle is probably necessary to maintain alignment for stereo correlation. The cup with "Silver Vale" is in the photo to give an idea of scale. This robot is like a very large RC truck, approximately 1/5 scale. From the floor to the tip of the wireless router antenna is about 15.6 inches. I'll need a rollbar of some kind to protect everything should the robot roll. With real vehicles, rolling tends to be very destructive. This is why SUVs, even with the extra armor, can be deathtraps due to the high center of gravity which makes them more prone to rolling. The electronics box is done except for the microcontroller and two compact flash cards the single board computers use as the boot device. So there is software work to do. I learned a good lesson about analog and especially power electronics over the last few days. The discipline of wasting money purchasing the wrong electronics and having to work through mistakes in thought experiments is not without value. Basically, analog electronics is like rocket science. Semiconductors have complex nonlinear device physics and even the simplest component like a transistor functions as multiple ideal components when operating in a real circuit. The analogous situation in software is when a system is highly data driven and distributed with many tiers. It is almost impossible to study the system and guarantee that it works in all situations. The OSMC (open source motor control) design uses bidirectional transient voltage suppressors between the motor terminals and ground to protect the MOSFETs. With a MOSFET, excessive voltage from the drain to source causes breakdown which destroys it. When I saw the OSMC design, the bidirectional TVS's seemed absolutely necessary. This is not true. A MOSFET has an inherent high current capability zener diode from the source to drain. So a full H-bridge does not absolutely require TVS or diodes to protect the MOSFETs from flyback voltages. Moreover, the OSMC design specified TVS's with a breakdown voltage at just about the breakdown voltage of the MOSFETs. The clamped voltage far exceeds the MOSFET breakdown voltage. So what gives? It didn't make sense to me at first. What is happening is that a real world design has many effects at play. Depending on the motors used, batteries, electronics enclosure, etc. a design may require many design elements. So the TVS's help protect the MOSFETs. They are one element of many. My design is pretty much bare-bones and lacks these protective elements. But it is also not a combat robot. It will be gently operated. This leads into my research for purchasing an oscilloscope. We are largely a digital generation. We tend to think in terms of absolute capacity. So if a circuit operates at 50 MHz, then we believe that everything of interest happens at that frequency. We only need to sample at a sufficiently high rate according to Nyquist to see all. This is not true. Low frequency electronics may have very high frequency transients. Just remember that the derivative of a step function is delta function. So when circuits switch, extremely high frequencies are involved. The mechanical analogy is designing solutions to fluid flow problems. This is rocket science. Very subtle changes might make the difference between a robust stable design and one that works most of the time but on rare occasions fails mysteriously. Electronics can be like this. So can software. So I decided not to use any TVS's or protection diodes. I'm just going to try running the H-bridges as they are now (the simplest solution - do nothing). I know this can work. If it doesn't and fuses blow (trivia - when MOSFETs fail, they short closed, not blow open, so fuses are absolute necessities), then I'll figure out what to do then. Perhaps it is then time to go with an off-the-shelf motor control solution. As for an oscilloscope, my brother suggested eBay. That has been very educational for me. If all you need is a 1970s or 1980s era analog scope, you can pick one up relatively cheap. One problem is that the market is flooded with broken and partially working uncalibrated test equipment. It's too expensive to have them fixed. So if you can fix a broken one, then you can do very well. I don't know enough to do that. I'm no longer set on the Bitscope. I'm now leaning towards something like what I used in school, a Tektronix DSO. I'm avoiding anything rack-mount and prefer luggability. There are a few TDS 220 and TDS 224 models for sale. But now we're talking $500 to $1000. I can get a new TDS 1002 for less than $1000. So while you do save quite a bit of money, there don't seem to be tremendous bargains. People know the value of good equipment. Some people can't write software without an IDE and debugger. I'm more backwards and can use primitive tools ok. I don't use anything graphical. I use command line debuggers and plain vi. This makes me think about how I will use an oscilloscope. Maybe I should go cheap with a primitive tool. My heart says otherwise. Tomorrow I'm going to Fry's to play with the Tektronix TDS model they have on display.