Hi all,
I'm looking for a few people to read and correct horrific mistakes, side-splitting howlers, and any other errata in my upcoming book, so I am posting here in hopes of locating a few candidates. If you are interested, please email me some information on your background and I'll work up a short list. I can't afford to provide cash compensation, but I can guarantee you a free copy of the book, when published (physical copies will probably be printed in Q4, 2004), as well as a mention in the acknowledgements. I can supply your review drafts either hardcopy or electronically in PDF format.
The book is kind of broad-ranging. It collects together most of the embedded-type information I needed to bring together in my submarine project in one handy location. There is a distinct leaning of information towards this particular project (mainly, where I discuss why I did certain things in a certain way), but this is not a book about "how to build a submarine"; it's applicable to many other projects. I've used the techniques here in industrial testing applications as well. The major thrust of the book is the premise that both of the following are difficult tasks:
- Creating a single-CPU project based around a 32-bit micro running embedded Linux. Embedded Linux is harder to use than "regular desktop" Linux. Realtime characteristics are not as good as a "to the metal" application. It's also often difficult to port drivers for off-the-shelf consumer peripherals, because these drivers are only tested in "desktop" environments.
- Creating a fantastically fine-tuned project around a small microcontroller and some proprietary or entirely embedded OS. Realtime characteristics are much better, but it's a REAL bear to add complicated stuff like WLAN, cameras, IrDA, etc to tiny systems like this. In some cases, it's also difficult to obtain small quantities of modules suitable for direct interface with embedded systems.
The path between this Scylla and Charybdis, therefore, is to use an x86-based SBC as a non-realtime master controller, running a stripped Linux distribution (which gives us a huge driver library for the non-realtime things such as LANs, mice, keyboards and cameras). This master controller interfaces to several tiny subcontrollers which have specific realtime functions. That way you don't have to wrestle with odd realtime OS extensions, and you still have access to all the cheap off-the-shelf peripherals your heart could desire. My book hopefully fills in the blanks in peoples' knowledge of how to construct such hybrid systems.
This is, if nothing else, an excellent way to build a breadboard prototype. Once you have everything jiggling like it should, you can make a realistic estimate of CPU horsepower, RAM, and so forth required for the device - and you can then go ahead and make a design optimized for speed, power consumption, cost, or whatever your criteria might be.
Topics covered in the book include:
- Creating a small (essentially non-realtime) Linux distribution for use on an x86-based master controller board. Examples on adding WLAN support and TCP/IP connectivity in general, and so forth. Specifically, the book talks about using an Advantech PCM-5820 (Geode) biscuit SBC, but information is given about several direct substitutes from other vendors, also. All of the code, with the exception of the board-specific kernel configuration, can be run directly on a regular Linux PC if desired.
- Creating bootable CompactFlash and CD-ROM media. Discussion of the whys and wherefores of flash, CD, hard disk systems.
- Creating realtime peripherals using AVR microcontrollers. Sample projects given in the book include "intelligent" stepper controller, DC motor controller with thermal feedback, light intensity sensors, temperature sensors, ultrasonic rangefinder, 3D accelerometer, and a few other projects. All of these projects interface over an SPI-type bus to the SBC's parallel port. Schematics, sourcecode and PCB layouts are also included.
- Interfacing to the above using the SBC's parallel port (code for Linux). The book includes a schematic and PCB layout for an interface that connects up to eight of the projects above simultaneously.
- Developing industrial-type GUIs on Linux. Some sample applications including simple introduction to machine vision concepts as well as info on developing dumb text-mode interfaces, slightly more complex graphical interfaces, and fullblown X interfaces.
The prerelease review version should be ready by the end of February. I have a deadline of the end of May to submit it, so I would like to get reviews in by the end of March to allow time for rewrites and ironing out any arguments.
(By the way: For people who have been following the sub project, (a) I know the web site is way out of date, I'm too busy writing the book to work on the HTML right now, and (b) The sub project doesn't work exactly as described in the book - the book is more general-purpose).
PS: I'm also looking around for someone who would be interested in marketing the PCBs as kit projects, so I can tell readers where they can buy ready-made boards.