AVR Freaks

Hi Denny,

Typically you burn a bootloader onto the chip early on, say in early production or on the test board when testing. You can use the test board, ISP programmer, JTAG, etc. I use the STK500 or a JTAG.

Most people program a USART-based bootloader, which is fairly easy to implement. Then you can reprogram the app code any time you want on the final board, using simple RS-232 from a PC, without special hardware.

If the bootloader is robust, you may never need to change it, so reprogramming the main code of your chip just requires connecting to the serial port on your final board (or however you communicate). If the bootloader needs reburning after installation, you can do so if you have an SPI, JTAG, etc. header built in, though you'll need the corresponding device. Alternatively for in-house/personal work, you can unplug the chip and put it into a test board. Reburning the bootloader via plain serial is a bit risky (IMHO), and rarely done (sensibly), since you have to rely on the chip itself programming over itself, and have no recourse should something go wrong (the term "brick" comes to mind).

Overall, the bootloader is a self-contained on-board program that makes use of the SPM instruction (on the AVR) to program itself; the other methods use external hardware to provide access to every byte on the chip. Thus, bootloader code is usually limited to programming the app area (non-boot area), and can't change fuses nor clear lock bits (though lock bits can be set).

SPI is nice, but it does require a dedicated connector and device for programming. There are other kinds of hardware-based programmers out there too (both serial and parallel). A bootloader is always a nice addition in any case, and costs nothing but some time to code or port (there are many examples out there). BTW, you generally have to compile a small program for the PC end too, so it can talk to the AVR bootloader from the PC.

Hope my verbosity has helped a bit...