On May 22, 2007, at 1:45 PM, Steve Franks wrote:
> We placed a mega128 and a Xilnx CPLD in a chain together. We
> programmed the xilinx with their tools several hundred times. We
> could take a fresh board with a mega128, program it 4-5 times with a
> jtagice, and everything would then start reading "0x00" - flash,
> eeprom, fuses, everything. The project quickly ran over schedule with
> us reworking pcb's after every 5 program cycles and the project was
> canceled. As a post-mortem, we actually were invited to send boards
> back to Norway for analysis, and after several months, the answer was,
> "you are correct, they are wiped internally, that is never supposed to
> happen, we're very sorry". Anyone who doubts my sincerity is (1)
> welcome to request a copy of my emails with atmel privately, and (2)
> should know that I still use Avr devices pretty much exclusively in my
> uC development, so am not attempting to discredit atmel here, just to
> say that they did not successfully do this particular thing (jtag
> chains). I make no claim that this is or is not an issue on any part
> other than the mega128. I have not gotten any indication to believe
> that it has been fixed in later devices, and I have not tried it
> since.
Wait, so you're saying that, using a JTAGICE Mk II, an ATmega128 was
wiped clean? As in, all the logic as well? I wouldn't have thought
that would be possible (i.e., it was etched in silicon). Or do you
mean just the flash portions?
--
Rick