russian,
Again, let me emphasize I know little about cars. Do you know if the processors generally used in automotive applications are JTAG compliant? If so, perhaps it would be possible to "take control" w/o completely replacing everything requisite to make the systems work. Assuming you do have to completely replace the on-board computer, would this "hack" be make/model specific, or would it have general application? I know plenty of car buffs who go to extraordinary efforts to boost performance for racing, but as far as I know everything they do is highly make/model specific and always involves large financial/time investments. I had one acquaintance years ago who worked extensively with modifying big block Chevy controllers for the marine industry. He claimed to be getting 1500hp+ from them primarily with firmware hacks (but I think he also did some marine "upgrades" to the cam/crank/pistons). I do know that he claimed to use the existing hardware; he just modified the "data tables". I cannot verify that any of his claims were true, but he was in high-demand and charged a lot for his services.
Anyway, while I have no real interest in modifying a particular engine/controller for a narrow purpose, I might be interested in participating in a more general hack, but a lot would depend on it NOT being make/model specific.
Your post is the second reference in this forum in the last few days to an ARM core processor. They must have some really good advertising, lol. I have not played with any of the recent ARM processors, but I did help on a project a decode or so ago involving one, and I am here to tell you that project got ugly in a hurry, and was way over my head. Perhaps there have been some developments that I am unaware of (happens all the time), but that is one beast of a processor, it could easily control an engine while decoding video and spending its idle time in sleep mode. Assuming you want to "control" a V12 doing 18k RPM, that's 1 revolution every 3.3mS. If there were 1000 events per revolution, that would be one event very 3.3uS. Assuming the processor needed 100 instructions per event you would only need a processor speed of 33nS per instruction. @168Mhz, your ARM is executing instructions @ ~6nS each, roughly 5 times faster than it would need to even in this extreme case. Since typical automotive engines are not V12s, rarely exceed 5kRPM and likely require less than 100 events per revolution the ARM will spend the vast majority of it's time in sleep mode, not that this is a bad thing, just saying that is a lot of computing power.
**I am not criticizing your choice of platform, just saying it is a real beast of a processor, and the last time I tangled with an ARM core I walked away broke sticky and confused, perhaps the high-level programming tools designed for it will make it a cake walk! I will look into them, the development board is certainly cheap enough**
Fish
