[Bulk] Re: [Diy_efi] Modular approach to EFI controllers

[ian]

On Mon, 2007-10-01 at 10:49 +0800, Mike wrote:
> 
> Is the FPGA that necessary, I mean, with a fast enough processor
> and the appropriate software it should be a chince, even if you had
> a second CPU doing nothing but  the timing and ion sensing etc,
> you are not going to have the spark events coincident so surely
> with a CPU even doing 4Mhz there is enough time, these beasts
> are cheap have good a/d and run 20Mhz etc... 

Oh yes. very necessary.

ok, so we overspecc'd it a bit so it can handle (in theory) a W16 engine
configuration with twin spark and twin injector multiport injection, but
still.

the math worked out upwards of 90,000 events per second at redline.

you *could* handle all that on a 20Mz CPU perhaps, but then if theres
one error in that and you lean out at 15,000 RPM or so its bye-bye
engine. much better to have a nice simple state machine do the job
predictably and without bothering the CPU.

if you actually want to _do_ anything with the ion sense data you're
looking at doing thousands of FFTs per second, which starts to really
rack up the amount of CPU grunt you need - and with heavy processing
would come high latency, or the need to design _very_ tight real time
processing.

Much easier to have the FPGA do the critical stuff and simply pass data
to the ARM for analysis and receive feedback to modify its controll
tables.