[Diy_efi] PC-laptop based efi

[Matthew Richard Friesen]

> The way I would suggest going about this is the divide and conquer
> method.  Build a few independent modules that do all the timing-critical
> functions such as spark timing and injection events (after all, you'd
> want this beast to be full sequential I'd imagine.)  These sub-systems
> wouldn't need to know anything about enrichment stragies or maps of any
> kind, let those higher level decisions happen at the CPU level where
> they are not as time critial.  

This is exactly the idea that i am going for. Have "stupid" modules that each 
do their little part, and have the pc telling them when they need to change. I 
must admit though that i am beginning think maybe it would be easier to start 
with the megasquirt, (probably the AVR, i'm not too fond of assembly, and i do 
have some experience in c ) and "adjust" it to my needs(wants). (Ahh.. the 
burdens of facing hard reality). Can the megasquirt handle sequetial 
injection? (if not, i might as well just hack a GM TBI, for all the trouble) 
If not how difficult would it likely be to add? 

Thanks for everyone's input, i hope you are all enjoying discussing this as 
much as i am.

-Matt







Quoting Brian Dessent <brian at dessent.net>:

> peterw wrote:
> > 
> > > Hi, i'm looking into the possibility of designing a efi control system
> > that
> > > would run on a laptop. I think that it would be of the multi-port
> variety.
> 
> > Speed and timing may bring you undone. In order to read sensors, lookup
> > tables, do calculations you will need a fast machine running a true real
> > time operating system, especially if you are doing lookups and calcs every
> > rev or every second rev. At 10,000RPM a rev takes about 6 milliseconds.
> This
> > may fine on a fast machine untill the OS wants to flush the disk cache or
> do
> > some other such system thing. Note Windows and Linux are not real time
> 
> I agree, you will pull your hair out trying to do this purely with a
> PC-like design.  You will make your life much harder by having all the
> time critical stuff percolate all the way up to the CPU.  At the very
> least you would want to run some custom bastardized FreeDOS or something
> running one big honkin' polling loop.  FORGET running Windows, it's out
> of the question.  RTLinux or QNX may be possible candidates thoough.
> 
> The way I would suggest going about this is the divide and conquer
> method.  Build a few independent modules that do all the timing-critical
> functions such as spark timing and injection events (after all, you'd
> want this beast to be full sequential I'd imagine.)  These sub-systems
> wouldn't need to know anything about enrichment stragies or maps of any
> kind, let those higher level decisions happen at the CPU level where
> they are not as time critial.  It sounds like the Ford EDIS-4/6/8 is a
> perfect example of this kind of implementation, and probably the reason
> why Bowling and Grippo are into it for Megajolt.
> 
> If you do this right you still retain all the advantages of a PC --
> mature, robust tools for development, logging, visualizing -- while
> still having the advange of most of the time sensitive tasks offloaded
> onto dedicated hardware.   I'd be a little worried about relying on a
> laptop or PC-104 chassis as the main ECU, I mean think of all the "my
> disk crashed and I was trying to run SCANDISK, officer!" jokes you could
> make.  All in all I'd say attack it more from the standpoint of "how
> could I build or modify an existing ECU so that it's in constant
> communication with the PC, with all internal variables exposed and
> accessable", rather than trying to bring the PC down to the level of the
> microcontroller.
> 
> Brian
> (of course I've never actually done any of this, but I've been down the
> same "wouldn't it be neat" trail of thought a million times...)
> 
> _______________________________________________
> Diy_efi mailing list
> Diy_efi at diy-efi.org
> http://www.diy-efi.org/mailman/listinfo/diy_efi
> 

 

_______________________________________________
Diy_efi mailing list
Diy_efi at diy-efi.org
http://www.diy-efi.org/mailman/listinfo/diy_efi