Using PC HW (& Ignition timing reference points)

[tom cloud]

>On Tue, 8 Oct 1996, tom cloud wrote:
>
>
>   ..[Mucho good stuff deleted]...
>> 
>> Now, understand, I really haven't thought about perzactly how to do all
>> these things.  Just wondering if anyone else has.  It is my experience
>> that being too much the purist (i.e. a hacker / fanatic trying to make
>> the confuser doo it all) puts you way up on the diminishing returns curve,
>> when the most expeditious (isn't that a new Ford product?) approach would
>> seem to be to use any and all tricks to get what we want.  [Now, if what
>> you want is to play wid da cornfuser, so be it.  As I've said before, I
>> wanna D-R-I-V-E !]
>
> Well put!  I have seen both extrems of design in IEEE hardware contests 
>(one design was *completely* analog -  to reprogram, change resistors), & 
>I think that there is a lot to be sair for a hybrid design.  It is 
>amazing what one can do with a R & a C & a CMOS gate!
>
>                Thor Johnson

well, to continue my tirade (feel like a preacher that's just gotten a few
strong "amens"):  clearly, if you've got lots of moola and major investment
in hardware, you'd have a serious development system, flow bench, dyno, etc.
but for us 'regular' folks ...

a simple R-C filter saves lots of code -- and one can still add
software filter routine.  A PC is a magnificent and inexpensive (I guess
I'm comparing to what an Intel MDS used to cost, not counting the PL-1
software, etc) development bed.  You can buy cheep A/D - D/A cards,
mock up the hardware, play wid it, build a few peripheral drivers
and test 'em (I'm talking the presettable counters for injector and
ignition timing, for example).  Seems to me you could easily have the
ignition timing be totally separate from the cpu exzept for the I/O
that loads the counters (they would be driven by a clock that would
either generate a fixed rate or a rate determined by rpm (a'la PLL)).
If the clock were a multiple of rpm, then the number loaded into the
counter would directly correspond to some advance angle.  If the period
were fixed, the advance would be a function of both the number and the
rpm.  The injectors could be done the same way, except the clock would
be fixed, as the number loaded should directly correspond to a
known amount of fuel per pulse.

I could ramble for hours about this.  What I'm searching for is the
'best' approach.  Doing it all cpu may be it??  I just know from
personal experience that noise reduction is a whole lot easier using
R-C or R-L than digital -- you're always gonna have aliasing probs,
esp with varying frequencies (e.g. rpm).  Clearly, your hardware
can cause serious limitations if it's designed without forethought.
In which case all cpu wudda been bedda.  I know that if this
were assigned to me as a design project, I'd probably do it all
cpu (except for the obvious transducers, etc) and then maybe
back up and design some peripheral circuits to simplify the design.
But, since it's a home project, I guess I think like the scrounge
that I am!


Tom Cloud <cloud at peaches.ph.utexas>