REPOSTED: Electronic spark timing ! PLEASE HELP !

[Jose Carlos Rublescki]

Sorry, I noticed my message was posted in HTML format. I decided to repost
as non-formatted text, as some of you may have missed it.
    Jose Rublescki - rublesck at ez-poa.com.br
----------------------------------------------------------------------------
---------------
It has recently been discussed in this list something about angular velocity
variations in the engine that would occur because of bumby roads. I have a
problem that is very related to this and I need help.

I have sucessfully designed an injection/igniton system based on 68HC11. The
injection part is working great. It is a sequential multipoint system and it
even has two injector for each of the (four) cylinders to handle turbo boost
situations. I use ethanol alcohol as fuel so I can use LOTS of boost and
need LOTS of fuel.

The problem comes with the ignition part. In my first system, the MCU
received  from the engine one pulse each 90-degrees of engine rotation (on
TDC's). For each 90 degree period the MCU counted the number of clock pulses
(2 Mhz clock), let's say n, and then it did simple aritmethics as follows:

90 DEGREES = n clock pulses
x  DEGREES = y pulses.

So if I wanted a spark to appear x DEGREES after the 90-degree mark, I
should just have to program the MCU to generate the spark after y clock
pulses.
(which means that I measured a 90-degree period, do math and applied the
result to the next period, assuming both periods would both be very similar
to each other).

I should work fine in theory, but the problem I found is that there was
sometimes too much variation from one 90-degree period to the next. I
thought I could devise the pattern of the variations. I tried to calculate
angular acelerations so that my system would predict what would be the
length of the next 90-degree period (and many other aproaches). All of this
failed. I came to the conclusion that the variations were random and that I
could not improve my system further. The amounts of variation I got on the
spark generation due to the variation of the measurements were big enough to
blow head gaskets of my turbo engine. (I didn't however, melted pistons as I
did when I was working on the injection and the mixture got inadvertedly
lean WITH lots of boost)

My BIG question is that there are here lots of cars made by Volkswagen where
the ECU only reads one pulse each 90-degree of engine rotation. Do they just
live with the fact that they will have spark timing variations or do they
have some misterious way (at least to me) make this calculation?

After I gave up using the system above cited, I decided to use a 40-teeth
wheel attached to the engine like FIAT and GM do here (only they have
different teeth count). I now get much better results, but they're still not
as great as I would expect. I now get pulses every 9-degree of engine
rotation. The process remains the same.

9 DEGREES = n clock pulses.
x DEGREES = y clock pulses. (figure out what y is and delay that value).

Thanks in advance,
Jose Rublescki
Dim-efi (Dio-efi) -- Do it myself efi (Do it ourselves efi)
 ---------------------------------------------------------------------------
------------