Ignition details

[Paul E. Campbell]

Okay, as per recommendations on this mailing list (DIY-EFI), I've decided to
tackle DIY ignition first.

The vehicle in question is an 88 Ford Bronco 2 with stock everything. The
engine is 2.9L V-6 with an "EEC-IV" type ignition (Ford specific terminology
there). The ignition box itself literally just takes commands from the EEC
module on when to fire and does the crankshaft timing filtering for the
benefit of the EEC. The EEC is of course a EFI module and some extra space
for Ford to continue computerizing the rest of their vehicles with (such as
the Ford Taurus stuff).

I've learned this much so far:

1. Regardless of WHEN you decide to fire off a spark, there are several things
which happen:

a. An initial high voltage ionizes the path across the electrodes in the plug.
This is usually on the order of 10-20kV.

b. The initial spark may go out (bad). Or it may start but burn very slowly
and cause backfiring as the still combusting gasses go out the exhaust port
(bad). Or it may go into growth mode (good).

c. Regardless of whether you use MSD, CD, or regular old ID, you normally
light the spark, then continue pouring energy into it. If you don't dump
in enough energy, it will either go out or be slow burning. If you have
enough, you'll get a good, sustained flame front moving at the maxium rate.

d. After the threshold point is reached, adding more energy is pointless and
just wastes energy. SAE 850075 confirms this. Another paper I have (but don't
have with me) suggests that this threshold is at the point where you can
detect an initial pressure rise in the cylinder due to thermal expansion.

2. SAE860485 says that using the spark plug as an ionization sensor does not
work at all. However, one bibliographic reference I have ("Engine Performance
monitoring by means of the spark plug", Proceedings of the Inst. of Mech.
Engineers, Part D, Journal of Automobile Engineering, vol. 209, #D2,
pp. 143-146, 1995) probable says you can do it. The local library doesn't
get this journal..can someone at least check for me? Also, the recent Saab
advertising suggests that this does work, in addition to measuring other
good things.

3. SAE750348 is a different type of ignition system not discussed on this
list. In that system, a standard ignition coil is used but a charge pump
continues to feed energy into the coil so you have a continuous spark (and
automatic relighting with that particular circuit as a side benefit). It
works like an inductive system but it can sustain a spark indefinitely so
it has the advantages that a capacitive multiple spark discharge system
would have.

Now, this part is conjecture:

1. Suppose we equip our ignition system with the appropriate sensors to
measure the conductivity across the plug. While a spark is lit, this could
be done by measuring the voltage/current/energy of the ignition system. During
the rest of the cycle, you could use a high frequency oscillator to get a
signal from the spark plug gap without lighting a spark.

2. First, we should see some base conductivity while the cylinder is drawing
fresh air.

3. We should see an increase in the conductivity when the injector fires
because the chamber is filled with a moist mixture of fuel and air.

4. When we light the spark, we should see a massive rise in conductivity
because there is an ionized path inside the cylinder.

5. As combustion occurs, we should see a distinct CHANGE in the conductivity.
This should be either an increase or a decrease (not sure which) because the
resulting exhaust gasses have a different conductivity. Once we see that
change occur, this should indicate that we have a sustained flame front that
is now moving outward away from the vicinity of the spark plug and we can
cease feeding energy into the ionized gap.

6. I'm stuck on #5. If I were a graduate student working on this, I'd just
take my CFR engine standard and try it out. But I'm not.

7. Undoubtably (judging by Saab's claims), you can do much more than just
detect a good flame front. The conductvity of the exhaust gasses for instance
should tell you how complete your burning was and the fact that you didn't
see a change in conductivity (see #5) by the time the exhaust valve is
opening says you got a misfire on your hands. A similar trick should tell you
about detonation and the conductivity of the air/fuel mixture should tell you
about leanness.

If someone can fill in the details I'm missing above (PLEASE!), then I will
start work on the following project:

1. Build a home made spark ignition system as follows:

a. On initial lighting, the system charges up a coil by acting as a charge
pump into the coil.

b. Once the spark is lit, the coil is treated as a voltage multiplier and
tank for an oscillator/transformer type ignition system.

c. The spark generator (a. and b.) are shut off at an appropriate time. This
system is similar to that described in SAE750348 but I'm think about some
modifications to it. For instance, the aforementioned high frequency
oscillator without sparking for sensor purposes.

2. The system already has to read the engine crank signals if it is going to
be a drop in replacement for the original Ford box. The actual crank signal
detector will probably use a PLL so that I can access my own higher resolution
timing signal.

3. The initial system will just follow the EEC blindly except that I'll set
the spark timing to a specific crank angle as far as stopping the spark.

4. Following that, I will look at the data as far as conductivity goes and
see if I can't optimize the system further, perhaps even as far as ignoring
the EEC's commands as far as spark timing and perhaps lying to the EEC as
far as crank angle goes (although at that point, I might as well just start
builidng my own EFI too). As far as processing this stuff goes, I would
consider the following:

a. The amount of data to be processed if this was implemented "directly"
(feed the input to a flash A/D converter with or without some sort of extra
programmable signal conditioning) is enormous. So instead, I would consider
doing everything according to crankshaft angle timing. And then program a
series of D registers to latch the data at specific time intervals. This could
be done by having a series of programmable triggers set to points on the PLL's
output. This would allow me to use a relatively pathetic microcontroller for
the actual data analysis and control. Those same triggers will already be
necessary for the spark duration control so this is no big thing.

As far as whether I'm biting off more than I can chew, I'll list the work I've
done over the past 3 years:

1. Designed and built a sensor system for measuring the slurry slump inside
a pilot scale autogenous mill grinding -8"+1" hematite ore by measuring the
conductivity of the slurry inside the mill (mostly the only technical problems
here were analyzing the data, getting the data off a rotating body, and
building a measurement device that could stand having several hundred pounds
of force striking it regularly).

2. Designed and built a power measurement box that was installed at a mine
on their 1500kW motor. This one wasn't so bad as far as interfacing goes, just
that it had to take a measurement every 5 minutes and store the data for
at least 3 months without any maintenance and survive power failures which
are frequent around here. The data logger stored the data in CRC protected
EEPROM and spewed the contents on command via RS-232 from a laptop.

3. Designed, modified, etc., the software and the electronics for one of the
prototypes of an online viscometer that measures mineral slurry viscosity
over a wide range (about 10 s^-1 to 6000 s^-1) of shear rates. The viscometer
had to collect the data, analyze it, clean itself out, etc.

4. I've got the transfer case from another one of my vehicles apart on my
bench. It's a Ford NP208 for a full size Bronco and one of the thrust washers
disintegrated in it (parts are in the mail right now..damn it, I want to go
four wheeling still this year and we already had 6" of snow).