O2 feedback

[garfield at pilgrimhouse.com]

On Thu, 18 Jun 1998 13:13:29 -0400, Chris Conlon <synchris at ricochet.net>
wrote:

>Take one injector signal, buffer it, feed it into an RC lowpass with
>RC = ~ 0.050 sec. This is a lame attempt to get a voltage corresponding
>to injector PW.

I think yer on the right track here, Chris. I expect this was one of
Scot's "yeah butts", that you manually wouldn't be able to provide the
feedback loop that the ECM expects on a change of INJ PW. Making up some
kinda INJ PW to O2 rich-lean "simulator" is probably needed.

The other part of the "engine model" that seems to be problematic is
modeling the relationship between LOAD and RPM. That's not quite as fast
a time constant as the one above, but it's still gonna be tedious at
best to try to manually mimic the engine's response to loading. You CAN
simulate the LOAD of course via MAP, which is easy to manually control,
but co-ordinating this with your manually controlled "RPM generator" is
the 'slightly harder' part.

If you think about the MONSTER that Dr. Pelican has spawned, from a
"modelling" perspective, this is a perfect place for something like a
PIC to model the engine. You DON'T need something "real time fastoid"
for "responding to IGN timing ticks off the crank wheel" level of timing
response, but you SURE would like some "mental horsepower" to give you
the right combo of inputs to the ECM when you wanna pretend you just
dropped anchor on the engine, OR you just put pedal to metal.

What you'd really like is a model that took as inputs, TPS and LOAD. NO,
I don't mean from the standpoint of how the *ECM* uses TPS (put that
outta yer mind for this discussion, pulleeze), but you'd like an engine
modeller to take roughly the same kinda inputs that the REAL engine
does; and what's that, but TPS?! Ya know, dudes, like a JOY STICK for
the throttle! Even with a fixed LOAD and TPS to run/control engine RPM
indirectly, you'd have something. Kinda like an engine model that's
stuck in a particular gear, with a particular friction load applied to
the wheels. You could do worse for learning from such a model, how the
ECM is functioning.

In fact, an "engine modeller" that took the same VE curves that the S-D
EFI controller is using, and provided the equivalent "harness response"
back to an ECM, might be enough of a hint to throw our "commercial
hobby" guys into a tizzy, thinkin bout how they could "turn this into a
cash cow product". Heh. NOTTTTT.

In the end, I think this is a GREAT idea Bruce has for messing with
various ECMs, and I'll bet if you pursued it smartly, you just might end
up with a "mirror" image in the form of an "engine modeller" to go along
with this range of ECMs, that would simplify some of these manual input
coordinations and faster-response time issues. In fact, rightly done,
with the right ancillary external circuits to handle the REAL fast
stuff, you could whip out a PIC gizmo that could be used to check out
just about ANY ECU/ECM on the bench. Woohee, wadaconcept.

>Some Toyota ECUs even have a signal output that indicates what the
>computer is trying to do to the mixture, i.e. richen or weaken it.
>Now *that* would make life *too* easy, eh?

Yeah, it sure would. BUT it also indicates that maybe the Toyota
engineering dept. thinks highly enough of the mechs they train, that
supplying this in the ECU is a useful feature for them. I'd sure
appreciate it; wouldn't we all!!

But YOU could easily build such an indicator as part of your PIC engine
modeller. Cuz if the ECM is increasing INJ PW against an "I'm lean"
indication from the O2 sensor, you can bet your britches the ECM is
intending to "enrichen". And if the ECM is STILL increasing INJ PW
against an "I'm rich" indication from the O2 sensor, it's probably in
open loop and going for POWER.

This engine/bench modelling idea has lots of potential. Don't
short-change it. Any time you can mount a harnessed set of electrics to
an engine with HIGH confidence level, you're WAY ahead of the game. And
the bench is ALWAYS an easier environment than the pits, any day.

Gar