Subject: first car hack
Date: Wed,  4 Jun 2003 18:19:09 +0000 (GMT)

So, I got this not-quite-new Subaru.  Being as it's a 2000, it's
computerized everything, and the engine is more of a welter of
little sensors and wired widgets than the Pooper could ever hope
to be.  *Misfire* sensors in each cylinder, f'krissake.  And of
course it has the usual annoying problem -- some set of running
characteristics is gratuitously *changed* by the engine computer
just as you crack the throttle open, leading to that seemingly
insurmountable lurching behavior no matter how carefully one's foot
is feathering the gas and clutch.  Transient G-force, we hates it.

In the Pooper, the Throttle Position Sensor [TPS] is a simple
switch, positioned just so, which closed a contact just before
the throttle flap arrives at fully closed.  The fix for this was
to simply pull the connector to same off, and it ran this way
for years.  Later on I wired a manual switch for this into the
back of the brain-box itself, giving the option for a lower and
more even idle.  When the throttle is fully closed, in general
an engine computer will come out of "closed-loop" mode, richen
the mix a little, retard the timing, and possibly do some weirdness
with air bypass solenoids.  The Pooper also uses this to completely
cut off injector opening if the thing is coasting > 15 mph, which
supposedly saves gas.  [In practice it doesn't matter.]

In the Subie, the throttle sensor is an actual pot, returning a
linear voltage sweep from 0.6 volts fully closed to almost 4 volts
fully open.  Pulling the connector doesn't work, since it also has
a power feed and a ground, and the brain-box detects if that little
harness has become open somehow and on goes the "check engine" light.
Well, I checked, and not only is the engine still there, it still
does that goddamn lurching.  Having wheedled a copy of the brain-box
connector pinout from the service manager where I got the car [boy,
was that pullin' teeth -- he must have said "proprietary" about six
times but finally caved], I saw what was probably needed, and threw
some basic electronics at the problem this afternoon.

The "radical change" appears to happen after a rise of about +.05
volts on the TPS input, and undoes itself at about +.03.  So it
has hysteresis, too.  Damn if this behavior didn't *feel* like a
physical switch at first, since it's so sudden and noticeable.
But evidently it's all the brain-box's fault, based on some
arbitrary thresholds in software.

Since the pinout sheets claimed that the throttle-closed and throttle-
open voltages could vary and indeed weren't quite what I was seeing
anyways, I figured that the brain-box is simply monitoring the input
and deriving a fresh "closed" baseline any time it dips anywhere
below its lowest reading to date.  No potentiometer mounted under a
hood is going to remain totally stable over time, so as the resistance
changes due to dirt, heating cycles, humidity, etc the computer must
adapt in some way.  So the theory was to simply lower its perceived
baseline just enough, so that its "radical change" threshold would
then be *below* the reading taken at true throttle-closed.  The
easiest way to do this would be leak a little bit of it to ground --
in fact, about 0.1 volt's worth or so.

The input's internal resistance appears to be 1K or a little less.
At wide open throttle [WOT], connecting a 1K resistor between that
line and ground brings it down to about 2 volts.  After splicing
into the right wires and playing with some resistor values, I
arrived at 3.3K seeming to do about the right thing, and threw
together a quick-n-dirty mounted switch to bring this in or out.

It seems to work.  Usage is to switch in the extra resistor when
starting up, and then open it before actually driving.  Response
is quite a bit smoother just off the peg now, leaving only the usual
expected effects to deal with: 1> there is a *lot* of drivetrain
lash since this is an all-wheel-drive thing and power is going out
through *three* differentials, and 2> the engine is somewhat
overpowered and appears to have relatively little flywheel, so it
still falls on its face a bit when the throttle closes.  I'm still
getting used to what the right speed-matching throttle "blip" is,
but now having that bottom-end lurch taken care of should help.

I'm still working on obtaining a full manual for the thing.  Neither
Helm or Haynes appear to have anything, and Chilton's stuff sounds
like it doesn't do the electricals much at all.  In the meantime,
I've been learning about the various types of onboard diagnostic
protocols, thus my recent question about OBD-II, and will probably
wind up having some widget that lets my laptop read and reset the
silly trouble codes.

_H*