ecu voltage problem

Tom Cloud cloud at peaches.ph.utexas.edu
Thu May 1 14:37:04 GMT 1997


 - 1st. as Fred mentioned ... beware of the glitzy places with
those weird guys stuck to the window (aka RatShack).  Please,
please don't buy stuff there for anything you have to depend
on !!!!!

there's lots smarter people on this list, but as you describe
this, interesting things come out -- esp. the fact that the
"fixed" pulse width you're getting 'just happens' to be the
same width as that from the crank sensor.  Does the injector
pulse occur at the same time too (then, you've built an
expensive crank pulse buffer  ;-)

being an electronics teacher, my first impulse is to start
with simple admonitions about wiring & shielding.  But I
hes-ti-tate so as not to offend.  Now that I've hes-ti-tated
a little, I'll continue on.  First off, I assume you know all
this, but just in case .....

(note that when you have a problem like you describe, you
get what I call "superstitious", and when everything works
great, one gets to where they become a little arrogant and
above the _rules_.  What I'll write here won't be exhaustive,
'cause I can't remember it all -- but others will fill in for
me.  Also, we don't normally follow all the rules ourselves --
just when things don't work   ;-)

 - only pick up power and ground from _one_ point.  Eschew
multiple grounds !!!  Example: if your crank sensor is connected
with a single wire and a shield connected to ground, and you
use both those wires (the shield is a wire) at your conditioning
circuit, and your conditioning circuit (probably your ecu??)
is also connected to 'ground', but at a different place,
you've created a _ground loop_.  The technically correct
way to hook that up is to connect the shield to ground
at only one end or the other -- thereby making it a _shield_
only and not a signal return.  Then, the ground path already
established by having the sensor grounded at one place
and the conditioning circuit grounded at another serves as
the only signal return.  The practical disadvantage to this
is that the iron engine, ground strap, steel chassis 'ground'
is a poor conductor and has lots of other electrical currents
running through it as well.  The solution -- though not
always easy -- is to make all transducers (sensors and actuators)
differential or floating so that both signal wires (hi and gnd)
make contact in only ONE place .... at your conditioning
circuit.

 - shielding.  as noted above, shielding does not have to
be a signal conductor.  In good microphones, for example,
the mike is differential (two wires) inside the shield.
The shield carries no signal current -- that's all carried
by the two wires within the shield, and they are not common
with any other circuit, so no circulating ground currents
(ground loops) can be picked up.

  note that shielding is for wiring that is both susceptible
to noise pickup (i.e. low levels or high Z) and to wiring that
emits noise (i.e. hi levels/voltages and hi-currents/lo-Z)

 -- Shielding can be accomplished by

    - position. i.e. keeping wiring away from ignition, starter,
      alternator, blower moter, and other sources of "noise"
      This can be done by routing wires such that some existing
      metal, like a fender or manifold, is placed between the
      noise source and your wiring.  It is also done by just
      keeping your distance (inverse square law) and crossing
      noxious wires, if you must, at right angles.

    - twisting.  wires twisted with a ground wire are offered
      a minimal amount of shielding.  Twisted wires have been
      shown to have a nominal impudence of 300 ohms.  Tighter
      twisting does not produce proportional increases in
      shielding, though it does make it better.

    - braided or foil shields (coax).  foil is the best (least
      leakage).

    - cable troughs/conduit (just remember that the wires in
      the conduit might also have to be shielded from each
      other.

 =====  end of shielding diatribe  =====

filtering:
 - capacitors.  disc ceramics and mica caps will attenuate
signals into the GHz range -- as will just running the wiring
next to a ground for a brief period (as in coax).  chip
tantalum caps are good to almost 20 MHz.  Mylars, polyester,
and other wound foil types are typically not good much above
2 or 3 MHz.  ["Good" means that the inductive properties of
a cap begin to overpower is capacitive reactance properties
above some frequency -- sometimes called "ESR" or effective
series resistance.  X-sub-C decreases with f, but at some
f, a cap's X-sub-C will begin to rise.  That's a phenomenon
associated with X-sub-L and is due to the way the cap is
made -- long pieces of foil are more inductive than silvered
discs of ceramic or mica.]

 - R-C networks.  placing a small R in series with the signal
wire and a cap to ground provides even more attenuation than
the cap by itself.

 - L-C networks.  even bettah than the R-C.

 - active filters.  simplest is the diode, zener or 'varistor'.
Used in conjuction with a series resistor as above they can
attenuate or clamp pulses without affecting their initial
rise time, which can be a problem with capacitive filtering.

 === end of diatribe -- can't think of anything else to say  ;-)

simply put, I suspect a wiring error ... maybe shielding or a
ground loop, maybe simpler than that.

>It is a truly bizarre glitch because the engine runs fine with our own 6V
>AC/DC Radio Shack power supply and the same chip ran fine in the other
>engine we have which is the same model but was mounted on a test stand.
>This indicates the code is capable of running the engine.  The motorola
>chip actually generates the pulse and sends it to injector drivers
>(National LM1949) which then open up the injectors.  The weird thing is
>that the input pulse to the drivers (coming from the motorola MCU) is the
>exact same length as the pulse from a crank sensor which happens to be 1
>ms.  The pulse from the crank sensor feeds into the MCU for RPM
>calculation and seems to work just fine regardless of power supply.  
>Sean

Tom Cloud <cloud at peaches.ph.utexas.edu>



More information about the Diy_efi mailing list