Grounds, Shielding, etc...

Ed Lansinger elansi01 at ctlw3908.mpg.gmpt.gmeds.com
Fri Aug 11 20:10:42 GMT 1995


Grant Beattie wrote:

>I need your expert advice on how to best shield electronics from
>noise/vibration generated in a vehicle so that the electronics can do it's
>job best.
>...
>For what it's worth my preliminary tests were less than stunning with the 
>CPU resetting itself intermittently.  For the time being I had housed it 
>in an unshielded plastic case and I think the ignition noise from the 
>heathen Briggs 5hp motor overwhelmed it.

There are a number of ways to skin this cat.  I'll present one I've used
successfully in the past.

1. Put the electronics in a metal box that is fully sealed.
2. ALL wires that go through the box MUST pass through their own
feedthrough bypass capacitors (2,000pF per cap is about what you want).
Check out Murata-Erie or Spectrum Electronics.  These companies make DB-25,
-15, etc. connectors with feedthroughs built in, although they can be
expensive.  You can also get discrete feedthrough caps from Murata.
3. [Ideal, but not necessary with the above] CPU board should have an
extensive ground plane.

>Should my cpu "box" be grounded to
>the chassis of the vehicle or should it be isolated?

I grounded my box to the '-' terminal on the battery.  The battery was
mounted to the chassis with insulating rubber standoffs.  The chassis and
engine were also connected to battery '-'.  All sensors and actuators got
ground from the box, though, not the chassis or engine.  This prevented
ground loops.


>Should various
>sensors have metal shielding around their cabling? 

With the setup just mentioned, I didn't need shielding.  I found it useful,
though, to put metal braid over important wires for physical protection.

>One example I can give you is that I have a magnet and hall sensor on the 
>rear (solid for karts) axle.  This requires me to run +5v, Gnd and signal 
>about 3' from the CPU to the sensor.  I'm using a cable which has a 
>ground shield and two inner conductors on which I'm placing +5v and 
>signal.  There is no connection whatsoever between this ground and the 
>chassis.  Is this good or evil?

Should work just fine, I did the same thing.

I've done fiber optics, too, which worked great but are problematic
for analog signals.

I went through hell working on this problem the first time I encountered
it.  If you have any wire coming through the wall of the box that isn't
bypassed right at the wall, it will rebroadcast the RF inside the box
before it gets to a capacitor.

The OEM's bring their wires in, send them to the board, and then every
pin has a surface mount cap on the other side of the board between that
pin and ground.  Then they put a metal shield over the exposed pins and
put the whole thing in an aluminum box.  The value of the cap is about
1,000pF and is of N7R material if I recall correctly.  It is still important
that there is no line of sight path between any of the wires coming
in and any traces or chips on the board.

Make sure your software gracefully handles resets nonetheless.  Problem
is, you're probably seeing conditions where the processor just goes
off into space and the watchdog can't even bring it back.  The feedthrough
caps will solve this problem.

-------------------------------------------------------
Ed Lansinger
General Motors Powertrain
Powertrain Control Center
Premium V Software & Calibration Group
Milford Proving Ground, Milford, MI
elansi01 at mpg.gmpt.gmeds.com  8-341-3049  (810) 684-3049
-------------------------------------------------------





More information about the Diy_efi mailing list