Automotive Circuit Protection - Part 2

[Garfield Willis]

On Mon, 26 Jun 2000 09:47:42 -0400 (EDT), Pat Ford <pford at qnx.com>
wrote:

>Hi Garfield, et al:
> Ive measured loaddump WAY above 70V, here is the scenerio, jump starting a car.
>Car A is running ~3000 rpm Vbat ~14.4, car B cranks draw the combined A and B 
>voltage down to ~9. Alternator is now pumping out Max power. When the starter 
>is released  on car B the alt in car A is still putting out max power, even 
>with the battery I measured +186V.
> I took the alt back to get it tested and was told that the magnetic field takes 
>time to decay, AND there is a small delay before the regulator can respond.

Yow, and as you say, "even with the battery"!? Ugh. As we'll get to in
Part 4, the SAE/ISO specs describe their model of the load dump event as
being a transient of a given max voltage, and duration/waveshape, so
they can thereby say what *energy* they're expecting to be contained in
the event.

But of course, they're only models; hopefully they work when designed to
in practice. Maybe the saving grace here is that these transients, if
they vary widely in peak voltage and waveshape, at least remain
somewhere in the range of level of energy, so that even if you design
for the 70V peak & standard SAE waveform, the actual transient while
different, is still protected by the bug-catcher because the bug is
overall rounds-bout the same energy level.

When you put a transorb across a bus, you're protected in terms of
voltage levels MUCH higher than the 70V found in the spec, simply
because the transorb avalanches above it's standoff voltage. So as long
as the threat doesn't exceed the energy limits of the transorb, and of
course as long as the threat doesn't reach voltages where it can jump
air gaps (namely "HV"), you're still protected.

What's worrisome about your experimental measurement would be if the
pulse on the ALT were same shape & duration as the SAE std event
"model", BUT ALSO at that much higher voltage level. That would be bad
news. :)  Maybe what this is saying is that the std. event model is just
the best they can do for a phenom that's hard to predict well.

Gar


----------------------------------------------------------------------------
To unsubscribe from gmecm, send "unsubscribe gmecm" (without the quotes)
in the body of a message (not the subject) to majordomo at lists.diy-efi.org