AVR-Chat

> From: Steve Hodge
> 
> Hmmm.   It's beginning to sound like a crap-shoot.    The device I'm
> specifically looking at protecting is an AVR uC, its I/O lines specifically.
> How would you do it, if I may ask?   The environment is a boat one,
> basically an automobile one (engines, pumps, motors) with the added features
> of local radar, VHF and HF radio transmissions, as well as wireless computer
> and instrument transmissions.  Lightning too.   Shielded cables are a given,
> as is spike suppression on inductive loads.

Hi Steve,

Preface: This is my first use of transorbs and my knowledge has come from this list,
the Electronics_101 list, the datasheets, as well as Internet reading and, as to
picking exact parts, to some extent using the time-honored flip-of-a-coin :-)

My application involves long cable runs outdoors and is subject to a lot of
(Florida) lightning.  Here is what I have had in the field on two prototype boards
successfully for over a year:

Nominal 12vdc Power input from a power supply over 1,000 feet away (in order from
input to regulator):
	- 5x10 mm, 250ma fast blow glass cartridge fuse 
	- LittelFuse SA15A Transorb to GND (Mouser# 576-SA15A)

Analog inputs connected to 100's of feet of aluminum rail, nominal signal voltage
0-5vdc (in order from input to regulator):
	- 5x10 mm, 100ma fast blow glass cartridge fuse
	- LittelFuse SAC7.0 Transorb to GND (Mouser# 576-SAC7.0) and
		5k resistor to unregulated V+ (rail bias)
	- in series with a 100k resistor (limit current to ADC pin and part of R/C
w/cap below)
	- .1 mfd cap to GND (to somewhat filter noise, I'm only interested 
		in fairly slow DC change - akin to sensing a switch)

The fuse protects the Transorb from major hits.  IIRC the transorbs can take a
"surge current" of 70A - significant, but that's for a very short time.

This has served me well on the prototype boards.  I had several cases when one or
more of the fuses blew due to nearby lightning strikes.

On one occasion the system appears to have taken a direct, or nearly direct
lightning strike.  One board was heavily blackened on the back (trace side) with
several vaporized traces.  The other board had one or two vaporized traces.  

Casualties:
	- Several lost traces
	- Most of the fuses were blown
	- A couple of Transorbs failed (shorted - which is good because it still
protects the rest of the circuit) 
	- One processor (PDIP ATMega16) was blown (oddly enough, it was on the board
with the least visible damaged)

A total of about $8 in parts and some wire to replace the blown traces and the
boards were back in service.

Considering the severity of the strike I was very happy with the results.

The big thing I learned from this was "pay attention to signal routing on the
board".  I expected that, when the fuse blew it would isolate the circuit enough
(due to the, now large, physical separation of the ends of the fuse) from high
voltage excursions, but, I forgot to pay attention to what was going on on the back
of the board.

From what I can tell, it looks like everything went as planned: the fuses (and a
couple of transorbs) sacrificed themselves to protect others - but - the (now
isolated) input passed close to other traces which allowed the (now released to
rise) spike voltage to jump across and attack the rest of the board.

Lesson:  Be sure that the input is separated from everything else by a physical
separation at least as large as created when the fuse blows!

I have no illusions of being able to protect the boards from direct strikes but this
seems to be working for me.

I have looked at the possibility of using the solid-state resettable fuses (I forget
the name for them at the moment) but rejected them because a) they don't provide the
desirable physical separation when blown, and b) I don't know if they are fast
enough to protect the transorb.

Note that I selected the smallest fuse I had on hand that would support the nominal
currents involved.  One thing I have to look into is some testing to pick a fuse
that is large enough to allow the transorb to take as much of a hit as possible
without failing (to having to replace unnecessarily blown fuses in cases where the
transorb could have absorbed the surge) but not so large that the transorbs fail.  

To do this one has to 'assume' (watch out, there's that word again ...) an expected
surge profile (short high voltage, long high current, etc.).  I'm just planning on
charging a large cap to maybe 1-2Kv and discharging it into the input (maybe via a
low value resistor, have to look at that).  I'll keep increasing the fuse size until
something fails and then decrease it by, let's say, 20% for a safety margin.

BTW:  I have an old defibrillator.  I wonder if that would make a good test source
:-)

I hope this is useful.
 
Cheers,

Chuck Hackett
"Good judgment comes from experience, experience comes from bad judgment"
7.5" gauge Union Pacific Northern (4-8-4) 844 http://www.whitetrout.net/Chuck