AVR-Chat

Thanks for the feedback, my replies below:

> From: wagnerj@proaxis.com
> ....
> Depending on where, in the world ... it could last a
> long time as you propose, or it could be toast by this time next year.

... or, in Florida, the same time tomorrow! :-)

This product is for the outdoor ride-on railroad hobby.  There are clubs all
over the globe but I dare say that 99% on this list have never seen one
(typically 7.x" gauge track) so it's not a big market.  If I make some
money, great but the endeavor combines my interests in machining,
electronics design/dev, software design/dev and Live Steam Railroads
(http://whitetrout.net/Chuck/844/) 
 
My test site is north of Tampa, Florida, USA.  Florida is known for having
lots of lightning so it's a good test site from that aspect :-)

I want to protect from natural ground (dirt, not "Gnd") voltage gradients,
lightning induced ground voltage gradients (i.e.: nearby lightning), and
voltages induced on the field wiring via "antenna" effects, etc.

As with all things, this is a compromise.  The hobby won't support a "NASA"
style approach, I need to implement enough protection to protect against
common occurrences.  In the case of very close/direct strikes all bets are
off.

I have done a lot of looking into this but I am no expert.
Lightning/surges/Gnd loops, etc. are tricky things and industry has spent
millions addressing it.

At the moment I have two ATMega16 based controllers in field testing.  They
each drive two 3-LED signal heads.  One of the heads is local, 10 to 30 feet
away from the controller, and the other is located approximately 800 feet
away connected by 4 conductors in a buried CAT-5 cable.

Analog inputs to the ATMega16 are connected to sections of aluminum rail
which are about 800 feet long.  From the rails, these signals pass through a
125 ma fuse to a junction which, in turn, connects to a) 5k resistor to
+12dc (supplies voltage to the rail), b) 7 volt TVS (Transorb) diode (clamps
spikes to 7v), c) 100k resistor feeding the A/D pin of the ATMega16 which is
also bypassed to Gnd with a .001 mfd cap.  I put in the fuse to protect the
TVS in very strong spikes but none have blown yet.  I contemplated a
gas-discharge, inductor, TVS network here but, so far, it doesn't seem to
need it.  So far I have had no failures/blown fuses ... time will tell.  

The prototype controller drives the LED lines with a ULN2003AN driver with a
1k resistor between the driver and the cable.  I put the current-limiting
resistor at the driver end to protect against field wiring shorts.  I have
had no driver failures so far.

The signal heads themselves consisted of 3 LEDs connected to the CAT-5.
Here I was getting consistent failures of LEDs particularly in the "remote"
heads.  I believe that what was happening was that spikes/surges in the LED
forward bias direction were clamped by the LED but surges in the reverse
direction were allowed to rise above the reverse breakdown voltage of the
LED.  To address this as cheaply as possible I added a general purpose diode
connected in the reverse direction across the LED to clamp "reverse" surges.
So far it has lived through several storms, again, we'll see.

So, the impetuous for my question was an attempt to reduce board real
estate/parts count at the controller end while still providing an acceptable
level of protection.  I was looking at:

1) Keep the current design (outboard driver) (pro: it works so far, con:
board space, parts count)

2) Connect the LED cable directly to the ATMega16 with the LED
current-limiting resistor value split between a resistor on the controller
(to limit surge current to the pin) and another on the signal head (to limit
surge current there) using the pin protection diodes within the ATMega16 to
protect the pin.  (pro: reduces real estate/parts count, con: exposes the
ATMega16 to surges, will it be enough?, will it induce bad things into the
V+ line?).

3) External protection diodes from the pin to V+/Gnd.  I would need to
ensure that the forward drop of these diodes is less than or equal to that
of the ATMega16 diodes. (pro: simple, con: will it be enough?, will it
induce bad things into the V+ line?) 

One thing to note is that, if a part does fail (ATMega16 or driver), the
cost of the part is not significant compared to the incontinence of the
outage and the time to repair (I don't want to use sockets).  In elapsed
time/effort, it takes almost as long to replace a driver as it does to
replace an ATMega16.  Thus, if I can reasonably protect the ATMega16 pins to
the desired reliability level then why not do that and eliminate the extra
driver chip?


> From: David VanHorn
> ....
> They do tell you not to put any current through the protection diodes.
> From a practical POV, I'd want to stay in the low microamp range..

> From: David Kelly
> ....
> Go look for an Atmel app note on direct reading voltages from AC power
> lines. ....
> 
> It went on with a disclaimer that there was no advertised spec as to
> what those diodes could handle but 1mA was thought to be safe, and
> directly connecting the CPU to AC was a bad idea if it also connected
> to humans.

So, it would seem that the internal diodes do not have any significant
current-carrying capability and are mainly for ESD protection.  Thus, option
2 above seems out so I'll stick with some kind of outboard protection.


> From: Mike Bronosky
> ....
> One sure fire way we over come this is the use of fiber optic cables.

Hard to push LED current down a fiber :-)

I suppose I could put the LEDs on the controller and have three 800 foot
fibers to the signal head :-)

> ....
> This is a link to Black Box the picture you will see is a protection
> device
> that is made by that company and marketed by Black Box as there own.
> Contact
> them and see if they have anything the will protect 5VDC lines. Or will
> provide you with their supplier.

The device you referenced lists at $66 - no way my market would support
that.

I have looked into a lot of this type of device and most of them use some
combination of gas-discharge tube, inductor, TVS (Transorb) diode.  I can
incorporate these in my application, it's just a matter of "how much is good
enough".

One thought I had was to have a header on the prototype controllers for each
external connection and have a small daughter board providing the
protection.  I could start with very simple protection and replace the
daughter boards with more sophisticated protection as needed.


> From: Donald H
> ....
> This is a no brainer !!!
> 
> Why is this discussion happening !!!

I am discovering that, when trying to develop a commercially viable product
(i.e.: max functionality for least cost) there are lots of issues that would
seem to be "no brainer"s but can bring up lots of tradeoff issues.

> You can use a 2n3904 and 24 Volts DC source and Limit the current in
> the LED, resistors are cheap.

That's basically what I'm doing at the moment except that my current power
distribution bus is 12vdc and, for other reasons, I'm looking at going to
5vdc LED drive which reduces the current-limiting resistor and increases the
potential surge current ... tradeoffs ... tradeoffs ...

> Driving more then 4 inches from a microprocessor pin is just lazy and
> foolish.

The controller needs to drive two displays that are 800 feet apart so the
LEDs must be driven by a total of 800 feet no matter where the controller is
located.  How would you do it?

I am looking at using a smaller controller at the "remote" head but this
brings up yet more tradeoff/cost issues and I need to keep it simple at the
moment because there are a host of other issues yet to be addressed in the
system.

> $.22 of parts takes care of any problem.
> 
> And if the line gets spiked, you replace a 15 cent transistor not a
> microprocessor.

No "IF" about it, the lines WILL get struck!  

The question is an old one in risk management: the expected "frequency of a
given failure" .vs. the "impact of that failure" (time, money, etc.).  

It's the challenge of providing protection that reduces the "frequency of a
given failure" to the point where the cost of providing that protection is
acceptable when compared to the "impact of that failure".

One problem is that I do not have reliable data as to what level of surge to
expect and how often do they happen.  If I knew that "only one surge every 5
years would rise above X induced voltage" it would be a "no brainer".  I
think an outage every 5 years is acceptable so I'd test to voltage level "X"
and move on to other issues.
 
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