[lug] OT--Quality ethernet switches

[Nate Duehr]

On Feb 25, 2004, at 2:18 PM, Stephen Queen wrote:

> I am working on an application where we are using ethernet
> in a harsh/noisy environement.

I would think that the folks building robotics for manufacturing 
facilities have also run into this problem.  Perhaps some research into 
what they do about it would be in order.

My guess is fiber optics.

Any place you have large electric motors turning on and off or 
heavy-duty arc welders operating (amongst other things) in close 
proximity to network gear you'd have this problem, so there have to be 
plenty of other folks who've tackled it.  You'd just have to hunt them 
down and perhaps you'd find a vendor that specializes in such gear.   I 
know one "fix" is probably lower data rates and completely different 
protocols other than Ethernet.  (Example: Your car probably has a CAN 
bus somewhere inside if it's relatively new.  You wouldn't want 
electrical problems keeping the anti-lock braking computer from talking 
to the traction-control computer!)

> I think we have a situation
> where the noise (emf) becomes so great that the ethernet
> connection becomes swamped with errors, and our applications
> program thinks its lost its connection.

Nothing is a given until proven in engineering... are you seeing CRC 
errors?  Lost connections?  What exactly are you seeing?  You could 
have a substandard cable plant that's allowing too much noise into the 
network from the information given.  All we can do is guess.

> I was also thinking
> that the ethernet switch that is receiving the data is an
> off the shelf Netgear switch, which is good in an office
> environment, could be changed out with a more robust switch.
> Does anyone know of a good switch for a hostile environment.

About the only thing you get electrically from a larger switch is a 
larger ground plane in the motherboard -- if the EMF is getting into 
the switch itself -- MOVE IT AS FAR AWAY from the EMF producing device 
as possible.

Also, are the network devices and whatever's causing all the EMF 
powered off the same AC bus?  You can buy isolation transformers with 
filtering that will attempt to mitigate such noise.

None of these are cheap solutions and none of them are appropriate 
without measurements and proof of exactly what's happening.

The cheapest solution is to isolate and get the EMF producing devices 
and the network cabling and devices as far away from each other as 
possible and to make sure you have an EXCELLENT ground in your 
facility.

In general if you're already in the soup this badly with EMF, it shows 
a lack of planning and true engineering up-front of the facility and no 
electrical design.  (But you probably knew that already.)  The fix will 
probably be more expensive than the prevention would have been.

Maybe I'm assuming the problem is bigger than it is.  But there's 
plenty of EE's coming out of college without jobs to go into.  Hire 
one!   ;-)  (At least EMF mitigation engineering can't be outsourced to 
another country very easily.)

And definitely do some hunting around for where manufacturers get their 
network gear.

Nate Duehr, nate at natetech.com