RE: Why do ISO-11898-2 and ARINC 825 never mention split termination?

Hi Richard,

There is a paper by Steve Corrigan on the TI site that show what happens
when the nodes are too close together.  I have scope photos of exactly that
effect.
Just before he passed away he was working on testing whether sharp bends in
the Thick DeviceNet cable would impact the signaling.  
The reflections from signals bouncing around the cable are only a problem if
they happen to coincide with a data bit in a message and the perturbation is
large enough to change the value.  And that may only happen if the message
from a different node happened at just the right time relative to other
nodes.  
Sounding like woo-woo stuff yet
It is and it isn't.   It's not really random in the sense that if the bits
are there at the right time and right place but reproducible with messaging
may be harder.  That's why it's best to follow the transmission line rules.

A good example is multipath interference in VHF radio.  The signal at the
receiver has come directly from the transmitter and also bounced off
something and arriving 180 degrees out of phase effectively reducing the
signal amplitude because added together they sum to zero.  So imagine you
have a periodic signal sending data.  It's arriving perfectly almost all the
time but every once in a while it's not there.  Look out the window and you
see the windmill, which is turning today has the blades at the right spot at
the right time to reflect the transmitter signal and create that multipath
interference.  This isn't a real example but just think how hard it would be
to find.  If the wind is light the blades turn more slowly and you might
never have the event.  Turn faster and same thing.   But for the moment that
the blades are in the right spot and the message is sent at that instant the
multipath occurs.

So it can be with the CAN messages.  

The answer about two 60 Ohm with a capacitor is far more complicated and has
a lot to do with electrical and magnetic interference.  Coupling noise from
the system through a capacitor onto the bus may not be the result you wanted
when the desire was to drain it away.

John Dammeyer

---------------------
From: [email protected]
[mailto:[email protected]] On Behalf Of Corey,
Richard
Sent: December-05-14 2:36 PM
To: '[email protected]'
Subject: [CANLIST] Why do ISO-11898-2 and ARINC 825 never mention split
termination?

ARINC 825 says:
 
3.2.3.1 Network Termination
A 120 Ω terminating resistor shall be installed on each end of a network
segment for impedance adaptation and to avoid disturbing reflections. 
 
The ISO specification 11898-2 also refers to termination as "a" resistor.  
 
Why are single resistors suggested or required instead of split terminations
(two 60 Ohm resistors plus a small capacitor)?
 
I thought the split termination was usually preferable and rejected some
high frequency noise from the bus. 
 
 
 
I see that minimum distances are required between the last node and a
terminator (1.5 meters), and also between each node (0.5 meters).  I've seen
that "minimum distance" suggestion elsewhere but seldom noticed any trouble
with desktop setups where these distances are inches or less.  (My
definition of "trouble" is any error frames at all, ever, except during
power cycles or hot plugging.)
 
We want to use CANbus to connect three adjacent cards inside a box, but it
seems crazy to stuff 3.5 meters of cable inside a small box (and make the
box big enough to fit!).  Does everyone who uses CANbus to connect adjacent
cards run big loops of cable around and around the cards to meet this
requirement?
 
I don't have any EE training, so all I understand about the reason for
inter-node distance and node-to-terminator distance is to reduce the risk of
"reflections".  Would it be safe to ignore those rules in an unchanging
network which will never have additional bus segments grafted onto it? 
Couldn't we adjust the distances until all the wave forms are clean, with
distances of a few inches at most, and then lock the geometry down and never
vary it?
 
Until I read about "minimum distance requirements", I never worried about it
in desktop setups.  I only got into trouble once.  I had a 1-inch CAN
connection between two cards in a cage that "just didn't work".  I
lengthened the loop and then the traces were clean and error frames went
away, so I stopped worrying.  I forget where the terminators were in that
rig (it was 10-12 years ago) but they were either within a few inches of the
two nodes or missing altogether at first. 
 
 
 
Rick Corey  

--
Archives and useful links: http://groups.yahoo.com/group/CANbus
Subscribe and unsubscribe at www.vector.com/canlist/
Report any problems to <[email protected]>