CAN Bus, Star wiring and termination.

(Only passing comment as I've no idea about CANbus) Cat5 cable has a nominal AC impedance of about 110 ohms. For signal use, the end of each cable run would normally expect to see a 100 ohm resistor across it to provide a matched cable termination to prevent reflections upsetting the data. I've absolutely no idea of what's involved with a CAN bus but do know CAT5 is CAT5. From what you describe, it seems a CAN bus does ask for termination. Signal wise, a 780ohm resistor across each cable end would seem to offer little benefit. I'm also puzzled why a 60ohm and not a 100ohm source impedance is needed. (Maybe CANbus chips are meant to drive 50ohm co-ax?). From a network point of view it's usual to wire at high impedance, from point-to-point, with a single 100 resistor at the furthest end of the data line. As you're at lowish data rates and lowish distances then yes, you might (if you feel lucky :) get away with 780ohms at the line ends or 60 ohm at star central. regards john

The CAN bus is basically an line terminated with 120 Ohm on both ends, stubs are limited to 50 cm IIRC. In your case the proposed star topology is not applicable to the CAN bus; I'd recommend changing the topology to a line (if you *have* to wire every display to the controller you'll have to bridge some CAN lines so that the mechanical star becomes an electrical line - a rather awkward 'solution') or change to a different bus.

You might get away with disregarding the CAN bus standard for short distances and low data rates but I'd advise against it.

If you *really* need a star topology as well stay within the CAN bus specifications you'll probably need a seperate CAN interface for each display...

I'm sure you've read it already but someone might be interested :

This AN covers some termination issues :

my 2 Yen,

Chris

Designed for cars, but not just for cars -- and when you find an 8-bit micro with Ethernet for less than $3 let me know.

For the CAN bus's collision detection/priority arbitration scheme to work a '0' bit transmitted by any one device on the bus has to be detected before the bit interval is over, which means that any transients on your cable need to have died down within about 1/2 to 2/3 of a bit time. 50m equates to 170ns in air, probably about 200ns on cat-5 (what _is_ the velocity factor?).

I have no idea what your termination scheme is going to do to your signal, but you could probably work it out mathematically or prototype it fairly quickly -- I think you'll find that you're more or less OK for the case where you're transmitting from the hub, but that any display on the end of things is going to see a nasty transient, which may or may not die down in time to make your data rate requirements.

If you can you may want to consider either 12 point-to-point links (in which case why bother with CAN?), or having each wire run out to a device and back, in which case (a) a single-point failure will take out your whole bus and (b) you'll need two twisted pairs to each display.

Several Motorl^W Freescale 68HC(S)08 derivatives. BTW, CAN bus drivers are also cheaper than Ethernet. And a powerful CANopen protocol stack can be done with much less footprint than Ethernet.

Oliver

....a rather stupid comparison. I wanted to say that CANopen is a powerful/versatile high level protocol with data abstraction etc. and even simpler/smaller to implement than the plain data transfer part of TCP(/IP) without any high level stuff.

Oliver