Tom Sefranek wrote:

> I have an application note from Analog devices that states that you can 
> only have 40 nodes on a bus.
> With the MINIMUM bus specification, I see 11 bits for bus node ID., so I 
> ASSUMED you could assign several THOUSAND nodes per bus.

You probably know this, but just to state it here: CAN is not a hardware
specification. The actual hardware layer is defined in other specs (ISO) or
none at all. In the end, it is the specific hardware interface (chip or
custom circuit) that defines the properties of the wire interface.

So this AN from AD may be valid for one of their CAN interface chips, but
doesn't necessarily say much about others.


> All nodes work well for the first 900 or so messages. then the lowest and
> highest ID nodes in each group of 6 stall, while the middle 4 (ID) nodes
> continue chugging on. The stalled nodes make some attempt to catch up,
> but eventually the software FIFOs fill up with unsent messages. It make
> no difference what ID I assign to the failing nodes.

> But the highest and lowest ID nodes are the most physically distant (6")
> in the star stub.

The two statements above seem to contradict each other: in one you say the
assigned ID makes no difference, in the other you say there's a clear
relationship between lowest/highest ID and physical distance. 

I agree that it doesn't really sound like only a wiring problem. Always the
same nodes that fail? How about changing their location in the ID sequence
(not anymore lowest/highest), and/or changing their location in the wiring?


> I know this is complicated, and if I knew then what I know now.... I
> would have designed it as all daisy chain. (But I read the spec.
> allowing nodes to be quite long off the bus relative to the 3 inches I
> am using in my nodes.)

Again, don't mistake the CAN spec for the wire interface spec. For the real
details, you need to look at the interface itself.

Gerhard

-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist