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

John ...

On Sun, Dec 7, 2014 at 3:30 PM, John Dammeyer <[email protected]>
wrote:

> Hi Dave,
>
> Long time…  Nice document.  Very comprehensive.    But misses the point
> about bit rate and bus length.
>

Thanks.  It is the companion document to
http://www.openlcb.org/trunk/specs/CanPhysicalS.pdf for OpenLCB, a public
specification for a Model Railroad (MRR) Local Control Bus.   A lot of it
is from Bob Jacobsen's direct experience with a lot of CAN systems, often
large ones, such as the CERN super-collider, whereCAN's limits were pushed
often to breaking, and required changes in bus configuration.


> I'd like to see a reference for lines 10..15 other than just ethereal app
> notes etc.  The data sheet 120 node maximum drive for the 82C250 for
> example is from the parallel combination of the 120 devices worst case
> input impedance in parallel with two 120 ohm resistors.   There is nothing
> scientific about reducing the number of nodes based on cable length or stub
> length.
>

The set of recommendations is a boiling down from all available known
sources, and Bob's experience -- i.e. it is our opinion, and as such is
just a guideline.  Moreover, it is specifically for CAN over CAT5 at 125
kbs, which is what we chose because it was as fast as we felt we could go
with acceptable bus length for our application (MRR).  Within these limits
one can expect a functional bus -- so these 'rules' are very conservative.
A longer bus, or more nodes is entirely possible, but one may have to
resort to experimentation and expensive equipment to tweak the bus.

Ultimately, the CAN bus length and the maximum number of nodes is limited
by many interacting factors: drive capability; cable matching; cable
characteristics such as resistance, inductance, and capacitance;
reflections; topology; node distribution (clumping), etc.


> Inductance and capacitance play a much larger role and to get that as low
> as possible usually means buying specialty cable which is thick, not very
> flexible and expensive.
>

Again, this is meant for a specific application and wire standardized on
CAT5.  I agree that one can do better than this, but as you say, one has to
choose the components more wisely, better drivers, better cable, good
termination, nodes not too close together and not clumped, etc.


> The problem I have with those magic rules is they then tend to get quoted
> without reference by someone else.  Then their document is quoted by
> someone else and pretty soon we have a general rule that everyone quotes
> without any underlying basis.  Then we see emails with people arguing that
> the maximum number of nodes on a bus is limited to 48 with no direct
> connection to your document; which has no reference either.
>

Well, we can't control what others choose to do with it.  The 'magic' rules
are meant to be simple for the average model railroader, who is not a
systems nor electronics engineer, and people should always reference, and
choose information wisely.

The main point was for the person that had asked about 'split termination',
and I had pointed him to this document for its references to termination
options, some of which I think you touched on.

Still other than that a good source for a lot of information.
>

Thanks,
David