>
> At 01:47 AM 12/18/96 -0800, you wrote:
>
> >Brian if you have www access you can get the application notes AN-409 and
> >AN-759 from National Semi. These are related to RS-485.
> >
> >You can go via http://www.natsemi.com
> >
> >However Maxim makes devices with 1/8 unit-load input impedance that
> >guarantees up to 256 transceivers on the bus. ie MAX1483
> >
> >
>
> I looked up the app-note for MAX1483.  It looks very much like what I want.
> I want a multi-drop bus with multiple transceivers over large distances too.
>
> My question is this:  How do you keep the network nodes from talking at the
> same time?  I understand how i2c works to resolve this with start and stop
> conditions, or how RTS, CTS handshaking works, but the MAX1483 doesn't have
> a clock  line or any other handshaking lines, just two twisted pair data
> lines.  Am I missing something or
> am I just dense?

Neither. That layer doesn't address flow control or contention issues. It's
the responsibility of the layer above: i.e. software. Some suggestions:

1) Primary-secondary polling. Have a master unit poll the secondaries.
Secondaries can only talk when spoken to. Not real good if real time
response needed.

2) Virtual token ring. Pass around a token, the node that has the token can
talk. When done pass the token to the next node. Again not real good for
real time response.

3) 1) with attention. Use a second twisted pair for an attention line so
that the seondaries can signal the primary for attention. Can use a timeout
scheme to prioritize either the type of activity that need attention or
to prioritize the nodes.

Bottom line: contention becomes a software issue. A separate clock proobably
isn't a good idea due to the propagation delays over very long distances.

In this instance, speed kills. If you keep it slow, then any of the above
techniques can work.

BAJ