AVR-Chat

> -----Original Message-----
> From: stevech11
> I'd think that the low speed confg would be fine.

From what I've read so far I believe it would be fine.  

After the application is more defined I'll have a better hand on message
sizes and xmit rates are.

Of course it depends on what one defines as low speed/high speed - I
remember when I was able to move from 300 BAUD to 1,200 BAUD and thought I
was in Heaven!

> -----Original Message-----
> From: dlc

> .... However, if you
> can deal with very low baud rates (which are less sensitive to timing
> and impedance issues), of say 9600 and lower, you can simply use any
> USART with "9th bit" addressing mode and CAN transceivers to get very
> long runs.  Pay close attention to ground issues on long runs since you
> can build up large ground potential differences that can damage
> devices.

I'm thinking of going this route (using CAN transceivers) on the initial
version (ATMega16) and implementing a simplified CAN-like bus arbitration
and message protocol.  If that works fine I'll leave it that way but if I
need more speed/sophistication I can transition to full CAN later.

< ... CAN
> networking is a broadcast bus, every node hears every message, each
> node typically sets filters to only listen to its own address.

That was my plan when I was considering RS-485 multi-drop-multi-master.  90%
of the traffic is consumed by at least two, and potentially many more, other
nodes.

> ....
> With your own USART networking scheme you have to/get to pick
> your own payload and protocol overhead.

Kind of like being your own boss - you get the benefits but you also get the
headaches :-)


> -----Original Message-----
> From: Mike Payson
> 
> Hmm... This sounds to me like a perfect application for a Zigbee
> network. The modules are a bit more expensive, but since you wouldn't
> have to run wires, the final cost should be roughly comparable if not
> substantially less. ....

As with many potential solutions it has pros & cons that I have looked at.
You'd have to have more info on the project, install sites, etc.  

In a nutshell: One of my other hobbies is ride-on, live steam railroading
(see: http://whitetrout.net/Chuck/844/photos and
http://whitetrout.net/Chuck/844).  The tracks are 7.5" gauge using aluminum
rails on treated wood ties.  The application I am working on is a robust,
automatic signal system for tracks around the country.  One of the major
things the system addresses is detecting a train shunting the rails.  This
becomes challenging in very wet weather where a 900 foot long "block" can
contain ~2,700 ties and they all conduct a bit, to make it worse, a block
that long starts to take on the characteristics of a very lossy transmission
line (w/distributed R & C) and it can get difficult to detect a "shunt" out
of the noise.  Oh, yea, 1,000's of feet of aluminum rail and cable-carrying
conduit makes a great antenna and lightning magnet and the system needs to
survive a nearby lightning strike (disregard DIRECT lightning strike, these
tracks don't have a NASA budget <g>).

The "nodes" I keep referring to are located at switches (detect point
position, control a point motor, etc.), connected to "blocks" of track (~10
to maybe ~1,000 feet long) to detect trains, at signal heads to drive the
signals, etc.

I really need to get a web page up on this thing but that's yet another
project - So many projects, so little time :-(  I think that the major
factors effecting RF .vs. hardwire in my case are:

1) Power
2) Node density

All the nodes need power and they are not near local power sources.  I could
provide solar chargers but that would dramatically increase cost and many of
the nodes are in areas with dense tree cover and batteries are not an option
in most situations.  I am using a #16 pair for power distribution (48vdc) in
a conduit along the track ROW with plenty of room for a CAT-5 cable.

I am getting CAT-5 for ~$55/1000, so if, in a particular area, the
inter-node distance is long enough or node density is low enough, Zigbee
might be cost effective.  I was planning on having network segment bridges
for various reasons and Zigbee might fit in there also.  But those issues
are for large installations and are down the road (track? <vbg>) at the
moment.

There are also a couple of cases where someone using my system might only
need a couple of nodes and might want to use RF w/batteries so they can take
in the equipment when not in use (vandalism), don't have to trench, etc.  I
see nothing in my design that would preclude this.

I have checked out the links you provided, thank you.  It's always good to
keep informed about potential solutions in case project factors change.

Thanks to all for the info ...
 
Cheers,

Chuck Hackett
"Good judgment comes from experience, experience comes from bad judgment"
7.5" gauge Union Pacific Northern (4-8-4) 844
http://www.whitetrout.net/Chuck