Modbus device ID

The company I work for is developing a controller that has three communications ports: two RS485 and one TCP. Modbus TCP and Modbus
RTU Master/Slave will be implemented. It has been proposed that each port have a different Modbus device ID, rather than one global ID.
I contend there is no benefit to this, and only adds confusion for integrators. Does anyone know of a precedent for this?
Thanks in advance.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Fri, 21 Dec 2007 05:53:05 -0800, btmull wrote:

This is a good question (the answer to which I can't help, since I've always built controllers from scratch, so have never used Modbus).
But I'm an obsessive net cop in one little way: you have multiply posted this to (at least) sci.engr.control and comp.arch.embedded. If you plan on using newsgroups much, please check the Wikipedia article on "cross posting". It mentions multiple posting, and its difference from cross posting.
This article is (IMHO) an excellent candidate for being cross posted to the groups to which it has been multiple posted.
--
Tim Wescott
Control systems and communications consulting
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Dec 21, 5:53am, snipped-for-privacy@yahoo.com wrote:

I agree with you as far as the RS4875 ports go. There is no need to have separate node IDs UNLESS the two ore more ports are on the same network and that would be stupid.
Modbus TCP will have an IP address which is different from a simple node address. I don't see how these can be the same UNLESS you are going to hard code the first 3 digits of the IP address to something like 192.168.1.xxx and make xxx the node address. I would find this to be stupid too. The most common way I see OEMs set up their IP addresses is 192.168.mmm.nnn where mmm is the machine center and nnn is the node within the machine center.
See http://www.robustdc.com/library/an029.pdf
Peter Nachtwey
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Fri, 21 Dec 2007 05:53:05 -0800 (PST), snipped-for-privacy@yahoo.com proclaimed to the world:

I don't have an opinion on your question, but I do want to suggest something. Make the I/O fully isolated from ground. Some modbus devices I have used in the past caused a lot of trouble because of ground loops developing and an isolator had to be added separately.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

As Peter pointed out, the Modbus TCP port will require a separate IP address. However, the Modbus messsage within the packet still has the slave ID, which can be used or ignored. Just to be clear, the IP address and Modbus ID of all ports will be user settable instead of hard coded, correct?
My feeling is that any integrator who would be confused by a device with two serial ports that can have different settings is simply too dumb to be an integrator. Further, having two RS-485 ports implies the possibility of connecting to two networks. If I had to juggle addressing on two networks to make sure that one or more devices each had the same address available on both, I would be highly annoyed. Stupidly annoying products tend to land on my DO NOT USE list.
As for precedents, here's 30 seconds worth of multiple serial ports that are separately configured. For the moment, I cannot think of a device where they aren't that way. - Allen-Bradley MicroLogix 1500 PLC - Numerous Modicon PLC's - Control Microsystems SCADAPack (up to four ports!) - GE/Fanuc 90-30 and 90-70 CMM module (even different protocols)
Mike
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

I would agree with all the respondents so far. If the device could be used to bridge between networks then you will have to be able to tolerate a form of Network->Node addressing, in which case you could have duplicated node IDs (each on different networks) or you could have duplicate network IDs with different node numbers (with different transports).
You might want to simplify your specification by avoiding one or the other (or both!) of those scenarios - but please make that clear to potential users!
If you implement addressing through programming/configuration, any chance you can display the network & node number in hardware next to the port? One advantage of little hexadecimal switches is that you don't need a laptop on the shop-floor to figure out what address a device has.
David.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here. All logos and trade names are the property of their respective owners.