ethernet point to point wiring

In article , Eddie wrote: :I am looking to replace an old serial RS485 multidrop point to point :communications line with ethernet but would like to retain the wiring :savings in a point to point layout. On RS485 we can connect 32 devices :with series point to point wiring. I have done some research that :indicates that if I used a switch at each device I would be limited to :a maximum of 7. Is this correct, or are new switches able to exceed :that number?

The ethernet standards for Spanning Tree Protocol place timing limitations that are such that with worst case timing scenarios, only 7 switches can be used.

If your devices are newer or your distances shorter, then -in practice- larger topoligies might work for you.

And if your topology is fixed and you can be fairly sure that no loops will be created even by accident, then you could turn spanning tree off.

:If so, how many can I realistically connect in a point to :point topology.

Using some routing instead of pure layer 2 switching.

Why try to fix something that is (probably) not broken?

RS485 may be "old" in the sense that it has been around for a while, but my question for you would be "Does it fit your current needs reliably, and can it also be reliably expanded to meet your future needs?"

If the answer to both questions is 'Yes,' why go through the expense, complexity, and hassle of "upgrading" to Ethernet?

This is curiosity on my part.

The only ethernet method that works on a bus like that, is the old coax based 10base5 or 10base2. They are considered obsolete now, so you'll have to search for used gear. Switches are available in various sizes. I've heard of 64 port models. With switches (or hubs), you can cut down on cabling by placing small switches close to groups of users and running a single cable back to a central switch.

BTW, cable is cheap and it doesn't cost much more to pull multiple cables at the same time.

I thought spanning tree was only an issue if multiple paths exist to any point.

Agreed. And if you decide you *do* need ethernet it sounds like you are looking at daisy-chaining switches down the line and only pulling one wire.

IMO that's worse than thinwire in the probablity of having mysterious time-killing problems that are solved only by wiring things properly the second time.

All you're saving is the cost of wire, and that's the cheapest part of the project unless you are dealing with distances close to the CAT5 limit.

/end of rant/

1. Eddie Oct 28, 8:30 pm show options

Newsgroups: comp.dcom.lans.ethernet From: "Eddie" - Find messages by this author Date: 28 Oct 2005 13:30:35 -0700 Local: Fri, Oct 28 2005 8:30 pm Subject: ethernet point to point wiring Reply | Reply to Author | Forward | Print | Individual Message | Show original | Report Abuse

Is I understand it the 7 hop limit was imposed along with a requirement that each bridge should forward each frame within one second (or drop it) as part of the STP timing budget.

Today these numbers seem somewhat laughable (1Gbps for 1 sec would require a 100M Byte buffer) however as far as I know (you should check for yourself since I have not read 802.1d for years) the standard has never been changed in this respect.

There seems to me no reason not to increase the number of bridge hops used if STP is not involved.

A couple of things to ponder.

Support - If you tell your support channel that you are not compliant with the standard then I strongly suspect that they will be reluctant to assist.

Probability of packet loss - If you do not recover from lost data then you should carefuly consider the increased probability of lost packets.

Maintainability - I do not know what you do at present however if you were to use low cost switches that cannot report or record errors or problems it will be very difficult to troubleshoot network failures.

Good luck.

Thanks to all for your comments. It appears that this solution won't do for our application. Our typical installation would involve 10-30 controllers on a single RS485 multi-drop cable in the 1000+ft total length range. With this layout and our polling master/slave protocol, Modbus/RTU serial at 19.2kbaud, we typically get 1-2 updates/second/controller if requesting several bytes of data from each controller. The Modbus/RTU serial protocol is only typically supported at speeds up to 38.4kbaud, so a simple baud rate increase isn't a soultion. A fairly seamless change for us would be to put all of these controllers onto ethernet running Modbus/TCP protocol which wraps the same serial command string into a TCP packet and uses ethernet for transport. The major issue for our isnstallations with ethernet is the home run wiring costs. Most of our wiring installations require electrical conduit with a final installed cost of ~$10/ft which is much larger than the cost of the cable itself. Therefore, anything we can do to decrease the total installed feet of cable/conduit is a big cost saver to our end users.

It sounds to me like the best solution for us if we want to utilize ethernet is to put some centrally located multiple port switches in the field and link them together to give the most effiecient layout.

Thanks again for your comments.

Ed

You didn't say how many stations you want to wire but if it's a small number, pull wire and do it right. It won't cost much.

If you are doing it for a large operation then do it right or you'll spend your days under desks and with unhappy users and then your successor will rewire the place, correctly.

I believe that would fall under this sentance of mine that you snipped:

"And if your topology is fixed and you can be fairly sure that no loops will be created even by accident, then you could turn spanning tree off."

In theory, there are no loops in our LAN. In practice, people accidently introduce loops... including me getting distracted about which cable end is which and so ending up plugging them both into the same stack.

When you reply to my messages, please quote something I actually said, instead of making it appear as though I said something I didn't.