Hi Folks,
I had always laboured under the belief that when you implement a network using coaxial ethernet cable, you basically had to have a single linear cable with pick-off at T-pieces along its length, and that branching coax structures were not on.
In particular, if you needed, say, to connect a laptop [C] to an ethernet between machines [A] and [B], and it wasn't feasible to extend the cable beyond [A] or [B] to reach [C] but the middle of the cable itself passed not too far from [C], then I thought that the only solution was to bring the cable in a long loop to [C] and connect to a T-piece in the middle of the cable. I.e.
[A]--------- -------------[B] | | | | | | | | | | \ / T | [C]
However, thinking about it a bit and then trying it out, I discover that you can in fact set up a branching coax network. The trick is to use spare T-pieces to implement the branching. Like this:
[A]----------T-------------[B] =I [T-piece at 90 degrees] | | [C]-----=I [T-piece on branch cable] | | [more machines on branch if you like] | =I [terminated T-piece to close branch]
The trick is the first T-piece at 90 degrees, whose lateral ("male") insert is plugged into the "female" socket of the T-piece doing its usual job on the main cable. The ("female") end of the "branch" cable is then put onto the other "male" insert of the first T-piece at 90 degrees. This piece of trickery ensures valid connection (and indeed functions just like the normal connection at the ethernet card).
And you will of course need a terminated T-piece at the far end of the "branch" cable in the usual way.
No doubt this could be extended to a whole tree of coax, though shortage of spare T-pieces (and spare machines to hook onto them to test it) prevents me from testing it.
Apologies if it's not new to you, but since I hadn't heard of it being done before I thought I'd share the idea.
Best wishes to all, Ted.
-------------------------------------------------------------------- E-Mail: (Ted Harding) Ted.Harding@nessie.mcc.ac.uk Fax-to-email: +44 (0)870 094 0861 [NB: New number!] Date: 26-Sep-04 Time: 16:25:41 ------------------------------ XFMail ------------------------------
[cross posting trimmed down to just ALUG]
On Sun, 26 Sep 2004 16:25:41 +0100 (BST), Ted Harding ted.harding@nessie.mcc.ac.uk wrote:
No doubt this could be extended to a whole tree of coax, though shortage of spare T-pieces (and spare machines to hook onto them to test it) prevents me from testing it.
Does anyone still use coax? I thought it was twisted pair and wireless all the way these days. I did vaguely hear of gigabit coax ...
Tim.
On Sun, Sep 26, 2004 at 04:25:41PM +0100, Ted Harding wrote:
Hi Folks,
I had always laboured under the belief that when you implement a network using coaxial ethernet cable, you basically had to have a single linear cable with pick-off at T-pieces along its length, and that branching coax structures were not on.
In particular, if you needed, say, to connect a laptop [C] to an ethernet between machines [A] and [B], and it wasn't feasible to extend the cable beyond [A] or [B] to reach [C] but the middle of the cable itself passed not too far from [C], then I thought that the only solution was to bring the cable in a long loop to [C] and connect to a T-piece in the middle of the cable. I.e.
[A]--------- -------------[B] | | | | | | | | | | \ / T | [C]
However, thinking about it a bit and then trying it out, I discover that you can in fact set up a branching coax network. The trick is to use spare T-pieces to implement the branching. Like this:
[A]----------T-------------[B] =I [T-piece at 90 degrees] | | [C]-----=I [T-piece on branch cable] | | [more machines on branch if you like] | =I [terminated T-piece to close branch]
The trick is the first T-piece at 90 degrees, whose lateral ("male") insert is plugged into the "female" socket of the T-piece doing its usual job on the main cable. The ("female") end of the "branch" cable is then put onto the other "male" insert of the first T-piece at 90 degrees. This piece of trickery ensures valid connection (and indeed functions just like the normal connection at the ethernet card).
You don't have to use a T-piece at 90 degrees, you can get a convertor... I have a bunch of them in a bag here, and a serious amount of coax cable.
More than happy to donate them to anyone that wants them, they were being thrown out a while ago from an office above where I work :)
Cheers,
Ted, did you do any network saturation testing?
I suspect this configuration will work fine as long as there is not a lot of concurrent traffic between many different machines.
The big problem with this setup is whether all stations will see collisions within the 10BaseX parameters - if not then you could get situations where throughput is very poor. However, for only a few stations and very restricted traffic patterns you might have on a home network it might work very well.
I run a mixed cat5/Coax network at home but have never dared break the rules so blatently - I'd be very interested in a later report of your experiences.
Interesting Jim
On Sun, 26 Sep 2004 Ted.Harding@nessie.mcc.ac.uk wrote:
Hi Folks,
I had always laboured under the belief that when you implement a network using coaxial ethernet cable, you basically had to have a single linear cable with pick-off at T-pieces along its length, and that branching coax structures were not on.
In particular, if you needed, say, to connect a laptop [C] to an ethernet between machines [A] and [B], and it wasn't feasible to extend the cable beyond [A] or [B] to reach [C] but the middle of the cable itself passed not too far from [C], then I thought that the only solution was to bring the cable in a long loop to [C] and connect to a T-piece in the middle of the cable. I.e.
[A]--------- -------------[B] | | | | | | | | | | \ / T | [C]
However, thinking about it a bit and then trying it out, I discover that you can in fact set up a branching coax network. The trick is to use spare T-pieces to implement the branching. Like this:
[A]----------T-------------[B] =I [T-piece at 90 degrees] | | [C]-----=I [T-piece on branch cable] | | [more machines on branch if you like] | =I [terminated T-piece to close branch]
The trick is the first T-piece at 90 degrees, whose lateral ("male") insert is plugged into the "female" socket of the T-piece doing its usual job on the main cable. The ("female") end of the "branch" cable is then put onto the other "male" insert of the first T-piece at 90 degrees. This piece of trickery ensures valid connection (and indeed functions just like the normal connection at the ethernet card).
And you will of course need a terminated T-piece at the far end of the "branch" cable in the usual way.
No doubt this could be extended to a whole tree of coax, though shortage of spare T-pieces (and spare machines to hook onto them to test it) prevents me from testing it.
Apologies if it's not new to you, but since I hadn't heard of it being done before I thought I'd share the idea.
Best wishes to all, Ted.
E-Mail: (Ted Harding) Ted.Harding@nessie.mcc.ac.uk Fax-to-email: +44 (0)870 094 0861 [NB: New number!] Date: 26-Sep-04 Time: 16:25:41 ------------------------------ XFMail ------------------------------
On 26-Sep-04 Jim Jackson wrote:
Ted, did you do any network saturation testing?
Not as such ... see below.
I suspect this configuration will work fine as long as there is not a lot of concurrent traffic between many different machines.
Well, that is basically the situation. I have two desktops, plus a laptop which is what I usually sit at, and it is a home network! (see also below)
The big problem with this setup is whether all stations will see collisions within the 10BaseX parameters - if not then you could get situations where throughput is very poor. However, for only a few stations and very restricted traffic patterns you might have on a home network it might work very well.
Essentially, the two desktops sit in a very small room at the top corner of the house, and an ethernet cable runs between them and from there downstairs, through a ventilation slot into the kitchen, where there's another little annex which I use as an office. The two desktops are each somewhat dedicated: [A] does all the email and archives, and [B] does printing. dialup and gateway and also has a fast processor and lots of RAM so is the machine for heavy stuff. The laptop [C] at the kitchen end is simply convenient, and mobile.
I normally connect the laptop to the kitchen cable, and according to taste migrate the laptop between the little table in the annex and the kitchen table which involves migrating that end of the cable as well. Easy enough, this bit.
However, for various reasons I wanted to also work in the lounge which is on the other side of the stairs. I could have unstapled the cable and fed it back through the slot to divert it there (with an extension), or alternatively patch in a long and awkward loop, but began to ponder alternatives.
Since there was already a T-piece on the cable where it starts down the stairs, simply serving to connect two lengths of coax cable, I thought of using this a pick-off point. So I did what I described: 10m of ethernet cable branching off laterally, and running into the lounge. As I said earlier, this seems to work fine.
The most severe test I've carried out is to run Windows on [B] using VMWare remotely, in an X window on the laptop. This is always demanding on the ethernet connection, since it seems even the smallest change to the Windows screen involves a huge amount of network traffic as it redraws the remote screen.
Even on a direct ethernet (no branches) there is perceptible sluggishness and, frankly, it's much better to go upstairs and sit at [B] when this is needed for extended work.
Doing the same thing on the laptop now on the branch-line, I found it almost as quick, and the most perceptible difference was slightly greater sluggishness with the mouse.
I perceived no difference at all in ordinary Linux apps running on a remote machine ([A] or [B]) in an X window on [C]. I did use this alternative arrangement for an extended period (a couple of hours).
However, the real test would be to connect up a fourth machine so that both branches are under load. But I only have three ...
I run a mixed cat5/Coax network at home but have never dared break the rules so blatently - I'd be very interested in a later report of your experiences.
Interesting Jim
Well, that's my experience so far! I suggest you try it out and see how you get on.
Best wishes, Ted.
-------------------------------------------------------------------- E-Mail: (Ted Harding) Ted.Harding@nessie.mcc.ac.uk Fax-to-email: +44 (0)870 094 0861 [NB: New number!] Date: 26-Sep-04 Time: 23:33:58 ------------------------------ XFMail ------------------------------
On Sun, 2004-09-26 at 16:25, Ted.Harding@nessie.mcc.ac.uk wrote:
Hi Folks,
Your solution will create lots of impeadance mis-matches in the cable...
The cable is 50ohms, and when two pieces are joined together with a T piece, the card driving them sees 25 ohms ( = two 50s in paralell), but it is designed to drive 25 ohm load.
If you connect a third piece of coax to a t-piece, the end of this piece now sees the 25 ohms presented by the other two in parallel (note this applies to all of the three cables). The 25 ohm mismatch will create reflections and will screw up the collision detection and could in exceptional cases prevent communications between some nodes.
Don't Do It
Peter
-
Brett Parker -
Jim Jackson -
Peter Onion -
Ted.Harding＠nessie.mcc.ac.uk -
Tim Green