Ethernet Networking Probs

[GearOff]

Hi folks,

This is a problem that has plagued me for some time now, but because it's not really a huge inconvenience I've just let it go.

Some time ago, I setup an Ethernet network at home, mostly for shared internet access, but since then the benefits of file transfer etc have been immeasurable and we couldn't live without it. The network is relatively small, with 4 machines on it, connected via an 8 port 10/100 Mototech switch. Cat-5 cabling is used throughout (RJ-45), and all the ethernet cards are 10/100.

The problem is that I am unable to run the network at 100Mb -- it's perfect at 10, but at 100 packet loss is unnacceptably high to the point where it does not work at all.

It isn't the switch, as I have replaced it (different brand etc) and it made no difference, it doesn't appear to be any one length of cable, as all the nodes are equally incapable of this speed, and it isn't the ethernet cards because we have 3 different types. I've tested it using two Intel cards in IBM Thinkpads which, given their price, would be completely capable of operating at that speed.

All machines run Windows 98 SE and TCP/IP is the protocol.

Any help would be much appreciated! I've never been quite satisfied with myself for letting this one get away from me :-)

Thanks in advance

------------------

[matelot]

GearOff - What packet size can you send reliably?

10/100 are the extremes, and I believe 100MB to be a pie-in-the-sky figure with perfect conditions.

Have you checked attenuation? Who made up the Cat 5 cables, or were they bought in?

Whgat about download speeds?

Just a few ideas.

------------------
Me, sweat? I'm that cool, it's condensation.

[edgewing]

I would be seriously looking at the cable. You shouldn't be getting packet loss over a relatively short run in your home. Either that or one of your ethernet cards might be a bit crook and busy shouting over the network?

LOL

[GearOff]

Thanks for your responses.

When set at 10Mbps I can send any packet size up to the maximum with 100% reliability. If I select 100Mbps line speed, regardless of packet size, the reliability is unpredictable and too low to make the network useable.

While running several ping tests under the same conditions with 16 byte packets, the results varied from 16% loss up to 80% loss.

The RJ-45 plugs were fitted by me (which is a worry) using the proper crimping tool etc. The actual cable runs were bought in. The writing on the side says it is Austel approved, verified Cat-5.

Is there a possibility that attaching the RJ-45 plugs isn't as easy as it appears, and that I may not be getting a good enough connection at each of the cable ends?

Also, is there a proven method for testing attenuation over a run?

I've fairly much ruled out the cards as being the problem -- for testing, I disconnect all but two of the nodes and I've tried various combinations. I figured testing it with the two Intel cards in the Thinkpads would be a better indication than the cheaper cards in the desktop PC's, but it is just as unreliable.

[ExSimGuy]

When you made up the cables, did you just connect the first twisted pair ro the first pair of pins, second to second etc? Or did you follow the correct sequence? Having the wrong wires twisted together could DEFINITELY cause serious problems at 100Mb which would not be there at 10!

------------------
---- "Per Ardua ad Mixas" ----
(Through hardship to the bars)

[GearOff]

From what I remember of wiring it up (it was some time ago now), I was under the impression that as long as the order of the wires was the same at both ends of the cable, it didn't matter which actual wires were associated with each pin. Is this not the case?

[AVPIN]

I think that ExSimGuy has it chaps

Gearoff - if you have a spare length of cable, strip back the insulation for about 100mm - you will see that the wires are twisted in discreet pairs. This is not obvious if you only strip enough for the plug termination.
The wire 'pairs' must be wired correctly - each pair forms a balanced transmission line to avoid cross-talk and noise pick-up.
If the pairs are mixed - cross talk is created - worse at high bandwidth.
It is impossible to check 'pairing' by doing a simple continuity check.

Good luck

AVPIN

[matelot]

GearOff

In theory, it would seem logical to make sure the same wires are at each end, irrespective of order. But these are TWISTED PAIRS to minimise magnetic fields, attenuation etc.

Conventional order is:

orange/white
orange
green/white
blue
blue/white
green
brown/white
brown.

Crimp damned tight after being absolutely sure they've gone in the plug in the correct order (good eyes needed!)

You can test attenuation and reliability of connection, but the kit ain't cheap! Get it right at home and it'll be right.

:)

------------------
Me, sweat? I'm that cool, it's condensation.

[R O Tiree]

ExSimGuy and AVPIN are definitely on the right track, I would say. Here's the theory:

OK, why pairing? Send a modulated signal up a long wire at 10 Mbps, even at a couple of volts and you generate a detectable electro-magnetic field. This will be picked up in the adjacent wires and a signal can be detected on those wires. The higher the frequency, the higher the strength of the field, which explains why this problem manifests itself at high net speeds. The way these networks work is as follows: a packet gets sent from one computer to another on the network. Think of the packet as a train. A train has an engine, some carriages and a caboose. Imagine that the IP address of the computer you are sending to is printed on the front of the train. Underneath that is printed the return address (ie the "from" computer"). The carriages are the data and the caboose is there to denote the end of the packet. Off it goes, round the network, and it visits each computer in turn. A computer sees the train approaching and reads the number off the front. If the computer address and train address do not match, the packet is ignored and allowed to pass through the card unmolested. If the address does match, the computer effectively reads the data (carriages) from the packet, reverses the addresses and sends the train on its way again. The "from" computer intercepts this train (because its address is now uppermost on the front), realises that the data has been received correctly, and it can now send another packet. From this you will see that there are "up" and "down" lines with almost identical signals flashing up and down the cable. If you twist pairs of wires together, the EMF in one wire is exactly counteracted by the EMF generated in its twin. Only 4 of the pins out of 8 in an RJ45 plug are used, so you've got to get the right ones.

I can't remember which ones they are, offhand, but this order always worked for me: hold the plug with the "open" end (the end the cable goes into) facing your chest, gold pins uppermost. Mentally label the pins 1 to 8, left to right. The standard cable we used had 4 twisted pairs as follows:

Orange + White/Orange (O-W/O)
Green + White/Green (G-W/G)
Blue + White/Blue (Bl-W/Bl)
Brown + White/Brown (Br-W/Br)

Insert the wires in the following order:

Pin Colour
--- ------
1 W/O
2 O
3 W/G
4 Bl
5 W/Bl
6 G
7 W/Br
8 Br

I know it doesn't look right at pins 3, 4, 5 and 6, but that is down to the internal wiring of the card. That's the way it is. If The colours of the wires in your cable don't match those above, find out, as AVPIN said, which cables are paired, then map those colour pairs to the table above. For example, if yellow and red are paired, map yellow to pin 1, red to pin 2, etc, etc, but you've got to get that "twist" at 3, 4, 5 and 6 right.

It is possible to tell how good your cables are with a tester. You may be able to hire one, or borrow one from your friendly local computer retailer/network guru. They are quite expensive, so be prepared to offer a check as deposit. He might allow you to bring your leads in for testing. On the other hand he might "generously" allow you to buy some of his fine wares at a grossly inflated price!! :)

[GearOff]

Gents, I think you've done it <G>

I'll get a chance to rewire a cable sometime over the next couple of days, and I'll let you all know how it goes. The significance of them being twisted pairs never occured to me, but what you're saying certainly makes sense.

At any rate, thank you all very much for your help, I'll hopefully get back to you with my tale of success over the next couple of days!

Thanks also for your incredibly detailed reply, Tiree, now I can finally say that I understand what actually goes on 'under the hood'. http://www.pprune.org/ubb/NonCGI/cool.gif

And they call this easy <G>

[This message has been edited by GearOff (edited 07 November 2000).]

[R O Tiree]

matelot - Snap :) :) We obviously read the same book!

[R O Tiree]

Gearoff - You will need to replace all the patch cables. If you only re-wire one of them, the data will get to the hub OK, but will get garbled as it travels up and down the next cable in sequence.

[matelot]

R O Tiree

Check, mate - I'm wiring up several of these over the next few days. Pain in the proverbial! :) :)

------------------
Me, sweat? I'm that cool, it's condensation.

[GearOff]

It worked. :)

Thanks again everyone that helped - it's very much appreciated. That'll teach me to RTFM next time!