(1 / 1)
Date: May 06, 1992 19:13
From: GAWD::SHEPPERD
To: @SYS$MAIL:HARDWARE
CC: SHEPPERD
This is a quick lesson about how to hookup ethernet cable and how NOT to hookup ethernet cable. First a little background so you can attach some reason to the madness: The cable is a transmission line and therefore abids by the rules established with transmission line theory. Transmission line theory says that, in an ideal case, if you were to connect a device at any point along the transmission line, it would appear to the device that the wire was infinitely long in either direction and that the wire would have a uniform characteristic impedance in the frequency range that the device intended to operate. In the real world, the ideal case is rarely achieved. Any little thing will likely change the impedance of the wire at that point. These include tight kinks, connectors, poor connections and "stubs". What happens when a signal runs across one of these changes in impedance is that some amount of it bounces back to the point of origin (and this bounce point becomes the point of origin for the reflected signal). This is also known as "ringing" and I'm sure you've all had plently of (bad) experience with dealing with that on circuit boards. The amount of signal reflected depends on how severe the impedance mismatch is. If it's severe enough (as is would be with a dead short or an open circuit), almost all of the signal will be reflected. In long lines transmission, a "ring" or bounce may take 100's of nanoseconds make the round trip, so for circuits that are expecting to deal with bit rates of 100ns, (as is ethernet), this can be rather nasty. Now for the meat of the issue: The standard established for ethernet cabling was designed such that inconsistencies in the cable (impedance mismatches) be minimized, yet at the same time allow for connectors and things so one can actually use the damn thing. To this end, the following rules must be adhered to when you find yourself needing to move equipment connected to the ethernet (if in doubt, don't mess with it): 1) The wiring topology ALWAYS must conform to: Terminator -----------------terminator (The "---" indicates ethernet cable). In our case, one of the terminated ends is in the computer room, so you don't need to worry about that one. If either of the terminators is removed or the wire is disconnected such that an end becomes unterminated, the segment will cease to function. This may also cause other segments (i.e. the entire network in the building) to stop working for a while. Please don't do this. 2) Taps are allowed in the cable, but no more that 29 (except for those fed from a twisted-pair, which are confined to 10) AND the tap must connect directly to the device. The taps we use are BNC T connectors and are known as invasive taps because they require the wire to be cut in two in order for them to be installed. I.e. Terminator ------T---...----T----terminator d d v v c c NEVER hookup a device like this: Terminator ------T----terminator | | | d v c The wire connecting the device to the ethernet cable is known as a "stub" which can cause an impedance mismatch of the worst kind (it can become a resonant circuit, depending on how long it is). Just say no to stubs. 3) Always assume there are other users on your segment. This is almost always true in our environment. "Party lines" are used throughout the building. There are maximum segment length limitations which you probably won't run into if you are simply rearranging a lab, but you may if you are moving to some other area of the building. We have some tools to determine and help with these things. ds
May 06, 1992