Debugging 10Base2
10Base2 ('coax') ethernet is a bus-type network. Unlike 10BaseT and 100BaseTX, where a single wire (segment) connects only two devices, one of which must be a hub or switch if more than two machines are to be networked, a 10Base2 segment can have many machines on it. When a machine transmits, the signal travels down the wire in both directions, and can be received by any machine on the segment. Unless the signal is grounded by a termination resistor on both ends of the segment, it will bounce off and interfere with other signals on the wire. Should the wire break anywhere in the segment, it will create two separate segments, both of which are missing a terminator on one end; a broken wire doesn't just prevent communication across the break, it renders the entire segment unusable. Such faults are difficult to diagnose systematically, as the entire network doesn't work unless the break is mended. A naive method of diagnosing a faulty 10Base2 segment is to work along the segment, replacing wires and T-pieces until the segment starts working again. There is a far better way to do this, however.
Shopping list
To diagnose the fault systematically, you will need a multimeter, and a BNC adaptor for it, or a commercial BNC tester.
It's pretty easy to make your own BNC adaptor, by soldering a pair of banana plugs onto one end of a 10Base2 lead (ensuring to connect the outside to the common/ground terminal, and the inside to the signal/live). You need either a male or female BNC adaptor, not both - a T-piece or cable converts one to the other, though make sure you test them first!.
It's possible to get by without a BNC adaptor and simply use probes, but it requires a steady hand (and the ability to use chopsticks can't hurt). To connect a multimeter to BNC, connect the common/ground terminal of the multimeter to the outside of the BNC plug, and the signal/live terminal of the multimeter to the inside of the plug (on the t-piece it is a hole, on a cable it is a spike).
A commercial BNC network tester does the same thing as a multimeter, except that it gives a reading of 'OK', 'short' and 'open', instead of a resistance.
Procedure:
If possible, shut down all the machines, or at least unplug the 'T-piece' from the back of all of them.
Start with the terminators. At each end of the segment, unplug the terminator, and measure both its resistance, and the resistance of the segment with only the far terminator connected. If either of the terminators measure other than 50 ohms, they are defective and should be replaced before starting the procedure again. If the terminators check out, but the wire measures something wildly different from 50 ohms with only the far terminator connected, there is a fault somewhere on the wire.
Starting in the middle of the broken segment, unplug the wire from one end of the T-piece, and connect the multimeter in its place. Set it to measure resistance, in the '0-1000 ohm' range. There are three possible readings:
- 50 ohms: This is the termination resistance. Current from the multimeter is travelling from the t-piece, through the (50 ohm) terminator, and back again without being shorted to ground. It is quite likely that the fault is not located in this part of the segment.
- 0-10 ohms: The signal wire is likely to be shorted to ground somewhere. A dodgy or damaged cable can do this, as can a damaged terminator.
- 1000 (or more) ohms: There is an open circuit - either the terminator is broken, or there is a break in the cable.
If the reading was anything but 50 ohms, you've narrowed down the location of the fault to the section you're testing. Briefly check the other side, in case there are two faults, and (assuming the other side checks out) replace the working segment with a
terminator, and repeat the testing procedure on a t-piece halfway along the broken side.
After a maximum of five tests, you should have been able to narrow the location of the fault down to a single broken wire, or a single broken t-piece. Wasn't that easier than replacing (up to thirty) wires until the segment wakes up again?
There are situations where this technique does not work - it will not detect reflections caused by a spur on the wire (where a long-enough 'T section' causes the same reflection behaviour as a missing terminator), nor will it detect if one of the computers' network cards is shorting the segment. It is possible (but more time-consuming) to detect such fault by the same 'divide and conquer' technique with the computers connected and switched on, using TCP/IP or ethernet ping between two computers on the subsegment instead of a multimeter or tester to see if it is working or not.
Armed with this technique, you can also debug christmas fairy lights without having to replace every single bulb - simply connect the common terminal of the multimeter to one pole of the mains plug (after you've unplugged it!), and use the signal terminal to probe the bulb sockets until you find an infinite resistance. In the UK, either check the fuse in the plug first, or connect the multimeter to the neutral pin of the plug.