Bitzenbytes.com

Ethernet, Round 3

Thinwire Ethernet

Thinwire coax gets its name by comparison to the other coaxial cable used for Ethernet which is, therefore, called Thickwire.  While RG-8 coaxial cable was surely specified to meet a demand, it's cumbersome nature probably led to research on how much you could "get away with" while using something lighter and less cumbersome.  The something lighter and less cumbersome is RG-58 A/U coaxial cable.  The construction of RG-58 A/U is similar to RG-8, albeit the diameter is smaller and the cable jacket is considerably thinner by comparison. 
 

I installed a lot of this stuff and it didn't take me long to discover that I could save a lot of time and money by NOT installing BNC cable ends on cable segments myself--much easier to call up the local custom cable making company and give them a list of the lengths that I needed.  BNC connector cable ends required a force fit, and I must admit to struggling with most of them.  For a network other than my own, pre-made cable made economic sense as well, unless it was a single cable length for a new workstation location.
 

Various manufacturers may call RG-58 A/U or equivalent by different nomenclature.   Beldon 9907 is a PVC clad coaxial cable, and Beldon 89907 is teflon clad coaxial cable for use in ceilings that also serve as air plenums.  All of these are considered to be "thinwire".  Any wire acceptable for an Ethernet installation should be marked IEEE 802.3 on the outer cable jacket.
 

BNC Tees and 50 ohm Terminators
  

Thinwire Ethernet supports a 10 MHz baseband signal on  RG-58 A/U coax (0.25 inches dia. nominal),  support segments up to 185 meters (606.8 feet or 607 feet) in length, and is referred to as 10Base-2. 
 

The "2" in the designation is a truncated 200, and the 200 comes from rounding up from 185 meters. Thin coax looks like the copper coaxial cabling that's often used to connect a VCR to a TV set, more or less.
 

Television installations typically use 75 ohm RG-59 media and the center conductor generally floats in a plastic insulated air space.  Note that RG-6 is an improved version of RG-59 and it is generally employed where channels in the higher frequencies (70-125) will be used. These coaxial cables cannot be used in place of RG-58 A/U.  The coax cabling used for ARCnet (92 ohm RG-62) cannot be used for 10Base-2 Ethernet either.
 

RG-58 A/U coaxial cable used for thinwire Ethernet, on the other hand, has thick foam insulation around the center conductor, and it is terminated by a 50 ohm resistor at each end.  The braided metal shield helps to block any outside interference from fluorescent lights, motors, and other computers. Stripping coax cable requires a special tool because each of the layers of the coax cable must be cut back to a different length.   A good connection requires use of the special stripper tool.  Likewise, the only way you'll properly crimp the male or female connectors on the center conductor is with a crimping tool made for this particular purpose.  Compared to "ThickWire", RG-58 A/U, along with being less expensive, is extremely flexible and is relatively easy to fish through stud-walls.  I'd never attempt to fish RG-8 through stud-walls.  One of these terminators (and only one) should be grounded.

A BNC connector is used to make the connection between the elements in a 10Base-2 network.  Not only is RG-58 A/U coaxial cable lighter and more flexible than RG-8 coaxial cable, but BNC connectors are a lot easier to connect than type N connectors.

The argument, over what the "BNC" in "BNC connector" means, will go on forever.  It has been variously defined as: British Navy Connector, British Naval Connector, Bulk Navy Connector, Bayonet Node Connector, Baby Neil Connector, Baby-N Connector, and more. My vote, for what "BNC" means, can be seen under the illustration shown here, and is based on the information that the two Amphenol engineers who invented the BNC connector were named Paul Neil and Carl Concelman.  The type N connector, mentioned in the previous article about ThickWire Ethernet was invented by Paul Neil.  But local culture will probably dictate what gets accepted as the correct nomenclature, right or wrong.
 

Commonly, BNC connector ends (always a plug on a cable segment) are crimped on with a special crimping tool, but there have also been designs that only require common tools likely to be available anywhere.
 

Mechanically, there are a few rules to follow when attaching RG-58 A/U coaxial cable. Each Network Interface Card (NIC) has a BNC connector. While you can attach RG-58 coaxial cable (coax) directly to that, this practice would be wrong.   If you did this, the network would be inefficient, if it worked at all.  The only thing that attaches to the BNC connector on a NIC is a "tee".  Nodes on a thinwire segment (of which there can only be 30) are "daisy-chained".   RG-58 A/U coax cable attaches between nodes directly to the tee on the NIC.  At each end of a daisy chain, there must be a terminator and one of them (only one) should be grounded. Grounding can be accomplished with a special terminator that has a pigtail or chain attached, or a ground clamp made for the purpose can be used.
 

The photo below shows typical components in a 10Base-2 network.  Some variation of what you see is how most 10Base-2 nodes are connected.

Daisy Chain

In a perfect world, when wiring with wall-plates, if the workstation is going to be within 3 meters of the wall-plate, subtract 6 meters, for each wall-plate, from the total length of the allowable run.  If you want to be realistic, subtract 12 meters.   This adds up faster than you might think.  Each wall-plate and station, meaning 2 barrel connectors and a TEE counts as three attachments to a segment.  By the rules, this rapidly reduces the number of stations that can be connected to each segment.
 

Each network card requires a BNC "Tee" connector.  One rule for BNC TEE connectors is that there can be no cable between the TEE and the NIC.   This is to make the connection between the "tap" and the transceiver conform to being less than the maximum allowed. Essentially, the TEE is the tap for thinwire; the transceiver is on the NIC.  For most thin-wire installations, the transceiver is built into the NIC.
 

The distance between the tap and the transceiver cannot be more than a certain amount.   This is why, for thicknet, the transceiver is as close to the transmission medium as possible.  There is no difference for thinwire installations.  Because the transceiver is on the NIC, the cable medium must be brought in proximity to the NIC.   For thickwire installations, the AUI cable conveys data between the network interface at the computer and the transceiver at the backbone media. 
 

AUI cable is not specified for thinnet installations.  Nevertheless, there are hosts still out there for which only an AUI interface is available.  There is no equivalent to the AUI cable for thinnet unless the host has a NIC with an AUI connector and an outboard thinwire transceiver is employed.  
 

For outboard thinwire AUI-to-thinnet transceivers, there may be a switch to enable or disable heartbeat (aka collision presence test, aka CPT, aka signal quality error, aka SQE.)  Before the IEEE 802.3 Ethernet standard (1985),  the Collision Presence Test (CPT) was the standard heartbeat for Ethernet II (1982).  After 802.3, the similar SQE TEST becomes the standard heartbeat.
 

My response to SQE is to specify a network interface card that does not require a "so called" heartbeat.  Turning on SQE has been known to create network transmission problems.  SQE is not needed for Ethernet transmission.  The SQE test signal is transmitted during the Interframe Gap. The function of an SQE test issued in the interframe gap is to confirm that the Network Interface Card would have seen a collision had a collision been detected by the transceiver.  The test is now considered to be unnecessarily redundant.
 

At each end of a coax run, there must be a 50 ohm terminating resistor.  Two PC's sitting right next to each other might only have 1 piece of coax running between them if they are the only two PC's on this network.  A finished office installation might employ wall plates for distribution.  No hubs are required.   This method of "stringing-on" one station after another is called "daisy-chaining."
 

Wall plates, essentially, employ butt (or barrel) connectors to extend the cable run, albeit there are some wall-plates with a special connector that must be crimped on.  While not a generally accepted practice, a barrel connector can be used to extend a cable length with a second length. As in most cases, there's a right way and a wrong way to do this.   In the illustration below, the "right" way would be on the right.
 

Two BNC Barrel Connectors:
The sinister barrel connector, not recommended, is on the left.

I've managed to make my own connections, but never achieved any notable speed when making them.  And No; the barrel connector in the photo above was not made by me.   It was my privilege to replace it with the one on the right.  After I made a floor plan, I, generally, called a cable shop and gave them my length specs and let them install the cable ends--they have equipment that makes my futzing with the crimper look downright foolish.
 

Each connecting element on a thin-wire run may add to attenuation.  Coax must not be kinked or made to negotiate tight radius turns--bends in the wire weaken (introduce attenuation) the signal.  At the Tee, the cable segment is always extended with a second cable.  Only a terminated "Tee" ends the cable segment.
 

There can be no extension between the TEE and the NIC.  The TEE must be connected directly to the NIC.

A Multiport Repeater is a hub-like device that can be employed for use with RG-58 coax cable; this device was the original device referred to as a "backbone in a box."   Logically, the eight thinwire segments are attached to a "backbone" via receivers.  This allows the bus topology to be configured as a star.  An AUI connector would allow this unit to be connected to a transceiver on a thick-wire backbone.
 

For the device illustrated, thinwire cable segments could be connected to each of eight ports.  A terminator was attached to any unused port.
 

It was originally DEC's intent, with the DEMPR example of the multiport repeater, that one workstation would be at the end of each DEMPR port.  This was expected to reduce the possibility of a "broken" daisy-chain.  Many is the 10Base-2 network that was brought down because somebody moved their office and took their workstation, previously in the middle of a daisy-chain, with them, along with the two cable segments used to connect the computer to the wall plate.  A barrel connector had to be inserted in the chain until the next user arrived in a recently vacated office.   While DEC's intention attempted to address the daisy-chain failure point problem, administrators quickly found that a centrally located DEMPR also allowed an efficient method of optimizing daisy-chain cable segments on the same subnet.  At least, this helped isolate the potential problem to a smaller segment.
 

Multiport repeaters came with their own rules.  Some had an internal terminator for each port and if no cable was attached, would recognize that and provide termination.  Others required physical termination at each port.  Each cable attached to a MultiPort repeater can be 185 meters long and each segment can have 29 stations attached, noting that any barrel connector inserted counts as a station.
 

No network segment, attached to a multiport repeater can be grounded; grounding for a segment is provided, internally, at the multiport repeater.
 

Normally, multiport repeaters would not be cascaded.  When cascaded, not more than two multiport repeaters could be placed between two communicating stations, and the cascade could not be attached to a thickwire backbone.
  

The cost of multiport repeaters essentially demanded that you have a lot of work stations daisy-chained to each port.  It was the only way to justify the purchase of a multiport repeater.  Even at that cost, when 10Base-T first arrived, a multiport repeater with many daisy-chained stations was still cheaper than the installation of individual 10Base-T ports for each station.  Eventually, of course, as the cost of installation for 10Base-T ports dropped dramatically, this was no longer true.
 

10Base-2 utilizes the same Manchester encoding used in 10Base-5 networks.   Ethernet design rules ensure that the transmission characteristics of the signal are maintained throughout the entire network.

More about Ethernet in Round 4