Artisoft Technical Bulletin 06.03.93 Engineering Position Paper on Ethernet Coaxial Cable There seems to be quite a bit of conflicting information regarding the types, values, and performance of the various type of coaxial cable which are available for use in "thinnet", [formerly called cheapernet] installations. In an attempt to clarify cables and terms, we will cover the basics of cable design and attempt to sort out various claims by cable vendors. Definitions of Cable Terminology Coaxial Cable A cable consisting of two cylindrical conductors with a common axis, separated by a dielectric. Dielectric Any insulating material between two conductors which permits electrostatic attraction and repulsion to take place. Dielectric Constant (K) The ratio of capacitance using the material in question as the dielectric, to the capacitance resulting when the material is replaced by air. Characteristic Impedance The impedance that, when connected to the output terminals of a transmission line of ANY length makes the line appear infinitely long. [Carefully consider this statement - understanding its meaning is the key to transmission line theory.] Decibel (dB) A unit to express differences of power levels. Used to express power gain in amplifiers or power loss in passive circuits or cables. RG/xx Cables Abbreviation for Radio Government, the original World War II military specifications for coax cable. Currently MIL-C17 defines quality of RG coax. If the cable meets current Mil specs, it will usually be marked RG/58 MIL C- 17. Termination A device which looks like pure resistance at the operating frequency of the driving source. This frequency is 10MHz in the case of Ethernet. Terminating a cable with its characteristic impedance causes the terminator to dissipate in the form of heat, all forward energy and causes no reflection of the wave back towards the source. The key here is "at the operating frequency". Velocity of Propagation The speed of an electrical signal down a length of cable compared to the speed of light in a vacuum, expressed as a percentage. It is the reciprocal of the square root of the dielectric constant of the cable insulation. Ethernet Transceiver A device which can "listen" to and "drive" a 10/5 MHz phase modulated signal of approximately 2 volts peak to peak onto the transmission line. The ideal receiver would look like an infinite impedance input. Since this is an impossible task, the distributed capacitance in a good design is kept to 4 P.F. [pico ferrets?] or less. This "bump" in capacitance in the transmission line at each node causes the characteristic impedance to change at that point in the cable. This change will cause some component of wave to be reflected back towards the source as the wave passes this point. When in transmit mode the transceiver has a source impedance of between 20 and 30 ohms. The transmit output stage in the transceiver is a Class A amplifier [stage is always between cut-off and saturation] an this maintains a constant source impedance over the full waveform. The transceiver also detects collisions [more than one transmitter active at a given time on the cable] by detecting a peak signal of over 2.2 volts. Since on a properly terminated cable, a transmitting node will put 2 volts peak to peak on the cable, if we see more than 2 volts it must be due to the mathematical sum of 2 or more transmitters on at once, hence a "collision". Determination of Cable Impedance The impedance of a coaxial cable is determined by the ratio of the INSIDE OF THE OUTER DIAMETER to the OUTSIDE DIAMETER OF THE INNER CONDUCTOR. For example, if we took a 2 inch diameter water pipe and put a 1/2 inch water pipe concentric inside of it, we would have a 50 ohm transmission line. If we used a 2 inch outer pipe and a 3/8 inner pipe it would be a 75 ohm transmission line, and if we used a 2 inch outer and a 1/4 inch inner pipe we would have a 93 ohm line. Cable Design Now that we know some working terms about coaxial cable, lets design one for an Ethernet application. If we use 2 inch copper pipe rather than iron water pipe, as in our example above, we would have less loss due to the better conductivity of copper. Better still, let silverplate the inner surface of the outer pipe and the outer surface of the inner pipe, since silver is about as good a conductor as we can use short of cryogenic type technologies. We would have a great transmission line, in fact, many radio and TV stations use "cable" exactly like our example to connect the transmitter to the antenna. Now we must face the trade-offs which will allow us to practically use a transmission line between nodes in an office environment. First of all, we need it to be flexible, so we have to fill the area between the conductors with a dielectric in order to maintain the center conductor exactly in the center of the outer conductor. We could use Teflon, it is stable, will take high temperatures, and is a low-loss dielectric material. The downside is cost. If we look at RG-142 it has the following specifications. o Solid silver plated center conductor o Teflon dielectric o A double braided silver plated outer conductor o 1/4" outside diameter (same as RG-58) o Cost: Approximately $18.50 per FOOT!!!! A great cable to use if you build spacecraft or F-15s on a cost plus government contract! So, lets look at the family of coaxial cables which use polyethylene as a dielectric and use bare copper or tinned copper wire as conductors. We trade off loss of signal for cost. There is a family of RG cables that are 1/2 inch in diameter such as RG-214 (silver plated conductors - $4.00/foot) and RG-213 (bare copper conductors - $1.20/foot) which would do a fine job for us, but 1/2 inch cable is not practical to run over our desk tops. Too bad, as due to the less loss of the larger cable, we could run longer lengths [thick ethernet] using the same transceivers, before the attenuation of the cable caused loss of signal between the nodes. Next we find RG-223, a nice cable, silver plated center conductor, polyethylene dielectric, double braid jacket [rated 99% coverage]. NOTE: Braided outer conductor will always have some little "windows" between the braid weave where signals can "escape" and other electrical signals can enter the cable. 95% coverage is good, anything below 85% coverage is worthless and only pretends to be coaxial cable. Well, our RG-223 certainly meets our specifications, but still costs $0.85/foot. RG-58 Cable Let's look at RG-58 cabling. Wow, it looks like there are many versions of RG-58, so we read the specifications a little closer. RG-58 TYPE Cable On version of RG-58 TYPE cable we find is Belden 9201, which looks good at first glance: loss is 3.9 dB per 100 meters at 10MHz, but WAIT! The shield coverage is 78%! They "forgot" for cost purposes to put a good outer conductor around our cable. We look further and find RG-58/U JAN-C17A [JAN stands for Joint Army Navy]. What do you know... it looks like the same cable as above but has the copper in the outer conductor to provide 95% coverage and meet the MIL specification. So, we have learned to beware of the word "TYPE" after a RG cable designation. I drive a car with Porsche "TYPE" tires, unfortunately it has a 1980 Honda Civic body. 50 ohm Impedance The above described cable [RG-58/U JAN-C17A] has 53.5 ohm impedance! What does this really mean? First of all, the actual impedance of "production" coaxial cable is + or - 5 ohms. This is due to the extrusion process which determines the thickness of the dielectric (spacing between inner and outer conductor). Since we have already determined that we want to terminate the cable in its nominal impedance we have to design a terminator which looks like 50 ohms at 10MHz. To get a pure resistance at that frequency, we need a resistor which does not have leads. A laboratory type 50 ohm terminator contains a circular deposit of carbon which connects the center conductor of a BNC connector to the shell (outer conductor) with minimal induced series inductance. It also costs $55.00 [each!]. So again, due to cost, we fall back to a leaded resistor soldered into the end of a BNC connector (looks like anything between 40 to 60 ohms at 10MHz, but costs less than $1.00. Let's say we do have a perfect 50 ohm terminator and a perfect 53.5 ohm cable. The VSWR [standing wave ratio] will be 1.07 to 1 and this value will NOT cause a problem on our transmission line. Belden lists their RG58 as 53.5 ohms but Alpha Wire, another high quality cable manufacturer lists their RG-58 as 50 ohms. Cable Manufacturers Many vendor's advertising will try to sell us cable "especially made for Ethernet". One vendor, for example. has a Technical Bulletin which promotes one of their best cables. First they tell us that "generic RG-58 TYPE cables are sometimes used for thinnet. These cables do not conform to all thinnet specifications and may place the entire network in jeopardy of performance reduction or system failure". Wow, scary stuff, but we are smarter now and realize the reference to RG-58 TYPE, and we know where they are coming from as we have already seen that they manufacturer a cheap "TYPE" version of RG-58 with only 78% outer braid coverage. The rest of the improvement claims in this bulletin are always referenced to RG-58 TYPE cable, never to RG-58 MIL C17. The next page of the bulletin explains attenuation to us. They tell us that "Lower attenuation allows for longer cable runs, maximum network size and less distortion". Quite a true statement, BUT when we check the column on the data sheet the cable in question, we find at 10MHz, 100 meters of cable has 4.3 dB of attenuation. Then if we flip over to RG-58 MIL C17, we find at 10 MHz the attenuation is 3.9 dB per 100 meters. Therefore, the MIL C17 is still the best! Next comes a discussion of "Velocity of Propagation". The description is fine, but with a 185 meter cable, the time of propagation from one end of the cable to the other is about 550 nanoseconds (billionths of a second) on .66% velocity of propagation and 490 nanoseconds on .80% velocity cable. Even assuming all the packets of Ethernet data were transferred from nodes at the extreme ends of the cable, we would have to transfer a billion divided by 60 packets before we could save a second of time. NOT significant! Next they tell us that higher velocity of propagation reduces attenuation. We should know not to believe this as we know that at 10 MHz [the frequency of Ethernet data] their own RG-58 MIL C17 at .66% propagation has less loss [attenuation] than the discussed cable at 80%. Next they mention "round trip propagation time". REALLY! Remember our earlier termination explanation? Once the signal passes the last node and arrives at the terminator, the terminator dissipates the energy and ideally NO signal ever is reflected back towards the source. It should look to the source as an infinite long transmission line, where the signal leaves, keeps going, and is never seen or heard from again! There is NO round-trip! Is this good cable - sure it is. But the above discussion is misleading. Real World Problems Let's turn to the real world and take a look at cable problems and their causes. The Number One Problem The number one problem in the real world is the BNC connector on the end of the cable. They come in many shapes and sizes, but can be divided into two classes - MIL UG-88 type and crimp on. MIL UG-88 Type Connectors There are 6 parts to this type of connector: 1] A center pin which is soldered on 2] An outer case 3] A rubber gasket [compressor] 4] A braid retaining collar 5] A slip washer 6] A center collar which holds the whole thing together When properly assembled, this connector is waterproof and will take a 40Lb pull without falling apart. We have seen this type of connector assembled 36 different ways, 35 of which short out or fall apart when a 5 Lb pull is applied! If not assembled correctly, a fine strand of braid will touch the center conductor when the connector is wiggled. Properly assembled with correct removal of the outer jacket [don't nick the outer strands, as they provide the strength to the unit], soldering the center pin [the polyethylene melts and lets the center conductor move off- center and touch the braid if too much heat is applied] and proper torque on the center collar when assembling the unit, this type plug will last a life- time. Crimp-On Connectors These types of connectors have good and bad points. Since connecting the center pin is usually the hardest part, some designs have the bare center conductor merely stick into a hollow pin, depending on pure mechanical contact to provide continuity of the connection. If it were not for humidity and corrosive things in our air, this might work on a long term basis. Unfortunately after a few weeks/months the surface of the copper turns a bit green and now we have one of the intermittent cables that starts working/not working when we wiggle it. The plug manufacturer gives us a collar with which we are to crimp the braid to the connector. Unfortunately, each different manufacturer has a slightly different size collar, and will sell you the proper crimping tool for 75 to 100 dollars. Using the correct crimping tool for the size collar does lead to a good connection. The problem appears in the form of a $4.95 "universal crimper". This tool can be depended upon to do a mechanically poor job of forming the collar onto the connector. Strands of the braid get torn off the cable or the dielectric gets smashed into the center conductor. So to sum up, if [on only if] all manufacturer procedures are followed and the correct tool is used to crimp the connection, a reliable connection can be obtained using "crimp-on" type BNC connectors. Other Problems Let's discuss some additional terms to better understand other factors that affect our Ethernet cable. Polyethylene A family of insulations derived from the polymerization of ethylene gas and characterized by outstanding electrical properties including high insulation resistance, low dielectric constant, and low dielectric loss across the frequency spectrum. Mechanically rugged, it resists abrasion and cold flow and is used as the dielectric material between the center conductor and the shield. Cold Flow Deformation of the insulation due to mechanical force or pressure [not due to heat softening] Polyvinylchloride A general purpose family of insulations whose basic constituent is polyvinylchloride or it's copolymer with vinyl acetate. Plasticizer, stabilizers, pigments, and fillers, are added in lesser quantity to improve mechanical and/or electrical properties of this martial. Used in the outer jacket of coaxial cables. Non-Contaminating PVC A polyvinylchloride formulation which does not cause migration of the plasticizer through the braid and into the dielectric. This contamination [chemical reaction] causes an increase in cable loss that may equal two to four times the original cable. A 100 foot length of this affected cable may show losses at 10 MHz of 12 dB rather than its original 4dB. Plasticizer Migration All flexible coaxial cables have a finite lifetime. In the case of Teflon dielectric cables this value may be 20 years or more. In the case of a very flexible cable [the manufacturer has used a PVC outer jacket with lots of plasticizers, which if exposed to heat or ultraviolet radiation, will migrate and contaminate the dielectric] this life may be measured in months. We have seen cable taken from walls with 2 foot sections which were stiff [not flexible as other parts of the cable]. This condition was traced to the fact that the cable [while installed] passed over or alongside hot air ducts. This is a classic case of plasticizers migrating, leaving the outer jacket stiff. When measured for loss, this cable showed a marked increase in signal attenuation. This property of coaxial cables is very important and yet it is brushed over by cable data sheets. Cold Flow Problems The next important failure mode we see in coaxial cables is due to cold flow of the dielectric. Picture if you will a cable run around a sharp corner [happens all the time when we pull cables through walls and over ceilings]. This stress will try to compress the outer jacket into the dielectric and try to reduce the distance between the center conductor and the outer conductor. We know by definition that this distance determines the impedance of the cable, so what happens is we end up with a 50 ohm cable which as 20 ohm points in it due to deforming of the interior shape. This problem is very noticeable in foam dielectric cables. Over time the stress point will cause the dielectric to cold flow around the center conductor so that removing the cable and rerouting it will not remove the damage done. This condition would seem to be a very strong point for Belden 9907 type cable, as it has a mini-rigid outer jacket and would greatly help in preventing this type of problem. We have pieces of RG-58 TYPE foam dielectric cable which when viewed on an oscilloscope, in the Time-Domain Reflectometer mode, will show "bumps" of impedance mismatched where it has been "folded, spindled, and mutilated". We have another piece which shows periodic small reflections at time intervals of approximately 1 meter. A visual inspection of this cable shows compression points about 1/4 inch wide where at one time in its life someone had used plastic cable ties to attach the cable to something along its length. Moisture Penetration Another detriment to cable life is moisture penetration of the area between the dielectric and the outer braid. Moisture will migrate between the braid as if the braid were a wick. This is a calculated risk if we use barrel or T BNC connectors to splice cable and the connection is between the walls or a ceiling where moisture is present. The fine copper wires of the braid will corrode and ultimately crumble apart, inducing an "open" in the outer shield. This is a very good reason to NEVER use a piece of cable that has damage in its PVC outer cover which exposes the braid area to the atmosphere. Conclusions 1] Know the quality of your coax cable. If it is not MIL C17F, JAN C17A or IEEE802.3 compliant then the manufacturer's data sheet must be examined to determine the quality of the cable. The term RG-58 TYPE is meaningless. 2] Insure the BNC connectors are installed per the manufacturer's data sheet with the proper tools to insure a good mechanical, electrical and waterproof connection. 3] If installation environment requirements [metal ducts with sharp corners, parallel runs with A.C. power cables, etc] exist, the cost of the grade of cable equal to Belden 9907 may be justified. This is to insure reliability of the network over time and not to improve performance. Artisoft, Inc. Technical Support Department 2202 N. Forbes Blvd Tucson, AZ 85745