Transmission Media and Network Devices

A transmission medium is the physical path along which a signal travels from the sender to the receiver. Everything else in this unit — protocols, devices, addresses — rides on top of the medium you choose, and a wrong choice cannot be fixed later by better software.

Media are of two kinds:

• Guided (wired) media — the signal is confined inside a solid conductor and guided along it: twisted pair cable, co-axial cable, optical fibre cable.
• Unguided (wireless) media — the signal is radiated into open space and spreads out: radio waves, microwaves, infrared waves.

Read those words carefully. Wireless communication is not communication without a medium — air, and in the case of satellites empty space, is the medium. It is called unguided because nothing forces the signal down one particular path.

The five things you always compare. Almost every question in this chapter is one of these five in disguise, so learn the axes and not just the cables.

Three terms students mix up. Bandwidth is strictly a range of frequencies, measured in hertz — a channel running from 1 MHz to 5 MHz has a bandwidth of 4 MHz. Data rate is the number of bits actually delivered per second, measured in bps. More bandwidth allows a higher data rate, but they are not the same quantity. Attenuation is the weakening of a signal with distance, measured in decibels (dB); it is the single reason repeaters exist.

The units trap. Link speeds are quoted in bits per second; file sizes are quoted in bytes. Since 1 byte = 8 bits, an 8 Mbps connection delivers roughly 1 MB every second. Forgetting the factor of 8 is the most common arithmetic mistake in this unit.

Worked example. A school in Nagpur has an 8 Mbps broadband line and wants to download a 900 MB software package. Its computer lab is 250 m of cable route away from the server room. Find (a) the download time and (b) whether one twisted-pair run can reach the lab.

(a) Convert the file into bits first: 900 MB × 8 = 7200 megabits. Time = 7200 Mb ÷ 8 Mbps = 900 seconds = 15 minutes. If you skip the ×8 you get 112.5 seconds — wrong by exactly a factor of eight, which is what the examiner is checking.

(b) A twisted-pair Ethernet segment is limited to 100 m by attenuation. 250 m cannot be a single run. Split as 100 m + 100 m + 50 m, the signal must be regenerated at the 100 m and 200 m marks — that is two repeaters (in practice, two switches in corridor cabinets). The alternative is one uninterrupted optical fibre run, for which 250 m is trivial. That trade-off — cheap copper with a short leash, versus costly glass with none — is what the next two sections are about.

Guided media are either copper or glass. Copper carries electricity; glass carries light. That one sentence predicts nearly every difference between them.

1. Twisted pair cable. Two insulated copper wires twisted around each other in a regular spiral, with several such pairs bundled inside one plastic sheath. A standard Ethernet cable contains four pairs and ends in an RJ-45 plug.

Why twist them at all? A current in a wire creates a magnetic field, and a wire lying next to it picks that field up — this unwanted coupling is called crosstalk. Outside noise from a motor or a power cable induces a voltage in the same way. When the two wires of a pair are twisted, each wire spends half the run closer to the noise source and half the run further from it, so both wires pick up almost exactly the same noise. The receiver reads only the difference between the two wires, and equal noise on both cancels out. More twists per metre means better cancellation, which is what the category numbers measure.

• UTP (Unshielded Twisted Pair) — no metal shield. Cheap, thin, flexible, easy to crimp. The default for LAN cabling everywhere.
• STP (Shielded Twisted Pair) — a metal foil or braid wrapped around the pairs. Costs more, is stiffer and must be properly earthed to work, but survives electrically noisy places such as a factory floor.

Category ratings worth memorising: Cat 5e gives 1 Gbps, Cat 6 gives 1 Gbps over the full 100 m (10 Gbps only for shorter runs of about 55 m), and Cat 6a gives 10 Gbps over the full 100 m.

2. Co-axial cable. "Co-axial" means the two conductors share the same axis — one sits inside the other.

Co-axial cable, from the inside out

   1. Copper conductor        carries the signal
   2. Insulating dielectric   keeps 1 and 3 apart
   3. Braided metal shield    blocks outside noise, acts as return path
   4. Plastic outer jacket    mechanical protection

The braided shield does two jobs at once: it is the return conductor, and it is a grounded metal cage that keeps outside noise out and the signal in. That is exactly why co-axial cable tolerates electrical noise and longer runs better than UTP does. The classic Ethernet forms were Thinnet (10BASE2), up to 185 m per segment and terminated with BNC connectors, and Thicknet (10BASE5), up to 500 m and tapped with N-type connectors; both ran at 10 Mbps. Co-axial lost the LAN to twisted pair because it is stiff, awkward to install and shares a single cable among all the machines on it. It survives today in cable television and cable broadband, where one high-bandwidth line feeds a whole street.

3. Optical fibre cable. A hair-thin strand of extremely pure glass called the core, surrounded by a second glass layer called the cladding whose refractive index is lower than the core's, all wrapped in a protective buffer and an outer jacket.

Optical fibre, cross-section

   +-----------------------------+
   |  Outer jacket               |
   |   +---------------------+   |
   |   |  Cladding   (n2)    |   |
   |   |   +-------------+   |   |
   |   |   | Core   (n1) |   |   |    n1 is greater than n2
   |   |   +-------------+   |   |
   |   +---------------------+   |
   +-----------------------------+

Data travels as pulses of light from an LED or a laser. When a ray inside the core strikes the core–cladding boundary at a shallow enough angle, the lower refractive index of the cladding causes total internal reflection: the light bounces back into the core instead of escaping, and keeps bouncing all the way to the far end. If the cladding's refractive index were equal to or greater than the core's, total internal reflection could not happen and the fibre would carry nothing. That single fact is a favourite one-mark question.

• Single-mode fibre — very narrow core, about 9 micrometres, driven by a laser. The light travels almost straight down the core. Used for backbone and long-distance links: tens of kilometres without a repeater.
• Multi-mode fibre — wider core, 50 or 62.5 micrometres, driven by an LED. Light takes many different paths, so rays arrive at slightly different times (modal dispersion) and the usable distance drops to a few hundred metres — perfectly adequate inside a building or across a campus, and cheaper.

Because the signal is light and not current, fibre is completely immune to electromagnetic interference, radiates no interference of its own, and conducts no electricity, so it is safe to run near heavy machinery and between buildings in a lightning-prone area. It is also the hardest medium to tap: there is no electrical field leaking out to pick up from outside, and getting at the light means bending or breaking the strand, which drops the received power at the far end and is detected at once.

Worked example. A multi-speciality hospital in Pune is being wired. Three separate requirements, three different answers:

1. Ward and office LAN, 40 computers, longest cable route 70 m, ordinary rooms. Use Cat 6 UTP. It is the cheapest option, 70 m is comfortably inside the 100 m limit, there is no unusual electrical noise, and any electrician can terminate it with RJ-45 plugs. Spending on fibre here would buy nothing.
2. Link from the main building to the radiology block, 400 m away, with the cable route passing the MRI and generator room. Use optical fibre. Copper fails twice over here: 400 m is four times the UTP limit, and the MRI and generator produce strong electromagnetic interference that would corrupt data on any copper cable, shielded or not. Fibre solves both, and because it is non-conductive it also avoids carrying an earth fault or lightning surge between two separate buildings.
3. Television feed to patient rooms. The building already has co-axial cable-TV wiring. Leave it alone — coax has plenty of bandwidth for video distribution and there is no reason to spend money replacing working infrastructure.

The lesson the board is testing: the "best" medium is not fibre. The best medium is the cheapest one that clears the distance and the noise.

Unguided media send the signal into open space. There is no cable to lay, which is their whole attraction, but there is also no cable to protect the signal, which is where all their problems come from. The three types are separated by frequency, and frequency explains everything else about them.

Hold on to one rule: higher frequency buys more bandwidth but loses penetration. Radio, at the bottom, goes through walls but carries less. Infrared, at the top, carries a lot but is stopped by a sheet of cardboard. Microwave sits in between and is the compromise that long-distance links are built on.

1. Radio waves. Radio waves are omnidirectional: an antenna radiates them in all directions, so the sending and receiving antennas do not have to be aligned or even in sight of each other. They diffract around obstacles and pass through ordinary walls, which is why an FM radio works indoors with the receiver lying flat on a table. Lower-frequency radio can also travel very long distances by reflecting off the ionosphere.

• Uses: AM and FM radio broadcasting, television broadcasting, cordless phones, walkie-talkies, and short-range indoor wireless networking.
• Limitations: because the signal spreads everywhere, anyone in range can receive it — poor security — and everyone using the same band interferes with everyone else. Most radio spectrum is therefore licensed; in India it is allocated by the Wireless Planning and Coordination wing of the Department of Telecommunications. The 2.4 GHz band is licence-free almost everywhere, which is exactly why Wi-Fi, Bluetooth and cordless phones all crowd into it and why they interfere with one another.

A boundary worth stating honestly. Wi-Fi and Bluetooth work at 2.4 GHz, which by the frequency table above lies inside the microwave band, yet textbooks list them under radio waves. Both statements have a reason: the frequency is microwave, but the behaviour is radio — a plain omnidirectional antenna, no dish, no alignment, and enough wall penetration to work indoors. In an exam, answer by behaviour. If the question mentions omnidirectional transmission or passing through walls, the answer is radio waves; if it mentions a narrow beam, a dish or line of sight, the answer is microwave.

2. Microwaves. Microwaves are unidirectional. They are focused into a narrow beam by a parabolic dish, so the transmitting and receiving dishes must be carefully aligned and must have a clear line of sight. They do not bend around hills and they are absorbed or reflected by walls, which is why a microwave link works over open ground but not through a building.

• Terrestrial microwave: dish antennas mounted on tall towers, on hilltops or on rooftops. Because the earth curves away, the towers must be within sight of each other, so a long route is built from a chain of repeater towers typically a few tens of kilometres apart. The limit is geometry, not transmitter power — taller towers see further.
• Satellite microwave: a communication satellite in geostationary orbit, 35,786 km above the equator, completes one orbit in a sidereal day of 23 hours 56 minutes — exactly the time the earth takes to turn once on its axis. It therefore keeps pace with the ground below, appears fixed in the sky, and a ground dish can be bolted in one position and never moved. It receives on an uplink frequency, shifts the signal to a different downlink frequency and retransmits — two different frequencies are used so that the powerful uplink does not drown the extremely faint downlink. One satellite covers a huge area, which makes this the natural medium for television broadcasting and for reaching remote or hilly places where no cable can go.
• Limitations: the height that gives a satellite its coverage also gives it delay. At the speed of light, the journey up and back down takes about 240 ms for a station directly below the satellite and up to about 280 ms near the edge of its coverage — roughly a quarter of a second per hop, no matter how fast the link is. That is why a satellite video call feels laggy even on a high-speed connection. Higher microwave frequencies are also absorbed by heavy rain ("rain fade"), which is why a satellite TV picture breaks up during a storm. Setting up a satellite link is very expensive, and any new tall building that blocks a terrestrial line of sight kills the link.

3. Infrared waves. Infrared sits just below visible light in frequency, and it behaves like light: it travels in straight lines, it reflects off walls and ceilings, and it cannot pass through any opaque object. An infrared link therefore needs a clear line of sight and works only over a few metres, inside one room.

Students write "cannot pass through walls" as a disadvantage and stop there. Say the other half, because that is where the marks are: being blocked by walls is also infrared's biggest advantage. Two consequences follow directly:

• Frequency reuse. An infrared system in one room cannot interfere with an identical system in the next room. That is precisely why every television in the country can use the same infrared codes and your remote still does not change your neighbour's channel.
• Security. An eavesdropper has to be physically inside the same room. A radio signal leaks into the street; an infrared signal does not.

Uses: television and air-conditioner remote controls, older laptop-to-phone IrDA links, some wireless mice and keyboards, and short-range in-room data transfer. Other limitation: sunlight is full of infrared, so an infrared link outdoors or in bright light is swamped by noise and becomes unreliable. No licence is needed to use it.

Worked example. A firm in Jaipur has two buildings 4 km apart. A busy road and a protected heritage zone lie between them, so no cable may be dug in. There is a clear line of sight from one rooftop to the other. Which wireless medium?

• Infrared — reject. Its usable range is a few metres inside a room, and outdoor daylight would swamp it. It is not even close.
• Radio waves — possible but poor. Radio would cover 4 km, but it radiates in every direction, so most of the power is wasted, the signal is easy to intercept from the road below, the licence-free bands are already crowded in a city, and the usable bandwidth per channel is lower.
• Terrestrial microwave — accept. Mount a parabolic dish on each rooftop and align them. 4 km is well inside a single line-of-sight hop, no digging is required, the narrow beam wastes little power and is hard to intercept, and the bandwidth is high enough to carry the office network.

State the conditions with the answer, because "suggest and justify" questions carry a mark for the caveat: the line of sight must stay clear, so a tall building put up later between the two roofs would break the link, and very heavy rain will degrade it.

Second worked example. Why does pressing the remote in your drawing room not change the television in the bedroom? Because infrared is stopped by the wall. The same property that limits infrared to one room is what allows every household in India to use identical remote codes with no interference at all.

These five sit at the edge of a network. They attach a machine to the medium or keep a signal alive on it. None of them decides where data should go — that is the next section.

1. Modem (MOdulator–DEModulator). A computer produces digital signals: sharp square pulses that are either 0 or 1. A traditional telephone line was engineered to carry the human voice, roughly 300 Hz to 3400 Hz of analog sound, and it simply cannot carry sharp digital pulses over any distance. A modem bridges the two worlds. At the sending end it modulates the bits onto an analog carrier wave by varying the wave's amplitude, frequency or phase. At the receiving end another modem demodulates the wave back into bits. Because both directions need converting, you need a modem at each end of the line.

• By placement: an internal modem is a card fitted inside the computer or built into the motherboard; an external modem is a separate box connected by a cable and has its own power supply and status lights.
• By line type: a dial-up modem works over an ordinary telephone line, occupies the voice channel so the phone cannot be used at the same time, and is limited to 56 kbps. A DSL/ADSL modem uses frequencies above the voice band on the same pair of copper wires, so the telephone and the internet work simultaneously and much faster. A cable modem works over the cable-television co-axial line.
• Layer: physical layer.

The single box your ISP installs at home usually contains a modem, a router, a small switch and a Wi-Fi access point in one case. Those are four different functions, and the board expects you to name them separately.

2. Ethernet card (NIC — Network Interface Card). A circuit board plugged into an expansion slot, or a chip built into the motherboard, that lets a computer attach to a wired network. It provides the RJ-45 socket, converts the computer's internal data into a serial stream of Ethernet frames and back again, and holds a permanent 48-bit MAC (Media Access Control) address written in by the manufacturer, for example 00:1A:2B:3C:4D:5E. The first three bytes identify the manufacturer and the last three identify the individual card — a scheme designed so that every card leaves the factory with a globally unique address. Common speeds are 10, 100 and 1000 Mbps. It works at the physical and data link layers.

Keep the two kinds of address straight: MAC is permanent, flat and burned into the hardware; IP is assigned by software and is hierarchical. A laptop has two MAC addresses because it has two cards — one Ethernet, one Wi-Fi.

3. RJ-45 (Registered Jack 45). An 8-pin modular connector crimped onto the end of a four-pair twisted-pair cable, which plugs into the socket on a NIC, hub, switch or router. Say this plainly in an exam: RJ-45 is a connector, not a device. It has no address, no power, no OSI layer and takes no decision; it only makes eight reliable electrical contacts.

The eight wires are arranged in one of two standard colour orders, T568A or T568B. Use the same standard at both ends and you get a straight-through cable (computer to switch). Use a different standard at each end and you get a crossover cable (computer to computer, or switch to switch on older equipment). Modern equipment detects and corrects this automatically, so crossover cables are rarely needed now. Do not confuse RJ-45 with RJ-11, the smaller 6-position jack on a telephone cable.

4. Wi-Fi card (wireless NIC / WLAN card). The wireless equivalent of the Ethernet card. Instead of a socket it has a radio transceiver and an antenna, and it follows the IEEE 802.11 family of standards. It associates with a wireless access point, which is normally cabled back to a switch. Like a wired NIC it has its own MAC address and works at the physical and data link layers. Wi-Fi is the trade name for equipment built to these standards; many textbooks expand it as "Wireless Fidelity".

• Bands: the 2.4 GHz band travels further and passes through walls better, but is crowded with Bluetooth devices, cordless phones and microwave ovens. The 5 GHz band has more channels and higher speed but a shorter range and is stopped more easily by walls.
• Standards: 802.11b gave 11 Mbps at 2.4 GHz; 802.11a gave 54 Mbps at 5 GHz and 802.11g gave 54 Mbps at 2.4 GHz; 802.11n works in both bands at hundreds of Mbps; 802.11ac works at 5 GHz at gigabit class; 802.11ax is marketed as Wi-Fi 6.

5. Repeater. An electronic device that receives a weak and distorted signal, regenerates it and retransmits it at full strength. It works at the physical layer.

The word regenerate is where the marks are. An amplifier boosts whatever reaches it — including all the noise the signal has picked up along the way — so after a few amplifiers the noise is as large as the signal itself. A repeater works on a digital signal: for each incoming pulse it decides whether it was a 0 or a 1, and then transmits a brand-new, clean, correctly timed pulse. Noise does not accumulate. That is why long digital links use repeaters and not amplifiers.

A repeater reads no addresses, filters nothing and takes no forwarding decision: whatever arrives on one side goes out on the other. Use it whenever a run exceeds the medium's limit — a twisted-pair link longer than 100 m, or a long fibre or microwave route. A Wi-Fi range extender is the same idea applied to radio.

Worked example. A college in Coimbatore runs a cable from the server-room switch to a new lab across the campus. The measured cable route is 140 m. Once connected, none of the lab computers shows a link, and the network staff have spent a day reconfiguring IP settings with no result.

Diagnosis. 140 m exceeds the 100 m limit of a twisted-pair Ethernet segment, so attenuation has weakened the signal below what the receiving card can read. This is a physical-layer problem. No amount of software configuration can fix it, which is exactly why the day was wasted.

Two valid fixes.

1. Install a repeater in a corridor cabinet at about the 70 m mark, splitting the run into 70 m + 70 m. Both halves are now within the limit and the signal is regenerated in the middle. In practice a small switch is used, since a switch regenerates the signal exactly as a repeater does and adds addressing intelligence for about the same money.
2. Replace the copper run with multi-mode optical fibre, with a fibre port or media converter at each end. 140 m is nothing for fibre, and it also removes any interference from the campus power cabling. This costs more today and is the better answer for the long term.

The rest of the lab. Each desktop needs a working Ethernet card, and each cable needs an RJ-45 plug at both ends to reach the switch. Staff laptops with no Ethernet socket join through their Wi-Fi cards via an access point cabled to the same switch. The college's internet line from the ISP terminates on a modem.

These four decide where data goes next, and the board asks about them constantly — usually as "differentiate between" or as "which device would you use here". Each step up this list means reading a bigger address and taking a smarter decision.

1. Hub. A hub is a multiport repeater working at the physical layer. When a frame arrives on any port, the hub regenerates the signal and sends it out of every other port. It keeps no table, reads no address and remembers nothing about who is connected where.

The consequences are severe. All the machines share a single collision domain, so only one may transmit at a time; if two transmit together their signals collide and both must retry. Communication is therefore half duplex and the total bandwidth is shared — eight machines on a 100 Mbps hub get about 12.5 Mbps each at best. Every machine also physically receives every frame, so a machine set to accept everything can read other people's traffic, which is a real security problem. Types: a passive hub only joins the wires and needs no power; an active hub regenerates and boosts the signal and needs power; an intelligent (manageable) hub is an active hub with management features. Hubs are obsolete in real networks; they are in your syllabus because they are the perfect contrast to a switch.

2. Switch. A switch works at the data link layer and keeps a MAC address table. It builds that table by itself. Every time a frame arrives, the switch reads the frame's source MAC address and records "this address is reachable on port n". It then reads the destination MAC address and looks it up: if it knows the port, it sends the frame out of that port only; if the address is not yet in the table, it floods the frame out of all other ports once, and learns the correct port from the reply.

MAC address table of the lab switch

   MAC address           Port
   00:1A:2B:3C:4D:5E      1
   00:1A:2B:3C:4D:7F      5
   4C:CC:6A:11:02:9B     12

This is the difference the board asks about most often, so keep it crisp:

A frame from PC1 on port 1, addressed to PC5 on port 5

  HUB, 8 ports
     in  : port 1
     out : ports 2,3,4,5,6,7,8   every machine is disturbed and
                                 six of them discard the frame

  SWITCH, 8 ports
     in  : port 1
     out : port 5 only           the destination MAC was found in
                                 the table against port 5

Because of this, each port of a switch is its own collision domain, each port can run full duplex at its full rated speed, and a machine no longer sees traffic meant for others. One caution: broadcast frames, addressed to FF:FF:FF:FF:FF:FF, are still sent out of every port. A switch does not break up a broadcast domain.

3. Router. A router works at the network layer and connects two or more different networks. It reads the destination IP address of each packet, looks it up in a routing table that lists which network is reachable through which interface and at what cost, and forwards the packet along the best available path. Because it works on IP addresses it does not care what lies underneath — one side can be fibre Ethernet and the other a wireless link.

A router also separates broadcast domains: each interface is a different network, and a broadcast on one side is not forwarded to the other. This is what stops a broadcast storm in one department from taking down the whole organisation. In a home or small office the same box normally also shares one public IP address among many private machines.

4. Gateway. A gateway connects two networks that use different protocols or entirely different architectures, and it translates between them. That is the whole distinction from a router, and it is the one students lose marks on:

• A router forwards between networks that already speak the same protocol — TCP/IP on both sides. It changes nothing about the packet's protocol; it only chooses the path.
• A gateway makes two networks that do not speak the same language understand each other. To translate it may have to work all the way up to the application layer, which makes it the slowest and most complex device here.

Examples: a VoIP gateway joining an office IP telephone network to the ordinary telephone network; a mail gateway converting between two different messaging systems; a gateway between an industrial control protocol on a factory floor and the TCP/IP office LAN.

A warning about the words "default gateway". In your computer's IP settings there is a field called Default Gateway, and it almost always holds the IP address of your ordinary router — it simply means "the way out of my network". The name is historical. If a question asks which device connects two networks using different protocols, the answer is gateway. If it just asks who forwards your packets towards the internet, the honest answer is your router.

Which device would you use? This table answers most one-mark scenario questions directly.

Worked example. A school in Lucknow has 30 computers in its lab, all plugged into a single 32-port hub. Students complain the network is slow whenever several machines are used at once. The school also has a small office network that must stay separate from the lab, a broadband line from its ISP, and an old biometric attendance controller that communicates using a protocol the LAN does not understand. A new annexe 130 m away must also be connected.

1. Why the lab is slow. All 30 machines are in one collision domain. They share the hub's bandwidth, only one may transmit at a time, and every attempt by two machines at once causes a collision and a retry. Replace the hub with a 32-port switch: collision domains go from 1 to 32, one per port, with 30 of them in use; each port runs full duplex at its full speed, and no machine can read another's traffic. The number of broadcast domains stays 1 in both cases — a switch does not change that.
2. Keeping office and lab apart. Connect each to a different interface of a router. That makes them two separate networks with two separate broadcast domains, while still allowing controlled traffic between them.
3. Internet. The ISP line terminates on a modem, whose output goes into the router's internet port. The router then shares the connection with both networks.
4. The attendance controller. Because it uses a different protocol, no router or switch can help. It needs a gateway acting as a protocol converter between the controller and the school LAN.
5. The annexe at 130 m. 130 m is beyond the 100 m twisted-pair limit, so either place a repeater (in practice an intermediate switch) about halfway, or run optical fibre to the annexe, which needs nothing in between.