Transmission Media | Local Area Networks

The protocol may be elegant and the topology perfectly designed, but in the end a wire, a fibre or a stretch of air carries the bits. This lecture sets the three families of media side by side, explains from first principles why the wires in a network cable are twisted, what degrades along the way, and how we manage to put dozens of conversations on the same physical medium.

1Recap4 min

What to keep in mind

A bit becomes a signal - voltage, light or a radio wave - through line coding (a digital signal) or through modulation (an analogue signal).
Communication is organised in layers. This lecture stays at layer 1, the physical layer.
The hard problem of the physical layer is not telling a 1 from a 0 but synchronisation: the receiver must know where a bit ends.
The symbol rate (baud) is physically limited by the medium. This lecture shows why it is limited.

Learning outcomes

Choose the right medium for a given distance and speed
Explain why the wires in a UTP cable are twisted and what its category means
Correctly make up straight-through, crossover and rollover cables
Explain the difference between single-mode and multimode fibre
Compare FDM, WDM and TDM
Calculate the latency of a link, with all of its components
Recognise attenuation, crosstalk and noise, and know what stops each

2Three families of media7 min

What "transmission medium" means

The medium is the physical thing through which the signal travels. A copper wire, a core of glass, the air in a room. Nothing more.

The fundamental division is between guided and unguided media. In a guided medium, the signal is constrained to follow a physical path - it goes where the wire takes it. In an unguided medium it propagates freely through space, and anyone within range can receive it.

This single difference explains nearly all the distinctions that follow, including those of security: to listen in on a cable you must get inside the building; to listen in on Wi-Fi it is enough to be in the car park.

Fig. 1 - A taxonomy of media. Modern local networks use, in practice, only three of the leaves of this tree: twisted pair for access, fibre for the backbone and microwave for mobility.

3Coaxial cable7 min

A central conductor, an insulating layer (the dielectric), a metallic braid for shielding and an outer jacket - all concentric, which is where the name comes from.

Fig. 2 - The structure of a coaxial cable. The shield has a dual role: it is part of the circuit and, at the same time, it isolates the signal from the outside world.

Type	Impedance	Historical use
RG-59	75 Ω	cable television, video surveillance
RG-58	50 Ω	Thin Ethernet (10BASE2)
RG-11	50 Ω	Thick Ethernet (10BASE5), long backbones

Ethernet began on coaxial cable but abandoned it - and it is worth understanding why, because the reason explains the shape of every modern network.

Why coaxial died in LANs The topology was a bus: a single cable with every computer hanging off it. There were two consequences, both serious. First, everybody shared the same medium, so there was a single collision domain for the whole network. Second, a single break in the cable - a loosened connector, a chair dragged across it - brought down the entire segment, and finding the spot was a nightmare.

Twisted pair brought the star topology: each host has its own cable running to a central device. A fault affects a single host and is located immediately.

4Twisted pair: why the wires are twisted13 min

The network cable you see in any office has eight wires, grouped into four pairs, and each pair is twisted with a slightly different pitch from the others. It is not decoration and it is not a by-product of manufacturing. It is the solution to a concrete problem.

Differential transmission, explained without formulae

A signal can be transmitted in two ways. Single-ended: one wire carries the signal and the receiver compares it with ground. Differential: two wires carry the same signal, but mirrored with respect to one another, and the receiver reads the difference between them.

The difference looks like a pointless complication - until interference appears. A motor starting up nearby, a power cable running alongside, a mobile phone: all of them induce a voltage in the wires. If the two wires of a pair are almost identically exposed, the interference induces almost the same voltage in both. And when the receiver takes the difference, the common voltage subtracts itself away and vanishes.

For them to be identically exposed, the wires must constantly change their position relative to each other. That is to say, they must be twisted. That is all there is to it.

Analogy Two witnesses describe the same event, but both were bribed with the same sum and in the same way. If you listen to them separately, the lie gets through. If you care only about the difference between their accounts, the common bribe drops out of the reckoning and only what genuinely differs remains.

The differing twist rates between pairs solve a second problem: crosstalk - the signal from one pair leaking into its neighbour. If every pair had the same pitch, they would stay "in phase" along the whole length of the cable and would couple strongly. With different pitches, the coupling cancels out on average.

Shielded or unshielded

Abbreviation	What it means	When it is used
UTP	Unshielded Twisted Pair - no shield at all	the overwhelming majority of office installations
FTP / ScTP	a foil shield over all the pairs together	environments with moderate interference
STP	an individual shield for each pair	industrial halls, hospitals, near power equipment

Badly executed shielding is worse than none A shield helps only if it is correctly bonded to ground, at both ends, through a connector and a patch panel that maintain its continuity. An unbonded shield, or one bonded at a single end, turns into an aerial that collects interference and injects it exactly where it should not go.

The rule of thumb: if the installation has no decent earthing, good UTP beats bad STP.

The categories

The "category" of a cable refers not to speed but to the frequency band it can carry with guaranteed parameters. Speed follows from the band plus the coding scheme - which is why Cat 5e carries 1 Gbps although it has the same band as Cat 5.

Category	Guaranteed band	Speed supported	Typical standard	Note
Cat 3	16 MHz	10 Mbps	10BASE-T	old telephone cabling
Cat 5	100 MHz	100 Mbps	100BASE-TX	effectively obsolete
Cat 5e	100 MHz	1 Gbps	1000BASE-T	far better crosstalk; the acceptable minimum today
Cat 6	250 MHz	1 Gbps (10 Gbps up to 55 m)	1000BASE-T / 10GBASE-T	the usual choice for new installations
Cat 6a	500 MHz	10 Gbps at 100 m	10GBASE-T	thicker and harder to pull
Cat 7 / Cat 8	600 MHz / 2 GHz	10 – 40 Gbps	short links in data centres	special connectors, short distances

The 100 metre limit does not depend on the category Whether you use Cat 5e or Cat 8, a twisted pair segment may not exceed 100 metres: 90 m of fixed cable plus 2 × 5 m of patch cords at the ends. The limit comes from the attenuation budget and the propagation time, not from the quality of the cable.

If you need more, you have exactly three options: an intermediate switch, a media converter to fibre, or fibre from the outset.

PoE - power over the network cable

Power over Ethernet carries current as well as data on the same twisted pair cable. A Wi-Fi access point, an IP camera or an IP telephone mounted in the ceiling thus receives both network and power from a single cable - with no power socket beside it.

The standards provide, in order, roughly 15 W (802.3af), 30 W (802.3at) and up to 90 W (802.3bt). It is one of the reasons copper has not disappeared, although fibre is superior in every other respect: fibre cannot carry current.

5The three classic wiring schemes10 min

The order of the wires in an RJ-45 plug is laid down by the TIA/EIA-568A and 568B standards. The difference between them is merely that the green and orange pairs swap places - but the combination of the two ends decides what kind of cable you end up with.

Fig. 3 - The wire order in the two standards. Pins 4, 5, 7 and 8 are identical; only 1, 2, 3 and 6 change - exactly the pairs Ethernet uses at 10 and 100 Mbps.

Cable type	End 1	End 2	When it is used
Straight-through	568B	568B	between devices at different layers: PC ↔ switch, switch ↔ router
Crossover	568A	568B	between devices at the same layer: PC ↔ PC, switch ↔ switch, router ↔ router
Rollover (console)	568B	completely reversed order (8→1)	from a computer to a device's console port

Why this rule exists

A computer transmits on pins 1–2 and receives on 3–6. A switch does exactly the opposite: it receives on 1–2 and transmits on 3–6 - because it was designed to connect to computers. Between the two, therefore, the wires run straight and every transmitter meets a receiver.

But between two computers, both would transmit on 1–2 and listen on 3–6: two mouths facing each other and two ears facing each other. The crossover cable swaps the pairs and repairs the situation. The same problem arises between two switches, for the same reason.

Why it rarely matters in practice any more Almost every device made in the last twenty years implements Auto-MDI/MDI-X: it detects by itself whether a crossover is needed and performs it electronically. You can connect two switches with a straight-through cable and it will work.

The rule is still worth knowing, for three reasons: Packet Tracer enforces it strictly, on old equipment it genuinely matters, and when a link "does not light up" the cable type remains one of the first suspects.

6Optical fibre13 min

How light stays inside a thread of glass

A fibre has a core of very pure glass, surrounded by cladding also made of glass but with a slightly lower refractive index. When light in the core reaches the boundary at a small enough angle to the wall, it does not escape - it is reflected completely back into the core. The phenomenon is called total internal reflection, and it is the same one that makes the surface of a swimming pool, seen from below, look like a mirror.

The ray of light thus advances in zig-zag through the core, kilometre after kilometre, without getting out.

Fig. 4 - In a multimode fibre, the rays that zig-zag travel further than the axial ones and so arrive later: a short pulse spreads out along the way. The phenomenon is called modal dispersion and it limits the distance. Single-mode fibre does not have the problem, because it has only one path.

Type	Core / cladding (µm)	Light source	Typical distance	Cost
Multimode, graded index	50/125 or 62.5/125	LED or VCSEL	hundreds of metres	cheap equipment
Single-mode	~9/125	laser	tens to hundreds of kilometres	expensive equipment

The wavelength windows

The attenuation of glass is not constant with wavelength. There are three "windows" in which it is minimal and in which, consequently, all optical equipment works.

Wavelength	Typical attenuation	Use
850 nm	~3.5 dB/km	multimode, short links, cheap components
1310 nm	~1.25 dB/km	single-mode, medium distances
1550 nm	~0.5 dB/km	single-mode, long distances; compatible with optical amplifiers

What a decibel is, briefly

dB is not an absolute unit but a ratio, expressed logarithmically. −3 dB means half the power. −10 dB means a tenth. −20 dB means a hundredth.

The advantage of the logarithmic scale: losses add up instead of multiplying. A path with 2 dB of fibre, 1 dB of connectors and 0.5 dB of splices has a total loss of 3.5 dB - a simple addition.

What fibre does better

Distances impossible for copper: tens of kilometres
It does not conduct current: no earthing problems, no potential difference between buildings
It neither emits nor picks up electromagnetic interference
It cannot be eavesdropped on without being physically touched
Enormous bandwidth, in practice unsaturated

What makes it awkward

It cannot carry power - so no PoE
Terminating it requires tools and training; it is not done on the fly
Limited bend radius: bent too tightly, the light escapes
A dirty connector can cost more than a kilometre of fibre

The golden rule of fibre The protective caps come off only at the moment of connection, and the surface is cleaned every single time. A speck of dust on a 9 µm core blocks a significant part of the light - and produces an intermittent fault, which is to say exactly the kind you spend three days hunting.

7The wireless medium9 min

Electromagnetic waves cover an enormous spectrum, but computer networks use a narrow slice of it: microwaves, roughly between 1 and 10 GHz. The choice is not accidental and follows from a trade-off worth keeping in mind whenever somebody proposes "a better band".

Property	Low frequencies	High frequencies
Propagation distance	large	small
Penetration of obstacles	good	poor
Interference	plenty - the band is crowded	little
Bandwidth available	small	large
Cost of equipment	low	high
Absorption in the atmosphere	low	rises with frequency

What this means in practice Low frequencies travel far and pass through walls, but the bandwidth available there is small and already occupied by others. High frequencies offer generous bandwidth but are stopped by the first wall.

That is why Wi-Fi at 2.4 GHz covers more ground than at 5 GHz, and 5 GHz more than 6 GHz - in exactly the order in which the available throughput increases. There is no "better" band; there is a band suited to the purpose.

Technology	Band	Typical application
AM radio	540 – 1600 kHz	medium-wave broadcasting, hundreds of kilometres
FM radio	88 – 108 MHz	commercial broadcasting
LoRaWAN	868 MHz (EU)	IoT sensors: kilometres, tiny power draw, low throughput
Zigbee, Bluetooth	2.4 GHz	automation, peripherals, short distances
Wi-Fi	2.4 / 5 / 6 GHz	wireless local networks

The ISM bands

Industrial, Scientific and Medical - portions of spectrum that may be used without a licence and without a fee, subject to certain power limits. In Romania the allocations are regulated by ANCOM through the National Table of Frequency Allocations.

The fact that the band is free explains at once the enormous success of Wi-Fi and the congestion within it: the same 2.4 GHz band is home to Wi-Fi, Bluetooth, Zigbee, remote controls, weather stations and the microwave oven.

8Multiplexing: several conversations on one medium11 min

A medium costs money. Fibre has to be pulled through ducts, spectrum has to be licensed, the channel has to be maintained. Consequently we nearly always want several communications to travel over it at the same time.

Multiplexing and demultiplexing

Multiplexing is the combining of several data streams into a single signal transmitted over a shared medium. Demultiplexing is the inverse operation, at the far end. The central question is what exactly gets divided: frequency, time or wavelength.

Method	Resource divided	Type	Example
FDM - Frequency Division Multiplexing	the frequency band	analogue	radio and TV channels, DSL
WDM - Wavelength Division Multiplexing	the wavelengths in a fibre	analogue (optical)	optical backbones; DWDM carries dozens of channels on one fibre
TDM - Time Division Multiplexing	time, in fixed slots	digital	E1/T1, a switch's internal bus

9The performance of a link13 min

Three quantities describe a link and they are constantly confused with one another - including in the advertising of Internet providers.

Bandwidth, throughput, latency

Bandwidth is the maximum theoretical capacity of the medium. The figure on the box.
Throughput is the amount of data actually transferred per unit of time - always lower, because of headers, collisions and retransmissions.
Latency is the time a bit needs to travel from one point to another.

Analogy An eight-lane motorway has large bandwidth: many cars fit on it at once. But if the speed limit is 30 km/h, the latency remains high - an individual car still arrives late. And if half the lanes are closed for roadworks, the throughput is far below the theoretical capacity.

High bandwidth and high latency coexist perfectly well: that is exactly what a satellite link looks like.

The four components of latency

Transmission time - how long it takes to "push" the frame onto the wire: size divided by rate. It depends on how big the frame is.
Propagation time - how long the signal needs to cross the medium: roughly 5 µs per kilometre, in copper as in fibre (≈ 2/3 of the speed of light in vacuum). It depends only on distance.
Device latency - the processing time in each switch and router along the path. It grows with the layer at which the device works.
Queueing time - it appears when the outgoing link is busy. The only component that varies strongly over time, hence the only one that produces jitter - and the one that ruins video calls.

Worked example

A 1518-byte Ethernet frame is transmitted over a 100 Mbps link, 100 m of Cat 5e UTP long. How long does it take until the last bit reaches the destination, ignoring the devices?

See the solution

Transmission: 1518 × 8 bits / 10⁸ bps = 12,144 / 10⁸ ≈ 121.4 µs

Propagation: 100 m × 5.56 ns/m ≈ 0.56 µs

Total ≈ 122 µs.

Notice the ratio: over short distances, transmission dominates propagation completely - by a factor of 200. On a transatlantic link the situation reverses: propagation of ~30 ms makes the transmission time of a single frame negligible.

This is exactly why transport protocols use sliding windows: on a long link, waiting for an acknowledgement of every frame before sending the next would waste 99 % of the capacity.

Mind the units - again Mbps is measured in powers of 10 (10⁶ bits per second), while MB is in powers of 2 (2²⁰ bytes). A 100 Mbps link transfers, at best, about 11.9 MB per second - not 100, not 12.5.

10What damages the signal on the way9 min

Four phenomena degrade a signal, whatever the medium. Each has a known remedy, and together they explain practically every numerical limit in this course.

Phenomenon	What happens	How it shows up	Remedy
Attenuation	the amplitude falls with distance, until it drops below the detection threshold	the link does not come up at all beyond a certain length	repeater or amplifier; respecting the maximum length
Crosstalk	the signal in one pair induces a voltage in the neighbouring pair	errors that increase with traffic	twisting with different pitches, individual shielding
Noise	external interference is superimposed on the signal	random errors, often correlated with a machine nearby	shielding, differential transmission, routes away from power cables
Dispersion	the pulse spreads out and overlaps the next one	a distance limit on multimode fibre	single-mode fibre, optical regeneration

The difference between a repeater and an amplifier

An amplifier magnifies everything: the signal and the accumulated noise alike. After a few amplifiers in cascade, the noise becomes larger than the signal.

A repeater does something else: it reads the bits, decides them (1 or 0) and regenerates them cleanly. Noise does not accumulate, because at every step the signal is rebuilt from scratch. That is why long chains use repeaters, not amplifiers.

Two devices deal with the borderline cases:

The repeater restores a degraded signal and extends the segment. There are electrical, optical and wireless variants. Beware: a repeater also extends the collision domain, so it is not free.
The media converter translates between different media - electrical ↔ optical, electrical ↔ wireless - and lets a fibre backbone terminate in a switch with copper ports. It is the standard solution when two buildings must be joined.

11Common mistakes4 min

"Cat 6 is faster than Cat 5e" Both carry 1 Gbps at 100 m. Cat 6 has a wider band and less crosstalk, hence more margin and the possibility of 10 Gbps over short distances - but on an ordinary gigabit link you will see no difference at all. The category buys margin and future-proofing, not immediate speed.
"I used shielded cable, so I am protected" A shield not properly bonded to ground at both ends behaves like an aerial and makes matters worse. Shielding makes sense only together with a correct earthing installation.
"The link is 150 m but it works, so it is fine" It may work today, under light traffic, and fail in six months when traffic grows or the temperature changes. The 100 m limit is not a suggestion. Beyond 100 m: an intermediate switch or fibre.
"High bandwidth means low latency" They are independent. A satellite link has high bandwidth and a latency of hundreds of milliseconds. For gaming and telephony, latency is what counts; for downloads, bandwidth is.
"Fibre is always better, so let us use it everywhere" Fibre does not carry power, it requires prepared terminations and expensive equipment at the ends. For 40 m to an office, UTP is the right choice. Fibre for backbones and distance; copper for access.

12Summary and glossary4 min

Media divide into guided and unguided, and from that single distinction follow the differences in distance, cost and security.
Twisting and differential transmission are the same idea seen from two angles: they make the common interference cancel itself out.
Fibre wins on distance and immunity; copper wins on cost, on simplicity, and because it can carry power.
Multiplexing divides frequency (FDM), time (TDM) or colour (WDM).
Latency has four components, and which of them dominates depends entirely on the scale of the network.

guided mediumthe signal follows a physical path: cable or fibre

UTP / STPtwisted pair cable without and with shielding

categorythe guaranteed frequency band of a twisted pair cable

568A / 568Bthe two standard wire orders in an RJ-45 plug

PoEelectrical power carried over the data cable

total internal reflectionthe phenomenon that keeps light inside the fibre core

modal dispersionpulse spreading caused by paths of differing length

dBa ratio expressed logarithmically; losses add up

optical budgettransmitted power minus every loss along the path

FDM / TDM / WDMsharing the medium by frequency, time or wavelength

latencyend-to-end time; it has four components

jitterthe variation of latency; it comes from queueing

crosstalksignal leaking from one pair into another

media convertertranslates between media: copper ↔ fibre ↔ wireless

13Self-check questions7 min

14Further reading and bibliography2 min

With the medium settled, the next lecture climbs a layer: how bits are organised into frames, how hosts identify one another, what happens when two speak at the same time, and why the switch replaced the hub.

TIA/EIA-568 - structured cabling for commercial buildings
ITU-T G.652 and G.657 - the characteristics of single-mode fibre
IEEE 802.3af / at / bt - the three generations of PoE
ANCOM - the Romanian National Table of Radio Frequency Allocations
Andrew S. Tanenbaum, Computer Networks, chapter 2 - transmission media and multiplexing