Telecommunications engineering turns information into signals, moves those signals across distance and reconstructs enough of the intended information for a receiver to use. It includes radio, optical fibre, copper, satellites, mobile networks, packet networks, modulation, coding, switching, routing, synchronisation and the standards that let equipment from different organisations interoperate.
A voice call appears simple: one person speaks and another hears. Underneath, sound becomes an electrical or digital representation, the representation is compressed or encoded, packets or radio symbols cross a network, timing is managed, errors are detected or corrected, and the receiver reconstructs audio quickly enough for conversation to feel natural.
ITU-T develops international standards describing how telecommunications networks operate and interwork, while ITU-R addresses radio systems, spectrum, propagation and satellite orbits. Singapore’s IMDA manages spectrum planning, allocation and assignment and publishes network standards for national telecommunications infrastructure. Source: ITU-T. Source: ITU-R Recommendations. Source: IMDA Spectrum Management.
Reading routes: begin with the child-friendly explanation; learn how information becomes a signal; work through decibels, a link budget and channel capacity; explore fibre, radio, packet networks, spectrum and resilience; then use the learning workshop.
The numerical links and network cases below are original teaching models. They are not radio-licensing advice, spectrum authorisations, network configuration instructions or guarantees of coverage.
Explain telecommunications to a child: the message has to survive the journey
Imagine writing “MEET AT 3” on a piece of paper and passing it through ten classmates. If each person copies the message, small errors can appear. A good communication system needs ways to represent the message clearly and detect when something went wrong.
Now imagine the message travelling as flashes of light through fibre or radio waves through the air. The receiver does not receive the original voice or picture directly. It receives a physical signal that represents information.
Telecommunications engineering asks how to create that representation, how to send it through a channel, how to separate it from noise and interference and how to reconstruct it at the other end.
The most important word is receiver. A signal has succeeded only when the receiver obtains enough correct information at the required time.
1. Information and signals are different things
Information is the message or distinction the receiver needs. A signal is the physical quantity used to carry that information.
A voice can become air pressure, microphone voltage, digital samples, radio modulation, optical pulses and finally speaker motion. The representation changes several times while the intended message remains connected through the chain.
Each conversion can add distortion, delay or error. Engineering therefore follows the complete route rather than assuming the message survives because one intermediate waveform looks clean.
This is closely connected to Electrical Engineering, which explains signals and sampling. Telecommunications extends that reasoning across channels, shared networks and long distances.
2. A channel is the physical environment through which a signal travels
A channel can be an optical fibre, copper pair, coaxial cable, free space, underwater acoustic path or another physical medium.
The channel changes the signal. It can attenuate some frequencies, delay different components differently, add noise and expose the signal to interference.
A useful channel model therefore includes more than distance. Frequency, environment, bandwidth, geometry and the receiving equipment all matter.
The engineering task is not to prevent every physical effect. It is to design a representation and receiver that achieve the required information transfer despite the effects that remain.
3. Bandwidth describes a range of frequencies or a communication resource
In physical signal analysis, bandwidth refers to a range of frequencies occupied or passed by a system. In everyday networking, people also use the word to mean data-carrying capacity.
Those meanings are related but not identical. A channel with more physical bandwidth can often support more data, but the result also depends on signal-to-noise ratio, coding, modulation and implementation.
A narrow voice channel and a high-rate video link therefore make different demands on the transmission system.
The useful habit is to state the units: hertz for frequency bandwidth, bits per second for data rate.
4. Noise is unwanted uncertainty added to the signal
Electronic components, thermal effects and the environment introduce noise. Interference can come from other intentional transmitters or unintended sources.
The receiver does not usually need a perfectly noise-free waveform. It needs enough separation between possible transmitted states to make decisions with acceptable error probability.
Signal-to-noise ratio therefore matters more than signal power alone. A strong signal buried in even stronger interference can be harder to decode than a weaker signal in a quiet channel.
Telecommunications engineering manages both sides: increase useful received signal when appropriate and reduce or tolerate unwanted disturbance.
5. Decibels make large power ratios easier to combine
For power ratios, decibels are defined as 10 log₁₀(P₂/P₁). A tenfold power ratio is 10 dB. A hundredfold ratio is 20 dB. Half power corresponds to approximately −3.01 dB.
Because multiplication of ratios becomes addition in logarithmic form, gains and losses in a link can be added in decibels.
Original example: a transmitter chain has +12 dB amplifier gain, −3 dB cable loss and +8 dB antenna gain. The net change is +17 dB relative to the chosen reference point.
This calculation does not provide absolute power until a starting level is supplied. Decibels describe a ratio or a level relative to a stated reference.
6. dBm expresses absolute power relative to one milliwatt
dBm is a power level referenced to 1 milliwatt. 0 dBm is 1 mW, 10 dBm is 10 mW, 20 dBm is 100 mW and 30 dBm is 1 W.
If a signal begins at 20 dBm and experiences a net −40 dB path change, the resulting ideal level is −20 dBm.
The arithmetic is simple because the logarithmic levels and gains share compatible units.
A power level by itself does not establish a usable link. Receiver sensitivity, noise, interference, bandwidth and required error performance still matter.
7. Worked link budget: account for every major gain and loss
Original teaching model: transmitter power = 20 dBm; transmit antenna gain = 5 dB; path loss = 90 dB; receive antenna gain = 3 dB; receiver cable loss = 2 dB.
Received power is 20 + 5 − 90 + 3 − 2 = −64 dBm.
Suppose a fictional receiver requires at least −72 dBm under the selected modulation and noise conditions. The simplified margin is 8 dB.
This is not a real radio design. Actual links include fading, interference, polarisation, implementation losses, regulatory power limits and environmental variation.
The educational value is the conservation-like accounting: every major gain and loss must have a place in the budget.
8. Margin is a buffer against uncertainty, not spare performance guaranteed forever
The 8 dB margin in the invented link is the difference between one predicted received level and one assumed threshold.
If rain, obstruction, antenna misalignment or interference introduces another 10 dB of effective loss, the modelled link no longer meets that threshold.
Engineering margin is therefore tied to the uncertainty and variation expected in the operating environment.
A system can have large margin in one location and almost none elsewhere. Coverage claims need spatial and temporal evidence rather than one successful test.
9. Modulation maps information onto a carrier or signal structure
Modulation changes a signal property—such as amplitude, frequency or phase—to represent information.
Digital modulation chooses among discrete signal states representing bits or groups of bits. More states can carry more bits per symbol but usually require the receiver to distinguish points that are closer together for a given power.
This creates a trade-off among spectral efficiency, required signal quality and implementation complexity.
The best modulation is therefore not the one with the largest number of states. It is the one compatible with the channel and service requirement.
10. A symbol is not necessarily one bit
Binary modulation can represent one bit per symbol. Four distinct signal states can represent two bits per symbol because 2² = 4. Sixteen states can represent four bits because 2⁴ = 16.
Original example: an ideal 16-state scheme operating at one million symbols per second represents four million raw bits per second before coding and overhead.
This arithmetic does not establish that the channel can support the required signal quality.
Higher-order modulation becomes useful only when the receiver can reliably distinguish the additional states under real noise and distortion.
11. Error-control coding adds structured redundancy
Communication systems often add extra bits that let the receiver detect or correct some errors.
This reduces the fraction of transmitted bits carrying new user information, but can dramatically improve useful reliability.
A code rate of 3/4 means that, in a simplified interpretation, three information bits are represented using four transmitted coded bits.
If the physical coded-bit rate is 40 Mbit/s, an ideal 3/4 code leaves 30 Mbit/s before other protocol overhead.
Redundancy is therefore not automatically waste. It can increase the amount of correct information delivered by making the channel more robust.
12. Shannon capacity gives a theoretical channel limit
For an ideal band-limited additive white Gaussian noise channel, Shannon’s capacity formula is C = B log₂(1 + S/N), where B is bandwidth in hertz and S/N is linear signal-to-noise ratio.
Original teaching model: take B = 1 MHz and S/N = 15, equivalent to about 11.76 dB.
Capacity is 1,000,000 × log₂(16) = 4 Mbit/s.
This is a theoretical upper bound under the stated channel model, not a claim that a practical modem will deliver exactly 4 Mbit/s.
It is useful because it exposes a fundamental trade-off: more bandwidth or better signal-to-noise ratio can increase theoretical information capacity.
13. Doubling power does not double capacity
Continue the same 1 MHz Shannon model. Increase S/N from 15 to 31, approximately doubling the signal power relative to noise.
Capacity becomes 1 MHz × log₂(32) = 5 Mbit/s.
The signal-to-noise ratio roughly doubled, but theoretical capacity rose from 4 to 5 Mbit/s, not to 8.
This logarithmic behaviour explains why communications engineering cannot rely indefinitely on increasing power.
Spectrum, coding, antennas, spatial reuse and network architecture can be equally important resources.
14. Multiplexing lets multiple information streams share infrastructure
Frequency-division multiplexing assigns different frequency ranges. Time-division multiplexing assigns different time slots. Wavelength-division multiplexing carries multiple optical wavelengths in fibre.
Modern packet networks share links statistically: packets from many users are interleaved as demand changes.
Sharing improves utilisation but creates coordination and congestion problems.
The correct multiplexing scheme depends on timing, isolation, scale and service requirements rather than one universal architecture.
15. Optical fibre carries information through guided light
Optical fibre guides light through a transparent structure designed to confine propagation. Transmitters convert electrical data into optical signals and receivers convert them back.
Fibre offers extremely high capacity and low attenuation over long distances, but real links still experience losses, dispersion, connector effects and equipment limits.
Multiple wavelengths can share one fibre, greatly expanding the information carried by the physical strand.
The network therefore includes optical amplifiers, switching, monitoring and protection as well as the glass itself.
16. Dispersion spreads pulses through time
Different spectral or propagation components can travel with different delays. A short transmitted pulse can therefore arrive spread out.
If neighbouring pulses overlap excessively, the receiver has more difficulty distinguishing symbols.
Higher data rates make timing margins smaller, so dispersion becomes more consequential.
Engineering responses can include suitable fibre, wavelength planning, modulation, equalisation or digital signal processing. The correct choice depends on the link distance and performance requirement.
17. Radio links share a physical environment
Radio waves propagate through an environment containing buildings, terrain, atmosphere and other transmitters.
Signals can reflect, diffract and scatter, producing multiple paths that arrive with different delays and phases.
A location that receives a strong signal at one frequency or orientation may experience deep fading after a small movement.
Wireless engineering therefore works statistically across space and time rather than assuming a simple inverse-distance relationship always determines the result.
18. Antennas connect electrical signals to electromagnetic waves
An antenna has direction-dependent behaviour. Gain represents how strongly it radiates or receives in a direction relative to a reference.
Higher directional gain can improve a link while reducing coverage in other directions and increasing sensitivity to alignment.
An antenna’s performance also depends on frequency, environment and nearby structures.
Changing an enclosure or mounting position can therefore change wireless performance even if the radio electronics remain unchanged.
19. MIMO uses several spatial signal paths
Multiple-input multiple-output systems use multiple antennas and signal processing to exploit spatial diversity or send multiple data streams.
Diversity can improve reliability when different signal paths fade independently enough. Spatial multiplexing can increase data rate when the channel provides sufficiently distinguishable paths.
The number of antennas does not automatically equal the number of useful independent streams.
Real performance depends on channel geometry, correlation, signal quality and receiver capability.
20. Spectrum is a shared and limited resource
Many services need access to radio-frequency spectrum: mobile communications, broadcasting, satellites, navigation, scientific services and short-range devices.
Because transmissions can interfere, spectrum use requires coordination and rules about frequency, power, geography and service type.
IMDA describes spectrum management in Singapore as planning, allocation, assignment and monitoring for efficient use, including investigation of interference. Source: IMDA.
The engineering design must therefore work inside the applicable regulatory allocation rather than treating an unused-looking frequency as automatically available.
21. Standards make interoperability possible
A handset built by one company must communicate with network equipment built by others. Packets need agreed formats. Timing and addressing need agreed meaning.
ITU-T describes international standards as defining how telecommunications networks operate and interwork. ITU-R recommendations address radio systems and spectrum-related technical standards. Source: ITU-T Recommendations. Source: ITU-R.
Standards do not make every network identical. They create interfaces within which vendors and operators can implement different internal designs.
A standards-compliant component can still fail a system requirement if configuration, capacity or environmental conditions are unsuitable.
22. Telecommunications standards continue to evolve
Mobile generations are long-lived engineering ecosystems rather than one release date. Research, requirements, radio interfaces, core-network architecture, device ecosystems and national deployments evolve through stages.
As of 2026, ITU is developing the IMT-2030 framework associated with future 6G systems. In March 2026, ITU reported that experts had agreed draft technical performance requirements, with formal approval expected in a later process. Source: ITU, IMT-2030 update.
This does not mean a complete global 6G network standard or deployment exists in 2026.
Telecommunications engineering must distinguish a research target, a standard, an implementation and an operating public service.
23. Packet switching breaks a message into separately transported units
Internet-style networks move packets that can share links with many other flows.
Packets may experience different delays or routes. Some can be lost and retransmitted depending on the protocol.
This sharing makes networks flexible and efficient, but it creates congestion and variable delay.
The application should therefore be designed around the service it actually needs. A file transfer can tolerate retransmission delay differently from a live conversation.
24. Routing chooses a path through the network
Routers exchange information and forward packets according to routing tables and policies.
The shortest geographic route is not necessarily the selected path. Operators consider topology, policy, capacity and availability.
A route can change after a failure while the user’s application continues operating.
This resilience depends on the network having a usable alternative and enough information to detect and reconverge around the failure.
25. Latency is a sum of several delays
Total communication delay can include propagation, transmission, processing and queueing.
Original example: a 1,500-byte packet contains 12,000 bits. On a 10 Mbit/s link, transmission time is 12,000/10,000,000 = 1.2 milliseconds.
Suppose propagation and processing together add 15 milliseconds and queueing adds an average 5 milliseconds. The simplified one-way latency is 21.2 milliseconds.
Increasing link speed to 100 Mbit/s reduces transmission time to 0.12 milliseconds, but total becomes 20.12 milliseconds if the other terms are unchanged.
This illustrates why “faster bandwidth” does not remove every source of latency.
26. Jitter matters when timing variation affects the application
Jitter is variation in packet delay. A voice or video receiver may use a buffer to absorb some timing variation.
A larger buffer can reduce interruptions caused by jitter while increasing end-to-end delay.
The correct buffer is therefore a trade-off, not an instruction to make it as large as possible.
Interactive applications care about timing more than bulk data transfer, which can often tolerate larger variation if total throughput remains adequate.
27. Congestion occurs when offered traffic exceeds usable capacity
A network link can serve only a finite amount of data. When arrivals exceed service capacity for long enough, queues grow.
Initially, more queueing increases delay. If buffers fill, packets can be dropped.
Transport protocols and applications can respond by reducing sending rates, but the effectiveness depends on the protocol and traffic mix.
A network that works well with average demand can still fail during a burst or shared event. Capacity planning therefore includes distributions and peaks, not only monthly averages.
28. Quality of service is about differentiated requirements
Not every packet has the same timing and loss sensitivity. A voice packet that arrives very late may be useless, while a file packet can still be valuable after retransmission.
Networks can classify and schedule traffic differently, but prioritisation does not create unlimited capacity.
If every service is marked highest priority, the classification stops distinguishing anything.
QoS engineering begins with actual application requirements, then uses appropriate queueing, capacity and policy controls.
29. Reliability has several layers
A fibre can fail. A router can fail. A power system can fail. A software configuration can fail. A route can exist but lack enough capacity.
Redundancy works only when the alternative does not share the same critical dependency.
Two cables in the same trench can look like two network paths while being vulnerable to one excavation event.
A resilient design maps physical, logical and operational dependencies rather than counting duplicate components.
30. Telecommunications resilience is a social infrastructure question
ITU notes that modern communication supports work, education, healthcare, payments, transport, safety and emergency information, making reliable connectivity important during disruption. Source: ITU, Digital Resilience and Spectrum.
Network resilience includes alternative paths, power, spare capacity, operational coordination, monitoring and restoration.
The service can fail even when radio coverage remains if identity, core-network, backhaul or power systems fail.
The engineering boundary should therefore follow the complete service from user device to the intended receiver.
31. Monitoring makes invisible network states observable
Operators monitor traffic, errors, delay, loss, signal quality and equipment state.
A single green device light cannot establish end-to-end service. The local equipment may be healthy while the upstream route is unavailable.
Useful monitoring connects symptoms to network layers and locations so operators can identify whether a problem is radio, fibre, routing, congestion, application or power.
Spectrum monitoring also helps identify interference and verify radiated signals, as described in IMDA’s spectrum-management framework. Source: IMDA.
32. Failure diagnosis follows the message route
| Observed symptom | Questions that narrow the mechanism |
|---|---|
| Strong radio signal but poor data rate | Is interference, modulation fallback, congestion or core-network capacity limiting performance? |
| High bandwidth but slow application | Do propagation, queueing, server processing or protocol round trips dominate? |
| Fibre link power is adequate but errors increase | Is dispersion, receiver quality, connector condition or equipment configuration involved? |
| One site loses service during a local power failure | Which radio, backhaul and core dependencies require power and backup? |
| Redundant route fails with the primary | Do both paths share a physical cable, building, power or upstream provider? |
| Packets arrive but voice sounds broken | Is jitter, loss, codec behaviour or timing buffer responsible? |
These are diagnostic questions, not instructions to interfere with operating networks or spectrum.
33. Repair should restore the route and correct the model
Suppose a network repeatedly fails because two “diverse” fibres share one physical duct. Repairing a broken fibre restores service once. Correcting the network model and creating genuinely independent routing addresses the hidden common dependency.
The records should then show the true physical path. Otherwise future resilience analysis continues from the same false assumption.
A software or configuration repair similarly needs version control, validation and monitoring after release.
Telecommunications systems remain dependable when physical infrastructure, logical topology and operational records describe the same network.
34. Learning workshop with worked answers
Question A: starting at 20 dBm, add +5 dB antenna gain, −90 dB path loss, +3 dB receive gain and −2 dB cable loss. Answer: −64 dBm.
Question B: if the receiver threshold is −72 dBm, what simplified margin remains? Answer: 8 dB.
Question C: a 16-state modulation carries how many bits per ideal symbol? Answer: log₂16 = 4 bits.
Question D: a coded-bit stream is 40 Mbit/s with code rate 3/4. What ideal information rate remains before other overhead? Answer: 30 Mbit/s.
Question E: Shannon model B = 1 MHz and S/N = 15. Answer: C = 1 MHz × log₂16 = 4 Mbit/s.
Question F: a 1,500-byte packet crosses a 10 Mbit/s link. Transmission time? Answer: 12,000/10,000,000 = 1.2 ms.
Question G: why can an internet connection with very high throughput still feel slow in conversation? Answer: latency and jitter can remain significant even when throughput is high.
35. A learning progression from waves to networks
Primary learners can pass encoded messages and discover why representation and error checking matter.
Secondary learners can use frequency, period, powers of ten, logarithms, binary representation and simple link budgets.
Advanced learners can add Fourier analysis, probability, coding, electromagnetic propagation, queueing, routing and network optimisation.
The transferable idea is always the same: identify what information must reach which receiver, through which physical and logical path, under which timing and reliability constraints.
36. Frequently asked questions
Is telecommunications only wireless?
No. Fibre, copper, radio, satellite and packet networks are all part of telecommunications engineering.
Does a stronger radio signal always mean faster data?
No. Interference, coding, modulation, network congestion and receiver capability can limit data rate even when signal strength is high.
Is bandwidth the same as speed?
No. Bandwidth can refer to frequency range, while user data speed is measured in bits per second. The two are related through channel and modulation design but are not synonyms.
Why do standards matter?
They let independently built devices and networks interoperate by agreeing on interfaces, formats and behaviour.
Is 6G already a finished global standard in 2026?
No. ITU’s IMT-2030 work is still developing technical requirements and the broader standardisation ecosystem. Research targets and final deployed standards are different stages.
Can this guide be used to transmit on any frequency?
No. Spectrum use is regulated. Use applicable IMDA or other national requirements for real transmissions and equipment.
37. Working glossary
Signal: a physical quantity carrying information. Channel: the medium or path through which a signal travels. Bandwidth: a defined frequency range or, in networking usage, an available data-carrying capacity.
SNR: signal-to-noise ratio. Modulation: mapping information into changes of a transmitted signal. Symbol: one transmitted state from a defined signalling alphabet. Coding: structured representation adding redundancy or other properties to support reliable communication.
Link budget: accounting of transmitter level, gains and losses to predict receiver level. Latency: elapsed communication time. Jitter: variation in delay. Throughput: useful data delivered per unit time.
Routing: selection of paths through a network. Multiplexing: sharing a communication resource among streams. Spectrum: the frequency resource used by radio services. Interoperability: ability of systems to communicate through agreed interfaces.
38. Evidence and scope
The link budget, Shannon calculation, coding and packet-delay examples are original. They do not describe a real licensed radio service, commercial network or guaranteed performance.
Primary references include ITU-T, ITU-T Recommendations, ITU-R Recommendations, ITU’s 2026 IMT-2030 update and resilience discussion; Singapore IMDA’s spectrum management and network standards.
The deeper answer: telecommunications engineering preserves distinctions across distance
A message can cross thousands of kilometres only because every layer preserves enough distinction for the next receiver: voltage states, optical symbols, radio phases, packet identities, addresses, timing and application meaning.
Telecommunications engineering works when the complete path—not merely one transmitter or cable—delivers the intended information with the required timing, reliability and interoperability.
Continue: Biomedical Engineering uses communication and sensing in health technology; Robotics Engineering depends on signals and networked control; Nuclear Engineering depends on instrumentation, control and reliable communication under demanding conditions. Return to the How X Works Hub for the complete subject map.