The essentials that underpin all telecom concepts.

Key Concepts:

• Linear algebra (vectors, matrices)
• Complex numbers
• Probability & statistics
• Fourier Series & Fourier Transform
• Laplace Transform
• Convolution

Why It Matters: Telecom revolves around mathematical modeling of signals.

Labs/Practice: Analyzed signal transformations in MATLAB.

Tools Used: MATLAB, Python (NumPy).

Lesson 1 – Foundations (Non-negotiable)

Telecommunications Engineering Self-Study – Asma’s Portfolio
February 2026 – Tunis

Why Foundations Matter So Much

Telecommunications is not primarily about hardware or protocols at first — it is about moving information reliably through a noisy, imperfect world.
Every later topic (modulation, coding, 5G, fiber, security attacks) builds directly on these ideas.

• If you don’t understand why noise kills bits → you won’t grasp BER curves.
• If complex numbers feel alien → QPSK, OFDM, and beamforming will remain mysterious.
• If bandwidth and SNR are fuzzy → Shannon’s limit and 5G spectral efficiency won’t click.

Think of this as learning the alphabet before writing novels.

Part A: Mathematics – The Language of Telecom

(≈60–70% of the “brain work”)

You don’t need to become a pure mathematician, but telecom uses very specific math tools almost daily.

1. Linear Algebra (Vectors & Matrices)

Signals and channels are often multi-dimensional.

• A simple radio signal uses two dimensions: In-phase (I) and Quadrature (Q) → that’s a 2D vector.
• MIMO (Multiple Input Multiple Output) in 4G/5G uses matrices: if you have 4 antennas transmit and 4 receive, the channel is a 4×4 matrix describing how each path mixes signals.
• Beamforming = applying a weight vector to steer energy.

Example
A transmitted symbol in QAM is a point in 2D space:
s = [I, Q]
Received signal: r = H × s + n (matrix multiplication + noise vector).
Linear algebra lets you invert or estimate H.

2. Complex Numbers

The single most important “trick” in telecom.

A sinusoid like cos(ωt) or sin(ωt) can be written compactly as the real or imaginary part of e^{jωt} (Euler’s formula: e^{jθ} = cosθ + j sinθ).

Why this matters hugely:

• Phase and amplitude become one complex number instead of two separate equations.
• Modulation (changing amplitude/phase) = multiplying by a complex scalar.
• Frequency shifts = multiplying by e^{j2πft}.
• Convolution in time ↔ multiplication in frequency (much easier with complex exponentials).

Quick example
In QPSK, the four symbols are at angles 45°, 135°, 225°, 315° → simply 1+j, -1+j, -1-j, 1-j (after scaling).
Adding noise → the received point moves in the complex plane; decision is nearest constellation point.

3. Probability & Statistics

Noise is random → everything probabilistic.

• AWGN = Additive White Gaussian Noise → noise samples ~ Normal(0, σ²).
• Bit Error Rate (BER) = probability a bit flips.
• Outage probability, fading statistics (Rayleigh/Rician distributions), error correction success rates.

Key mindset
In telecom we rarely say “this will work”; we say “this will work with 99.999% probability” or “BER < 10^{-5}”.

4. Fourier Series & Fourier Transform

Any (reasonable) signal = sum of sinusoids at different frequencies.

• Time domain: s(t) (waveform you see on oscilloscope)
• Frequency domain: S(f) (spectrum analyzer view) → shows which frequencies carry power
• Bandwidth = range where most energy lives
• Filtering = multiplying in frequency domain (easy) instead of convolving in time
• OFDM (core of Wi-Fi, LTE, 5G) splits wideband channel into many narrow subcarriers → each flat-fading and easy to equalize

Intuition example
A square wave looks sharp in time → many high-frequency harmonics in frequency. Sharp edges = wide bandwidth.

5. Laplace Transform

Generalizes Fourier for stability analysis (mostly analog/control side, but useful).

• Used to analyze filters, system poles/zeros.
• In telecom mostly seen indirectly (filter design, PLLs in radios).

6. Convolution

The mathematical operation behind every linear filter.

Output y(t) = x(t) * h(t) (input convolved with impulse response).

• In frequency domain → Y(f) = X(f) × H(f) (multiplication = much faster)
• Digital filters (FIR/IIR) = discrete convolution

Practice tip
A matched filter (optimal for detecting known signal in AWGN) is literally the time-reversed conjugate of the signal → convolution peaks at correct timing.

Quick self-check questions (try mentally):

• Why can we represent a carrier signal as Re{A e^{j(ωt + φ)}} instead of A cos(ωt + φ)?
• If I have two signals x(t) and y(t), what does x(t) * y(t) represent in frequency domain?