EM Field Guide
Electromagnetism for Electronic Engineers

Electric charge is a fundamental property of matter, quantised in integer multiples of the elementary charge e = 1.602×10⁻¹⁹ C. Charge is conserved (cannot be created or destroyed, only transferred). Coulomb's Law gives the force between two point charges: it follows an inverse-square law identical in form to gravity, but enormously stronger and bipolar (like charges repel; unlike attract). In electronics, Coulomb's law governs: ESD events (charges accumulating on IC pins), capacitor plate attraction, and the field fringing at PCB trace edges. The force between two electrons separated by 1 nm is F ≈ 230 nN — negligible at circuit scale, but the collective force of trillions of electron-ion pairs is what drives all electrical phenomena.

The electric field E maps the force that would act on a positive unit test charge at every point in space. It is a vector field — at each point it has both magnitude and direction. The field lines of E point from positive charges toward negative charges and never cross. The critical engineering relationship is E = -∇V: the electric field is the negative gradient of potential — E points from high voltage toward low voltage, with magnitude equal to the rate of voltage change with distance. Between two parallel plates with voltage V and gap d: E = V/d [V/m]. The dielectric breakdown field of air is ~3 MV/m — this is why high-voltage traces must be spaced far apart on PCBs.

Gauss's Law (Maxwell's first equation) states that the total electric flux through any closed surface equals the enclosed free charge divided by permittivity. The power of Gauss's Law lies in its use with high-symmetry geometries where the E-field is constant on a chosen Gaussian surface, making the integral trivial. For a long coaxial cable (inner conductor radius a, outer conductor radius b), Gauss's Law immediately gives the radial E-field and thus the capacitance per unit length: C/L = 2πε/ln(b/a) [F/m]. D = εE — the displacement field D accounts for both free charges and bound charges in dielectrics, and is the quantity whose flux is related solely to free charge.

Electric potential V is the potential energy per unit charge — a scalar field, far easier to work with than the vector field E. Voltage between two points is the work done per unit charge to move a positive test charge from A to B against the electric field: V_AB = -∫_A^B E·dl. Since E is conservative (in electrostatics), this line integral is path-independent — only the endpoints matter. Equipotential surfaces are always perpendicular to E-field lines. The gradient relationship E = -∇V is the key computational bridge: find V first (scalar, easier), then differentiate to get E. On a PCB, every net is (ideally) an equipotential surface; parasitic resistances and inductances create unwanted potential gradients across them.

Capacitance is a purely geometric property (for a given dielectric): it depends on the shape and size of the conductors and the material between them, not on the charge or voltage. Any two conductors separated by an insulator form a capacitor — this is why parasitic capacitance is everywhere in electronics (trace-to-trace, pin-to-ground, bonding wire capacitance). For a parallel-plate capacitor: doubling the area doubles C; halving the gap doubles C; doubling εᵣ doubles C. In IC technology, capacitance per unit area scales as C/A = ε₀εᵣ/t_oxide — a 3 nm SiO₂ gate oxide (εᵣ ≈ 3.9) gives C/A ≈ 11.5 fF/μm². The transition to high-k dielectrics (HfO₂, εᵣ ≈ 25) allows thicker physical oxides with the same capacitance.

When a dielectric is placed in an electric field, bound charges within the material shift slightly — positive nuclei move with the field, negative electrons against it — creating electric dipoles. This polarisation P reduces the net field inside the material and increases the charge that can be stored for a given voltage. The relative permittivity εᵣ (also called dielectric constant) quantifies this effect. In PCB design, εᵣ of the substrate material determines trace impedance, signal velocity, and coupling between adjacent traces. FR4 has εᵣ ≈ 4.2–4.8 (varies with frequency and temperature); Rogers RO4003C has εᵣ = 3.55 ±0.05 — critical for controlled-impedance RF designs. Dielectric loss is characterised by the loss tangent tan δ — FR4 tan δ ≈ 0.02 at 1 GHz; RO4003C tan δ = 0.0027. High loss tangent means signal attenuation and heating in the substrate.

The energy stored in a capacitor is distributed throughout the electric field between the plates — this is a fundamental insight of field theory. The energy density u = ½εE² allows the total energy to be computed for any geometry by integrating over all space. For engineers, the lumped-circuit form U = ½CV² is more practical. A 10 μF capacitor charged to 400 V (common in switching power supplies) stores U = ½ × 10×10⁻⁶ × 160000 = 0.8 J — sufficient to cause serious injury or ignite flammable material on discharge. Capacitive energy storage in power converters, camera flash circuits, and pulsed power systems all rest on this equation.

The microscopic form of Ohm's Law, J = σE, states that current density (current per unit cross-sectional area) is proportional to the local electric field. This is more fundamental than the circuit-level V = IR because it applies point-by-point in the conductor — including non-uniform geometries. Current flows in the direction of E (for positive carriers). Copper: σ = 5.96×10⁷ S/m; silicon (intrinsic): σ ≈ 4.4×10⁻⁴ S/m — a difference of 11 orders of magnitude. At high frequencies, the effective cross-section of a conductor shrinks due to the skin effect, increasing resistance — this is why RF conductors are hollow tubes or silver-plated.

The continuity equation is the mathematical statement of charge conservation: charge cannot be created or destroyed, so any decrease in charge within a volume must produce a net current flowing out of that volume. In a DC steady state, ∂ρ/∂t = 0, so ∇·J = 0 — current flow lines have no sources or sinks (Kirchhoff's Current Law is a consequence of this). In a transient, charge can accumulate or deplete locally. The continuity equation, together with Gauss's Law, implies that any free charge in a conductor decays with a time constant τ = ε/σ — for copper, τ ≈ 10⁻¹⁹ s (effectively instantaneous). KCL is the circuit-level expression of this fundamental law.

Two quantities describe magnetic fields: B (magnetic flux density, or "the B-field") is the fundamental field quantity — it determines the force on moving charges. H (magnetic field intensity) is the "source field" — it is related to the free currents that produce the field, and is the quantity that satisfies Ampère's Law in simple form. In free space and non-magnetic materials, B = μ₀H. In magnetic materials, B = μ₀(H + M) = μ₀μᵣH where M is the magnetisation. Engineers typically work with H for magnetic circuit analysis (analogous to Kirchhoff's voltage law for reluctance circuits) and with B for force and flux calculations. B field lines always form closed loops (no magnetic monopoles: ∇·B = 0).

The Biot–Savart Law is the magnetic analogue of Coulomb's Law for electric fields — it gives the B-field contribution from each infinitesimal current element, which must then be integrated over the entire current path. For an infinite straight wire carrying current I, the field circles the wire with magnitude B = μ₀I/(2πr) — at r = 10 mm from a 1 A wire, B = 20 μT. This has direct EE consequences: parallel PCB traces carrying opposing currents (differential pairs) have fields that cancel in the far field, dramatically reducing EMI radiation. Conversely, a high-current bus bar creates a B-field that induces voltages in nearby loops — a common source of interference.

Ampère's Law is the magnetic analogue of Gauss's Law — it relates the line integral of H around any closed path to the net current threading through any surface bounded by that path. For high-symmetry geometries, it makes field calculations trivial. Toroidal inductor: N turns carrying current I on a toroid of mean radius r: H = NI/(2πr) — the field is entirely confined within the core. This is why toroidal inductors have negligible external radiation — and why they are preferred in audio and power applications where EMI matters. Coaxial cable: outside the outer conductor, the fields from inner and outer conductors cancel exactly — coax is inherently self-shielding.

Magnetic flux Φ is the total magnetic flux density integrated over an area — it quantifies "how much B passes through" a surface. The law ∇·B = 0 (Maxwell's second equation) states there are no magnetic monopoles — B-field lines always form closed loops. Every field line that enters a closed surface must exit it. This has practical consequences: in transformer design, all flux lines threading the primary must thread the secondary (in an ideal transformer) — any flux that doesn't, "leakage flux", contributes to leakage inductance. In magnetic circuit analysis: Φ = NI/ℜ where ℜ [A/Wb] is reluctance — the magnetic analogue of resistance.

Faraday's Law (Maxwell's third equation) is arguably the most important equation in electrical engineering: any changing magnetic flux through a closed loop induces an EMF in that loop. This is the operating principle of every transformer, generator, and electric motor. For a lumped inductor: v = L·di/dt is a direct consequence (L = dλ/di). The negative sign (Lenz's Law) means the induced EMF opposes the change producing it. In high-speed PCB design, changing currents in one trace create changing B-fields that induce voltages in adjacent loops — this is inductive crosstalk, one of the primary signal integrity failure modes. Even a 1 mm × 1 mm loop with 1 A at 1 GHz generates significant induced interference in nearby circuits.

Lenz's Law is the physical content of the minus sign in Faraday's Law: the induced EMF always drives a current whose magnetic field opposes the change in the original flux. It is a consequence of energy conservation — if induction aided the change, it would create a runaway positive feedback. Lenz's Law governs: back-EMF in motors (a running motor opposes its driving voltage — the current drawn decreases as speed increases); inductive kickback (a suddenly interrupted inductor current induces a high-voltage spike opposing the interruption — hence flyback diodes); eddy current braking; and the snubber circuit requirement across relay coils and transformer windings.

Self-inductance L is the flux linkage per unit current in the same coil — a measure of how strongly a circuit "fights" changes in its own current. Mutual inductance M is the flux linkage in coil 2 due to unit current in coil 1. The coupling coefficient k (0 = no coupling, 1 = perfect coupling) characterises how much of coil 1's flux links coil 2. In transformer design, maximising k (tight coupling, good core material, minimal leakage) improves efficiency; leakage inductance L_lk = (1-k²)L limits performance. On a PCB, two parallel traces have mutual inductance that causes inductive crosstalk — a rising current in one induces a voltage in the other proportional to M·dI/dt.

Magnetic energy is stored in the B-field throughout space, with density u = B²/(2μ). For a 100 μH inductor carrying 10 A: U = ½ × 100×10⁻⁶ × 100 = 5 mJ. This energy must go somewhere when current is interrupted — if no path is provided (flyback diode), it appears as a destructive voltage spike. In a buck converter, the inductor stores energy during the on-time and releases it during the off-time — the entire energy conversion cycle depends on U = ½LI². In a flyback converter (isolated SMPS), the transformer primary stores energy that is later transferred to the secondary — it operates as a coupled inductor, not a true transformer.

Ferromagnetic materials (iron, nickel, cobalt and their alloys) have enormously high μᵣ because the magnetic domains align with applied H. Above the Curie temperature (768°C for iron), this ordering is lost. The B-H curve (hysteresis loop) is the core characterisation of a magnetic material: saturation flux density B_sat, remanence B_r (residual field when H = 0), and coercivity H_c (field needed to demagnetise). Soft magnetic materials (high μᵣ, low H_c) are used for transformers and inductors — they're easy to magnetise and demagnetise with low losses. Hard magnetic materials (high H_c, high B_r) are used for permanent magnets. The area enclosed by the hysteresis loop equals the energy dissipated per cycle — key to core loss calculations in power magnetics.

The Lorentz Force is the total electromagnetic force on a charged particle — the sum of the electric force qE and the magnetic force qv×B. The magnetic force is always perpendicular to the velocity and does no work — it can only change direction, not speed. This is why magnetic fields steer electrons in CRT displays and mass spectrometers. Force on a current-carrying conductor follows from the Lorentz force on its charge carriers: a straight conductor of length L carrying current I in a uniform field B experiences force F = BILsin(θ). This is the operating principle of every electric motor, loudspeaker, and galvanometer. Hall sensors measure B by the transverse voltage developed across a current-carrying conductor due to the Lorentz force on its carriers.

Maxwell's pivotal contribution was adding the displacement current ∂D/∂t to Ampère's Law. Without it, Ampère's Law was inconsistent in time-varying situations (e.g., a charging capacitor: current flows in the wire but not between the plates — yet the original Ampère's Law would give different H depending on which surface you chose). The displacement current J_D = ε∂E/∂t flows "through" the capacitor gap and resolves this inconsistency. More profoundly, the symmetry between Faraday's Law (changing B creates E) and the new Ampère's Law (changing E creates B) predicts that self-sustaining electromagnetic waves can propagate through empty space — and from the predicted speed c = 1/√(ε₀μ₀), Maxwell identified light as an electromagnetic wave.

These four equations, together with the Lorentz Force Law and the constitutive relations (D = εE, B = μH, J = σE), provide a complete description of all classical electromagnetic phenomena. From these equations one can derive: the wave equation (and predict light), Snell's law, the skin effect, transmission line behaviour, antenna radiation, and every other EM phenomenon an engineer encounters. The integral forms are best for computing fields in symmetric geometries or understanding boundary conditions. The differential (vector) forms are best for deriving wave equations and understanding field structure. Every calculation in RF engineering ultimately traces back to these four equations.

Taking the curl of Faraday's Law and substituting the Ampère-Maxwell Law yields the electromagnetic wave equation — a second-order PDE whose solutions are transverse electromagnetic (TEM) waves propagating at speed c. The E and H fields are always perpendicular to each other and to the direction of propagation. In a medium with relative constants εᵣ and μᵣ, the phase velocity becomes v_p = c/√(εᵣμᵣ). For FR4 (εᵣ ≈ 4.5, μᵣ = 1): v_p = c/√4.5 ≈ 0.471c — signals travel at 47% of light speed. At 1 GHz on FR4, λ = v_p/f ≈ 141 mm — any trace longer than ~14 mm (λ/10) must be treated as a distributed element.

The wave impedance η is the ratio of E to H in a plane wave — it is a property of the medium, not the wave. In free space, η₀ = 377 Ω. This number is crucial for EMC engineering: the impedance of the electromagnetic field in space determines how much of an incident wave reflects from a conductive shield. The mismatch between 377 Ω (free space) and the near-zero impedance of a conductor is what makes metallic shielding so effective. For a good conductor with conductivity σ: η_c = √(jωμ/σ) = (1+j)/σδ where δ is skin depth — and |η_c| ≪ 377 Ω, giving near-total reflection. The 377 Ω also appears in antenna theory: a perfect half-wave dipole has radiation resistance of 73 Ω — a significant fraction of η₀.

Polarisation describes the orientation of the E-field vector in the plane perpendicular to propagation. Linear polarisation: E oscillates along a fixed direction — standard for most antennas (vertical, horizontal). Circular polarisation (CP): E rotates, tracing a circle — RHCP or LHCP, used in GPS (L1/L2 bands), satellite comms, and RFID. The key advantage of CP: a circularly polarised wave can be received by a linearly polarised antenna in any orientation without a 3 dB polarisation loss (worst case for linear-to-linear mismatch). Polarisation mismatch between transmit and receive antennas causes polarisation loss factor (PLF): two orthogonal linear antennas have 100% loss (infinite dB) — complete signal null. This is exploited in polarisation diversity and cross-polarisation isolation in antenna design.

The Poynting Vector S = E × H gives the instantaneous power per unit area flowing in an EM field — it points in the direction of energy flow. Integrating S over a closed surface gives the net power entering or leaving that volume. The Poynting theorem (electromagnetic energy conservation) states: power flowing out of a volume = rate of decrease of stored EM energy + power dissipated as heat. Critically, power in a wire doesn't flow inside the copper — it flows in the E and H fields surrounding the wire, guided into the conductor. A coaxial cable with V = 100 V and I = 10 A: the 1 kW of power flows in the dielectric space between inner and outer conductors, not in the copper. Poynting vector analysis also determines far-field radiated power from antennas.

At DC, current distributes uniformly across a conductor's cross-section. At higher frequencies, the induced eddy currents (from Faraday's Law) oppose current flow in the interior, confining it to a thin skin of depth δ near the surface. The current density decays exponentially: J(x) = J_0·e^(-x/δ). At one skin depth, current is 37% (1/e) of the surface value. AC resistance increases as R_AC/R_DC = r/(2δ) for a round wire of radius r ≫ δ. In practice: at 1 GHz, only the outermost 2 μm of a copper conductor carry significant current — hence silver plating (slightly better conductivity) on RF connectors and waveguides, and surface roughness becoming critical (micro-roughness increases effective path length, increasing attenuation).

When an EM wave hits a boundary between two media, part reflects (reflection coefficient Γ) and part transmits (τ). The driving quantity is the impedance mismatch between the two media. This is the exact electromagnetic analogue of voltage reflections on mismatched transmission lines. At a perfect conductor: η₂ → 0, Γ → -1 (total reflection with phase reversal). Air-to-FR4 (η₁=377Ω, η₂=177Ω): Γ = (177-377)/(177+377) = -0.36 — 13% of power reflects at normal incidence. This is why RF absorbers use graded impedance transitions (pyramid-shaped foam) — gradual impedance change from η₀ to near-zero minimises reflections through many small steps rather than one large mismatch.

Snell's Law applies to EM waves at dielectric interfaces — the phase matching condition requires the tangential component of the wave vector to be continuous across the boundary, yielding the familiar sine relationship. The refractive index n = √(εᵣμᵣ). Total Internal Reflection (TIR) occurs when a wave in a denser medium hits a less dense medium at an angle exceeding θ_c — all power is reflected. TIR is the basis of optical fibre: light confined in high-n glass core by lower-n cladding. In microwave engineering, substrate-bound surface waves can undergo TIR at the PCB edge, coupling into the space above the board and creating spurious radiation. Careful substrate thickness control and edge effects management are needed in millimetre-wave PCB designs.

When a signal trace is electrically long (length > λ/10), it must be modelled as a distributed network of R, L, G, C per unit length. The Telegrapher's Equations describe voltage and current propagation along such a structure. Solutions are forward- and backward-travelling waves: V(z,t) = V⁺e^(γz) + V⁻e^(-γz). The propagation constant γ = α + jβ contains both attenuation (α, in Np/m or dB/m) and phase shift (β, in rad/m). For a lossless line (R' = G' = 0): γ = jβ = jω√(L'C') — pure phase shift. The phase velocity v_p = ω/β = 1/√(L'C'). All signal integrity analysis in modern high-speed digital design starts from the RLGC model.

Z₀ is the impedance seen looking into an infinite transmission line — the ratio of voltage to current for a travelling wave in one direction. It is determined entirely by the line geometry and dielectric, not by the terminations or source. Standard values: 50 Ω (RF, test equipment, coax), 75 Ω (cable TV, CATV), 100 Ω (differential pairs, Ethernet), 90 Ω (USB 2.0 differential). On a PCB, Z₀ can be controlled by adjusting trace width w and/or dielectric height h. PCB manufacturers typically offer ±10% Z₀ control on standard processes, ±5% for controlled-impedance. Widening the trace lowers Z₀ (more C', less L'); thicker dielectric raises Z₀ (less C' relative to L').

When a transmission line is terminated with a load impedance Z_L ≠ Z₀, the mismatch causes a reflection. The reflection coefficient Γ quantifies this: |Γ| = 0 is a perfect match; |Γ| = 1 is total reflection (open circuit: Z_L → ∞, Γ = +1; short circuit: Z_L = 0, Γ = -1). VSWR (Voltage Standing Wave Ratio) is the ratio of maximum to minimum voltage amplitude along the line, caused by the superposition of forward and reflected waves. VSWR = 1.0 is perfect; VSWR = 2.0 means |Γ| = 0.333, Return Loss = 9.5 dB, 11% of power reflected. Typical antenna system requirement: VSWR < 2:1 (Return Loss > 9.5 dB). Sensitive receiver front-ends may require VSWR < 1.5:1 (RL > 14 dB).

The general input impedance formula shows that a transmission line transforms impedance — the impedance seen at the input depends on the load, line impedance, and electrical length. Two special cases: λ/4 line (quarter-wave transformer): transforms Z_L to Z₀²/Z_L — a short circuit becomes an open circuit and vice versa, and it can match any real impedance to Z₀ with intermediate Z₀ = √(Z_s·Z_L). λ/2 line: repeats the load impedance at the input — transparent. These are the fundamental tools of narrowband RF matching. λ/4 transformers are used to match 50 Ω lines to 73 Ω dipoles: Z₀_transformer = √(50×73) = 60.4 Ω.

The Smith Chart is a graphical representation of the complex reflection coefficient Γ plotted in polar form, with overlaid contours of constant resistance (r) and constant reactance (x). It is the single most useful tool for RF matching and transmission line analysis. Reading a Smith Chart: Circles of constant r pass through the right edge (r = 0 to ∞); arcs of constant x pass through the right edge. Moving along a lossless line corresponds to rotating around the centre at 2β per unit length, clockwise (toward generator). Adding a series inductance moves along constant-r circles upward; adding a series capacitance moves downward; adding a shunt element moves on the admittance chart (rotated 180°). Modern VNAs display S11 directly on the Smith Chart — mastery of the Smith Chart is essential for any RF engineer.

Maximum power is transferred from source to load when the load impedance is the complex conjugate of the source impedance. For real impedances, this means Z_L = Z_S (same value). L-network matching: Two reactive elements (inductor and capacitor) can match any real Z_S to any real Z_L over a narrow band. The Q of the network determines bandwidth: Q = f₀/BW. Higher Q = narrower match, lower IL in band, but more sensitive to component variations. Pi and T networks offer one extra degree of freedom (three reactive elements) allowing Q to be independently chosen. Wideband matching uses cascaded sections or transmission-line techniques. LNA input matching is critical: noise figure is minimised at the noise-optimal impedance, which may differ from the conjugate match — the designer must balance gain and noise.

The lumped-circuit model (Kirchhoff's Laws, V = IR, Z = jωL, Z = 1/jωC) assumes that the physical dimensions of every component are negligible compared to the wavelength of the signal — that voltage and current are uniform throughout each element. When this breaks down (element size > λ/10), distributed effects dominate: the phase of voltage and current vary along the element, and reflections and standing waves appear. The λ/10 boundary is a primary design driver in modern electronics: at 5G mmWave frequencies (28–39 GHz), even a 1 cm trace is 2–3λ long on FR4 and must be modelled as a transmission line. This forces massive changes in package design, PCB layout, and the very concept of what constitutes a component.

S-parameters measure network behaviour in terms of travelling power waves rather than open/short-circuit voltages and currents — making them ideal for RF because they don't require impractical perfect opens or shorts at high frequency, and they directly relate to the matched load conditions under which RF circuits operate. S11 = input return loss; S21 = forward insertion gain (or loss); S12 = reverse isolation; S22 = output return loss. A VNA (Vector Network Analyser) directly measures all S-parameters versus frequency. Typical amplifier specs: S21 = +20 dB (gain), S11 = -15 dB (input RL), S22 = -12 dB (output RL), S12 = -30 dB (isolation). For a reciprocal, passive network: S12 = S21. For a lossless network: the S-matrix is unitary.

A metallic hollow tube can guide microwave energy in discrete modes characterised by their field configurations. Unlike TEM transmission lines (coax, microstrip), waveguides support only TE or TM modes (not both field components transverse) and have a cutoff frequency f_c — below which signals cannot propagate. The dominant TE₁₀ mode in a rectangular waveguide has cutoff at f_c = c/(2a); practical operation is from ~1.25f_c to ~1.9f_c to avoid multimode propagation. Advantages over coax at microwave frequencies: much lower attenuation (no centre conductor skin effect), higher power handling, excellent shielding. Standard waveguide bands: WR-90 (X-band, 8.2–12.4 GHz), WR-28 (Ka-band, 26.5–40 GHz). Widely used in radar, satellite comms, and point-to-point microwave links.

A closed metallic box (resonant cavity) supports discrete resonant modes at frequencies determined by its dimensions. At resonance, standing waves form — stored energy oscillates between E and H fields with Q limited only by conductor losses (skin-effect dissipation). Cavity Q values of 10,000–100,000 are achievable — far superior to lumped LC resonators (Q ≈ 50–500). Cavities are used in: microwave filters (bandpass filters with cavities of defined frequencies), klystron and magnetron tubes, cavity-backed antennas, particle accelerator cavities (Q > 10⁹ in superconducting versions). PCB power plane resonance (a low-Q cavity formed by power and ground planes) is a significant noise source in digital designs above 100 MHz.

Any accelerating charge radiates electromagnetic energy. In antenna terms, time-varying currents (dI/dt ≠ 0) radiate because accelerating charges produce changing E and H fields that propagate outward as EM waves. The Hertzian dipole (infinitesimal current element Il) is the fundamental radiating source. Its radiation resistance R_rad = 80π²(l/λ)² scales as (l/λ)² — a short antenna (l ≪ λ) has very small R_rad and is thus inefficient: most input power is dissipated as heat in the wire resistance, not radiated. A half-wave dipole (l = λ/2) has R_rad = 73 Ω — well-matched to a 75 Ω feedline. This is why antenna size matters: electrically short antennas require matching networks and have inherently narrow bandwidth.

Directivity D measures how concentrated the radiated power is in the peak direction compared to an isotropic (equal radiation in all directions) radiator, assuming all input power is radiated. Gain G further includes radiation efficiency η_eff — losses in the antenna reduce gain below directivity. An isotropic antenna (theoretical) has G = 0 dBi. A half-wave dipole: D = 2.15 dBi. A 3-element Yagi: G ≈ 8 dBi. A parabolic dish of diameter D at wavelength λ: G = η·(πD/λ)² ≈ 55–65% efficiency. A 1 m dish at 10 GHz (λ = 30 mm): G ≈ 0.6 × (π×1/0.03)² = 65,800 = 48.2 dBi. High gain = narrow beam = must be accurately pointed. A 30 dBi antenna has a half-power beamwidth of roughly 1.8°/√(G_linear).

The Friis equation gives received power in a free-space link between two antennas of known gain separated by distance R. The term (λ/4πR)² is the free-space path loss factor — power spreads over an ever-larger sphere as it propagates. Link budget analysis: start with transmit power, add gains, subtract path loss and all other losses (cable losses, atmospheric absorption, polarisation mismatch), and compare to receiver sensitivity. Link margin = received power − sensitivity threshold. Wi-Fi example: P_t = 20 dBm, G_t = G_r = 2 dBi, R = 20 m, f = 2.4 GHz: FSPL = 62 dB, P_r = 20+2+2-62 = -38 dBm. Receiver sensitivity = -90 dBm. Margin = 52 dB — comfortable.

The electromagnetic field around an antenna has three distinct regions. In the reactive near field (immediately around the antenna), energy oscillates back and forth without net radiation — stored reactive energy dominates. Coupling a probe into this region significantly loads the antenna (changes its impedance). In the Fresnel region, radiation dominates but the phase of the wavefront varies significantly over any aperture — the radiation pattern is distance-dependent. In the far field (Fraunhofer region), the wavefront is approximately plane, radiation pattern is distance-independent, and E/H = η₀. Antenna gain patterns are always specified in the far field. Near-field communications (NFC, 13.56 MHz) deliberately exploits the reactive near field — two coupled inductors exchanging energy, not propagating waves.

Electromagnetic Compatibility is the discipline of ensuring that electronic equipment neither produces interference that disrupts other equipment (emissions) nor is disrupted by ambient interference (immunity). Key standards: FCC Part 15 (USA), CISPR 32/EN 55032 (EU) for radiated and conducted emissions; IEC 61000-4 series for immunity (ESD, EFT, surge, radiated field). Class B limits (residential) are stricter than Class A (industrial). A single high-frequency digital clock on an otherwise quiet PCB can cause FCC test failure — hence EMC design must begin at schematic stage, not as an afterthought. The most common emission sources in digital electronics: clock harmonics radiated from traces, common-mode currents on cables, DC-DC converter switching noise.

Shielding works by two mechanisms: reflection (impedance mismatch between free space at η₀ and the low-η conductor) and absorption (attenuation within the conductor skin depth). For a good conductor at high frequencies, absorption dominates — at 1 GHz, just 0.1 mm of copper absorbs the wave to effectively zero. The weakest link principle: a single unsealed aperture (slot, cable penetration, connector gap) dominates the shield effectiveness. A slot of length L acts as a slot antenna — its SE drops to SE_slot = 20·log₁₀(λ/2L). A 10 mm slot at 1 GHz (λ = 300 mm): SE_slot = 20·log₁₀(15) = 23.5 dB — limiting the shield to 23.5 dB regardless of wall thickness. This is why EMC enclosures use conductive gaskets at seams.

A "ground" is not zero-impedance at high frequencies — it has inductance and resistance, and any current flowing through it creates a voltage. Ground loops occur when a signal has multiple return paths at different potentials — any noise current in the shared ground path appears as an interference voltage on the signal. Single-point ground (star ground) works well at low frequencies — each circuit returns to a single point, preventing shared-impedance coupling. Solid ground plane works best at high frequencies — extremely low inductance, provides a direct low-impedance return path for every trace above it. Never use a single-layer PCB with a ground trace (not a plane) for any circuit above 1 MHz. The return current under a microstrip trace mirrors the signal current — any break in the ground plane forces the return current to detour, increasing loop area and radiation.

Crosstalk is the unwanted coupling of signals between adjacent traces. Capacitive (electric) coupling occurs between traces via fringing electric fields — a voltage on the aggressor drives a displacement current through C_m into the victim. Inductive (magnetic) coupling occurs via shared magnetic flux — a changing current in the aggressor induces EMF in the victim loop via mutual inductance M. The two mechanisms can add or subtract at different ends of the line. Differential signalling (LVDS, USB, HDMI, PCIe) dramatically reduces crosstalk by cancellation: the two lines carry opposite-polarity signals, and their fields largely cancel in the far field. Design rules: 3W spacing (traces spaced by 3× their width) reduces capacitive coupling by ~70%; route sensitive traces perpendicular to aggressors on adjacent layers.

Controlling trace impedance is essential for signal integrity above 100 Mb/s. Reflections from impedance discontinuities cause ringing, overshoot, and data errors. Microstrip (trace on outer layer above ground plane) has some field in air — effective εᵣ ≈ (εᵣ+1)/2; signals propagate slightly faster than in full dielectric. Stripline (trace embedded between two ground planes) has field entirely in dielectric — well-defined εᵣ, lower crosstalk, no radiation, but slower propagation and more loss. Design process: specify Z₀ (typically 50 Ω single-ended, 100 Ω differential); provide dielectric stackup to PCB manufacturer; they adjust trace width to hit target. Via design (antipads, via stubs) must also be controlled — a via stub acts as a shunt stub resonator, notching the frequency response.

The most important signal integrity principle: every signal current has a return current — and the return current path is as important as the signal path. At low frequencies, return current takes the path of least resistance. Above about 1 MHz (depending on geometry), it takes the path of least inductance — which is directly beneath the signal trace in the ground plane. The signal/return pair forms a tightly coupled transmission line with the smallest possible loop area, minimising inductance and radiation. Any discontinuity in the reference plane (slot, cutout, layer change without local bypass capacitor) forces return current to detour, creating a larger loop area, higher inductance, and increased radiation and crosstalk. Never route a high-speed signal across a split plane.

The power and ground planes of a PCB form a parallel-plate transmission line resonator. At resonant frequencies, voltage standing waves across the planes can cause large noise voltages on the supply rail — this shows up as a high peak in the PCB's power delivery network (PDN) impedance. For a 100×150 mm FR4 board: lowest resonance f₁₀ = v_p/(2a) = (c/√4.5)/(2×0.15) ≈ 472 MHz. At this frequency, the PDN impedance can peak to tens of ohms — causing power supply noise that appears as EMI and timing jitter. Solutions: distribute decoupling capacitors across the board (not just near ICs) to create a low-impedance bypass at resonant frequencies; use thin dielectrics (high C' per unit area) between power and ground layers; use embedded capacitance layers.

A decoupling capacitor placed near an IC pin provides a local charge reservoir — it supplies the instantaneous current surge when the IC switches, preventing the supply voltage from drooping while the slower board-level power supply responds. But a real capacitor has equivalent series inductance (ESL) (package inductance + via inductance + trace inductance) which limits its effectiveness at high frequency. Above the self-resonant frequency, the capacitor behaves as an inductor — useless for decoupling. Best practice: use multiple capacitor values in parallel (10 μF + 100 nF + 1 nF) to maintain low impedance across a wide frequency range; place smallest caps closest to IC power pins; minimise via and trace inductance by using multiple vias and wide traces; use a solid, uninterrupted ground/power plane pair directly under the IC.

Eddy currents are induced currents (Faraday's Law) that flow circularly within a conductive core material in response to a time-varying B-field. They produce ohmic heating (I²R losses) and create a field that opposes the inducing B-field. Reduction techniques: laminate the core (silicon steel sheets 0.3–0.5 mm thick separated by insulation) — eddy currents are limited to the thin lamination width; use ferrite cores (electrically insulating — essentially zero σ, so eddy current loss is negligible). At SMPS frequencies (>20 kHz), laminated silicon steel becomes lossy; ferrite cores are mandatory. The Steinmetz equation captures the frequency and flux density dependence of total core loss — essential for transformer design and thermal management.

The ideal transformer transforms voltage by turns ratio n, current by 1/n, and impedance by n². In a real transformer, three loss mechanisms and three parasitic elements must be modelled: magnetising inductance L_m (finite core permeability — requires finite magnetising current); leakage inductance L_lk (flux that doesn't link both windings — limits bandwidth and causes voltage spikes on switching); winding resistance (copper loss, skin and proximity effect); core loss (hysteresis and eddy current). The proximity effect is particularly important in high-frequency transformer design: eddy currents induced by adjacent winding layers concentrate current near the surface, increasing effective resistance. Litz wire (many fine individually insulated strands) reduces proximity effect losses.

Ferrite beads are lossy inductors — the complex permeability of ferrite has both real (μ') and imaginary (μ'') components. At low frequencies, the bead is inductive (μ' dominates, low loss). Above the bead's SRF, μ'' dominates — the bead becomes resistive and absorbs RF energy as heat rather than reflecting it. This is why ferrite beads are superior to inductors for EMI suppression: reflected noise can cause problems elsewhere; absorbed noise is simply dissipated. Common-mode chokes (two windings wound together on a ferrite core) present high impedance to common-mode currents (both conductors carrying current in the same direction) while presenting near-zero impedance to differential-mode current (normal signal). Used on every USB cable, Ethernet cable, and HDMI — to suppress common-mode noise without affecting the differential signal.

ESD is the sudden flow of charge between two objects at different electrostatic potentials. Gate oxides of 3-5 nm thickness can withstand only ~10-15 V before dielectric breakdown — an ESD event of kilovolts is catastrophically destructive. HBM (Human Body Model) simulates a person touching a device: 100 pF discharged through 1.5 kΩ, producing ~1.3 A peak current lasting ~150 ns. CDM (Charged Device Model) is the fastest and most destructive: the device itself accumulates charge, then discharges when grounded — rise times < 200 ps. Protection strategies: TVS diodes (transient voltage suppressors) clamp voltage; ESD-protected I/O structures on-chip; PCB layout keeping ESD paths away from sensitive signals; grounding straps in manufacturing; ANSI/ESD S20.20 handling procedures. A good ESD protection diode must clamp faster than the CDM pulse — sub-nanosecond response is required.

Thermal noise arises from the random thermal motion of electrons in any conductor — it is a direct consequence of the equipartition theorem of statistical mechanics, and is an electromagnetic phenomenon (fluctuating currents create fluctuating fields). It is unavoidable and white (flat spectrum to very high frequencies). At T = 290 K (room temp): noise power = kT = 4×10⁻²¹ W/Hz = -174 dBm/Hz. This sets the absolute floor for any RF receiver. A 50 Ω resistor generates V_n = √(4×1.38e-23×290×50×B). In a 1 MHz BW: V_n = 0.9 μV_rms. Noise Figure (NF) quantifies how much an amplifier degrades SNR: NF = SNR_in/SNR_out in dB. A 3 dB NF means half the SNR is lost (noise doubled). LNA design targets NF < 1 dB in many applications.

The boundary conditions are the field-theoretic equivalent of KVL and KCL — they enforce field continuity at the interface between two different media, derived directly from the integral forms of Maxwell's equations. Tangential E is continuous — if it weren't, an infinite force would act on surface charges. Normal B is continuous — from ∇·B = 0, no flux can terminate at an interface. Tangential H is discontinuous by the surface current K — in a perfect conductor, K can be nonzero, allowing E_t = 0 inside while H_t is nonzero outside. Normal D is discontinuous by surface charge density ρ_s. These conditions are the starting point for deriving the Fresnel reflection/transmission coefficients, waveguide modal fields, and the behaviour of antennas in the presence of ground planes.

In sinusoidal steady state at angular frequency ω, all field quantities vary as e^(jωt). The time-derivative operator ∂/∂t becomes multiplication by jω, converting Maxwell's PDEs into algebraic equations in the complex amplitudes (phasors). This enormously simplifies analysis — impedance, admittance, gain, and phase shift are all phasor ratios. The propagation constant γ = α + jβ, the complex permittivity ε_c = ε' - jε'' = ε(1 - j·tan δ), and the complex permeability μ_c all arise naturally. The loss tangent tan δ = ε''/ε' = σ/ωε characterises material lossiness. All S-parameter, impedance, and spectrum analyser measurements are phasor-domain measurements. Vector network analyser (VNA) calibration and error correction algorithms operate entirely in complex phasor space.

In electrostatics, a charge at position r' creates an instantaneous potential at observation point r. In electrodynamics, information travels at speed c — the potential at r and time t is determined by the retarded state of the source at time t_ret = t - |r-r'|/c (the time it took light to travel from r' to r). This finite propagation speed is the origin of radiation: a charge that was at one position at time t_ret has since moved — its field at r still "remembers" its earlier position and the discontinuity between old and current positions drives a radiation field that falls as 1/r (while static fields fall as 1/r²). This is why only accelerating charges radiate (a uniformly moving charge has a retarded potential pattern that exactly cancels the near-static field — Feynman's derivation). All antenna analysis — from simple dipoles to phased arrays — ultimately traces back to these retarded potential integrals.

The Larmor formula gives the total power radiated by a non-relativistically accelerating point charge. It confirms: only accelerating charges radiate. A charge at rest: no radiation. A charge moving at constant velocity: no radiation (though it has a field). A charge undergoing sinusoidal oscillation (current in an antenna): radiates continuously. This is the microscopic origin of antenna radiation. In RF engineering: the oscillating current in an antenna conductor accelerates electrons sinusoidally — each electron radiates according to Larmor, and the collective radiation of ~10²³ electrons/cm³ adds coherently to produce the macroscopic antenna field. Relativistic extension (Liénard formula) predicts synchrotron radiation (charges bent by magnetic fields) and is relevant for particle accelerators and certain RF tube devices.

Classical Maxwell equations are extraordinarily accurate for all practical electronics, but break down in several domains. Quantum tunnelling: at gate oxide thicknesses below ~3 nm, electrons tunnel quantum-mechanically through the "classically forbidden" region — direct tunnel current becomes the primary leakage mechanism in sub-7 nm CMOS. Classical EM predicts zero current here; QM gives the correct non-zero tunnelling probability. Photon statistics: at very high frequencies (optical, X-ray) or very weak fields, light must be treated as photons (quantum optics). At THz frequencies, quantum noise (shot noise, vacuum fluctuations) becomes significant for high-sensitivity receivers. Casimir force: quantum fluctuations of the EM vacuum produce an attractive force between closely spaced conductors — relevant in MEMS devices at nanometre gaps. For all practical electronics from DC to 100+ GHz, classical Maxwell equations provide a complete and exact description.