Network theorems and circuit analysis

In short

Network analysis is the set of laws and theorems that let you find every voltage and current in a lumped circuit of sources, resistors, inductors and capacitors without solving it from first principles each time. Kirchhoff's two laws are the foundation; mesh and nodal analysis turn them into simultaneous equations; superposition splits a multi-source problem into one-source problems; Thevenin and Norton collapse a large network into a two-element equivalent at the terminals of interest, and maximum power transfer then fixes the best load. It is the first subject in every ECE curriculum (AICTE "Network Theory"/"Circuit Theory", UGC CBCS) because amplifier biasing, filters, transmission lines, power electronics and even device modelling all reduce to these methods. In exams it carries heavy weight in GATE EC ("Networks, Signals and Systems") and EE ("Electric Circuits"), ESE, ISRO, DRDO, BARC and BSNL JTO papers, and in placement technical rounds at BEL, HAL and similar PSUs, Texas Instruments, Qualcomm, Analog Devices India, Cadence and Synopsys, where the standard screen is "find the Thevenin equivalent" or "find the current in this branch".

Explained

Circuit elements, sources and sign conventions

• Lumped vs distributed: a lumped model assumes the element's size is far smaller than the signal wavelength, so voltage and current depend on time only - valid until circuit dimensions approach λ/10, beyond which transmission-line analysis is needed.
• Passive elements: resistor (dissipative, v = Ri), inductor (stores energy in a magnetic field, v = L di/dt), capacitor (stores energy in an electric field, i = C dv/dt). Energy: ½Li² and ½Cv².
• Classification of networks: linear/non-linear (obeys homogeneity and additivity or not), passive/active (cannot deliver net energy / can), bilateral/unilateral (resistor vs diode), lumped/distributed, time-invariant/time-varying, reciprocal/non-reciprocal.
• Independent sources: ideal voltage source has zero internal resistance and fixes v regardless of i; ideal current source has infinite internal resistance and fixes i regardless of v. A voltage source is never short-circuited, a current source never open-circuited, in a physical sense.
• Dependent (controlled) sources: four types - VCVS, VCCS, CCVS, CCCS. They model transistors and op-amps. In every theorem, dependent sources are never deactivated; only independent sources are.
• Passive sign convention: current entering the + terminal means p = vi is power *absorbed*; a negative result means the element delivers power. Tellegen's theorem guarantees Σ vk ik = 0 over all branches (power balance), a useful answer check.

Kirchhoff's laws and network reduction

• KCL (charge conservation): the algebraic sum of currents entering any node, or any closed surface (a supernode), is zero. Gives n − 1 independent equations for n nodes.
• KVL (energy conservation): the algebraic sum of voltages around any closed loop is zero. Gives b − n + 1 independent equations for b branches.
• Series: same current; Req = ΣR, Leq = ΣL, 1/Ceq = Σ(1/C). Parallel: same voltage; 1/Req = Σ(1/R), Ceq = ΣC, 1/Leq = Σ(1/L).
• Voltage divider applies only to series (same-current) elements; current divider only to parallel (same-voltage) elements. Loading a divider with a finite load invalidates the simple ratio.
• Source transformation: a V source with series R ⇄ an I = V/R source with the same R in parallel. Equivalent only at the external terminals - internal power dissipation differs. Not possible for an ideal source (R = 0 or ∞).
• Star-delta (Y-Δ) transformation removes the bridge resistor that blocks series-parallel reduction, e.g. an unbalanced Wheatstone bridge.
• Symmetry and balanced bridge: if R1/R2 = R3/R4 the bridge is balanced, the detector branch carries zero current and can be removed (open) or shorted - both give the same Req.

Mesh and nodal analysis

Mesh (loop) current analysis - based on KVL, best when there are few meshes and mostly voltage sources; valid only for planar circuits.

1. Assign a circulating mesh current (say clockwise) to each of the b − n + 1 meshes.
2. Write KVL per mesh: (self-resistance)×(own current) − Σ(mutual resistance)×(adjacent current) = Σ driving EMFs.
3. A current source shared by two meshes gives a supermesh: write KVL around the outer loop and add the constraint equation I1 − I2 = Is.
4. In matrix form [R][I] = [V]; [R] is symmetric for networks with no dependent sources.

Nodal analysis - based on KCL, best when there are few nodes and mostly current sources; works for non-planar circuits and is what SPICE uses.

1. Choose a reference (ground) node; label the other n − 1 node voltages.
2. Write KCL per node: (sum of conductances at the node)×V_node − Σ(conductance to neighbour)×V_neighbour = sum of currents injected.
3. A voltage source between two non-reference nodes gives a supernode: write KCL for the pair and add the constraint V1 − V2 = Vs. If one end is grounded, that node voltage is simply known.
4. In matrix form [G][V] = [I]; [G] is the conductance matrix.
5. Which to choose: compare (b − n + 1) with (n − 1) and pick the smaller count; a modified (mesh/nodal) approach with dependent sources needs one extra controlling-variable equation each.

The network theorems

• Superposition: in a *linear* network the response to several independent sources is the sum of the responses to each acting alone, with others deactivated (V sources shorted, I sources opened). Dependent sources stay active. Does not apply to power, because power is quadratic in v and i.
• Thevenin's theorem: any linear two-terminal network is equivalent, at those terminals, to a single voltage source Vth = open-circuit voltage in series with Rth = resistance looking in with all independent sources deactivated.
• Norton's theorem: the same network is equivalent to a current source IN = short-circuit current in parallel with RN = Rth. Thevenin and Norton are source transformations of each other.
• Finding Rth when dependent sources are present: you cannot "look in and reduce", because the dependent source responds. Use either (a) Rth = Voc/Isc, or (b) deactivate all independent sources, apply a 1 V test source and compute Rth = 1/I (or apply 1 A and read V).
• Maximum power transfer: a source of Vth, Rth delivers maximum power to RL = Rth, with Pmax = Vth²/4Rth and efficiency only 50% - which is why it is used in signal/communication matching (antennas, amplifier output stages) and never in power distribution, where high efficiency matters. In AC, the condition is ZL = Zth\* (conjugate match); if only |ZL| is adjustable with fixed angle, set |ZL| = |Zth|.
• Reciprocity theorem: in a single-source, linear, bilateral, passive network, the ratio of response to excitation is unchanged if the positions of excitation and response are interchanged. Fails if a dependent source or any unilateral element is present.
• Millman's theorem: several parallel voltage sources with series resistances collapse to one source Veq = Σ(VkGk)/ΣGk with Req = 1/ΣGk - a fast shortcut for parallel branch problems.
• Substitution theorem: any branch can be replaced by any other combination having the same terminal v and i without disturbing the rest of the network.
• Compensation theorem: a change ΔZ in a branch carrying I can be accounted for by inserting an opposing source of value I·ΔZ with all other sources deactivated - the basis of small-change sensitivity analysis.
• Duality: series ⇄ parallel, R ⇄ G, L ⇄ C, V source ⇄ I source, KVL ⇄ KCL, mesh ⇄ node. A dual network has the same equation form, so one solution serves both.

Network topology (graph methods)

• Replace every branch by a line: a graph with b branches and n nodes. A tree is a connected subgraph touching all nodes with no loop; it has exactly n − 1 twigs; the remaining b − n + 1 branches are links (co-tree).
• Independent equations: KCL gives n − 1, KVL gives b − n + 1, total b - exactly the number of unknown branch currents.
• Incidence matrix A (n − 1 × b, entries +1/−1/0) expresses KCL as A·Ib = 0.
• Tie-set (loop) matrix B: each link defines one fundamental loop; KVL becomes B·Vb = 0, and branch currents follow from Ib = Bᵗ·IL.
• Cut-set matrix Q: each twig defines one fundamental cut-set; Q·Ib = 0 and Vb = Qᵗ·Vt. The orthogonality relations A·Bᵗ = 0 and Q·Bᵗ = 0 hold.
• Number of possible trees of a graph = det(A·Aᵗ) - occasionally asked numerically in GATE.

AC steady-state analysis and power

• For sinusoidal steady state, replace v(t) = Vm sin(ωt + φ) by the phasor V = Vrms∠φ, and elements by impedances: ZR = R, ZL = jωL, ZC = 1/(jωC) = −j/(ωC). Every DC theorem then carries over with complex arithmetic - only maximum power transfer changes form, to ZL = Zth\*.
• Admittance Y = 1/Z = G + jB. Series impedances add; parallel admittances add.
• Power: S = V I\* = P + jQ (VA); P = VI cos φ (W, real); Q = VI sin φ (VAR, lagging positive for inductive); power factor = cos φ = P/S. Only R consumes average power; L and C exchange reactive power.
• Indian practice and units: SI throughout (V, A, Ω, S, W, VAR, VA, H, F); supply reference 230 V ± 6%, 50 Hz ± 3% single-phase and 400 V three-phase as per BIS/CEA practice, and instrument accuracy classes per IS 1248 for indicating instruments. A real meter loads the circuit (voltmeter in parallel, finite resistance; ammeter in series, finite resistance), so Thevenin resistance decides measurement error - this is the standard lab viva link.

Key points

• KCL - algebraic sum of node currents is zero; KVL - algebraic sum of loop voltages is zero; both follow from conservation of charge and energy.
• Independent equations - n − 1 node equations and b − n + 1 loop equations for a graph of n nodes, b branches.
• Mesh vs nodal - pick whichever gives fewer unknowns; nodal needs a reference node, mesh needs a planar circuit.
• Supernode / supermesh - the fix for a floating voltage source and a shared current source respectively; each needs a constraint equation.
• Source transformation - V–R series ⇄ I–R parallel, equivalent externally only; ideal sources cannot be transformed.
• Superposition - valid for linear circuits and for v and i only, never for power; dependent sources are never switched off.
• Thevenin - Vth = Voc, Rth = resistance with independent sources killed; one source plus one resistance replaces the whole linear network.
• Norton - IN = Isc, RN = Rth; related by Vth = IN·Rth.
• Rth with dependent sources - use Voc/Isc or a 1 V/1 A test source; simple reduction is wrong.
• Maximum power transfer - RL = Rth (DC) or ZL = Zth\* (AC); efficiency at match is 50%.
• Reciprocity - holds only for linear, passive, bilateral, single-source networks.
• Millman - parallel voltage branches collapse to Veq = ΣVkGk/ΣGk, Req = 1/ΣGk.
• Star-delta - Y to Δ and back; the standard escape from an irreducible bridge.
• Tellegen - Σ vk ik = 0 for any network obeying KCL and KVL; use it to verify a solved answer.
• Duality - interchange series/parallel, R/G, L/C, V/I, mesh/node to get a second solved circuit free.
• Voltage/current dividers - dividers apply only to series/parallel sets respectively; loading changes the ratio.
• Phasors - ωL and 1/ωC convert any AC problem into a DC-like one with complex numbers.
• Power triangle - S² = P² + Q², pf = P/S = cos φ.

Formulas and examples

Example 1 - Mesh analysis with two sources. Mesh 1: a 12 V source in series with R1 = 2 Ω and the shared branch R2 = 2 Ω. Mesh 2: the shared R2 = 2 Ω, R3 = 2 Ω and a 6 V source aiding mesh-2 clockwise current.

• Mesh 1: (2 + 2)I1 − 2I2 = 12 → 4I1 − 2I2 = 12
• Mesh 2: −2I1 + (2 + 2)I2 = 6 → −2I1 + 4I2 = 6
• Determinant = 16 − 4 = 12. I1 = (12×4 + 2×6)/12 = 60/12 = 5 A; I2 = (4×6 + 2×12)/12 = 48/12 = 4 A.
• Current in the shared 2 Ω = I1 − I2 = 1 A (direction of I1).
• Check by power: sources deliver 12×5 + 6×4 = 84 W; resistors dissipate 5²×2 + 4²×2 + 1²×2 = 50 + 32 + 2 = 84 W ✓ (Tellegen).

Example 2 - Thevenin, Norton and maximum power. A 24 V source in series with 6 Ω, a 12 Ω shunt across it, then 4 Ω in series out to terminals A-B.

• Vth = open-circuit voltage = 24 × 12/(6 + 12) = 16 V (no drop across 4 Ω at open circuit).
• Rth = (6 ∥ 12) + 4 = (6×12/18) + 4 = 4 + 4 = 8 Ω. Norton: IN = 16/8 = 2 A with 8 Ω parallel.
• For maximum power, RL = 8 Ω: I = 16/16 = 1 A, P = 1²×8 = 8 W = Vth²/4Rth = 256/32 ✓.
• With RL = 32 Ω instead: I = 16/40 = 0.4 A, P = 0.16×32 = 5.12 W - less, confirming the match condition.

Example 3 - Superposition. A 10 V source through R1 = 5 Ω to node X; R2 = 5 Ω from X to ground; a 2 A current source injecting into X.

• 10 V alone (current source opened): Vx1 = 10 × 5/(5 + 5) = 5 V.
• 2 A alone (voltage source shorted): R = 5 ∥ 5 = 2.5 Ω → Vx2 = 2 × 2.5 = 5 V.
• Vx = 5 + 5 = 10 V, so current in R2 = 10/5 = 2 A.
• Nodal check: (Vx − 10)/5 + Vx/5 = 2 → 2Vx = 20 → Vx = 10 V ✓.
• Power is *not* superposable: P(R2) actual = 10²/5 = 20 W, whereas 5²/5 + 5²/5 = 10 W - wrong.

Example 4 - Delta-star to crack a bridge. Find R between A and D: A-B = 6 Ω, A-C = 18 Ω, B-C = 12 Ω, B-D = 4 Ω, C-D = 8 Ω.

• Convert delta A-B-C (sum = 6 + 12 + 18 = 36) to a star at N: RA = 6×18/36 = 3 Ω, RB = 6×12/36 = 2 Ω, RC = 12×18/36 = 6 Ω.
• Now N-B-D = 2 + 4 = 6 Ω in parallel with N-C-D = 6 + 8 = 14 Ω → 6×14/20 = 4.2 Ω.
• R_AD = 3 + 4.2 = 7.2 Ω. (If a 36 V source were applied, total current = 36/7.2 = 5 A.)

Example 5 - AC phasor analysis and power. 230 V (RMS), 50 Hz across R = 30 Ω in series with L = 0.1273 H.

• XL = 2π × 50 × 0.1273 ≈ 40 Ω; Z = 30 + j40, |Z| = √(900 + 1600) = 50 Ω, φ = tan⁻¹(40/30) = 53.13° lagging.
• I = 230/50 = 4.6 A, phasor I = 4.6∠−53.13° A.
• P = I²R = 21.16 × 30 = 634.8 W; Q = I²X = 21.16 × 40 = 846.4 VAR; S = 230 × 4.6 = 1,058 VA; pf = 30/50 = 0.6 lag.
• Check: √(634.8² + 846.4²) = 1,058 VA ✓.

Where it is used

• Semiconductor and EDA industry: small-signal equivalent circuits in Cadence Spectre/HSPICE are solved by modified nodal analysis; design engineers at Texas Instruments India, Analog Devices, NXP and Qualcomm Bengaluru use Thevenin reduction daily for bias networks and loading analysis.
• Analog and power-supply design: output impedance of a regulator, divider loading in feedback networks, source-impedance matching in RF front ends and antenna matching at ISRO, DRDO labs and BEL.
• Measurement and instrumentation: bridge circuits (Wheatstone, Kelvin, Maxwell) for strain gauges, RTDs and load cells; Thevenin resistance decides voltmeter loading error - core to instrumentation roles at BHEL, L&T and NPCIL.
• Power and energy systems: per-phase equivalent circuits, short-circuit (fault) calculation as a Thevenin source behind an impedance, used by state transmission utilities and in POWERGRID/NTPC engineer roles.
• Testing and field service: fault-finding on PCBs and panels by node-voltage reasoning and source deactivation; AMIE/diploma lab practicals on verification of Thevenin, Norton, superposition and maximum power transfer.
• Exam and recruitment: GATE EC/EE networks section, ESE, ISRO/BARC/DRDO scientist entries, BSNL JTO and most campus aptitude-plus-technical screens.

Interview questions

• Q: State KCL and KVL and the physical law behind each. A: KCL says the algebraic sum of currents at a node is zero (conservation of charge); KVL says the algebraic sum of voltages around a closed loop is zero (conservation of energy).
• Q: How do you find Thevenin resistance when the circuit contains a dependent source? A: You cannot simply deactivate and reduce - either compute Rth = Voc/Isc, or kill only the independent sources, apply a 1 V (or 1 A) test source at the terminals and take Rth = V/I.
• Q: Why can superposition not be used for power? A: Power depends on the square of voltage or current, so it is non-linear; superposition applies only to linear quantities, i.e. v and i.
• Q: At maximum power transfer the efficiency is only 50% - so why is it used? A: In signal and communication systems the goal is maximum power into the load (receiver, antenna, speaker), not efficiency; in power distribution we deliberately keep RL ≫ Rth for high efficiency.
• Q: What is the AC condition for maximum power transfer? A: ZL = Zth\* (conjugate match) - equal resistance and cancelling reactance; if only the magnitude of ZL can be varied, set |ZL| = |Zth|.
• Q: When is nodal analysis preferred over mesh? A: When the node count minus one is smaller than the mesh count, when current sources dominate, or when the circuit is non-planar - which is why simulators use nodal.
• Q: What is a supernode and when do you need one? A: A closed surface enclosing two nodes joined by a voltage source; KCL is written for the pair and the source equation V1 − V2 = Vs supplies the missing relation.
• Q: Can you convert an ideal voltage source into a current source? A: No - source transformation needs a finite series resistance; an ideal source has zero (or infinite) internal resistance so the equivalent current would be infinite.
• Q: Why does the reciprocity theorem fail for a transistor amplifier? A: It contains a dependent source and is unilateral, so excitation and response cannot be interchanged without changing the transfer ratio.
• Q: For a graph with 7 branches and 5 nodes, how many independent KCL and KVL equations exist? A: KCL: n − 1 = 4; KVL: b − n + 1 = 3; total 7, matching the 7 unknown branch currents.
• Q: Why is star-delta conversion needed at all? A: Because a bridged network has no two elements purely in series or parallel; converting one delta to a star (or vice versa) breaks the bridge and restores series-parallel reduction.
• Q: How would you measure Thevenin parameters of a real black box on the bench? A: Measure the open-circuit voltage with a high-impedance voltmeter, then load it with a known resistor and read the terminal voltage; Rth = R(Voc/VL − 1). Avoid a true short, which may damage the source.

Common mistakes

• Applying the voltage-divider ratio to a divider that has a load across one resistor, or a current divider to elements that are not actually in parallel.
• Deactivating dependent sources along with independent ones in superposition or while finding Rth - the single most common GATE trap.
• Adding power contributions from separate superposition cases, or using RMS and peak values interchangeably in AC power calculations.
• Sign errors: forgetting that the mutual-resistance terms in a mesh matrix are negative, or mixing up polarity when traversing a source in KVL.
• Writing a KCL equation at the reference node (it is dependent on the others) or using more node equations than n − 1.
• Treating "maximum power transfer" as the design goal in a power-delivery problem, where 50% loss is unacceptable.
• Omitting the constraint equation for a supernode or supermesh, leaving the system under-determined.

Related topics

Network analysis: transients, resonance and two-port networks · Semiconductor diodes and applications · BJT and MOSFET · Amplifiers and feedback · Operational amplifiers · Active filters and op-amp applications