ESE ECE Prelims common mistakes
ESE ECE Prelims · Understand the traps and how to avoid them.
In Network Theory, applying superposition or Thevenin's theorem to a circuit containing diodes or dependent sources without first checking linearity, then reporting a single numerical answer as if the circuit were linear.
Coaching material drills nodal and mesh analysis on linear resistor networks, so candidates pattern-match every circuit to those methods and forget that superposition requires linearity and that dependent sources must stay active when computing Thevenin resistance.
What to do instead
Before choosing a method, list every element and mark nonlinear (diodes, transistors) and dependent sources. For Thevenin resistance with dependent sources, apply a test source and use R_th = V_test/I_test; only use the series/parallel shortcut when no dependent sources exist. Practise two or three mixed circuits per revision session and explicitly write why each method is or is not valid.
In Electronic Devices and Circuits, memorising a single value such as 0.7 V for every diode and 0.2 V for every Schottky, and ignoring the operating point when deciding whether a diode is on or off in a multi-diode circuit.
Standard textbook constants get treated as universal truths, and candidates rarely practise the guess-and-verify procedure needed when several diodes share a node.
What to do instead
Adopt an explicit algorithm: assume a state for each diode, solve, then check consistency of the assumed current direction and voltage polarity; if inconsistent, flip the state and repeat. Keep the exponential diode equation in mind to justify when the constant-drop model is acceptable (forward current well above saturation). Work at least five multi-diode problems with full state tables.
In Analog and Digital Circuits, computing the operating point of a BJT amplifier but then using the wrong small-signal parameter (for example r_e versus r_pi, or h_fe versus beta) and forgetting that the hybrid-pi r_pi equals beta/gm, leading to gain answers off by a factor of beta.
Two parameter sets are taught side by side, and candidates mix them under time pressure without tracking units or definitions.
What to do instead
Write a one-page conversion sheet (gm = I_C/V_T, r_pi = beta/gm, r_e = V_T/I_E, alpha = beta/(beta+1)) and derive each from the others once by hand. In every amplifier problem, first fix the DC operating point, then label the small-signal model with the chosen parameters before substituting numbers, and sanity-check gain magnitude against gm*R_C.
In Signals and Systems, stating the Fourier transform or Laplace region of convergence from a memorised table without checking whether the signal is right-sided, left-sided, or two-sided, and therefore giving an unstable system as stable.
Tables are learned as pure transform pairs, and the ROC is treated as an afterthought even though it determines uniqueness and stability.
What to do instead
For every transform problem, sketch the signal support first and write the ROC alongside the transform. Apply the stability test explicitly: a causal LTI system is stable only if all poles lie strictly in the left half-plane (or inside the unit circle for discrete time). Redo three transform pairs for each sidedness case until the ROC becomes automatic.
In Communication Systems, calculating a figure of merit, noise figure, or required bandwidth using the wrong reference (for example using the input SNR at the detector instead of the RF front end, or mixing decibels and linear ratios) and reporting an answer that is dimensionally plausible but numerically wrong.
Noise figure and SNR chains involve cascaded stages and dB arithmetic that candidates practise only as isolated formulas, and Friis's formula is often misapplied by adding noise figures instead of converting to noise factors.
What to do instead
Build every link-budget or noise-chain problem as a table of stages with gain and noise factor in linear units, apply Friis's formula, then convert the final result to dB only at the end. Keep a rule card: add powers in linear, add dB in log, never add noise figures directly. Check the answer against the best-case single-stage noise figure.
In Electromagnetics and Antennas, substituting distances or lengths in centimetres or millimetres into formulas such as phase constant beta = 2*pi/lambda, or the Friis transmission equation, without converting to metres, and also forgetting the 4*pi or 120*pi factors in free-space formulas.
Problems mix units deliberately, and candidates compute quickly under exam pressure with the first number that appears.
What to do instead
Adopt a habit of writing every given quantity with its unit and converting to SI at the top of the solution before any formula. Keep a formula card that includes the constant (eta_0 = 120pi ohm, lambda = c/f, path loss = 20log10(4pid/lambda)) and do a dimensional check on the final answer. Practise five problems written deliberately in mixed units.
In Control Systems, judging stability only from the open-loop transfer function or from the number of right-half-plane poles, without applying the Routh-Hurwitz criterion to the closed-loop characteristic equation or checking the Nyquist encirclement count.
Block-diagram reduction is taught first and stability is often tested on the open-loop gain, so candidates skip forming 1 + G(s)H(s) = 0.
What to do instead
Always form the closed-loop characteristic polynomial before any stability claim, then apply Routh-Hurwitz, noting the special cases of a zero row and a row of zeros. When a Nyquist or Bode question appears, state the encirclement or gain/phase margin criterion explicitly. Solve one stability problem per method (Routh, Nyquist, root locus) each week and compare conclusions.
In Measurements and Instrumentation, ignoring the loading effect of the instrument on the circuit, for example assuming an ideal voltmeter or ammeter, or neglecting the meter's internal resistance when computing the measured value.
Ideal-instrument assumptions are used in circuit courses, and candidates carry them into measurement questions where the whole point is the loading error.
What to do instead
For every measurement problem, redraw the circuit with the instrument's internal resistance included (voltmeter in parallel with its resistance, ammeter in series with its resistance) and compute the actual reading. Compare it with the ideal reading to state the percentage error. Memorise typical orders of magnitude for voltmeter and ammeter resistances and for shunt and multiplier design.
In Microprocessors, Microcontrollers and Computer Fundamentals, confusing the roles of the program counter, stack pointer, and instruction register, or mixing up the number of T-states and machine cycles for an instruction when computing execution time and delay loops.
Architecture is learned as a list of registers, and timing is memorised as isolated numbers rather than derived from the fetch-execute sequence.
What to do instead
Draw the fetch-execute cycle once with each register's role at every step. For timing, build a small table of instruction, machine cycles, T-states, and clock periods, then compute a delay loop by summing T-states per iteration and multiplying by the loop count and clock period. Verify by writing the loop in assembly and counting cycles explicitly.
In General Studies and Engineering Aptitude (Paper I), spending disproportionate time on long numerical derivations and leaving comprehension, ethics, and project-management items unattempted, or answering negative-marking items by guessing without elimination.
The paper mixes general aptitude with engineering fundamentals, and candidates trained on technical papers misjudge the time budget and the scoring value of short factual items.
What to do instead
Do a timed mock with a strict per-section cap (for example, one pass for all short items, a second pass for numericals) and record the time spent per section. Practise elimination on multiple-choice items with negative marking and set a personal rule to skip an item after a fixed time. Review the score breakdown to confirm the short items are being captured before attempting long derivations.