ISRO Scientist Electrical common mistakes
ISRO Scientist Electrical · Understand the traps and how to avoid them.
Applying superposition or Thevenin's theorem to a circuit containing dependent sources by 'turning off' the dependent source along with the independent ones, then computing the wrong equivalent resistance.
Candidates memorise 'kill the sources' as a blanket rule and never internalise that dependent sources are controlled quantities, not excitations, so they cannot be zeroed. This is reinforced by the many textbook drill problems that only use independent sources.
What to do instead
Adopt the rule: when a dependent source is present, compute R_th by applying a test source (1 V or 1 A) at the port with all independent sources killed and the dependent source left intact, then R_th = V_test/I_test. Practise five such networks (one with a current-controlled voltage source, one with a VCCS) until the test-source method is automatic, and always write the controlling variable's defining equation before solving.
Predicting transformer efficiency or regulation from nameplate kVA alone, or assuming maximum efficiency occurs at full load, without locating the load fraction where copper loss equals iron loss.
The efficiency-versus-load curve peaks at the load where variable I²R loss equals constant core loss, i.e. at fraction x = sqrt(P_i/P_cu,FL). Candidates who memorise 'max efficiency at full load' get numerical answers wrong whenever the problem gives separate core and full-load copper losses.
What to do instead
Always extract P_i (core, from open-circuit test) and P_cu,FL (from short-circuit test) first. Compute x = sqrt(P_i/P_cu,FL) and state the load in kVA as x·S_rated. For regulation use the exact phasor form with per-unit R and X, and check the sign for leading power factor, where regulation can be negative.
Solving a DC shunt motor speed-control or braking problem by plugging numbers into the back-emf equation without first checking whether flux is constant, and mixing up the direction of armature current during regenerative versus plugging braking.
The relation N ∝ (V − I_a R_a)/φ is simple, so candidates skip the assumption audit. In field-weakening and in braking, φ and the sign of I_a change, and the same formula then gives the wrong speed or torque direction.
What to do instead
Before substituting, write down explicitly which quantities are held constant, and draw the machine with armature and field marked, arrows for E_b and I_a. For braking, verify that in plugging the supply and back emf aid each other (I_a = (V + E_b)/R_a) while in regenerative braking E_b exceeds V and current reverses. Do one worked example of each braking mode.
Computing per-unit or actual line flows and losses on a transmission line without converting the given impedance to the correct base, or ignoring the difference between sending-end and receiving-end voltage when using the short-line approximation.
Base conversion (Z_base = V_base²/S_base, and the change-of-base formula for given per-unit values) is mechanical but easy to skip when the problem states impedance in ohms and power in MVA at a different voltage level; candidates also apply V_s ≈ V_r formulas outside their valid regime.
What to do instead
Standardise a three-step routine: choose S_base and V_base per zone, compute Z_base for each zone, then convert every impedance to per unit before any calculation. For lines, decide short (<80 km), medium (nominal π or T), or long (ABCD with hyperbolic functions) from length and state it. Keep sending-end and receiving-end quantities in separate columns of your working.
Answering protection coordination and relay-setting questions by picking the 'fastest' or 'most sensitive' option rather than checking selectivity, CT ratio, plug setting multiplier, and time-multiplier setting against the grading margin.
Candidates treat relay questions as qualitative recall ('differential is best') and avoid the arithmetic. The exam, however, tests PSM = fault current in relay coil / (CT ratio × plug setting) and the resulting time from the relay characteristic, plus the requirement that the downstream relay clear before the upstream one.
What to do instead
For every coordination question, draw the single-line diagram, mark CT ratios and fault levels at each bus, and compute PSM and operating time for the primary and backup relay. Confirm the grading margin (typically 0.3–0.4 s) is satisfied. Memorise the purpose of each scheme (differential for transformer/generator unit protection, distance zones 1/2/3 for lines) and the settings each requires.
Determining closed-loop stability of a feedback system from the open-loop transfer function by counting poles of G(s)H(s) in the right half-plane, or misreading the Routh array when a zero appears in the first column or an entire row vanishes.
The Routh–Hurwitz criterion applies to the characteristic equation 1 + G(s)H(s) = 0, not to G(s)H(s) itself. Candidates also forget the special-case rules (replace a zero by ε, or differentiate the auxiliary polynomial from the preceding row) and so mis-sign the number of right-half-plane roots.
What to do instead
Always form the characteristic polynomial 1 + G(s)H(s) explicitly before building the Routh array. Drill the two special cases separately. Cross-check the result with the Nyquist criterion (encirclements of −1) or a root-locus sketch, and for gain-margin questions read the margin off the Bode plot at the phase-crossover frequency, not the gain-crossover frequency.
Analysing a phase-controlled rectifier or chopper with purely resistive-load formulas when the load is inductive, thereby ignoring discontinuous conduction, the freewheeling diode's effect, and the difference between average and RMS values used for power.
Average output voltage formulas for R-load are the ones most often memorised, but with an R-L load the current may not reach zero before the next thyristor fires, changing the integration limits. Candidates then compute average voltage correctly but RMS (hence power and form factor) incorrectly.
What to do instead
For each converter, first decide continuous versus discontinuous conduction by comparing the extinction angle with the firing interval, then set integration limits accordingly. Compute V_avg and V_rms as separate integrals, and only use V_rms²/R for load power. Note explicitly what the freewheeling diode does to the output (clamps it to zero, removes negative voltage).
Treating an op-amp circuit as ideal in every problem - assuming infinite gain, zero input current, and virtual short - even when the question supplies finite open-loop gain, input bias current, or slew rate and asks for the resulting error.
The ideal-op-amp shortcut works in most drill problems, so candidates apply it reflexively. Non-ideal parameters are precisely what the exam tests in error-budget and bandwidth questions, and the virtual-short assumption fails once finite gain is specified.
What to do instead
Before solving, scan the problem statement for any finite parameter (A_ol, I_b, V_os, slew rate, GBW). If present, use the finite-gain model: V_out = A_ol(V+ − V−), solve the node equations without assuming V+ = V−, and compute the error term separately. For slew-rate questions, check that the required dV/dt = 2πfV_pk does not exceed the slew rate; for bandwidth, use the gain-bandwidth product to find the closed-loop corner.
Applying a Laplace or Fourier transform to a signal or system without first checking the region of convergence or the existence conditions, and confusing the unilateral and bilateral transform for initial-condition problems.
Candidates learn transform pairs as a lookup and skip the ROC, which matters for stability and causality questions and for distinguishing signals with the same algebraic transform. Using the bilateral transform on a circuit with initial stored energy also drops the initial-condition terms.
What to do instead
For every transform problem, state the ROC alongside the transform and use it to answer causality/stability. For circuits and differential equations with nonzero initial conditions, use the unilateral Laplace transform and include the initial-condition terms (e.g. sL·i(0) and C·v(0)/s). Verify by checking that the ROC of a stable system contains the jω axis.
Solving numerical-methods and linear-algebra items by executing the algorithm blindly - e.g. Gauss elimination without partial pivoting, or Newton–Raphson without checking the derivative is nonzero - and reporting an answer without verifying convergence or the residual.
These are taught as procedures, and candidates reproduce steps under time pressure without the diagnostic habits that catch a wrong pivot, a divergent iteration, or an ill-conditioned matrix. The exam often plants a case where the naive method fails.
What to do instead
Before iterating, check the convergence condition (|φ'(x)| < 1 for fixed point, nonzero f' for Newton, diagonal dominance for Gauss–Seidel). Use partial pivoting whenever a pivot is small relative to the column. After obtaining an answer, substitute it back to compute the residual and confirm the error tolerance. For matrix questions, check the determinant and rank before claiming invertibility or a unique solution.
In C and data-structure questions, predicting output by reading code linearly and ignoring operator precedence, integer division, short-circuit evaluation, pointer arithmetic scaling, or the difference between array indexing and pointer dereference.
Candidates trace code the way they read prose, left to right, and mentally evaluate expressions without the precedence table. Pointer arithmetic is especially error-prone because p + 1 advances by sizeof(type), not one byte.
What to do instead
Rewrite each expression with explicit parentheses according to the C precedence order before evaluating, and track every variable's type and value in a table line by line. For pointers, annotate the address and the pointed-to value separately. For data structures, draw the tree or list after each operation rather than tracking it in your head, and state the complexity of the operation the question asks about.