Measurements and instrumentation

In short

Measurements and instrumentation is the engineering of getting a trustworthy number out of a physical quantity: the sensor that converts it, the signal conditioning that cleans it, the display or ADC that reads it, and the error analysis that says how much the number can be believed. For an ECE student it ties together bridges, transducers, oscilloscopes, digital voltmeters and data-acquisition systems, and it supplies the vocabulary — accuracy, precision, resolution, sensitivity, calibration, traceability — used in every test and validation job. GATE IN and GATE EE examine it every year (bridge balance, error propagation, DVM counts, CRO measurements, transducer characteristics); it is not a separate section of the GATE EC syllabus, but university ECE syllabi devote a full paper to it under AICTE model curriculum. In Indian industry it is the core skill of test engineers at BEL, BHEL, ISRO/VSSC, DRDO labs, Keysight/Tektronix/Yokogawa India, automotive and power-utility testing, and of every NABL-accredited calibration lab working to traceability from CSIR-NPL, New Delhi.

Explained

Measurement system, standards and traceability

• A measurement system has three stages: primary sensing element (transducer) → signal conditioning / variable conversion (amplify, filter, linearise, convert) → output / terminating stage (indicator, recorder, ADC, controller).
• Methods: *direct* comparison with a standard; *indirect* (via a calibrated chain); *null/deflection*. Null methods (bridges, potentiometers) are more accurate because the detector need only be sensitive, not calibrated.
• Hierarchy of standards: *international* (SI, maintained by BIPM) → *primary* (national, at CSIR-NPL under the Legal Metrology Act, 2009) → *secondary* (accredited labs) → *working* (bench standards).
• Traceability is an unbroken documented chain of calibrations to a national standard; in India calibration labs are accredited by NABL (under Quality Council of India) to ISO/IEC 17025. Trade measurement is governed by the Legal Metrology Act, 2009 and the Legal Metrology (Packaged Commodities) Rules, 2011.
• SI base units (post-2019 redefinition): all seven are defined from fixed constants — the kilogram from the Planck constant *h*, the ampere from the elementary charge *e*, the kelvin from the Boltzmann constant *k*, the second from the ¹³³Cs hyperfine transition. The physical kilogram prototype is no longer the definition.
• Calibration compares an instrument against a standard and records deviation; it does not by itself improve the instrument. *Adjustment* or *correction* does.

Errors: types, analysis and combination

• Error e = measured − true value; correction = −e. Relative error = e / true value.
• Gross errors: human — misreading, wrong connection, arithmetic. Reduced by repetition and care, not by statistics.
• Systematic errors (bias, repeatable): *instrumental* (friction, zero offset, wear, loading), *environmental* (temperature, humidity, stray fields), *observational* (parallax, reaction time). Detected by comparison with a better standard; removable by correction.
• Random errors: residual, unexplained scatter; treated statistically.
• Statistics: arithmetic mean, deviation d = xᵢ − x̄, standard deviation σ, variance σ², probable error 0.6745σ, standard error of the mean σ/√n. Random error falls as 1/√n — averaging 100 readings reduces it only ten-fold, not hundred-fold.
• Limiting (guarantee) error: manufacturer's stated accuracy, usually ± % of full-scale, so the *percentage error of the reading* grows as the reading falls — read near full scale.
• Uncertainty: modern practice reports *Type A* (statistical) and *Type B* (judgement) components combined into a combined standard uncertainty, then an expanded uncertainty U = k·u_c with coverage factor k = 2 (~95%), per ISO/IEC Guide 98 (GUM). NABL certificates always quote U and k.
• Loading error: an instrument changes the circuit it measures. A voltmeter of resistance R_v across R shunts it; an ammeter adds burden. Figure of merit: ohms-per-volt sensitivity = 1/I_fsd.

Static and dynamic characteristics

• Accuracy: closeness to the true value. Precision: closeness of repeated readings to one another. Precise ≠ accurate — a biased instrument can be very repeatable.
• Resolution: smallest detectable change. Threshold: smallest input that produces a detectable output from zero. Sensitivity = Δoutput/Δinput (static sensitivity = slope of the calibration curve).
• Dead zone / dead band: range of input producing no output. Hysteresis: different output for the same input on increasing and decreasing runs. Backlash: lost motion in gears/linkages.
• Linearity: maximum departure from the best-fit straight line, expressed as % of full scale. Drift: zero drift (whole curve shifts), span/sensitivity drift (slope changes), zonal drift.
• Repeatability (same conditions, short term) vs reproducibility (different observer/instrument/place/time).
• Dynamic behaviour: instruments modelled as zero-order (potentiometer, pure gain), first-order (thermometer, RC — step response 1 − e^−t/τ, τ = time constant; 63.2% in one τ, 95% in 3τ), second-order (galvanometer, accelerometer, U-tube — ω_n and ζ govern overshoot, settling, and bandwidth).
• Damping: under-damped (ζ<1, overshoot and ringing), critically damped (ζ=1), over-damped (ζ>1, slow). Practical instruments use ζ ≈ 0.6–0.7 for the flattest frequency response and fastest settling to within a few percent.
• Dynamic errors: *speed-of-response* lag, *fidelity*, *dynamic error* for sinusoidal input = amplitude ratio departure from unity plus phase lag.

Analogue instruments and measurement of R, L, C

• Essential torques: *deflecting* (from the measurand), *controlling* (spring or gravity — gives a definite scale), *damping* (air friction, fluid friction, eddy-current — kills oscillation without affecting steady deflection).
• PMMC: deflection θ ∝ I (uniform scale), DC only, very sensitive (µA fsd), eddy-current damped. Extended by shunt (ammeter) and series multiplier (voltmeter). A rectifier-type AC meter uses a PMMC with a bridge rectifier; it responds to *average* value but is scaled to read RMS of a sinewave (form factor 1.11) — so it mis-reads non-sinusoidal waveforms.
• Moving iron (attraction/repulsion): θ ∝ I², non-uniform (cramped) scale, reads true RMS, works on AC and DC, cheap and robust — the common panel meter.
• Electrodynamometer: torque ∝ product of two currents — used for wattmeters (true power, includes power factor) and as a transfer instrument between AC and DC.
• Induction type: energy meters; rotating disc, braking magnet; register ∝ ∫vi dt (kWh).
• Resistance:
  • Low (<1 Ω): Kelvin double bridge, four-terminal (potential leads separate from current leads) to eliminate lead and contact resistance.
  • Medium (1 Ω–100 kΩ): Wheatstone bridge, substitution, ohmmeter.
  • High (>100 kΩ to MΩ/GΩ): megohmmeter (Megger), loss-of-charge method, guarded (three-terminal) methods to divert surface leakage.
• AC bridges (general balance Z₁Z₄ = Z₂Z₃ — two conditions: magnitude and phase, so two elements must be adjustable):
  • Maxwell's inductance–capacitance bridge: medium-Q inductors (1<Q<10); fails for very high Q (impractically large R) and very low Q.
  • Hay's bridge: high-Q inductors (Q>10), capacitor in series with resistance.
  • Anderson's bridge: low-Q inductors and accurate L against a standard C.
  • De Sauty / modified De Sauty: capacitance comparison. Schering bridge: capacitance and dissipation factor / tan δ of insulation and cables at high voltage — the standard HV insulation diagnostic.
  • Wien bridge: frequency measurement (and the classic RC oscillator network).
• Q-meter: uses series resonance; Q = V_C/V_in; measures coil Q, L, C and self-capacitance.
• Potentiometer (DC): null method using a standard cell; the reference technique for calibrating voltmeters, ammeters (via standard resistor) and wattmeters.
• Instrument transformers: CT (series, secondary 1 A or 5 A, never open-circuit the secondary) and PT/VT (parallel, secondary 110 V) extend AC measurement to high current/voltage and provide isolation; they introduce ratio error and phase-angle error, which matter for metering accuracy class.

Digital instruments, CRO and data acquisition

• Digital voltmeter (DVM) types: *ramp*, *staircase-ramp*, *successive-approximation* (fast, used in DAQ), *integrating / dual-slope* (slow, very accurate), *sigma-delta* (modern 6½-digit bench meters).
• Dual-slope ADC: integrates the input for a fixed time T₁, then de-integrates a reference until zero in T₂; V_in = V_ref·T₂/T₁. Component tolerances of R and C cancel, and if T₁ is an integer number of mains periods (20 ms at 50 Hz, or multiples), 50 Hz hum is rejected — hence high NMRR. CMRR is improved by guarding/floating input.
• Digit specification: a 3½-digit meter counts 0–1999 (2,000 counts); 4½ digits = 0–19999. Resolution = 1/(counts); accuracy is quoted as ±(% of reading + n counts).
• Digital multimeter measures V(dc/ac), I, R, continuity, diode, capacitance, frequency; true-RMS DMMs handle distorted waveforms, average-responding ones do not.
• Digital frequency meter / universal counter: *frequency mode* counts input cycles in a known gate time — ±1 count error, poor at low frequency; *period mode* counts clock pulses in one input period — better at low frequency. Crossover where the two errors are equal: f = √(f_clock/T_gate) — about 1 kHz for a 1 MHz clock and a 1 s gate. Errors: ±1 count, time-base error, trigger error.
• CRO: CRT (electron gun, deflection plates, phosphor screen) or modern DSO. Blocks: vertical amplifier with attenuator and delay line, horizontal time-base (sawtooth sweep), trigger circuit, Z-axis blanking.
  • Vertical sensitivity V/div, time-base s/div; bandwidth is the −3 dB point; rise time t_r ≈ 0.35/BW. Rule of thumb: probe/scope bandwidth ≥ 3–5× the signal's highest significant harmonic.
  • Lissajous figures give frequency ratio (f_vertical/f_horizontal = horizontal tangencies/vertical tangencies) and phase (from the ellipse intercepts).
  • DSO adds sampling (real-time and equivalent-time), memory depth, FFT and maths; sampling theorem demands f_s > 2f_max or aliasing gives false low-frequency traces.
  • Probes: 10× passive probe raises input impedance and must be compensated against scope input capacitance (adjust for a flat square-wave corner); active and differential probes for high frequency and isolation.
• Spectrum analyser (superheterodyne — RBW, VBW, span) shows amplitude vs frequency; logic analyser shows many digital channels against a clock; vector network analyser measures S-parameters.
• Data acquisition system (DAS): sensor → signal conditioning (bridge excitation, amplification, isolation) → anti-aliasing filter → multiplexer → sample-and-hold → ADC → processor/storage (a per-channel sample-and-hold ahead of the multiplexer is used only when channels must be sampled simultaneously). Key design issues: sampling rate, channel count, throughput, common-mode rejection, isolation, grounding (single-point ground), shielding and guarding against 50 Hz pickup.
• Signal conditioning for ECE: instrumentation amplifier (three op-amps, high CMRR, high input Z, gain set by one resistor) for bridge and thermocouple outputs; isolation amplifier for medical and mains-referenced measurement; cold-junction compensation for thermocouples; 4–20 mA current loop for industrial transmission (immune to lead resistance, 4 mA = live zero so a broken wire is detectable).

Transducers and sensors

• Classification: *active* (self-generating — thermocouple, piezoelectric, photovoltaic, tachogenerator) vs *passive* (need excitation — strain gauge, RTD, LVDT, capacitive); *analogue* vs *digital*; *primary* vs *secondary*.
• Resistive: potentiometer (displacement); strain gauge — gauge factor G_f = (ΔR/R)/ε, metal foil G_f ≈ 2, semiconductor 50–200 (higher sensitivity, poorer temperature stability); RTD (Pt-100, α ≈ 0.00385 /°C, nearly linear, −200 to 850 °C); thermistor (NTC, highly non-linear, β equation, very sensitive).
• Inductive: LVDT — primary plus two opposed secondaries; output ∝ core displacement with phase indicating direction, essentially infinite resolution, needs phase-sensitive demodulation; null is never exactly zero (residual voltage). Variable reluctance, eddy-current proximity sensors.
• Capacitive: change in area, distance or dielectric — pressure, level, humidity, touch screens, MEMS accelerometers; high resolution but sensitive to cable capacitance and stray fields.
• Piezoelectric (quartz, PZT): charge ∝ force; excellent for dynamic measurement (vibration, shock, knock sensors, ultrasonic) but cannot measure static quantities because the charge leaks; needs a charge amplifier.
• Thermoelectric: thermocouple (Seebeck effect, mV output, needs cold-junction compensation; types K, J, T, R/S/B). Laws of thermocouples: homogeneous circuit, intermediate metals, intermediate temperatures.
• Hall effect: V_H ∝ B·I/t — current sensors, position, brushless-motor commutation.
• Optical: photodiode, phototransistor, LDR, optical encoder (incremental/absolute, Gray code), fibre sensors.
• Pressure/flow/level: Bourdon tube, diaphragm, bellows, piezoresistive MEMS; orifice/venturi (Δp ∝ Q²), rotameter, turbine, electromagnetic flowmeter (conductive liquids), ultrasonic/Coriolis mass flow; level by float, capacitive, ultrasonic, differential pressure.
• MEMS and smart sensors: silicon accelerometers and gyroscopes, pressure sensors, microphones — now with built-in ADC and I²C/SPI digital output; backbone of phones, automotive airbag/ESP and IoT nodes.
• Selection criteria: range, sensitivity, accuracy, linearity, dynamic response, operating environment, excitation, output type, cost and life.

Key points

• Accuracy vs precision - accuracy is closeness to truth; precision is repeatability; they are independent.
• Resolution vs threshold - resolution is the smallest *change* detected anywhere on scale; threshold is the smallest input detected *from zero*.
• Null method - a bridge or potentiometer balanced to zero needs only a sensitive detector, so it beats a deflection method on accuracy.
• Limiting error - specified as ± % of full scale, so percentage error of the reading rises steeply at low readings.
• Random error reduction - standard error of the mean = σ/√n; systematic error is unaffected by averaging.
• Three torques - deflecting, controlling and damping; without a controlling torque the pointer has no definite scale.
• PMMC - linear scale, DC only, most sensitive; moving iron - square-law scale, true RMS, AC and DC.
• Rectifier AC meter - average-responding but RMS-calibrated for sinewaves (form factor 1.11); errs on distorted waveforms.
• Kelvin double bridge - low resistance by four-terminal sensing, eliminating lead and contact resistance.
• Maxwell's bridge for medium Q, Hay's for high Q, Anderson's for low Q, Schering for tan δ of HV insulation.
• AC bridge balance needs both magnitude and phase conditions - hence two adjustable elements.
• Dual-slope DVM - ratio of two times, so R and C tolerances cancel; integrating over 20 ms rejects 50 Hz mains hum.
• Digit count - 3½ digits = 2,000 counts = 0.05% resolution; accuracy quoted as ±(% reading + counts).
• Counter errors - ±1 count plus time-base error; use period mode at low frequency, frequency mode at high.
• CRO rise time t_r ≈ 0.35/bandwidth; measured rise time is the RSS of signal and scope rise times.
• Sampling theorem - f_s > 2f_max, with an anti-aliasing filter before the sampler, else aliases appear as genuine signals.
• Instrumentation amplifier - high CMRR and input impedance for low-level bridge, strain-gauge and thermocouple signals.
• Piezoelectric sensors are dynamic-only; LVDT gives magnitude and direction; 4–20 mA loops give a live zero.
• Traceability - unbroken chain to CSIR-NPL standards; Indian calibration labs accredited by NABL to ISO/IEC 17025, reporting uncertainty U with k = 2.

Formulas and examples

Example 1 — Limiting error and why you read near full scale. A 0–150 V voltmeter is specified ±1.0% of full scale. Readings of 150 V and 30 V are taken.

• Absolute limiting error = 0.01 × 150 = ±1.5 V at any reading.
• At 150 V: % error of reading = 1.5/150 × 100 = ±1.0%.
• At 30 V: % error of reading = 1.5/30 × 100 = ±5.0%.
• Conclusion: the same instrument is five times worse at one-fifth scale — choose a range so the reading lies in the upper two-thirds.

Example 2 — Loading error of a voltmeter. A 10 V source feeds two 50 kΩ resistors in series; the voltage across the lower resistor is measured with a voltmeter of sensitivity 1,000 Ω/V on the 10 V range.

• True voltage = 10 × 50/(50+50) = 5.00 V.
• Voltmeter resistance R_v = 1,000 × 10 = 10 kΩ.
• 50 kΩ ∥ 10 kΩ = (50×10)/60 = 8.333 kΩ.
• Measured V = 10 × 8.333/(50 + 8.333) = 83.33/58.333 = 1.4286 V.
• Error = 1.4286 − 5 = −3.571 V → −71.4%. A higher-sensitivity meter (20,000 Ω/V → R_v = 200 kΩ) gives 10 × 40/(50+40) = 4.444 V, i.e. −11.1% error; a DMM at 10 MΩ gives 4.9875 V, −0.25%.

Example 3 — Error propagation in a derived quantity. Power is found from P = V²/R with V = 100 V ±1% and R = 50 Ω ±2%.

• Nominal P = 100²/50 = 200 W.
• Worst-case: δP/P = 2(δV/V) + (δR/R) = 2(1%) + 2% = ±4% → P = 200 ± 8 W, i.e. 192 to 208 W.
• If the errors are independent random (RSS): √(2² + 2²) = ±2.83% → ±5.66 W. Exams usually want the worst-case figure unless "probable/random" is stated.

Example 4 — Maxwell's bridge and the Q limit. A Maxwell inductance–capacitance bridge balances with R₁ = 2.4 kΩ in parallel with C₁ = 0.1 µF, R₂ = 1 kΩ, R₃ = 500 Ω, at 1 kHz.

• L_x = R₂R₃C₁ = 1,000 × 500 × 0.1×10⁻⁶ = 0.05 H (50 mH).
• R_x = R₂R₃/R₁ = (1,000 × 500)/2,400 = 208.3 Ω.
• Q = ωR₁C₁ = 2π × 1,000 × 2,400 × 0.1×10⁻⁶ = 1.508 (check: ωL/R = 6,283×0.05/208.3 = 1.508 ✓).
• Q is low–medium, so Maxwell's bridge is appropriate. For Q = 100 at the same frequency, R₁ = Q/(ωC₁) = 100/(6,283×10⁻⁷) = 159 kΩ — a large, non-ideal variable resistor, which is exactly why Hay's bridge is used for high-Q coils.

Example 5 — DVM resolution and counter accuracy. (a) A 4½-digit DVM on the 20 V range, specified ±(0.05% of reading + 2 counts), reads 12.345 V.

• Counts = 20,000; 1 count (LSB) = 20/20,000 = 1 mV.
• Error = 0.0005 × 12.345 + 2 × 0.001 = 0.00617 + 0.002 = ±8.17 mV ≈ ±0.066% of reading.

(b) A counter with a 1 s gate and a time base accurate to ±1 ppm reads 50,000 Hz.

• ±1 count error = ±1/50,000 = ±20 ppm = ±0.002%; time base ±1 ppm → total ≈ ±21 ppm ≈ ±1.05 Hz.
• With a 10 s gate, N = 500,000 → ±1 count = ±2 ppm; total ≈ ±3 ppm. Longer gate, better resolution.

Example 6 — Strain gauge quarter bridge. A foil gauge, G_f = 2.0, R = 350 Ω, in a quarter bridge excited at 5 V, sees a strain of 500 µε.

• ΔR/R = G_f·ε = 2.0 × 500×10⁻⁶ = 1×10⁻³ → ΔR = 0.35 Ω.
• V_o = V_ex·G_f·ε/4 = 5 × 1×10⁻³/4 = 1.25×10⁻³ V = 1.25 mV — hence the need for an instrumentation amplifier of gain ≈ 1,000 to reach a 1.25 V ADC input.

Where it is used

• Test and measurement industry: Keysight, Tektronix, Rohde & Schwarz, Yokogawa, Fluke and Indian firms such as Scientific Mes-Technik and Aplab — applications and calibration engineers, ATE development.
• Public-sector R&D and manufacturing: BEL, BHEL, HAL, ECIL, ISRO/VSSC and DRDO labs run large instrumentation groups for telemetry, environmental test (vibration, thermal-vacuum) and avionics test benches; entrance through GATE and their own technical rounds.
• Power sector and utilities: CT/PT metering, static energy meters and smart meters to IS 13779 / IS 16444, meter testing under CBIP/CEA metering regulations, insulation diagnostics by Schering bridge and tan δ at state utilities and in cable factories (Polycab, Finolex).
• Process and manufacturing automation: 4–20 mA transmitters, PLC/SCADA/DCS inputs, flow and level measurement in refineries, fertiliser and pharma plants; instrumentation technicians and control engineers at L&T, Honeywell, Emerson, Yokogawa India.
• Automotive and electronics production: in-circuit and functional test, LCR and HiPot testers, sensor calibration (MAP, knock, wheel speed), EMI/EMC labs for CE/BIS compliance; ICAT and ARAI test facilities.
• Calibration and quality roles: NABL-accredited labs, in-house calibration cells maintaining traceability to CSIR-NPL, writing uncertainty budgets and ISO/IEC 17025 procedures; also medical instrumentation (ECG, patient monitors) where isolation and CMRR are safety-critical.

Interview questions

• Q: What is the difference between accuracy and precision? A: Accuracy is how close a reading is to the true value; precision is how closely repeated readings agree. An instrument with a zero offset can be highly precise and badly inaccurate.
• Q: Why is a null method more accurate than a deflection method? A: At balance the detector reads zero, so only its sensitivity matters, not its calibration or linearity; the result depends on accurately known standard components.
• Q: A voltmeter is specified ±1% of full scale. Is the error 1% at every reading? A: No. The absolute error is fixed at 1% of full scale, so as a percentage of the reading it grows — at 10% of scale it is 10%. Always read in the upper part of the range.
• Q: Why does a dual-slope DVM reject 50 Hz noise so well? A: It integrates the input over a fixed time chosen as an integer multiple of the mains period (20 ms), so the average of the interfering sinewave over that window is zero.
• Q: Why does a rectifier-type AC voltmeter read wrongly on a square wave? A: It responds to the average rectified value but its scale is multiplied by the sinewave form factor 1.11; a square wave has form factor 1.0, so the reading is about 11% high.
• Q: Why is the Kelvin double bridge used instead of a Wheatstone bridge for low resistance? A: It uses four-terminal connections with a second pair of ratio arms, so lead and contact resistances — comparable to the unknown itself — do not enter the balance equation.
• Q: When would you choose Hay's bridge over Maxwell's? A: For high-Q coils (Q > 10); Maxwell's bridge would need an impractically large parallel resistance, since R₁ = Q/(ωC₁).
• Q: What does the Schering bridge measure and why does the power industry care? A: Capacitance and dissipation factor tan δ at high voltage; a rising tan δ indicates moisture ingress or deteriorating insulation in cables, bushings and transformers.
• Q: What does a CRO bandwidth of 100 MHz imply about rise time? A: t_r ≈ 0.35/BW = 3.5 ns; a signal with a faster edge will be visibly slowed, and the displayed rise time is roughly the RSS of signal and scope rise times.
• Q: Why is an anti-aliasing filter needed before the ADC, and not after? A: Once sampled below the Nyquist rate, high-frequency content folds down and is indistinguishable from real low-frequency signal; filtering must happen in the analogue domain before sampling.
• Q: Why can a piezoelectric transducer not measure static pressure? A: The generated charge leaks away through the crystal and amplifier input resistance, so only time-varying inputs give a sustained output; static measurement needs a piezoresistive or capacitive sensor.
• Q: Why is 4–20 mA preferred over 0–5 V for field transmitters? A: Current is unaffected by lead resistance and voltage drop, it is immune to most noise pickup, and the 4 mA live zero lets a broken wire be distinguished from a genuine zero reading.
• Q: What is traceability, and who maintains India's national standards? A: An unbroken documented chain of calibrations back to a national or international standard with stated uncertainties; in India CSIR-NPL, New Delhi maintains the national standards, and NABL accredits calibration labs to ISO/IEC 17025.

Common mistakes

• Treating "± % of full scale" as "± % of reading", which badly understates error at low readings.
• Assuming averaging many readings removes all error — it reduces random error as 1/√n but leaves systematic bias untouched.
• Ignoring loading: using a 1,000 Ω/V analogue voltmeter across a high-impedance node and trusting the number.
• Adding errors arithmetically when the question asks for probable (RSS) error, or vice versa.
• Forgetting that a moving-iron meter reads true RMS while a rectifier-PMMC meter is only RMS-calibrated for sinewaves.
• Writing one balance equation for an AC bridge; two (real and imaginary) are required, so two elements must be adjustable.
• Leaving a CT secondary open-circuited (dangerous high voltage) or an uncompensated 10× probe on a square wave and then "diagnosing" overshoot that the probe created.
• Confusing resolution with accuracy — a 6½-digit display can still be out by 1% if uncalibrated.
• Mixing units: mV with V, µε with ε, µF with F in bridge formulas, or forgetting ω = 2πf.

Related topics

Control systems basics · Digital signal processing: filters and FFT · Analog and digital electronic circuits · Electrical machines and power systems measurement · Wireless sensor networks · Electromagnetic compatibility and shielding