Active filters and op-amp applications

In short

An active filter uses op-amps with resistors and capacitors — no inductors — to shape a signal's frequency response, and the same op-amp blocks also perform integration, differentiation, instrumentation amplification, precision rectification, logarithmic compression and level detection. Because an op-amp provides gain, high input impedance and near-zero output impedance, active filters can be cascaded without loading one another, which is impossible with passive RC sections. The topic is heavily examined in GATE EC/EE (Analog Circuits), in university Linear Integrated Circuits papers, and in lab vivas on filter design. In industry it appears wherever a sensor signal must be conditioned before an ADC — Bharat Electronics, ISRO, DRDO labs, Tejas Networks, Kaynes, Syrma SGS, medical electronics (ECG front-ends) and energy-meter design.

Explained

Filter fundamentals and why "active"

• Filter - a network whose gain depends on frequency. Passband passes signals, stopband rejects them, and the boundary is the cut-off frequency f_c where gain falls 3 dB (to 0.707) below the passband value.
• Types by band: low-pass (LPF), high-pass (HPF), band-pass (BPF), band-stop/band-reject (notch), and all-pass (flat gain, frequency-dependent phase).
• Order (n) = number of independent reactive elements (poles). Roll-off is 20n dB/decade (6n dB/octave) outside the passband. A first-order section gives 20 dB/decade, second-order 40 dB/decade.
• Active advantages: gain greater than unity, no bulky/lossy inductors, buffered cascading, easy tuning, small size at low frequency (a 1 Hz filter needs no henry-sized coil).
• Active limitations: needs a DC supply; limited by the op-amp's gain-bandwidth product, slew rate, noise and offset; cannot handle high power or high voltage; above a few MHz passive LC designs win.
• Practical design rules: op-amp GBW ≳ 100·Q·f₀; slew rate SR ≥ 2πfV_p to avoid distortion; keep resistors in the 1 kΩ–1 MΩ range (low values load the op-amp, high values add thermal noise and offset error through bias current); use metal-film resistors (±1%) and polyester/polypropylene capacitors (±5% or better) because f₀ and Q depend on component tolerance.

Filter approximations (the "shape" of the response)

• Butterworth (maximally flat): flattest possible passband, monotonic, mild roll-off near f_c; the default choice. |A| = A₀/√(1+(f/f_c)²ⁿ).
• Chebyshev (equal-ripple): allows ripple (0.1–3 dB) in the passband to get a sharper transition for the same order; poorer step response (overshoot/ringing).
• Bessel (maximally flat delay): nearly constant group delay, so pulses pass without overshoot; the slowest roll-off. Used for data/pulse channels.
• Elliptic (Cauer): ripple in both bands plus transmission zeros — the steepest transition for a given order, used when a close-in interferer must be killed.
• Normalised Butterworth polynomials (for cascade design): n=2 → s²+1.414s+1; n=3 → (s+1)(s²+s+1); n=4 → (s²+0.765s+1)(s²+1.848s+1). The required stage Qs are therefore 1/1.414 = 0.707 for n = 2 and 1.307, 0.541 for n = 4. An odd-order filter is built as one first-order section plus second-order sections.

Standard filter circuits

• First-order LPF: non-inverting buffer fed by an RC low-pass, or an inverting amplifier with C across R_f. Gain A₀ = −R_f/R₁, f_c = 1/(2πR_fC).
• First-order HPF: C in series with the input resistor; f_c = 1/(2πR₁C), gain −R_f/R₁ above f_c.
• Sallen-Key (VCVS) second order: two RC sections with positive feedback from a non-inverting amplifier of gain K = 1 + R_b/R_a. Low component spread, low output impedance, Q set by K — but Q is sensitive to gain error, so avoid Q above about 10 in a single stage.
• Multiple-feedback (MFB / infinite-gain) second order: inverting topology, Q less sensitive to op-amp gain, inverts the signal; the usual choice for narrow band-pass.
• Band-pass: wide-band (BW > f₀, i.e. Q < 1) is made by cascading an HPF and an LPF; narrow-band (Q > 1) needs a single MFB or state-variable section.
• Notch: (a) twin-T network around an op-amp, f_n = 1/(2πRC), Q raised by feedback; (b) summing a band-pass output back from the input (band-reject = all − band-pass); (c) state-variable/biquad for tunable notches.
• State-variable (KHN/biquad): two integrators plus a summer — three or four op-amps give LP, HP and BP outputs simultaneously, with f₀ and Q independently adjustable; the basis of the UAF42-type universal filter.
• All-pass: |gain| = 1 at all frequencies, phase φ = −2·tan⁻¹(f/f₀) for first order; used for delay equalisation and phase shifting.
• Switched-capacitor filter (MF10, LTC1060, MAX7400): resistors replaced by capacitors switched at a clock rate f_clk; cut-off becomes f_c = f_clk/50 or /100, so the filter is digitally tunable and fully integrable — at the cost of clock feedthrough and the need for an anti-aliasing pre-filter.

Linear op-amp applications

• Inverting integrator: C in the feedback path. v_o = −(1/R₁C)∫v_i dt. A square wave in gives a triangular wave; a triangular wave gives a parabolic one. It is a 90° phase lag and a −20 dB/decade slope.
  • Practical integrator: DC offset and bias current charge C and saturate the output, so a large R_f (typically 10R₁ or more) is shunted across C. True integration holds above f_a = 1/(2πR_fC); below it the circuit is a plain inverting amplifier of gain −R_f/R₁.
• Differentiator: C in series with the input, R_f in feedback. v_o = −R_fC·dv_i/dt. Gain rises at +20 dB/decade, so it amplifies high-frequency noise and can oscillate; a small series R₁ (and small C_f across R_f) limits gain above f_b = 1/(2πR₁C) and stabilises it.
• Summing / difference / averaging amplifier: v_o = −R_f(v₁/R₁+v₂/R₂+…); with all R equal and R_f = R/n it averages. Difference amplifier v_o = (R₃/R₂)(v₂−v₁) only if the two resistor ratios match exactly.
• Instrumentation amplifier: two buffer op-amps with a shared gain resistor R_G feeding a difference amplifier. A_v = (1 + 2R/R_G)(R₃/R₂). Very high input impedance, gain set by one resistor, CMRR of 100–120 dB (AD620, INA128, LM324-based student versions) — the standard front-end for strain gauges, load cells, thermocouples and ECG electrodes.
• V-to-I and I-to-V converters: a transimpedance amplifier (photodiode into the virtual ground, v_o = −I·R_f) converts current to voltage; the Howland current pump and the floating-load configuration give a load-independent current, used for 4–20 mA transmitters.
• Log and antilog amplifiers: a diode or BJT in the feedback path gives v_o = −ηV_T·ln(v_i/(I_S R)); the antilog circuit swaps the positions. Used for compression, multiplication/division of analogue signals and dB-reading meters; strongly temperature dependent, so matched transistors and tempco compensation are needed.
• Analogue multiplier / divider: log + sum + antilog, or Gilbert cell; also voltage-controlled amplifiers.

Non-linear and switching op-amp applications

• Comparator: open-loop op-amp (or a dedicated comparator like LM311/LM339); output goes to ±V_sat depending on which input is larger. Noise on a slowly changing input causes multiple transitions near the threshold.
• Schmitt trigger (regenerative comparator): positive feedback through R₁–R₂ creates two thresholds. For the inverting Schmitt with the divider between output and the + input, V_UT/V_LT = ±V_sat·R₂/(R₁+R₂); hysteresis = V_UT − V_LT. It cleans noisy signals, squares up sine waves and sets switching levels in on/off controllers.
• Window comparator / level detector: two comparators plus a gate, flags whether a signal lies between two limits — used in protection and sorting circuits.
• Precision rectifier: a diode inside the feedback loop; the loop gain divides out the 0.6–0.7 V forward drop, so millivolt signals are rectified accurately. Half-wave (one diode, one op-amp) and full-wave/absolute-value (two op-amps) versions are the front end of true-RMS and average-reading AC meters.
• Peak detector: diode charging a hold capacitor through a buffer; holds the maximum value, with a reset switch. Used in vibration and audio level meters.
• Sample-and-hold: analogue switch (JFET/CMOS) plus hold capacitor plus FET-input buffer; freezes the input during ADC conversion. Errors: acquisition time, droop (leakage/C_H), pedestal and feedthrough.
• Clippers, clampers and voltage-limiters using zeners in the feedback path; current-to-frequency and voltage-to-frequency converters; gyrator (simulated inductor) to realise LC-like responses without coils.

Specifications, standards and Indian practice

• Anti-aliasing filter: every sampled-data system needs an LPF with cut-off below f_s/2; the required stopband attenuation fixes the order, which is why 4th- and 6th-order Butterworth or switched-capacitor filters are common ahead of ADCs.
• 50 Hz notch: India's mains and grid frequency is 50 Hz (operated by the regional load despatch centres under the CEA/CERC grid code, nominal band around 49.90–50.05 Hz), so biomedical and instrumentation front-ends here use 50 Hz notches, not the 60 Hz ones shown in many American textbooks.
• Biomedical: ECG channels follow IEC 60601-2-25/IS equivalents — typically 0.05–150 Hz bandwidth with a 50 Hz notch and a high-CMRR instrumentation amplifier; patient-safety requirements come from IEC 60601-1 (adopted as IS 13450 series by BIS).
• Power quality and metering: harmonic measurement instruments follow IEC 61000-4-7 and limits IEC/IS 61000-3-2; smart energy meters (IS 16444) and class-1 meters (IS 13779) use anti-aliasing and anti-tamper filtering ahead of their metering ADCs.
• Audio and communication: bass/treble (tone control) networks, de-emphasis, pre-emphasis, channel band-pass filters and DTMF detection filters; measurement weighting filters (A-weighting) in sound-level meters conform to IEC 61672.
• Device choices in Indian labs and products: 741 (single, slow, 0.5 V/µs), LM358/LM324 (low-cost dual/quad, single-supply), TL072/TL082 (JFET input, low bias current — good for integrators and filters), OP07 (precision, low offset), OPA/ADA parts for low noise, AD620/INA128 for instrumentation.

Key points

• Active filter - RC network plus op-amp; gives gain and buffering, needs no inductor.
• Cut-off frequency - where gain is 3 dB (0.707×) below passband; defines passband edge.
• Order n - fixes roll-off at 20n dB/decade and the achievable stopband attenuation.
• Quality factor Q - f₀/BW for a band-pass, 1/(2ζ) generally; Q = 0.707 gives a maximally flat (Butterworth) second-order response, Q = 0.5 is critically damped, Q > 1 peaks near f₀.
• Butterworth - flat passband; Chebyshev - sharper but rippled; Bessel - best pulse response; Elliptic - steepest transition.
• Sallen-Key - non-inverting second-order section; equal-component form has f₀ = 1/(2πRC) and Q = 1/(3−K).
• Multiple-feedback - inverting second-order section; preferred for high-Q band-pass, less sensitive to op-amp gain.
• State-variable/biquad - LP, HP and BP outputs together, with independent f₀ and Q tuning.
• Notch filter - twin-T or band-reject summing; in India tuned to 50 Hz for mains hum.
• Switched-capacitor filter - cut-off set by a clock (f_clk/100 or /50); digitally tunable, needs its own anti-aliasing filter.
• Integrator - v_o = −(1/RC)∫v_i dt; needs a shunt R_f to stop DC saturation.
• Differentiator - v_o = −RC·dv_i/dt; needs an input series resistor to stop noise amplification and instability.
• Instrumentation amplifier - A_v = (1+2R/R_G)(R₃/R₂); high Z_in, high CMRR, single-resistor gain control.
• Precision rectifier - diode inside the feedback loop removes the 0.7 V drop, so millivolt AC can be rectified.
• Schmitt trigger - positive feedback gives two thresholds; hysteresis = 2V_sat·R₂/(R₁+R₂) immunises against noise.
• Sample-and-hold - holds the analogue value during conversion; key errors are droop and acquisition time.
• Op-amp limits - GBW, slew rate, input offset and bias current set the highest usable frequency and the DC accuracy of every block above.
• Anti-aliasing - mandatory LPF below f_s/2 ahead of any ADC.

Formulas and examples

Example 1 — first-order active LPF. Design an inverting LPF with DC gain −10 and f_c = 1 kHz, then find the gain at 2 kHz.

• Pick R₁ = 10 kΩ → R_f = 100 kΩ (gain −10).
• C = 1/(2πR_f f_c) = 1/(2π × 10⁵ × 10³) = 1/(6.2832 × 10⁸) = 1.59 nF (use 1.5 nF and trim R_f).
• At 2 kHz: f/f_c = 2, |A| = 10/√(1+4) = 10/2.236 = 4.47 → 20·log₁₀4.47 = 13.0 dB (i.e. 7 dB below the 20 dB passband gain).

Example 2 — order selection and Sallen-Key design. A Butterworth LPF must be within 3 dB up to 1 kHz and at least 40 dB down at 4 kHz. Then design the section for f_c = 2 kHz.

• n ≥ log₁₀[(10⁴ − 1)/(10^0.3 − 1)]/(2·log₁₀4) = log₁₀(9999/0.9953)/(2 × 0.6021) = 4.002/1.2041 = 3.32 → n = 4 (two cascaded second-order sections, Q = 1.307 and 0.541).
• Equal-component Sallen-Key at f_c = 2 kHz with C = 10 nF: R = 1/(2πf_cC) = 1/(2π × 2000 × 10⁻⁸) = 1/1.2566×10⁻⁴ = 7.96 kΩ (use 8.2 kΩ).
• Butterworth Q = 0.707 → K = 3 − 1/0.707 = 1.586; with R_a = 10 kΩ, R_b = 0.586 × 10 k = 5.86 kΩ (5.6 kΩ + 270 Ω).
• Attenuation of this one section at 8 kHz (f/f_c = 4): 10·log₁₀(1 + 4⁴) = 10·log₁₀257 = 24.1 dB below its passband gain of 1.586.

Example 3 — narrow band-pass (MFB). Design f₀ = 1 kHz, Q = 5, centre gain −10, C = 10 nF.

• ω₀ = 2π × 1000 = 6283.2 rad/s; ω₀C = 6.2832 × 10⁻⁵ S.
• R₁ = Q/(|A₀|ω₀C) = 5/(10 × 6.2832×10⁻⁵) = 7.96 kΩ.
• R₃ = 2Q/(ω₀C) = 10/6.2832×10⁻⁵ = 159 kΩ.
• R₂ = Q/[(2Q²−|A₀|)ω₀C] = 5/(40 × 6.2832×10⁻⁵) = 1.99 kΩ.
• Check: R₁‖R₂ = 1591.6 Ω; f₀ = 1/(2πC√(R₁‖R₂ × R₃)) = 1/(2π × 10⁻⁸ × √(1591.6 × 159155)) = 1.00 kHz ✓; A₀ = −159155/(2 × 7957) = −10.0 ✓; BW = f₀/Q = 200 Hz.
• Op-amp needs GBW ≥ 100 × 5 × 1 kHz = 500 kHz — a 741 (≈1 MHz) is marginal; TL072 (3 MHz) is safe.

Example 4 — practical integrator. R₁ = 10 kΩ, C = 0.1 µF, R_f = 100 kΩ, supply ±15 V.

• R₁C = 10⁴ × 10⁻⁷ = 1 ms.
• A +1 V step applied for 5 ms: v_o = −(1/1 ms) × 1 V × 5 ms = −5 V (ramp, still inside saturation).
• A 1 kHz square wave of ±1 V: each half period is 0.5 ms, so the triangular output swings 1 × 0.5 ms/1 ms = 0.5 V peak-to-peak (±0.25 V).
• Lower frequency limit f_a = 1/(2πR_fC) = 1/(2π × 10⁵ × 10⁻⁷) = 15.9 Hz; below this the block behaves as an amplifier of gain −10, which is what stops drift from saturating it.

Example 5 — Schmitt trigger. V_sat = ±13 V, R₁ = 100 kΩ (output to + input), R₂ = 10 kΩ (+ input to ground).

• V_UT = +13 × 10/(110) = +1.18 V; V_LT = −1.18 V; hysteresis = 2.36 V.
• A sine wave of 5 V peak at 1 kHz therefore produces a clean square wave; input noise up to ±1.18 V peak cannot cause false triggering.
• To shift the switching band, return R₂ to a reference V_R instead of ground: thresholds become V_R·R₁/(R₁+R₂) ± V_sat·R₂/(R₁+R₂).

Where it is used

• Instrumentation and test: strain-gauge and load-cell bridges, thermocouple and RTD conditioning, vibration monitoring — instrumentation amplifier plus LPF plus 50 Hz notch; roles in BEL, ECIL, L&T, Honeywell and Yokogawa India design teams.
• Medical electronics: ECG/EEG/EMG front-ends (0.05–150 Hz band-pass, 50 Hz notch, high CMRR), pulse oximeters (transimpedance amplifier for the photodiode), at BPL Medical, Trivitron, Skanray and Wipro GE.
• Energy metering and power quality: anti-aliasing filters and precision rectifiers in energy meters (IS 13779, IS 16444 smart meters) and harmonic analysers (IEC 61000-4-7) at Secure Meters, Genus, HPL.
• Audio and consumer: tone controls, crossover networks for active speakers, mic preamplifiers, DTMF and tone detection — Sound-level meters use A-weighting filter networks.
• Communication and aerospace: channel band-pass filters, IF and baseband filtering, loop filters in PLLs, telemetry anti-aliasing filters in ISRO and DRDO payload electronics.
• Job roles: analogue/mixed-signal design engineer, signal-conditioning and sensor-interface engineer, hardware test and validation engineer, PCB and EMC engineer at ESDM companies such as Kaynes, Syrma SGS, Centum, SFO Technologies and Cyient.

Interview questions

• Q: Why are active filters preferred over passive LC filters at audio frequencies? A: They avoid bulky lossy inductors, provide gain greater than one, and their low output impedance lets stages be cascaded without interaction.
• Q: What does the Q of a filter mean physically? A: For a band-pass it is f₀/bandwidth, i.e. the selectivity; generally it measures how underdamped the poles are — Q = 0.707 is maximally flat, higher Q peaks near f₀ and rings on a step input.
• Q: Why is a second-order equal-component Sallen-Key given a gain of 1.586? A: Because Q = 1/(3 − K), so K = 1.586 gives Q = 0.707, the Butterworth value; a unity-gain equal-component section would give only Q = 0.5.
• Q: How do you get a fourth-order Butterworth filter? A: Cascade two second-order sections with the same f_c but Q = 1.307 and Q = 0.541, from the factors of the fourth-order Butterworth polynomial — not two identical Q = 0.707 stages.
• Q: Why does a practical integrator need a resistor across the capacitor? A: Input offset voltage and bias current act as a DC input that the capacitor integrates until the output saturates; the shunt resistor limits DC gain and sets the lowest frequency of true integration.
• Q: Why is an op-amp differentiator unstable? A: Its gain rises with frequency at +20 dB/decade, so it amplifies high-frequency noise and the extra phase lag of the feedback network can make the loop oscillate; a series input resistor (and small feedback capacitor) limits the high-frequency gain.
• Q: What makes a "precision" rectifier precise? A: The diode sits inside the feedback loop, so the op-amp's loop gain divides the 0.7 V forward drop down to microvolts referred to the input, letting millivolt signals be rectified accurately.
• Q: Difference between a comparator and a Schmitt trigger? A: A comparator has one threshold and can chatter on a noisy input; a Schmitt trigger uses positive feedback to create two thresholds (hysteresis), so once it switches the input must move back by the hysteresis before it switches again.
• Q: Why is an instrumentation amplifier better than a plain difference amplifier? A: It has very high input impedance on both inputs, its CMRR does not depend on matching the source resistances, and gain is set by one resistor without disturbing the matched ratios.
• Q: In India, why is the notch filter in biomedical equipment tuned to 50 Hz? A: The Indian supply and grid frequency is 50 Hz, so mains interference picked up by electrodes and leads appears at 50 Hz and its harmonics.
• Q: How does a switched-capacitor filter set its cut-off frequency? A: Resistors are replaced by capacitors switched at a clock rate, so the equivalent time constant depends on the clock; f_c is a fixed fraction of f_clk (commonly f_clk/100 or f_clk/50) and is therefore digitally tunable.
• Q: A filter is designed for f₀ = 50 kHz and Q = 10 but shows a sagging peak. What would you check first? A: The op-amp gain-bandwidth — Q·f₀ should be at least two orders below GBW (here ≥ 50 MHz by the 100·Q·f₀ rule); a 741 cannot sustain that loop gain, so use a faster amplifier.

Common mistakes

• Treating the −3 dB cut-off as the frequency where the signal is fully removed; attenuation at f_c is only 3 dB whatever the order.
• Cascading two identical Q = 0.707 sections and calling it a fourth-order Butterworth — the required Qs are 1.307 and 0.541.
• Omitting the shunt resistor on an integrator, or the series resistor on a differentiator, and then blaming the op-amp for drift or oscillation.
• Forgetting that the inverting LPF's f_c is set by the feedback pair R_f and C, not by the input resistor, and mixing kΩ with Ω or nF with µF in 1/(2πRC).
• Ignoring slew rate and GBW: a design that works at 1 kHz with a 741 fails at 100 kHz or at high Q.
• Using a comparator where hysteresis is needed, then trying to cure the resulting multiple transitions with capacitors.
• Assuming a difference amplifier gives good CMRR with 5% resistors; CMRR is limited by resistor ratio mismatch, so 0.1% parts or an instrumentation-amplifier IC are required.

Related topics

Operational amplifiers · Amplifiers and feedback · Data converters: ADC and DAC · Oscillators and power supplies · Network analysis: transients, resonance and two-port networks · Signals and systems: Laplace transform and frequency response