MEGR 3171  |  UNC Charlotte Mechatronics 2
Dr. Roger Tipton
Mechanical Engineering & Engineering Science
Back to Course Hub
Week 5 — Module 1: Precision Measurement & Signal Science

Active Filters & Data Acquisition

Designing analog filters to shape signals before digitization, and configuring DAQ systems to correctly capture those signals without aliasing artifacts.

Learning Objectives

1. Active Filter Fundamentals

Active filters use op-amps with RC feedback networks to achieve filtering without inductors. Advantages over passive RC filters: no signal attenuation in the passband, higher input impedance, lower output impedance, and ability to cascade stages without loading.

Filter Characterization Parameters

Cutoff Frequency (fc)

The -3 dB frequency: where output power drops to half (-3 dB = 0.707 of input voltage). The boundary between passband and transition band.

Filter Order (n)

Determines the roll-off slope. Each pole adds -20 dB/decade. A 2nd-order filter rolls off at -40 dB/decade beyond fc.

Butterworth Response

Maximally flat passband (no ripple). Moderate roll-off. Best choice when flat frequency response matters more than transition sharpness.

Chebyshev Response

Steeper roll-off than Butterworth at the cost of passband ripple. Used when sharp transition between pass and stop bands is critical.

Sallen-Key 2nd-Order Low-Pass (Butterworth)

Cutoff Frequency
f_c = 1 / (2π · R · C) (for equal R, equal C design) For Butterworth Q = 0.707: Select R1 = R2 = R, C1 = 2C, C2 = C (or use tables) Roll-off above f_c: -40 dB/decade (2nd order)

2. The Nyquist-Shannon Sampling Theorem

To faithfully reconstruct a continuous signal from its digital samples, the sampling rate must be at least twice the highest frequency component present in the signal. Sampling below this rate causes aliasing.

Nyquist Criterion
f_s ≥ 2 · f_max (Nyquist criterion) f_Nyquist = f_s / 2 (highest representable frequency) In practice: use f_s = 5 to 10 × f_max to leave margin for realistic (imperfect) anti-aliasing filter roll-off.

Aliasing

When a signal component above f_Nyquist is sampled, it "folds" back into the spectrum and appears as a false low-frequency component. Aliased frequencies cannot be distinguished from real signals after digitization. The only cure is an anti-aliasing filter applied before the ADC.

Exam Key Point Anti-aliasing filters must have their cutoff frequency set below f_Nyquist = f_s / 2, and they must attenuate out-of-band signals to below the ADC's quantization noise floor by f_Nyquist.

3. Data Acquisition Architecture

A complete DAQ chain: Anti-aliasing filter → Sample-and-hold (S/H) → Multiplexer (MUX) → Analog-to-Digital Converter (ADC) → Digital bus.

The sample-and-hold circuit freezes the analog value during the ADC conversion time (aperture time). The multiplexer sequentially connects multiple channels to a single ADC — note that multiplexed channels are not simultaneously sampled.

Quantization Error and ENOB
Quantization step size: Δ = V_FSR / 2^N Quantization noise (RMS): e_q = Δ / √12 ENOB = (SNR_measured - 1.76) / 6.02 where SNR is in dB. Ideal N-bit ADC: SNR = 6.02N + 1.76 dB

ENOB (Effective Number of Bits) reflects actual ADC performance degraded by noise, nonlinearity, and distortion. A 16-bit ADC may only achieve 14 ENOB in practice.

Practice Problems

Problem 1 — Anti-Aliasing Filter A DAQ system samples at 10 kHz. An accelerometer signal contains components up to 3 kHz. Specify the minimum required cutoff frequency for an ideal brick-wall anti-aliasing filter. With a real 2nd-order Butterworth, what cutoff would you choose and why?

Nyquist frequency = f_s / 2 = 5 kHz. An ideal filter would cut off exactly at 5 kHz to prevent aliasing.

However, a 2nd-order Butterworth has a gradual roll-off. At 5 kHz a signal component must be attenuated to below quantization noise. Setting f_c = 3 kHz (equal to signal bandwidth) leaves 1 octave (3 to 5 kHz) for the filter to attenuate. At 2 octaves above f_c, a 2nd-order filter attenuates by about 40 dB × 2 = 80 dB, which is typically sufficient.

Ideal brick-wall: f_c = 5 kHz. Real 2nd-order Butterworth: f_c ≈ 3 kHz. Increasing sampling rate to ≥30 kHz is preferred in practice.
Problem 2 — Quantization A 10-bit ADC with a 5 V full-scale range digitizes a signal. Calculate the quantization step size and RMS quantization noise voltage.

Δ = V_FSR / 2^N = 5.0 / 2^10 = 5.0 / 1024 ≈ 4.88 mV

e_q (RMS) = Δ / √12 = 4.88 / 3.464 ≈ 1.41 mV

Step size ≈ 4.88 mV, RMS quantization noise ≈ 1.41 mV
Problem 3 — Aliasing A 400 Hz sine wave is sampled at 500 Hz. At what frequency does the alias appear in the digitized spectrum?

Nyquist = 500/2 = 250 Hz. The 400 Hz signal is above Nyquist. Alias frequency = |f_signal - f_s| = |400 - 500| = 100 Hz. Alternatively: alias = f_s - f_signal = 500 - 400 = 100 Hz.

Alias appears at 100 Hz. The 400 Hz component is indistinguishable from a 100 Hz signal in the digitized data.