Short notes

Impulse Invariance Method Short Notes

Before the bilinear transform became the dominant method, DSP textbooks used impulse invariance to design the digital version of a Butterworth LPF for a speech codec — the idea is appealing: make the digital filter's impulse response a sampled version of the analog prototype's impulse response. Applied to a 4th-order Chebyshev analog filter for an 8 kHz sampled telephone speech system, the method works well because speech energy above 4 kHz is negligible, so the aliasing inherent in the method does not cause audible degradation. Whenever the signal is genuinely band-limited to well below fs/2, impulse invariance is valid and straightforward to apply.

ECE, EI

How it works

Design procedure: expand H_a(s) in partial fractions as Σ Ak/(s−sk). The digital filter H(z) = Σ Ak·T/(1−e^(sk·T)·z⁻¹), where T is the sampling period. Each s-plane pole at sk maps to a z-plane pole at e^(sk·T). Poles in the left-half s-plane (Re(sk) < 0) map inside the unit circle, preserving stability. The frequency response of the digital filter is H(e^(jω)) ≈ (1/T)·Σ Ha(j(ω−2πk)/T), showing that the digital response is the sum of shifted copies of the analog response — this aliasing is unavoidable and worsens as the analog filter's stopband attenuation is insufficient before the first alias appears at ω = π.

Key points to remember

Aliasing is the fundamental limitation of impulse invariance: the method cannot be used for high-pass or band-stop filters because their magnitude response does not decay to zero before fs/2, causing severe aliasing of the stopband into the passband. Impulse invariance preserves the shape of the impulse response and matches the analog frequency response well at frequencies well below π/T. The bilinear transform is preferred in modern design precisely because it eliminates aliasing at the cost of frequency warping. Both methods give the same filter order as the analog prototype. For a Butterworth analog prototype, the partial fraction expansion always has N distinct poles, each contributing one first-order section; pairing complex conjugate poles gives biquad (second-order section) implementations.

Exam tip

The examiner always asks you to compare impulse invariance and bilinear transform on four criteria — aliasing, frequency warping, suitable filter types, and stability preservation — write these four rows as a table and fill each cell concisely before writing any derivation.

More Digital Signal Processing notes