ButterHighpass#

Description#

Digital high-pass Butterworth IIR filter, designed via the standard analog-prototype + lp2hp + bilinear pipeline.

\[ \text{ButterHighpass}(N, W_c) \]

  • order (int, \(\ge 1\)): filter order. Higher order = steeper roll-off, larger group delay.

  • cutoff_freq (float in \((0, 1)\)): normalised cutoff. 1 is the Nyquist frequency (matches scipy.signal.butter's Wn convention).

NaN handling#

Policy: ignore. A NaN in any input at index t causes the function to skip that step: output at t is NaN and internal state is unchanged. Subsequent finite samples are processed as if step t had not occurred.

Implementation Details#

Bit-exact match to scipy.signal.butter(order, cutoff, btype='highpass') + scipy.signal.lfilter. Verified to ~1e-12 in tests/test_signal.py.

from screamer import ButterHighpass
hp = ButterHighpass(order=4, cutoff_freq=0.1)
out = hp(noisy_signal)

Uses the same IIRFilter engine as the existing Butter low-pass class -- O(order) per step.