Something went wrong. Try again.
[READ-ONLY] Mirror of https://github.com/jmrplens/phonometry. Standards-conformant acoustics library for Python: measurement and metrology, docs.jmrp.io/phonometry
Something went wrong. Try again.
1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283# Copyright (c) 2026. Jose M. Requena-Plens"""Advanced audio processing tests including Pink Noise spectral analysis."""
from typing import cast
import matplotlib.pyplot as pltimport numpy as np
from phonometry import octavefilter
def generate_pink_noise(samples: int) -> np.ndarray: """ Generate pink noise (1/f noise) using the Voss-McCartney algorithm. """ num_cols = 16 rng = np.random.default_rng(42) # Seeded for reproducibility array = np.empty((samples, num_cols)) array.fill(np.nan) array[0, :] = rng.random(num_cols) array[:, 0] = rng.random(samples)
# Update random values at geometric intervals for i in range(1, samples): for j in range(1, num_cols): if i % (2**j) == 0: array[i, j] = rng.random() else: array[i, j] = array[i - 1, j]
pink = np.sum(array, axis=1) pink = pink - np.mean(pink) return cast(np.ndarray, pink / np.max(np.abs(pink)))
def test_pink_noise_flatness() -> None: """ Verify that Pink Noise produces a flat spectrum on a fractional octave analyzer.
**Purpose:** Pink noise has equal energy per octave (or fractional octave). This is a fundamental property used for calibrating audio systems. If we analyze pink noise with a 1/3 octave filter bank, the resulting SPL values should be approximately constant across all bands.
**Verification:** - Generate 2 seconds of pink noise using the Voss-McCartney algorithm. - Analyze it with a 1/3 octave filter. - Calculate the deviation of the SPL values from the mean.
**Expectation:** - The SPL spectrum should be relatively flat. - The maximum deviation from the mean SPL (ignoring edge bands due to filter transient/boundary effects) should be small (< 3 dB). """ fs = 48000 duration = 2.0 # Reduced for faster test samples = int(fs * duration) x = generate_pink_noise(samples)
spl, freq = octavefilter(x, fs=fs, fraction=3, order=6, limits=[20, 20000])
mean_spl = np.mean(spl)
# Check flatness in central bands (ignoring edges) valid_spl = np.array(spl)[2:-2] deviation = np.max(np.abs(valid_spl - np.mean(valid_spl)))
# Plot results for visual verification (optional in CI) _, ax = plt.subplots(figsize=(10, 6)) ax.semilogx(freq, spl, "b-o", label="Measured SPL", markerfacecolor="white") ax.axhline(float(mean_spl), color="r", linestyle="--", label="Mean SPL", alpha=0.7) ax.set_title("Pink Noise 1/3 Octave Spectrum (Flatness Check)", fontweight="bold") ax.set_xlabel("Frequency [Hz]") ax.set_ylabel("Level [dB]") ax.grid(True, which="both", alpha=0.3) ax.legend() plt.savefig("tests/pink_noise_test.png", dpi=150) plt.close()
assert deviation < 3.0, f"Spectrum deviation too high: {deviation:.2f} dB"