lasp/test/test_cppslm.py

98 lines
2.5 KiB
Python
Raw Permalink Normal View History

#!/usr/bin/python3
import numpy as np
from lasp import cppSLM
from lasp.filter import SPLFilterDesigner
2023-01-12 19:31:55 +00:00
def test_cppslm1():
"""
Generate a sine wave
"""
fs = 48000
2023-01-12 19:31:55 +00:00
omg = 2 * np.pi * 1000
slm = cppSLM.fromBiquads(fs, 2e-5, 1, 0.125,
np.array([[1., 0, 0, 1, 0, 0]]).T)
t = np.linspace(0, 10, 10 * fs, endpoint=False)
# Input signal with an rms of 1 Pa
in_ = np.sin(omg * t) * np.sqrt(2)
# Compute overall RMS
2023-01-12 19:31:55 +00:00
rms = np.sqrt(np.sum(in_**2) / in_.size)
# Compute overall level
2023-01-12 19:31:55 +00:00
level = 20 * np.log10(rms / 2e-5)
# Output of SLM
out = slm.run(in_)
2023-01-12 19:31:55 +00:00
# Output of SLM should be close to theoretical
# level, at least for reasonable time constants
# (Fast, Slow etc)
2023-01-12 19:31:55 +00:00
assert (np.isclose(out[-1, 0], level))
def test_cppslm2():
"""
Generate a sine wave, now A-weighted
"""
fs = 48000
2023-01-12 19:31:55 +00:00
omg = 2 * np.pi * 1000
filt = SPLFilterDesigner(fs).A_Sos_design()
2023-01-12 19:31:55 +00:00
slm = cppSLM.fromBiquads(fs, 2e-5, 0, 0.125,
filt.flatten(), # Pre-filter coefs
np.array([[1., 0, 0, 1, 0, 0]]).T # Bandpass coefs
)
t = np.linspace(0, 10, 10 * fs, endpoint=False)
# Input signal with an rms of 1 Pa
in_ = np.sin(omg * t) * np.sqrt(2)
# Compute overall RMS
2023-01-12 19:31:55 +00:00
rms = np.sqrt(np.sum(in_**2) / in_.size)
# Compute overall level
2023-01-12 19:31:55 +00:00
level = 20 * np.log10(rms / 2e-5)
# Output of SLM
out = slm.run(in_)
# Output of SLM should be close to theoretical
# level, at least for reasonable time constants
# (Fast, Slow etc)
2023-01-12 19:31:55 +00:00
assert np.isclose(out[-1, 0], level, atol=1e-2)
def test_cppslm3():
fs = 48000
2023-01-12 19:31:55 +00:00
omg = 2 * np.pi * 1000
filt = SPLFilterDesigner(fs).A_Sos_design()
2023-01-12 19:31:55 +00:00
slm = cppSLM.fromBiquads(fs, 2e-5, 0, 0.125,
filt.flatten(),
np.array([[1., 0, 0, 1, 0, 0]]).T)
t = np.linspace(0, 10, 10 * fs, endpoint=False)
in_ = 10 * np.sin(omg * t) * np.sqrt(2) + np.random.randn()
# Compute overall RMS
2023-01-12 19:31:55 +00:00
rms = np.sqrt(np.sum(in_**2) / in_.size)
# Compute overall level
2023-01-12 19:31:55 +00:00
level = 20 * np.log10(rms / 2e-5)
# Output of SLM
out = slm.run(in_)
2023-01-12 19:31:55 +00:00
Lpeak = 20 * np.log10(np.max(np.abs(in_) / 2e-5))
Lpeak
slm.Lpeak()
2023-01-12 19:31:55 +00:00
assert np.isclose(out[-1, 0], slm.Leq()[0], atol=1e-2)
assert np.isclose(Lpeak, slm.Lpeak()[0], atol=2e0)
if __name__ == '__main__':
test_cppslm1()
test_cppslm2()
test_cppslm3()