Skip to content

Commit

Permalink
Add unit tests for Resampler
Browse files Browse the repository at this point in the history
  • Loading branch information
@mhostetter
mhostetter committed Nov 12, 2023
1 parent 712c2dd commit 84b5675
Showing 1 changed file with 329 additions and 0 deletions.
329 changes: 329 additions & 0 deletions tests/dsp/test_resampler.py
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,329 @@
import numpy as np

import sdr


def debug_plot(y, y_truth):
import matplotlib.pyplot as plt

plt.figure()
sdr.plot.time_domain(y, label="Test")
sdr.plot.time_domain(y_truth, label="Truth")
plt.legend()
plt.show()


# def test_non_streaming_rate():
# mode = "rate"
# N = 50
# x = np.random.randn(N) + 1j * np.random.randn(N) # Input signal
# r = np.random.randint(3, 7) # Interpolation rate

# fir = sdr.Interpolator(r)
# y = fir(x, mode)

# # The output should align with the input. Every r-th sample should match.
# np.testing.assert_array_almost_equal(y[::r], x)


# def test_non_streaming_full():
# mode = "full"
# N = 50
# x = np.random.randn(N) + 1j * np.random.randn(N) # Input signal
# r = np.random.randint(3, 7) # Interpolation rate

# fir = sdr.Interpolator(r)
# y = fir(x, mode)

# xr = np.zeros(N * r, dtype=complex)
# xr[::r] = x[:]
# y_truth = scipy.signal.convolve(xr, fir.taps, mode=mode)

# # Given the polyphase decomposition, the polyphase output is slightly shorter
# np.testing.assert_array_almost_equal(y, y_truth[: y.size])


# def test_streaming():
# N = 50
# x = np.random.randn(N) + 1j * np.random.randn(N) # Input signal
# r = np.random.randint(3, 7) # Interpolation rate

# fir = sdr.Interpolator(r, streaming=True)

# d = 10 # Stride
# y = np.zeros(N * r, dtype=complex)
# for i in range(0, N, d):
# y[i * r : (i + d) * r] = fir(x[i : i + d])

# xr = np.zeros(N * r, dtype=complex)
# xr[::r] = x[:]
# y_truth = scipy.signal.convolve(xr, fir.taps, mode="full")[0 : N * r]

# np.testing.assert_array_almost_equal(y, y_truth)


# def test_streaming_match_full():
# N = 50
# x = np.random.randn(N) + 1j * np.random.randn(N) # Input signal
# r = np.random.randint(3, 7) # Interpolation rate

# fir1 = sdr.Interpolator(r)
# y_full = fir1(x, mode="full")

# fir2 = sdr.Interpolator(r, streaming=True)
# d = 10 # Stride
# y_stream = np.zeros_like(y_full)
# for i in range(0, N, d):
# y_stream[i * r : (i + d) * r] = fir2(x[i : i + d])
# y_stream[(i + d) * r :] = fir2.flush()

# np.testing.assert_array_almost_equal(y_full, y_stream)


def test_7_5_kaiser():
"""
Matlab:
>> x = cos(pi/4*(0:19)');
>> fir = dsp.FIRRateConverter(7, 5);
>> y = fir(x);
"""
x = np.cos(np.pi / 4 * np.arange(20))
y_truth = np.array(
[
0,
-0.000221371923965,
0.000574219878925,
-0.000488015063336,
-0.000845633199504,
0.003230097703218,
-0.004016421787642,
0,
0.008762301554564,
-0.015547832246794,
0.009668969889023,
0.013408679740684,
-0.042210721138260,
0.048500035713907,
0,
-0.129289109639484,
0.426798688747611,
1.065732496320718,
0.822489045403122,
0.270415487739051,
-0.189056975480333,
-0.707106781186547,
-0.992176816644380,
-0.928552056386907,
-0.628285559938961,
-0.115259551360420,
0.438835907100540,
0.844204496412674,
]
)

fir = sdr.Resampler(7, 5, streaming=True)
y = fir(x)

# debug_plot(y, y_truth)
np.testing.assert_almost_equal(y, y_truth)


def test_7_5_linear():
"""
Matlab:
>> x = cos(pi/4*(0:19)');
>> fir = dsp.FIRRateConverter(7, 5, 'linear');
>> y = fir(x);
"""
x = np.cos(np.pi / 4 * np.arange(20))
y_truth = np.array(
[
0.142857142857143,
0.857142857142857,
0.832632446392313,
0.505076272276105,
0.000000000000000,
-0.505076272276105,
-0.832632446392313,
-0.958158111598078,
-0.748948669588469,
-0.303045763365663,
0.202030508910442,
0.707106781186547,
0.916316223196156,
0.874474334794235,
0.606091526731327,
0.101015254455221,
-0.404061017820884,
-0.790790557990390,
-1.000000000000000,
-0.790790557990391,
-0.404061017820884,
0.101015254455221,
0.606091526731325,
0.874474334794234,
0.916316223196156,
0.707106781186547,
0.202030508910442,
-0.303045763365662,
]
)

fir = sdr.Resampler(7, 5, "linear-matlab", streaming=True)
y = fir(x)

# debug_plot(y, y_truth)
np.testing.assert_almost_equal(y, y_truth)


def test_7_5_zoh():
"""
Matlab:
>> x = cos(pi/4*(0:19)');
>> fir = dsp.FIRRateConverter(7, 5, 'zoh');
>> y = fir(x);
"""
x = np.cos(np.pi / 4 * np.arange(20))
y_truth = np.array(
[
1.000000000000000,
1.000000000000000,
0.707106781186548,
0.000000000000000,
0.000000000000000,
-0.707106781186547,
-1.000000000000000,
-0.707106781186548,
-0.707106781186548,
-0.000000000000000,
0.707106781186547,
0.707106781186547,
1.000000000000000,
0.707106781186548,
0.000000000000000,
0.000000000000000,
-0.707106781186547,
-1.000000000000000,
-1.000000000000000,
-0.707106781186547,
-0.000000000000000,
0.707106781186547,
0.707106781186547,
1.000000000000000,
0.707106781186547,
0.707106781186547,
0.000000000000001,
-0.707106781186546,
]
)

fir = sdr.Resampler(7, 5, "zoh", streaming=True)
y = fir(x)

# debug_plot(y, y_truth)
np.testing.assert_almost_equal(y, y_truth)


def test_5_7_kaiser():
"""
Matlab:
>> x = cos(pi/4*(0:20)');
>> fir = dsp.FIRRateConverter(5, 7);
>> y = fir(x);
"""
x = np.cos(np.pi / 4 * np.arange(20))
y_truth = np.array(
[
0.000069187197637,
-0.000581099350198,
0.001968738750881,
-0.004794480014143,
0.010241960638799,
-0.019969377248886,
0.037900876807341,
-0.079345721289059,
0.286007581662442,
0.964071849392977,
-0.034189330257715,
-0.873227486664808,
-0.818051016895984,
0.160578164604500,
0.949478110003271,
]
)

fir = sdr.Resampler(5, 7, streaming=True)
y = fir(x)

# debug_plot(y, y_truth)
np.testing.assert_almost_equal(y, y_truth)


def test_5_7_linear():
"""
Matlab:
>> x = cos(pi/4*(0:20)');
>> fir = dsp.FIRRateConverter(5, 7, 'linear');
>> y = fir(x);
"""
x = np.cos(np.pi / 4 * np.arange(20))
y_truth = np.array(
[
0.200000000000000,
0.824264068711929,
0.000000000000000,
-0.824264068711928,
-0.765685424949238,
0.141421356237309,
0.882842712474619,
0.707106781186548,
-0.282842712474619,
-0.941421356237309,
-0.565685424949238,
0.424264068711928,
1.000000000000000,
0.424264068711929,
-0.565685424949237,
]
)

fir = sdr.Resampler(5, 7, "linear-matlab", streaming=True)
y = fir(x)

# debug_plot(y, y_truth)
np.testing.assert_almost_equal(y, y_truth)


def test_5_7_zoh():
"""
Matlab:
>> x = cos(pi/4*(0:20)');
>> fir = dsp.FIRRateConverter(5, 7, 'zoh');
>> y = fir(x);
"""
x = np.cos(np.pi / 4 * np.arange(20))
y_truth = np.array(
[
1.000000000000000,
0.707106781186548,
0.000000000000000,
-1.000000000000000,
-0.707106781186548,
0.707106781186547,
1.000000000000000,
0.707106781186548,
-0.707106781186547,
-1.000000000000000,
-0.000000000000000,
0.707106781186547,
1.000000000000000,
0.000000000000001,
-0.707106781186546,
]
)

fir = sdr.Resampler(5, 7, "zoh", streaming=True)
y = fir(x)

# debug_plot(y, y_truth)
np.testing.assert_almost_equal(y, y_truth)

0 comments on commit 84b5675

Please sign in to comment.