scipy 1.16.2__cp312-cp312-win_arm64.whl

scipy/signal/tests/test_fir_filter_design.py

@@ -0,0 +1,851 @@
+import math
+import numpy as np
+
+from numpy.testing import assert_warns
+from pytest import raises as assert_raises
+import pytest
+
+import scipy._lib.array_api_extra as xpx
+from scipy._lib._array_api import (
+    xp_assert_close, xp_assert_equal, assert_almost_equal, assert_array_almost_equal,
+    array_namespace, xp_default_dtype
+)
+from scipy.fft import fft, fft2
+from scipy.signal import (kaiser_beta, kaiser_atten, kaiserord,
+    firwin, firwin2, freqz, remez, firls, minimum_phase, convolve2d, firwin_2d
+)
+
+skip_xp_backends = pytest.mark.skip_xp_backends
+xfail_xp_backends = pytest.mark.xfail_xp_backends
+
+
+def test_kaiser_beta():
+    b = kaiser_beta(58.7)
+    assert_almost_equal(b, 0.1102 * 50.0)
+    b = kaiser_beta(22.0)
+    assert_almost_equal(b, 0.5842 + 0.07886)
+    b = kaiser_beta(21.0)
+    assert b == 0.0
+    b = kaiser_beta(10.0)
+    assert b == 0.0
+
+
+def test_kaiser_atten():
+    a = kaiser_atten(1, 1.0)
+    assert a == 7.95
+    a = kaiser_atten(2, 1/np.pi)
+    assert a == 2.285 + 7.95
+
+
+def test_kaiserord():
+    assert_raises(ValueError, kaiserord, 1.0, 1.0)
+    numtaps, beta = kaiserord(2.285 + 7.95 - 0.001, 1/np.pi)
+    assert (numtaps, beta) == (2, 0.0)
+
+
+class TestFirwin:
+
+    def check_response(self, h, expected_response, tol=.05):
+        xp = array_namespace(h)
+        N = h.shape[0]
+        alpha = 0.5 * (N-1)
+        m = xp.arange(0, N) - alpha   # time indices of taps
+        for freq, expected in expected_response:
+            actual = abs(xp.sum(h * xp.exp(-1j * xp.pi * m * freq)))
+            mse = abs(actual - expected)**2
+            assert mse < tol, f'response not as expected, mse={mse:g} > {tol:g}'
+
+    def test_response(self, xp):
+        N = 51
+        f = .5
+
+        # increase length just to try even/odd
+        h = firwin(N, f)  # low-pass from 0 to f
+        self.check_response(h, [(.25,1), (.75,0)])
+
+        h = firwin(N+1, f, window='nuttall')  # specific window
+        self.check_response(h, [(.25,1), (.75,0)])
+
+        h = firwin(N+2, f, pass_zero=False)  # stop from 0 to f --> high-pass
+        self.check_response(h, [(.25,0), (.75,1)])
+
+        f1, f2, f3, f4 = .2, .4, .6, .8
+        h = firwin(N+3, [f1, f2], pass_zero=False)  # band-pass filter
+        self.check_response(h, [(.1,0), (.3,1), (.5,0)])
+
+        h = firwin(N+4, [f1, f2])  # band-stop filter
+        self.check_response(h, [(.1,1), (.3,0), (.5,1)])
+
+        h = firwin(N+5, [f1, f2, f3, f4], pass_zero=False, scale=False)
+        self.check_response(h, [(.1,0), (.3,1), (.5,0), (.7,1), (.9,0)])
+
+        h = firwin(N+6, [f1, f2, f3, f4])  # multiband filter
+        self.check_response(h, [(.1,1), (.3,0), (.5,1), (.7,0), (.9,1)])
+
+        h = firwin(N+7, 0.1, width=.03)  # low-pass
+        self.check_response(h, [(.05,1), (.75,0)])
+
+        h = firwin(N+8, 0.1, pass_zero=False)  # high-pass
+        self.check_response(h, [(.05,0), (.75,1)])
+
+    def mse(self, h, bands):
+        """Compute mean squared error versus ideal response across frequency
+        band.
+          h -- coefficients
+          bands -- list of (left, right) tuples relative to 1==Nyquist of
+            passbands
+        """
+        w, H = freqz(h, worN=1024)
+        f = w/np.pi
+        passIndicator = np.zeros(len(w), bool)
+        for left, right in bands:
+            passIndicator |= (f >= left) & (f < right)
+        Hideal = np.where(passIndicator, 1, 0)
+        mse = np.mean(abs(abs(H)-Hideal)**2)
+        return mse
+
+    def test_scaling(self, xp):
+        """
+        For one lowpass, bandpass, and highpass example filter, this test
+        checks two things:
+          - the mean squared error over the frequency domain of the unscaled
+            filter is smaller than the scaled filter (true for rectangular
+            window)
+          - the response of the scaled filter is exactly unity at the center
+            of the first passband
+        """
+        N = 11
+        cases = [
+            ([.5], True, (0, 1)),
+            ([0.2, .6], False, (.4, 1)),
+            ([.5], False, (1, 1)),
+        ]
+        for cutoff, pass_zero, expected_response in cases:
+            h = firwin(N, cutoff, scale=False, pass_zero=pass_zero, window='ones')
+            hs = firwin(N, cutoff, scale=True, pass_zero=pass_zero, window='ones')
+            if len(cutoff) == 1:
+                if pass_zero:
+                    cutoff = [0] + cutoff
+                else:
+                    cutoff = cutoff + [1]
+            msg = 'least squares violation'
+            assert self.mse(h, [cutoff]) < self.mse(hs, [cutoff]), msg
+            self.check_response(hs, [expected_response], 1e-12)
+
+    def test_fs_validation(self):
+        with pytest.raises(ValueError, match="Sampling.*single scalar"):
+            firwin(51, .5, fs=np.array([10, 20]))
+
+
+class TestFirWinMore:
+    """Different author, different style, different tests..."""
+
+    def test_lowpass(self, xp):
+        width = 0.04
+        ntaps, beta = kaiserord(120, width)
+        cutoff = xp.asarray(0.5)
+        kwargs = dict(cutoff=cutoff, window=('kaiser', beta), scale=False)
+        taps = firwin(ntaps, **kwargs)
+
+        # Check the symmetry of taps.
+        assert_array_almost_equal(taps[:ntaps//2], xp.flip(taps)[:ntaps//2])
+
+        # Check the gain at a few samples where
+        # we know it should be approximately 0 or 1.
+        freq_samples = xp.asarray([0.0, 0.25, 0.5-width/2, 0.5+width/2, 0.75, 1.0])
+        freqs, response = freqz(taps, worN=xp.pi*freq_samples)
+
+        assert_array_almost_equal(
+            xp.abs(response),
+            xp.asarray([1.0, 1.0, 1.0, 0.0, 0.0, 0.0]), decimal=5
+        )
+
+        taps_str = firwin(ntaps, pass_zero='lowpass', **kwargs)
+        xp_assert_close(taps, taps_str)
+
+    def test_highpass(self, xp):
+        width = 0.04
+        ntaps, beta = kaiserord(120, width)
+
+        # Ensure that ntaps is odd.
+        ntaps |= 1
+
+        cutoff = xp.asarray(0.5)
+        kwargs = dict(cutoff=cutoff, window=('kaiser', beta), scale=False)
+        taps = firwin(ntaps, pass_zero=False, **kwargs)
+
+        # Check the symmetry of taps.
+        assert_array_almost_equal(taps[:ntaps//2], xp.flip(taps)[:ntaps//2])
+
+        # Check the gain at a few samples where
+        # we know it should be approximately 0 or 1.
+        freq_samples = xp.asarray([0.0, 0.25, 0.5 - width/2, 0.5 + width/2, 0.75, 1.0])
+        freqs, response = freqz(taps, worN=np.pi*freq_samples)
+
+        assert_array_almost_equal(xp.abs(response),
+                                  xp.asarray([0.0, 0.0, 0.0, 1.0, 1.0, 1.0]), decimal=5)
+
+        taps_str = firwin(ntaps, pass_zero='highpass', **kwargs)
+        xp_assert_close(taps, taps_str)
+
+    def test_bandpass(self, xp):
+        width = 0.04
+        ntaps, beta = kaiserord(120, width)
+        kwargs = dict(
+            cutoff=xp.asarray([0.3, 0.7]), window=('kaiser', beta), scale=False
+        )
+        taps = firwin(ntaps, pass_zero=False, **kwargs)
+
+        # Check the symmetry of taps.
+        assert_array_almost_equal(taps[:ntaps//2], xp.flip(taps)[:ntaps//2])
+
+        # Check the gain at a few samples where
+        # we know it should be approximately 0 or 1.
+        freq_samples = xp.asarray([0.0, 0.2, 0.3 - width/2, 0.3 + width/2, 0.5,
+                                   0.7 - width/2, 0.7 + width/2, 0.8, 1.0])
+        freqs, response = freqz(taps, worN=np.pi*freq_samples)
+
+        assert_array_almost_equal(xp.abs(response),
+                xp.asarray([0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0]), decimal=5)
+
+        taps_str = firwin(ntaps, pass_zero='bandpass', **kwargs)
+        xp_assert_close(taps, taps_str)
+
+    def test_bandstop_multi(self, xp):
+        width = 0.04
+        ntaps, beta = kaiserord(120, width)
+        kwargs = dict(cutoff=xp.asarray([0.2, 0.5, 0.8]), window=('kaiser', beta),
+                      scale=False)
+        taps = firwin(ntaps, **kwargs)
+
+        # Check the symmetry of taps.
+        assert_array_almost_equal(taps[:ntaps//2], xp.flip(taps)[:ntaps//2])
+
+        # Check the gain at a few samples where
+        # we know it should be approximately 0 or 1.
+        freq_samples = xp.asarray([0.0, 0.1, 0.2 - width/2, 0.2 + width/2, 0.35,
+                                   0.5 - width/2, 0.5 + width/2, 0.65,
+                                   0.8 - width/2, 0.8 + width/2, 0.9, 1.0])
+        freqs, response = freqz(taps, worN=np.pi*freq_samples)
+
+        assert_array_almost_equal(
+            xp.abs(response),
+            xp.asarray([1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0]),
+            decimal=5
+        )
+
+        taps_str = firwin(ntaps, pass_zero='bandstop', **kwargs)
+        xp_assert_close(taps, taps_str)
+
+    def test_fs_nyq(self, xp):
+        """Test the fs and nyq keywords."""
+        nyquist = 1000
+        width = 40.0
+        relative_width = width/nyquist
+        ntaps, beta = kaiserord(120, relative_width)
+        taps = firwin(ntaps, cutoff=xp.asarray([300, 700]), window=('kaiser', beta),
+                        pass_zero=False, scale=False, fs=2*nyquist)
+
+        # Check the symmetry of taps.
+        assert_array_almost_equal(taps[:ntaps//2], xp.flip(taps)[:ntaps//2])
+
+        # Check the gain at a few samples where
+        # we know it should be approximately 0 or 1.
+        freq_samples = xp.asarray([0.0, 200, 300 - width/2, 300 + width/2, 500,
+                                   700 - width/2, 700 + width/2, 800, 1000])
+        freqs, response = freqz(taps, worN=np.pi*freq_samples/nyquist)
+
+        assert_array_almost_equal(xp.abs(response),
+                xp.asarray([0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0]), decimal=5)
+
+    def test_array_cutoff(self, xp):
+        taps = firwin(3, xp.asarray([.1, .2]))
+        # smoke test against the value computed by scipy==1.5.2
+        xp_assert_close(
+            taps, xp.asarray([-0.00801395, 1.0160279, -0.00801395]), atol=1e-8
+        )
+
+    def test_bad_cutoff(self):
+        """Test that invalid cutoff argument raises ValueError."""
+        # cutoff values must be greater than 0 and less than 1.
+        assert_raises(ValueError, firwin, 99, -0.5)
+        assert_raises(ValueError, firwin, 99, 1.5)
+        # Don't allow 0 or 1 in cutoff.
+        assert_raises(ValueError, firwin, 99, [0, 0.5])
+        assert_raises(ValueError, firwin, 99, [0.5, 1])
+        # cutoff values must be strictly increasing.
+        assert_raises(ValueError, firwin, 99, [0.1, 0.5, 0.2])
+        assert_raises(ValueError, firwin, 99, [0.1, 0.5, 0.5])
+        # Must have at least one cutoff value.
+        assert_raises(ValueError, firwin, 99, [])
+        # 2D array not allowed.
+        assert_raises(ValueError, firwin, 99, [[0.1, 0.2],[0.3, 0.4]])
+        # cutoff values must be less than nyq.
+        assert_raises(ValueError, firwin, 99, 50.0, fs=80)
+        assert_raises(ValueError, firwin, 99, [10, 20, 30], fs=50)
+
+    def test_even_highpass_raises_value_error(self):
+        """Test that attempt to create a highpass filter with an even number
+        of taps raises a ValueError exception."""
+        assert_raises(ValueError, firwin, 40, 0.5, pass_zero=False)
+        assert_raises(ValueError, firwin, 40, [.25, 0.5])
+
+    def test_bad_pass_zero(self):
+        """Test degenerate pass_zero cases."""
+        with assert_raises(ValueError, match="^Parameter pass_zero='foo' not in "):
+            firwin(41, 0.5, pass_zero='foo')
+        with assert_raises(ValueError, match="^Parameter pass_zero=1.0 not in "):
+            firwin(41, 0.5, pass_zero=1.)
+        for pass_zero in ('lowpass', 'highpass'):
+            with assert_raises(ValueError, match='cutoff must have one'):
+                firwin(41, [0.5, 0.6], pass_zero=pass_zero)
+        for pass_zero in ('bandpass', 'bandstop'):
+            with assert_raises(ValueError, match='must have at least two'):
+                firwin(41, [0.5], pass_zero=pass_zero)
+
+    def test_fs_validation(self):
+        with pytest.raises(ValueError, match="Sampling.*single scalar"):
+            firwin2(51, .5, 1, fs=np.array([10, 20]))
+
+
+@skip_xp_backends(cpu_only=True, reason="firwin2 uses np.interp")
+class TestFirwin2:
+
+    def test_invalid_args(self):
+        # `freq` and `gain` have different lengths.
+        with assert_raises(ValueError, match='must be of same length'):
+            firwin2(50, [0, 0.5, 1], [0.0, 1.0])
+        # `nfreqs` is less than `ntaps`.
+        with assert_raises(ValueError, match='ntaps must be less than nfreqs'):
+            firwin2(50, [0, 0.5, 1], [0.0, 1.0, 1.0], nfreqs=33)
+        # Decreasing value in `freq`
+        with assert_raises(ValueError, match='must be nondecreasing'):
+            firwin2(50, [0, 0.5, 0.4, 1.0], [0, .25, .5, 1.0])
+        # Value in `freq` repeated more than once.
+        with assert_raises(ValueError, match='must not occur more than twice'):
+            firwin2(50, [0, .1, .1, .1, 1.0], [0.0, 0.5, 0.75, 1.0, 1.0])
+        # `freq` does not start at 0.0.
+        with assert_raises(ValueError, match='start with 0'):
+            firwin2(50, [0.5, 1.0], [0.0, 1.0])
+        # `freq` does not end at fs/2.
+        with assert_raises(ValueError, match='end with fs/2'):
+            firwin2(50, [0.0, 0.5], [0.0, 1.0])
+        # Value 0 is repeated in `freq`
+        with assert_raises(ValueError, match='0 must not be repeated'):
+            firwin2(50, [0.0, 0.0, 0.5, 1.0], [1.0, 1.0, 0.0, 0.0])
+        # Value fs/2 is repeated in `freq`
+        with assert_raises(ValueError, match='fs/2 must not be repeated'):
+            firwin2(50, [0.0, 0.5, 1.0, 1.0], [1.0, 1.0, 0.0, 0.0])
+        # Value in `freq` that is too close to a repeated number
+        with assert_raises(ValueError, match='cannot contain numbers '
+                                             'that are too close'):
+            firwin2(50, [0.0, 0.5 - np.finfo(float).eps * 0.5, 0.5, 0.5, 1.0],
+                        [1.0, 1.0, 1.0, 0.0, 0.0])
+
+        # Type II filter, but the gain at nyquist frequency is not zero.
+        with assert_raises(ValueError, match='Type II filter'):
+            firwin2(16, [0.0, 0.5, 1.0], [0.0, 1.0, 1.0])
+
+        # Type III filter, but the gains at nyquist and zero rate are not zero.
+        with assert_raises(ValueError, match='Type III filter'):
+            firwin2(17, [0.0, 0.5, 1.0], [0.0, 1.0, 1.0], antisymmetric=True)
+        with assert_raises(ValueError, match='Type III filter'):
+            firwin2(17, [0.0, 0.5, 1.0], [1.0, 1.0, 0.0], antisymmetric=True)
+        with assert_raises(ValueError, match='Type III filter'):
+            firwin2(17, [0.0, 0.5, 1.0], [1.0, 1.0, 1.0], antisymmetric=True)
+
+        # Type IV filter, but the gain at zero rate is not zero.
+        with assert_raises(ValueError, match='Type IV filter'):
+            firwin2(16, [0.0, 0.5, 1.0], [1.0, 1.0, 0.0], antisymmetric=True)
+
+    def test01(self, xp):
+        width = 0.04
+        beta = 12.0
+        ntaps = 400
+        # Filter is 1 from w=0 to w=0.5, then decreases linearly from 1 to 0 as w
+        # increases from w=0.5 to w=1  (w=1 is the Nyquist frequency).
+        freq = xp.asarray([0.0, 0.5, 1.0])
+        gain = xp.asarray([1.0, 1.0, 0.0])
+        taps = firwin2(ntaps, freq, gain, window=('kaiser', beta))
+        freq_samples = xp.asarray([0.0, 0.25, 0.5 - width/2, 0.5 + width/2,
+                                                        0.75, 1.0 - width/2])
+        freqs, response = freqz(taps, worN=np.pi*freq_samples)
+        freqs, response = xp.asarray(freqs), xp.asarray(response)
+        assert_array_almost_equal(
+            xp.abs(response),
+            xp.asarray([1.0, 1.0, 1.0, 1.0 - width, 0.5, width]), decimal=5
+        )
+
+    @skip_xp_backends("jax.numpy", reason="immutable arrays")
+    def test02(self, xp):
+        width = 0.04
+        beta = 12.0
+        # ntaps must be odd for positive gain at Nyquist.
+        ntaps = 401
+        # An ideal highpass filter.
+        freq = xp.asarray([0.0, 0.5, 0.5, 1.0])
+        gain = xp.asarray([0.0, 0.0, 1.0, 1.0])
+        taps = firwin2(ntaps, freq, gain, window=('kaiser', beta))
+        freq_samples = np.array([0.0, 0.25, 0.5 - width, 0.5 + width, 0.75, 1.0])
+        freqs, response = freqz(taps, worN=np.pi*freq_samples)
+        freqs, response = xp.asarray(freqs), xp.asarray(response)
+        assert_array_almost_equal(
+            xp.abs(response),
+            xp.asarray([0.0, 0.0, 0.0, 1.0, 1.0, 1.0]), decimal=5
+        )
+
+    @skip_xp_backends("jax.numpy", reason="immutable arrays")
+    def test03(self, xp):
+        width = 0.02
+        ntaps, beta = kaiserord(120, width)
+        # ntaps must be odd for positive gain at Nyquist.
+        ntaps = int(ntaps) | 1
+        freq = xp.asarray([0.0, 0.4, 0.4, 0.5, 0.5, 1.0])
+        gain = xp.asarray([1.0, 1.0, 0.0, 0.0, 1.0, 1.0])
+        taps = firwin2(ntaps, freq, gain, window=('kaiser', beta))
+        freq_samples = np.array([0.0, 0.4 - width, 0.4 + width, 0.45,
+                                    0.5 - width, 0.5 + width, 0.75, 1.0])
+        freqs, response = freqz(taps, worN=np.pi*freq_samples)
+        freqs, response = xp.asarray(freqs), xp.asarray(response)
+        assert_array_almost_equal(
+            xp.abs(response),
+            xp.asarray([1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0]), decimal=5
+        )
+
+    @skip_xp_backends("jax.numpy", reason="immutable arrays")
+    def test04(self, xp):
+        """Test firwin2 when window=None."""
+        ntaps = 5
+        # Ideal lowpass: gain is 1 on [0,0.5], and 0 on [0.5, 1.0]
+        freq = xp.asarray([0.0, 0.5, 0.5, 1.0])
+        gain = xp.asarray([1.0, 1.0, 0.0, 0.0])
+
+        taps = firwin2(ntaps, freq, gain, window=None, nfreqs=8193)
+        alpha = 0.5 * (ntaps - 1)
+        m = xp.arange(0, ntaps, dtype=freq.dtype) - alpha
+        h = 0.5 * xpx.sinc(0.5 * m)
+        assert_array_almost_equal(h, taps)
+
+    def test05(self, xp):
+        """Test firwin2 for calculating Type IV filters"""
+        ntaps = 1500
+
+        freq = xp.asarray([0.0, 1.0])
+        gain = xp.asarray([0.0, 1.0])
+        taps = firwin2(ntaps, freq, gain, window=None, antisymmetric=True)
+
+        flip = array_namespace(freq).flip
+        dec = {'decimal': 4.5} if xp_default_dtype(xp) == xp.float32 else {}
+        assert_array_almost_equal(taps[: ntaps // 2], flip(-taps[ntaps // 2:]), **dec)
+
+        freqs, response = freqz(np.asarray(taps), worN=2048)    # XXX convert freqz
+        assert_array_almost_equal(abs(xp.asarray(response)),
+                                  xp.asarray(freqs / np.pi), decimal=4)
+
+    @skip_xp_backends("jax.numpy", reason="immutable arrays")
+    def test06(self, xp):
+        """Test firwin2 for calculating Type III filters"""
+        ntaps = 1501
+
+        freq = xp.asarray([0.0, 0.5, 0.55, 1.0])
+        gain = xp.asarray([0.0, 0.5, 0.0, 0.0])
+        taps = firwin2(ntaps, freq, gain, window=None, antisymmetric=True)
+        assert taps[ntaps // 2] == 0.0
+
+        flip = array_namespace(freq).flip
+        dec = {'decimal': 4.5} if xp_default_dtype(xp) == xp.float32 else {}
+        assert_array_almost_equal(taps[: ntaps // 2],
+                                  flip(-taps[ntaps // 2 + 1:]), **dec
+        )
+
+        freqs, response1 = freqz(np.asarray(taps), worN=2048)  # XXX convert freqz
+        response1 = xp.asarray(response1)
+        response2 = xp.asarray(
+            np.interp(np.asarray(freqs) / np.pi, np.asarray(freq), np.asarray(gain))
+        )
+        assert_array_almost_equal(abs(response1), response2, decimal=3)
+
+    def test_fs_nyq(self, xp):
+        taps1 = firwin2(80, xp.asarray([0.0, 0.5, 1.0]), xp.asarray([1.0, 1.0, 0.0]))
+        taps2 = firwin2(80, xp.asarray([0.0, 30.0, 60.0]), xp.asarray([1.0, 1.0, 0.0]),
+                        fs=120.0)
+        assert_array_almost_equal(taps1, taps2)
+
+    def test_tuple(self):
+        taps1 = firwin2(150, (0.0, 0.5, 0.5, 1.0), (1.0, 1.0, 0.0, 0.0))
+        taps2 = firwin2(150, [0.0, 0.5, 0.5, 1.0], [1.0, 1.0, 0.0, 0.0])
+        assert_array_almost_equal(taps1, taps2)
+
+    @skip_xp_backends("jax.numpy", reason="immutable arrays")
+    def test_input_modyfication(self, xp):
+        freq1 = xp.asarray([0.0, 0.5, 0.5, 1.0])
+        freq2 = xp.asarray(freq1)
+        firwin2(80, freq1, xp.asarray([1.0, 1.0, 0.0, 0.0]))
+        xp_assert_equal(freq1, freq2)
+
+
+@skip_xp_backends(cpu_only=True)
+class TestRemez:
+
+    def test_bad_args(self):
+        assert_raises(ValueError, remez, 11, [0.1, 0.4], [1], type='pooka')
+
+    def test_hilbert(self):
+        N = 11  # number of taps in the filter
+        a = 0.1  # width of the transition band
+
+        # design an unity gain hilbert bandpass filter from w to 0.5-w
+        h = remez(11, [a, 0.5-a], [1], type='hilbert')
+
+        # make sure the filter has correct # of taps
+        assert len(h) == N, "Number of Taps"
+
+        # make sure it is type III (anti-symmetric tap coefficients)
+        assert_array_almost_equal(h[:(N-1)//2], -h[:-(N-1)//2-1:-1])
+
+        # Since the requested response is symmetric, all even coefficients
+        # should be zero (or in this case really small)
+        assert (abs(h[1::2]) < 1e-15).all(), "Even Coefficients Equal Zero"
+
+        # now check the frequency response
+        w, H = freqz(h, 1)
+        f = w/2/np.pi
+        Hmag = abs(H)
+
+        # should have a zero at 0 and pi (in this case close to zero)
+        assert (Hmag[[0, -1]] < 0.02).all(), "Zero at zero and pi"
+
+        # check that the pass band is close to unity
+        idx = np.logical_and(f > a, f < 0.5-a)
+        assert (abs(Hmag[idx] - 1) < 0.015).all(), "Pass Band Close To Unity"
+
+    def test_compare(self, xp):
+        # test comparison to MATLAB
+        k = [0.024590270518440, -0.041314581814658, -0.075943803756711,
+             -0.003530911231040, 0.193140296954975, 0.373400753484939,
+             0.373400753484939, 0.193140296954975, -0.003530911231040,
+             -0.075943803756711, -0.041314581814658, 0.024590270518440]
+        h = remez(12, xp.asarray([0, 0.3, 0.5, 1]), xp.asarray([1, 0]), fs=2.)
+        atol_arg = {'atol': 1e-8} if xp_default_dtype(xp) == xp.float32 else {}
+        xp_assert_close(h, xp.asarray(k, dtype=xp.float64), **atol_arg)
+
+        h = [-0.038976016082299, 0.018704846485491, -0.014644062687875,
+             0.002879152556419, 0.016849978528150, -0.043276706138248,
+             0.073641298245579, -0.103908158578635, 0.129770906801075,
+             -0.147163447297124, 0.153302248456347, -0.147163447297124,
+             0.129770906801075, -0.103908158578635, 0.073641298245579,
+             -0.043276706138248, 0.016849978528150, 0.002879152556419,
+             -0.014644062687875, 0.018704846485491, -0.038976016082299]
+        atol_arg = {'atol': 3e-8} if xp_default_dtype(xp) == xp.float32 else {}
+        xp_assert_close(
+            remez(21, xp.asarray([0, 0.8, 0.9, 1]), xp.asarray([0, 1]), fs=2.),
+            xp.asarray(h, dtype=xp.float64), **atol_arg
+        )
+
+    def test_fs_validation(self):
+        with pytest.raises(ValueError, match="Sampling.*single scalar"):
+            remez(11, .1, 1, fs=np.array([10, 20]))
+
+    def test_gh_23266(self, xp):
+        bands = xp.asarray([0.0, 0.2, 0.3, 0.5])
+        desired = xp.asarray([1.0, 0.0])
+        weight = xp.asarray([1.0, 2.0])
+        remez(21, bands, desired, weight=weight)
+
+
+@skip_xp_backends(cpu_only=True, reason="lstsq")
+class TestFirls:
+
+    def test_bad_args(self):
+        # even numtaps
+        assert_raises(ValueError, firls, 10, [0.1, 0.2], [0, 0])
+        # odd bands
+        assert_raises(ValueError, firls, 11, [0.1, 0.2, 0.4], [0, 0, 0])
+        # len(bands) != len(desired)
+        assert_raises(ValueError, firls, 11, [0.1, 0.2, 0.3, 0.4], [0, 0, 0])
+        # non-monotonic bands
+        assert_raises(ValueError, firls, 11, [0.2, 0.1], [0, 0])
+        assert_raises(ValueError, firls, 11, [0.1, 0.2, 0.3, 0.3], [0] * 4)
+        assert_raises(ValueError, firls, 11, [0.3, 0.4, 0.1, 0.2], [0] * 4)
+        assert_raises(ValueError, firls, 11, [0.1, 0.3, 0.2, 0.4], [0] * 4)
+        # negative desired
+        assert_raises(ValueError, firls, 11, [0.1, 0.2], [-1, 1])
+        # len(weight) != len(pairs)
+        assert_raises(ValueError, firls, 11, [0.1, 0.2], [0, 0], weight=[1, 2])
+        # negative weight
+        assert_raises(ValueError, firls, 11, [0.1, 0.2], [0, 0], weight=[-1])
+
+    @skip_xp_backends("dask.array", reason="dask fancy indexing shape=(nan,)")
+    def test_firls(self, xp):
+        N = 11  # number of taps in the filter
+        a = 0.1  # width of the transition band
+
+        # design a halfband symmetric low-pass filter
+        h = firls(11, xp.asarray([0, a, 0.5 - a, 0.5]), xp.asarray([1, 1, 0, 0]),
+                  fs=1.0)
+
+        # make sure the filter has correct # of taps
+        assert h.shape[0] == N
+
+        # make sure it is symmetric
+        midx = (N-1) // 2
+        flip = array_namespace(h).flip
+        assert_array_almost_equal(h[:midx],  flip(h[midx+1:])) # h[:-midx-1:-1])
+
+        # make sure the center tap is 0.5
+        assert math.isclose(h[midx], 0.5, abs_tol=1e-8) 
+
+        # For halfband symmetric, odd coefficients (except the center)
+        # should be zero (really small)
+        hodd = xp.stack((h[1:midx:2], h[-midx+1::2]))
+        assert_array_almost_equal(hodd, xp.zeros_like(hodd))
+
+        # now check the frequency response
+        w, H = freqz(np.asarray(h), 1)
+        w, H = xp.asarray(w), xp.asarray(H)
+        f = w/2/xp.pi
+        Hmag = xp.abs(H)
+
+        # check that the pass band is close to unity
+        idx = xp.logical_and(f > 0, f < a)
+        assert_array_almost_equal(Hmag[idx], xp.ones_like(Hmag[idx]), decimal=3)
+
+        # check that the stop band is close to zero
+        idx = xp.logical_and(f > 0.5 - a, f < 0.5)
+        assert_array_almost_equal(Hmag[idx], xp.zeros_like(Hmag[idx]), decimal=3)
+
+    def test_compare(self, xp):
+        # compare to OCTAVE output
+        taps = firls(9, xp.asarray([0, 0.5, 0.55, 1]),
+                    xp.asarray([1, 1, 0, 0]), weight=xp.asarray([1, 2]))
+        # >> taps = firls(8, [0 0.5 0.55 1], [1 1 0 0], [1, 2]);
+        known_taps = [-6.26930101730182e-04, -1.03354450635036e-01,
+                      -9.81576747564301e-03, 3.17271686090449e-01,
+                      5.11409425599933e-01, 3.17271686090449e-01,
+                      -9.81576747564301e-03, -1.03354450635036e-01,
+                      -6.26930101730182e-04]
+        atol_arg = {'atol': 5e-8} if xp_default_dtype(xp) == xp.float32 else {}
+        known_taps = xp.asarray(known_taps, dtype=xp.float64)
+        xp_assert_close(taps, known_taps, **atol_arg)
+
+        # compare to MATLAB output
+        taps = firls(11, xp.asarray([0, 0.5, 0.5, 1]),
+                     xp.asarray([1, 1, 0, 0]), weight=xp.asarray([1, 2]))
+        # >> taps = firls(10, [0 0.5 0.5 1], [1 1 0 0], [1, 2]);
+        known_taps = [
+            0.058545300496815, -0.014233383714318, -0.104688258464392,
+            0.012403323025279, 0.317930861136062, 0.488047220029700,
+            0.317930861136062, 0.012403323025279, -0.104688258464392,
+            -0.014233383714318, 0.058545300496815]
+        known_taps = xp.asarray(known_taps, dtype=xp.float64)
+        atol_arg = {'atol': 3e-8} if xp_default_dtype(xp) == xp.float32 else {}
+        xp_assert_close(taps, known_taps, **atol_arg)
+
+        # With linear changes:
+        taps = firls(7, xp.asarray((0, 1, 2, 3, 4, 5)),
+                     xp.asarray([1, 0, 0, 1, 1, 0]), fs=20)
+        # >> taps = firls(6, [0, 0.1, 0.2, 0.3, 0.4, 0.5], [1, 0, 0, 1, 1, 0])
+        known_taps = [
+            1.156090832768218, -4.1385894727395849, 7.5288619164321826,
+            -8.5530572592947856, 7.5288619164321826, -4.1385894727395849,
+            1.156090832768218]
+        known_taps = xp.asarray(known_taps, dtype=xp.float64)
+        xp_assert_close(taps, known_taps)
+
+    def test_rank_deficient(self, xp):
+        # solve() runs but warns (only sometimes, so here we don't use match)
+        x = firls(21, xp.asarray([0, 0.1, 0.9, 1]), xp.asarray([1, 1, 0, 0]))
+        w, h = freqz(np.asarray(x), fs=2.)
+        w, h = map(xp.asarray, (w, h))   # XXX convert freqz
+        absh2 = xp.abs(h[:2])
+        xp_assert_close(absh2, xp.ones_like(absh2), atol=1e-5)
+        absh2 = xp.abs(h[-2:])
+        xp_assert_close(absh2, xp.zeros_like(absh2), atol=1e-6, rtol=1e-7)
+        # switch to pinvh (tolerances could be higher with longer
+        # filters, but using shorter ones is faster computationally and
+        # the idea is the same)
+        x = firls(101, xp.asarray([0, 0.01, 0.99, 1]), xp.asarray([1, 1, 0, 0]))
+        w, h = freqz(np.asarray(x), fs=2.)
+        w, h = map(xp.asarray, (w, h))   # XXX convert freqz
+        mask = xp.asarray(w < 0.01)
+        h = xp.asarray(h)
+        assert xp.sum(xp.astype(mask, xp.int64)) > 3
+        habs = xp.abs(h[mask])
+        xp_assert_close(habs, xp.ones_like(habs), atol=1e-4)
+        mask = xp.asarray(w > 0.99)
+        assert xp.sum(xp.astype(mask, xp.int64)) > 3
+        habs = xp.abs(h[mask])
+        xp_assert_close(habs, xp.zeros_like(habs), atol=1e-4)
+
+    def test_fs_validation(self):
+        with pytest.raises(ValueError, match="Sampling.*single scalar"):
+            firls(11, .1, 1, fs=np.array([10, 20]))
+
+class TestMinimumPhase:
+
+    @pytest.mark.thread_unsafe
+    def test_bad_args(self):
+        # not enough taps
+        assert_raises(ValueError, minimum_phase, [1.])
+        assert_raises(ValueError, minimum_phase, [1., 1.])
+        assert_raises(ValueError, minimum_phase, np.full(10, 1j))
+        assert_raises((ValueError, TypeError), minimum_phase, 'foo')
+        assert_raises(ValueError, minimum_phase, np.ones(10), n_fft=8)
+        assert_raises(ValueError, minimum_phase, np.ones(10), method='foo')
+        assert_warns(RuntimeWarning, minimum_phase, np.arange(3))
+        with pytest.raises(ValueError, match="is only supported when"):
+            minimum_phase(np.ones(3), method='hilbert', half=False)
+
+    def test_homomorphic(self):
+        # check that it can recover frequency responses of arbitrary
+        # linear-phase filters
+
+        # for some cases we can get the actual filter back
+        h = [1, -1]
+        h_new = minimum_phase(np.convolve(h, h[::-1]))
+        xp_assert_close(h_new, np.asarray(h, dtype=np.float64), rtol=0.05)
+
+        # but in general we only guarantee we get the magnitude back
+        rng = np.random.RandomState(0)
+        for n in (2, 3, 10, 11, 15, 16, 17, 20, 21, 100, 101):
+            h = rng.randn(n)
+            h_linear = np.convolve(h, h[::-1])
+            h_new = minimum_phase(h_linear)
+            xp_assert_close(np.abs(fft(h_new)), np.abs(fft(h)), rtol=1e-4)
+            h_new = minimum_phase(h_linear, half=False)
+            assert len(h_linear) == len(h_new)
+            xp_assert_close(np.abs(fft(h_new)), np.abs(fft(h_linear)), rtol=1e-4)
+
+    @skip_xp_backends("dask.array", reason="too slow")
+    @skip_xp_backends("jax.numpy", reason="immutable arrays")
+    def test_hilbert(self, xp):
+        # compare to MATLAB output of reference implementation
+
+        # f=[0 0.3 0.5 1];
+        # a=[1 1 0 0];
+        # h=remez(11,f,a);
+        h = remez(12, [0, 0.3, 0.5, 1], [1, 0], fs=2.)
+        k = [0.349585548646686, 0.373552164395447, 0.326082685363438,
+             0.077152207480935, -0.129943946349364, -0.059355880509749]
+        h = xp.asarray(h)
+        k = xp.asarray(k, dtype=xp.float64)
+        m = minimum_phase(h, 'hilbert')
+        xp_assert_close(m, k, rtol=5e-3)
+
+        # f=[0 0.8 0.9 1];
+        # a=[0 0 1 1];
+        # h=remez(20,f,a);
+        h = remez(21, [0, 0.8, 0.9, 1], [0, 1], fs=2.)
+        k = [0.232486803906329, -0.133551833687071, 0.151871456867244,
+             -0.157957283165866, 0.151739294892963, -0.129293146705090,
+             0.100787844523204, -0.065832656741252, 0.035361328741024,
+             -0.014977068692269, -0.158416139047557]
+        h = xp.asarray(h)
+        k = xp.asarray(k, dtype=xp.float64)
+        m = minimum_phase(h, 'hilbert', n_fft=2**19)
+        xp_assert_close(m, k, rtol=2e-3)
+
+
+class Testfirwin_2d:
+    def test_invalid_args(self):
+        with pytest.raises(ValueError,
+                           match="hsize must be a 2-element tuple or list"):
+            firwin_2d((50,), window=(("kaiser", 5.0), "boxcar"), fc=0.4)
+
+        with pytest.raises(ValueError,
+                           match="window must be a 2-element tuple or list"):
+            firwin_2d((51, 51), window=("hamming",), fc=0.5)
+
+        with pytest.raises(ValueError,
+                           match="window must be a 2-element tuple or list"):
+            firwin_2d((51, 51), window="invalid_window", fc=0.5)
+
+    def test_filter_design(self):
+        hsize = (51, 51)
+        window = (("kaiser", 8.0), ("kaiser", 8.0))
+        fc = 0.4
+        taps_kaiser = firwin_2d(hsize, window, fc=fc)
+        assert taps_kaiser.shape == (51, 51)
+
+        window = ("hamming", "hamming")
+        taps_hamming = firwin_2d(hsize, window, fc=fc)
+        assert taps_hamming.shape == (51, 51)
+
+    def test_impulse_response(self):
+        hsize = (31, 31)
+        window = ("hamming", "hamming")
+        fc = 0.4
+        taps = firwin_2d(hsize, window, fc=fc)
+
+        impulse = np.zeros((63, 63))
+        impulse[31, 31] = 1
+
+        response = convolve2d(impulse, taps, mode='same')
+
+        expected_response = taps
+        xp_assert_close(response[16:47, 16:47], expected_response, rtol=1e-5)
+
+    def test_frequency_response(self):
+        """Compare 1d and 2d frequency response. """
+        hsize = (31, 31)
+        windows = ("hamming", "hamming")
+        fc = 0.4
+        taps_1d = firwin(numtaps=hsize[0], cutoff=fc, window=windows[0])
+        taps_2d = firwin_2d(hsize, windows, fc=fc)
+
+        f_resp_1d = fft(taps_1d)
+        f_resp_2d = fft2(taps_2d)
+
+        xp_assert_close(f_resp_2d[0, :], f_resp_1d,
+                        err_msg='DC Gain at (0, f1) is not unity!')
+        xp_assert_close(f_resp_2d[:, 0], f_resp_1d,
+                        err_msg='DC Gain at (f0, 0) is not unity!')
+        xp_assert_close(f_resp_2d, np.outer(f_resp_1d, f_resp_1d),
+                        atol=np.finfo(f_resp_2d.dtype).resolution,
+                        err_msg='2d frequency response is not product of 1d responses')
+
+    def test_symmetry(self):
+        hsize = (51, 51)
+        window = ("hamming", "hamming")
+        fc = 0.4
+        taps = firwin_2d(hsize, window, fc=fc)
+        xp_assert_close(taps, np.flip(taps), rtol=1e-5)
+
+    def test_circular_symmetry(self):
+        hsize = (51, 51)
+        window = "hamming"
+        taps = firwin_2d(hsize, window, circular=True, fc=0.5)
+        center = hsize[0] // 2
+        for i in range(hsize[0]):
+            for j in range(hsize[1]):
+                xp_assert_close(taps[i, j], 
+                                taps[center - (i - center), center - (j - center)], 
+                                rtol=1e-5)
+
+    def test_edge_case_circular(self):
+        hsize = (3, 3)
+        window = "hamming"
+        taps_small = firwin_2d(hsize, window, circular=True, fc=0.5)
+        assert taps_small.shape == (3, 3)
+
+        hsize = (101, 101)
+        taps_large = firwin_2d(hsize, window, circular=True, fc=0.5)
+        assert taps_large.shape == (101, 101)
+
+    def test_known_result(self):
+        hsize = (5, 5)
+        window = ('kaiser', 8.0)
+        fc = 0.1
+        fs = 2
+
+        row_filter = firwin(hsize[0], cutoff=fc, window=window, fs=fs)
+        col_filter = firwin(hsize[1], cutoff=fc, window=window, fs=fs)
+        known_result = np.outer(row_filter, col_filter)
+
+        taps = firwin_2d(hsize, (window, window), fc=fc)
+        assert taps.shape == known_result.shape, (
+            f"Shape mismatch: {taps.shape} vs {known_result.shape}"
+        )
+        assert np.allclose(taps, known_result, rtol=1e-1), (
+            f"Filter shape mismatch: {taps} vs {known_result}"
+        )