def test_fftfilt_nd():
np.random.seed(123)
x = np.random.randn(2**16) + 1j * np.random.randn(2**16)
b = np.random.randn(128)
oned = fftfilt.fftfilt(b, x)
nd = fftfilt.fftfilt_nd(b, x[:, None]).squeeze()
assert np.allclose(oned, nd)
x = x.reshape((2**12, 2**4))
oned = np.zeros_like(x)
for k in range(16):
oned[:, k] = fftfilt.fftfilt(b, x[:, k])
nd = fftfilt.fftfilt_nd(b, x)
assert np.allclose(oned, nd)
def test_fftfilt_nd():
np.random.seed(123)
x = np.random.randn(2 ** 16) + 1j * np.random.randn(2 ** 16)
b = np.random.randn(128)
oned = fftfilt.fftfilt(b, x)
nd = fftfilt.fftfilt_nd(b, x[:, None]).squeeze()
assert np.allclose(oned, nd)
x = x.reshape((2 ** 12, 2 ** 4))
oned = np.zeros_like(x)
for k in range(16):
oned[:, k] = fftfilt.fftfilt(b, x[:, k])
nd = fftfilt.fftfilt_nd(b, x)
assert np.allclose(oned, nd)
def test_decimating_fir():
np.random.seed(123)
x = np.random.randn(2**16) + 1j * np.random.randn(2**16)
dfir = decimating_fir.DecimatingFIR(downsample_factor=16, num_taps=1024)
gold = fftfilt.fftfilt(dfir.coefficients.ravel(), x)[15::16]
result = dfir.process(x)
assert np.allclose(gold, result)
def test_decimating_fir():
np.random.seed(123)
x = np.random.randn(2**16) + 1j*np.random.randn(2**16)
dfir = decimating_fir.DecimatingFIR(downsample_factor=16,num_taps=1024)
gold = fftfilt.fftfilt(dfir.coefficients.ravel(),x)[15::16]
result = dfir.process(x)
assert np.allclose(gold,result)
def mean(self):
if self._mean is None:
if self._lpf_data is None:
self._lpf_data = fftfilt.fftfilt(
lpf, self.data)[len(lpf):] * self.wavenorm
self._mean = self._lpf_data.mean(0, dtype='complex')
return self._mean
def low_pass_fir(data,
num_taps=256,
cutoff=1 / 256.,
nyquist_freq=1.0,
decimate_by=1):
taps = scipy.signal.firwin(num_taps, cutoff / nyquist_freq)
result = fftfilt(taps, data)[num_taps:]
result = result[::decimate_by].copy(
) # add .copy to ensure we separate this from the full sized original data
return result
def mean_error(self):
if self._std is None:
if self._lpf_data is None:
self._lpf_data = fftfilt.fftfilt(lpf,self.data)[len(lpf):]*self.wavenorm
# the standard deviation is scaled by the number of independent samples
# to compute the error on the mean.
real_error = self._lpf_data.real.std(0)/np.sqrt(self._lpf_data.shape[0]/len(lpf))
imag_error = self._lpf_data.imag.std(0)/np.sqrt(self._lpf_data.shape[0]/len(lpf))
self._std = real_error + 1j*imag_error
return self._std
def deglitch_block(ts,thresh=5):
tsl = np.roll(np.abs(fftfilt(scipy.signal.firwin(16,1/16.),ts)),-8)
mask = medmadmask(tsl,thresh=thresh)
mask[:-50] = mask[:-50] | mask[50:]
mask[50:] = mask[50:] | mask[:-50]
nmask = mask.sum()
# print "rejecting",nmask/float(ts.shape[0])
out = ts.copy()
try:
out[mask] = np.array(random.sample(ts[~mask],nmask))
except ValueError:
print "more masked values than samples to draw from!"
return out
def mean_error(self):
if self._std is None:
if self._lpf_data is None:
self._lpf_data = fftfilt.fftfilt(
lpf, self.data)[len(lpf):] * self.wavenorm
# the standard deviation is scaled by the number of independent samples
# to compute the error on the mean.
real_error = self._lpf_data.real.std(0) / np.sqrt(
self._lpf_data.shape[0] / len(lpf))
imag_error = self._lpf_data.imag.std(0) / np.sqrt(
self._lpf_data.shape[0] / len(lpf))
self._std = real_error + 1j * imag_error
return self._std
def deglitch_block(ts, thresh=5):
tsl = np.roll(np.abs(fftfilt(scipy.signal.firwin(16, 1 / 16.), ts)), -8)
mask = medmadmask(tsl, thresh=thresh)
mask[:-50] = mask[:-50] | mask[50:]
mask[50:] = mask[50:] | mask[:-50]
nmask = mask.sum()
# print "rejecting",nmask/float(ts.shape[0])
out = ts.copy()
try:
out[mask] = np.array(random.sample(ts[~mask], nmask))
except ValueError:
print "more masked values than samples to draw from!"
return out
def low_pass_fir(data, num_taps=256, cutoff=1/256.,nyquist_freq=1.0,decimate_by=1):
taps = scipy.signal.firwin(num_taps,cutoff/nyquist_freq)
result = fftfilt(taps,data)[num_taps:]
result = result[::decimate_by].copy() # add .copy to ensure we separate this from the full sized original data
return result
def lpf256(ts):
return fftfilt(scipy.signal.firwin(256,1/256.),ts)
def lpf256(ts):
return fftfilt(scipy.signal.firwin(256, 1 / 256.), ts)
def mean(self):
if self._mean is None:
if self._lpf_data is None:
self._lpf_data = fftfilt.fftfilt(lpf,self.data)[len(lpf):]*self.wavenorm
self._mean = self._lpf_data.mean(0,dtype='complex')
return self._mean
