def decodefft(finf,data, dropheights = False):
#output: decoded data with the number of heights reduced
#two variables are added to the finfo class:
#deco_num_hei, deco_hrange
#data must be arranged:
# (channels,heights,times) (C-style, profs change faster)
#fft along the entire(n=None) acquired heights(axis=1), stores in data
num_chan = data.shape[0]
num_ipps = data.shape[2]
num_codes = finf.subcode.shape[0]
num_bauds = finf.subcode.shape[1]
NSA = finf.num_hei + num_bauds - 1
uppower = py.ceil(py.log2(NSA))
extra = int(2**uppower - finf.num_hei)
NSA = int(2**uppower)
fft_code = py.fft(finf.subcode,n = NSA,axis=1).conj()
data = py.fft(data,n=NSA,axis=1) #n= None: no cropped data or padded zeros
for ch in range(num_chan):
for ipp in range(num_ipps):
code_i = ipp % num_codes
data[ch,:,ipp] = data[ch,:,ipp] * fft_code[code_i,:]
data=py.ifft(data,n=NSA,axis=1) #fft along the heightsm
if dropheights:
return data[:,:-extra-(num_bauds-1),:]
else:
return data[:,:-extra,:]
def decodefft(finf, data, dropheights=False):
#output: decoded data with the number of heights reduced
#two variables are added to the finfo class:
#deco_num_hei, deco_hrange
#data must be arranged:
# (channels,heights,times) (C-style, profs change faster)
#fft along the entire(n=None) acquired heights(axis=1), stores in data
num_chan = data.shape[0]
num_ipps = data.shape[2]
num_codes = finf.subcode.shape[0]
num_bauds = finf.subcode.shape[1]
NSA = finf.num_hei + num_bauds - 1
uppower = py.ceil(py.log2(NSA))
extra = int(2**uppower - finf.num_hei)
NSA = int(2**uppower)
fft_code = py.fft(finf.subcode, n=NSA, axis=1).conj()
data = py.fft(data, n=NSA,
axis=1) #n= None: no cropped data or padded zeros
for ch in range(num_chan):
for ipp in range(num_ipps):
code_i = ipp % num_codes
data[ch, :, ipp] = data[ch, :, ipp] * fft_code[code_i, :]
data = py.ifft(data, n=NSA, axis=1) #fft along the heightsm
if dropheights:
return data[:, :-extra - (num_bauds - 1), :]
else:
return data[:, :-extra, :]
def f( t, *args ):
for i,arg in enumerate(args): params[ free_params[i] ] = arg
tshift = params[-1]
ideal = fmodel( t, *args )
irf = cspline1d_eval( self.irf_generator, t-tshift, dx=self.irf_dt, x0=self.irf_t0 )
convoluted = pylab.real(pylab.ifft( pylab.fft(ideal)*pylab.fft(irf) )) # very small imaginary anyway
return convoluted
def main():
pylab.ion();
ind = [0,];
ldft = [0,];
lfft = [0,];
lpfft = [0,]
# plot a graph Dft vs Fft, lists just support size until 2**9
for i in range(1, 9, 1):
t_before = time.clock();
dsprocessing.dspDft(rand(2**i).tolist());
dt = time.clock() - t_before;
ldft.append(dt);
print ("dft ", 2**i, dt);
#pylab.plot([2**i,], [time.clock()-t_before,]);
t_before = time.clock();
dsprocessing.dspFft(rand(2**i).tolist());
dt = time.clock() - t_before;
print ("fft ", 2**i, dt);
lfft.append(dt);
#pylab.plot([2**i,], [time.clock()-t_before,]);
ind.append(2**i);
# python fft just to compare
t_before = time.clock();
pylab.fft(rand(2**i).tolist());
dt = time.clock() - t_before;
lpfft.append(dt);
pylab.plot(ind, ldft);
pylab.plot(ind, lfft);
pylab.plot(ind, lpfft);
pylab.show();
return [ind, ldft, lfft, lpfft];
def fast_fracdiff(x, d):
T = len(x)
np2 = int(2**np.ceil(np.log2(2 * T - 1)))
k = np.arange(1, T)
b = (1, ) + tuple(np.cumprod((k - d - 1) / k))
z = (0, ) * (np2 - T)
z1 = b + z
z2 = tuple(x) + z
dx = pl.ifft(pl.fft(z1) * pl.fft(z2))
return np.real(dx[0:T])
def _ConvFft(signal, FilterKernel):
"""
Convolution with fft much faster approach
works exactly as convolve(x,y)
"""
ss = numpy.size(signal);
fs = numpy.size(FilterKernel)
# padd zeros all until they have the size N+M-1
signal = numpy.append(signal, numpy.zeros(fs+ss-1-ss));
FilterKernel = numpy.append(FilterKernel, numpy.zeros(fs+ss-1-fs));
signal = pylab.real(pylab.ifft(pylab.fft(signal)*pylab.fft(FilterKernel)));
return signal[:fs+ss-1];
def fast_fracdiff(x, cols, d):
for col in cols:
T = len(x[col])
np2 = int(2**np.ceil(np.log2(2 * T - 1)))
k = np.arange(1, T)
b = (1, ) + tuple(np.cumprod((k - d - 1) / k))
z = (0, ) * (np2 - T)
z1 = b + z
z2 = tuple(x[col]) + z
dx = pl.ifft(pl.fft(z1) * pl.fft(z2))
x[col + "_frac"] = np.real(dx[0:T])
return x
def _ConvFft(signal, filterkernel):
"""
Convolution with fft much faster approach
works exatcly as convolve(x,y)
"""
ss = numpy.size(signal)
fs = numpy.size(filterkernel)
# padd zeros all until they have the size N+M-1
signal = numpy.append(signal, numpy.zeros(fs + ss - 1 - ss))
filterkernel = numpy.append(filterkernel, numpy.zeros(fs + ss - 1 - fs))
signal = pylab.real(pylab.ifft(
pylab.fft(signal) * pylab.fft(filterkernel)))
return signal[:fs + ss - 1]
def calcFourier(self, yout, tspan):
'''
Calculate FFT and return magnitude spectrum and frequencies
freqs === Python array
aMags === numpy array
rMags === numpy array
'''
ActEC = yout[:, 3]
RepEC = yout[:, 9]
freqs = fftfreq(len(tspan), tspan[1] - tspan[0])
aMags = array(abs((fft(ActEC))))
rMags = array(abs((fft(RepEC))))
return freqs, aMags, rMags
def compute_response(self, **kargs):
"""Compute the window data frequency response
:param norm: True by default. normalised the frequency data.
:param int NFFT: total length of the final data sets( 2048 by default. if less than data length, then
NFFT is set to the data length*2).
The response is stored in :attr:`response`.
.. note:: Units are dB (20 log10) since we plot the frequency response)
"""
from pylab import fft, fftshift, log10
norm = kargs.get('norm', self.norm)
# do some padding. Default is max(2048, data.len*2)
NFFT = kargs.get('NFFT', 2048)
if NFFT < len(self.data):
NFFT = self.data.size * 2
# compute the fft modulus
A = fft(self.data, NFFT)
mag = abs(fftshift(A))
# do we want to normalise the data
if norm is True:
mag = mag / max(mag)
response = 20. * log10(mag) # factor 20 we are looking at the response
# not the powe
#response = clip(response,mindB,100)
self.__response = response
def PlotFftNAbs(y,
dt=0.1
): # plot the grah of fft (amplitude) of y with command plot
"""
Normalized version plot of FFT
equal abs(FFT) * 2/y.length
plot also just half spectrum (just positive frequencies),
and from frequencys 1:N/2
the mean (DC component) is removed before fft
"""
N = numpy.size(y)
#remove the DC component just the mean (could use a detrend function)
y = y - numpy.mean(y)
fund = 1 / (N * dt) # freq fundamental, as outras sao multiplas dessa
f = range(N)
for k in range(N - 1):
f[k + 1] = fund * k #multiplas da fundamental
yn = pylab.fft(y)
# perfeito. se vpce colocar um seno de 30 e outro de 60, com aplitude 1 e 2 no espectro
# de amplitude voce vai encontrar exatamente 1 e 2
absyn = 2 * numpy.abs(yn) / numpy.size(yn)
pylab.plot(f[0:numpy.size(f) / 2], absyn[0:numpy.size(f) / 2])
return [f, absyn]
def PlotPowersSpectrumdB(y, dt=0.1):
"""
Try of power spectral density estimation
not windowed no Welch yet.
plot also just half spectrum (just positive frequencies),
and from frequencys 0:N/2
the mean (DC component) is removed before fft
"""
# plot the grah of fft (amplitude) of y with command plot
N = numpy.size(y)
#remove the DC component just the mean (could use a detrend function)
y = y - numpy.mean(y)
fund = 1 / (N * dt) # freq fundamental, as outras sao multiplas dessa
f = range(N)
for k in range(N - 1):
f[k + 1] = fund * k #multiplas da fundamental
yn = pylab.fft(y)
power = yn * numpy.conj(yn) / (numpy.sqrt(numpy.size(yn)))
power = numpy.abs(power)
# create decibel scale Lb = log10(P1/P0) where P0 is the maximum value
P0 = max(power)
power = numpy.log10(power / P0)
#ignore Dc and the negative frequencies
pylab.plot(f[0:numpy.size(f) / 2], power[0:numpy.size(f) / 2])
def performFourierDataReturn(self, wav_name, input_location=None):
if input_location == None:
input_location = ''
if wav_name.endswith('.wav'):
wav_file_name = os.path.join(
os.path.join(__location__, input_location), wav_name)
else:
wav_file_name = os.path.join(
os.path.join(__location__, input_location), wav_name) + '.wav'
sampFreq, snd = wavfile.read(wav_file_name)
# convert 16 bit values to be between -1 to 1
snd /= 2.0**15
# get average of data if more than 1 channels
if self.channels > 1:
s1 = self.getAverageData(snd)
else:
s1 = snd.T[0]
n = len(s1)
p = fft(s1)
nUniquePts = int(ceil((n + 1) / 2.0))
p = p[0:nUniquePts]
p = abs(p)
p /= float(n)
p **= 2
if n % 2 > 0:
p[1:len(p)] = p[1:len(p)] * 2
else:
p[1:len(p) - 1] = p[1:len(p) - 1] * 2
return sampFreq, nUniquePts, n, p
def hacerfft(channel):
tamanho = len(channel)
fdata = fft(channel)
print('> FFT realizada')
nUniquePts = int(ceil((tamanho)/2))
fdata = fdata[0:nUniquePts]
fdata = abs(fdata)
fdata = fdata/float(leng)
fdata = fdata**2
if leng % 2 > 0:
fdata[1:int(ceil(len(fdata)))] = fdata[1:int(ceil(len(fdata)))] * 2
else:
fdata[1:int(ceil(len(fdata)))-1] = fdata[1:int(ceil(len(fdata)))-1] * 2
freqArray = arange(0, nUniquePts, 1.0)*(fs/tamanho)
plot(freqArray/1000, 10*log10(fdata[0:tamanho:1]))
xlabel('Frequency (kHz)')
ylabel('Power (dB)')
plt.show()
print('> FFT graficada')
return fdata
def time_fft(data2, samplerate=100., inverse=False,hann=False):
'''
IF N_PARAMS() EQ 0 then begin
print, 'time_fft, data, samplerate=samplerate, inverse=inverse,hann=hann'
return, -1
ENDIF
'''
data=data2
if hann:
w1 = pl.hanning(len(data))
data = data*w1
#frequency axis:
freqs = pl.arange(1+len(data)/2)/float(len(data)/2.0)*samplerate/2. # wut.
if len(data) % 2 == 0 : freqs = pl.concatenate((freqs, -freqs[1:(len(freqs)-1)][::-1]))
if len(data) % 2 != 0 : freqs = pl.concatenate((freqs, -freqs[1:len(freqs)][::-1]))
response = pl.fft(data)
if inverse : response = pl.ifft(data)
out = {'freq': freqs, 'real': response.real, 'im': response.imag, 'abs': abs(response)}
return out
def plot_signal(s,sr,start=0,end=440,N=32768):
'''plots the waveform and spectrum of a signal s
with sampling rate sr, from a start to an
end position (waveform), and from a start
position with a N-point DFT (spectrum)'''
pl.figure(figsize=(8,5))
pl.subplot(211)
sig = s[start:end]
time = pl.arange(0,len(sig))/sr
pl.plot(time,sig, 'k-')
pl.ylim(-1.1,1.1)
pl.xlabel("time (s)")
pl.subplot(212)
N = 32768
f = pl.arange(0,N/2)
bins = f*sr/N
win = pl.hanning(N)
scal = N*pl.sqrt(pl.mean(win**2))
sig = s[start:start+N]
window = pl.fft(sig*win/max(sig))
mags = abs(window/scal)
spec = 20*pl.log10(mags/max(mags))
pl.plot(bins,spec[0:N//2], 'k-')
pl.ylim(-60, 1)
pl.ylabel("amp (dB)", size=16)
pl.xlabel("freq (Hz)", size=16)
pl.yticks()
pl.xticks()
pl.xlim(0,sr/2)
pl.tight_layout()
pl.show()
def hacerfft(channel, texto=None):
fdata = fft(channel)
print('\t> FFT realizada')
nUniquePts = int(ceil((leng) / 2))
fdata = fdata[0:nUniquePts]
fdata = abs(fdata)
fdata = fdata / float(leng)
fdata = fdata**2
if leng % 2 > 0:
fdata[1:int(ceil(len(fdata)))] = fdata[1:int(ceil(len(fdata)))] * 2
else:
fdata[1:int(ceil(len(fdata))) -
1] = fdata[1:int(ceil(len(fdata))) - 1] * 2
freqArray = arange(0, nUniquePts, 1.0) * (fs / leng)
plot(freqArray / 1000, 10 * log10(fdata[0:leng:1]))
if texto == None:
xlabel('Frequency (kHz)')
else:
xlabel('Frequency (kHz) | ' + str(texto))
ylabel('Power (dB)')
plt.show()
print('\t> FFT graficada')
return fdata, freqArray
def hacerfft(channel, freq): # en Hz
fdata = fft(channel)
print('> FFT realizada: ' + str(freq))
nUniquePts = int(ceil((N) / 2))
fdata = fdata[0:nUniquePts]
fdata = abs(fdata)
fdata = fdata / float(N)
fdata = fdata**2
if N % 2 > 0:
fdata[1:int(ceil(len(fdata)))] = fdata[1:int(ceil(len(fdata)))] * 2
else:
fdata[1:int(ceil(len(fdata))) -
1] = fdata[1:int(ceil(len(fdata))) - 1] * 2
freqArray = arange(0, nUniquePts, 1.0) * (fs / N)
plot(freqArray, 10 * log10(fdata[0:int(N):1]))
xlabel('Frequency (Hz)')
ylabel('Power (dB)')
plt.show()
print('> FFT graficada ' + str(freq))
return fdata
def compute_response(self, **kargs):
"""Compute the window data frequency response
:param norm: True by default. normalised the frequency data.
:param int NFFT: total length of the final data sets( 2048 by default. if less than data length, then
NFFT is set to the data length*2).
The response is stored in :attr:`response`.
.. note:: Units are dB (20 log10) since we plot the frequency response)
"""
from pylab import fft, fftshift, log10
norm = kargs.get('norm', self.norm)
# do some padding. Default is max(2048, data.len*2)
NFFT = kargs.get('NFFT', 2048)
if NFFT < len(self.data):
NFFT = self.data.size * 2
# compute the fft modulus
A = fft(self.data, NFFT)
mag = abs(fftshift(A))
# do we want to normalise the data
if norm is True:
mag = mag / max(mag)
response = 20. * log10(mag) # factor 20 we are looking at the response
# not the powe
#response = clip(response,mindB,100)
self.__response = response
def __init__(self, path, name=None):
if name is None:
path, name = os.path.split(path)
self.name = name
self.path = path
self.nowPlaying = False
# Read wav file
self.sampFreq, self.tDomain = wavfile.read(path + '/' + name)
# Get frequency domain
fDomain = pylab.fft(self.tDomain)
nPoints = len(self.tDomain)
nUniquePoints = int(math.ceil((nPoints+1)/2.0))
fDomain = abs(fDomain[0:nUniquePoints])
# Normalize frequency domain
fDomain = fDomain/float(nPoints)
fDomain = fDomain ** 2
if nPoints % 2 > 0:
fDomain[1:len(fDomain)] = fDomain[1:len(fDomain)] * 2
else:
fDomain[1:len(fDomain)-1] = fDomain[1:len(fDomain)-1] * 2
# Get frequency values to go with magnitudes
time = nPoints / float(self.sampFreq)
frequencies = np.arange(nUniquePoints) / time
# Store frequency domain as list of (frequency, magnitude)
self.fDomain = zip(frequencies, fDomain)
# Store sizes of arrays for later
self.nPoints = nPoints
self.nUniquePoints = nUniquePoints
def STA(pks, I, V, sampr, delay):
currents = []
voltages = []
out = {}
for i in range(len(pks)):
if i == 0:
if pks[i] > sampr * (delay + 1):
currents.append(I[int(pks[i] - sampr):int(pks[i])])
voltages.append(V[int(pks[i] - sampr):int(pks[i])])
else:
if (pks[i] - pks[i - 1]) > sampr:
currents.append(I[int(pks[i] - sampr):int(pks[i])])
voltages.append(V[int(pks[i] - sampr):int(pks[i])])
if len(currents):
currents = np.array(currents)
avgI = np.mean(currents, axis=0)
f_current = (fft(avgI) / len(avgI))[0:int(len(avgI) / 2)]
Freq = np.linspace(0.0, sampr / 2.0, len(f_current))
out = {
'currents': currents,
'voltages': voltages,
'f_current': f_current,
'Freq': Freq
}
return out
def PlotPowersSpectrumdB(y, dt=0.1):
"""
Try of power spectral density estimation
not windowed no Welch yet.
plot also just half spectrum (just positive frequencies),
and from frequencys 0:N/2
the mean (DC component) is removed before fft
"""
# plot the grah of fft (amplitude) of y with command plot
N = numpy.size(y)
#remove the DC component just the mean (could use a detrend function)
y = y - numpy.mean(y)
fund = 1/(N*dt) # freq fundamental, as outras sao multiplas dessa
f=range(N)
for k in range(N-1):
f[k+1]=fund*k #multiplas da fundamental
yn = pylab.fft(y)
power = yn*numpy.conj(yn)/(numpy.sqrt(numpy.size(yn)))
power = numpy.abs(power)
# create decibel scale Lb = log10(P1/P0) where P0 is the maximum value
P0 = max(power)
power = numpy.log10(power/P0);
#ignore Dc and the negative frequencies
pylab.plot(f[0:numpy.size(f)/2], power[0:numpy.size(f)/2])
def fft_smoothing(input_signal, sigma):
"""smooth the fast transform Fourier
Params :
input_signal : audio signal
sigma : relative to the length of the output signal
Returns :
a shorter and smoother signal
"""
size_signal = input_signal.size
#shorten the signal
new_size = int(floor(10.0 * size_signal * sigma))
half_new_size = new_size // 2
fftx = fft(input_signal)
short_fftx = []
for ele in fftx[:half_new_size]:
short_fftx.append(ele)
for ele in fftx[-half_new_size:]:
short_fftx.append(ele)
apodization_coefficients = generate_apodization_coeffs(
half_new_size, sigma, size_signal)
#apply the apodization coefficients
short_fftx[:half_new_size] *= apodization_coefficients
short_fftx[half_new_size:] *= flipud(apodization_coefficients)
realifftxw = ifft(short_fftx).real
return realifftxw
def zonalWN_fourier(signal, lats, lons):
"""
Function to compute dominant zonal wavenumber at a latitude circle at each time step
INPUT:
signal (array like):
Input signal array. Shape must be [time, lev, lat, lon].
lats: array with latitude values.
lons: array with longitude values.
OUTPUT:
maxWN_mrunmean (array like) :
Convolved signal with Hann window of constant width. Shape is [time, lev, lat, lon].
"""
notime = signal.shape[0]
nolev=signal.shape[1]
# loop over all latitudes between lat1 and lat2 (descending in lat)
n = (len(lons)+1)//2
A = np.zeros([notime, nolev, len(lats), len(lons)], dtype='complex') # complex array to hold the fourier transf of the input signal
power = np.zeros([notime, nolev, len(lats),n])
#k=np.arange(1,n,1) # zonal wavenumber
maxWN = np.zeros([notime,nolev, len(lats)])
#loop over time, isentropes and latitudes
for t in range (0,notime):
for l in range (0,nolev):
for ilat in range(0,len(lats)):
A[t,l,ilat,:] = pl.fft(signal[t,l,ilat,:]) # fourier transform of the signal (v) at each latitude
for kk in range(0,n): # exclude wavenumber 0
power[t,l,ilat,kk] = 4*(A[t,l,ilat,kk].real**2 + A[t,l,ilat,kk].imag**2)
maxWN[t,l,ilat] = power[t,l,ilat].argmax(axis=0) # the index at which the maximum in the power spectrum occurs, which is the maximum wavenumber
### Smoothing of the signal ###
# fill the remaining latutues with the boundary values
ext_lat=10 # extension of the additional boundaries in degrees of latitude
res=int(abs(lats[1]-lats[0])) # resolution in lats
N = ext_lat//res # window for the moving average in terms of index in the lats array!!! so N=5 is an extended 10 deg boundary for a 2*2 resolution
# now apply some smoothing to avoid jumps between different integers wavenumers
maxWN_mrunmean=np.zeros([notime, nolev, len(lats)])
for t in range (0,notime):
for l in range (0,nolev):
maxWN_mrunmean[t,l,:] = np.convolve(maxWN[t,l,:], np.ones((N,))/N, mode='same')
maxWN_mrunmean=np.where(maxWN_mrunmean<1,1,maxWN_mrunmean)
return maxWN_mrunmean
def compute_fft(self, start, end):
dur = end - start
fft = pl.fft(self.data[start:end])
real_range = np.ceil((dur+1)/2.0)
fft = fft[0:real_range]
fft = abs(fft)
return fft * np.hanning(len(fft))
def compute_fft(self, start, end):
dur = end - start
fft = pl.fft(self.data[start:end])
real_range = np.ceil((dur+1)/2.0)
fft = fft[0:real_range]
fft = abs(fft)
return fft * np.hanning(len(fft))
def QuasiSpace(self,ns,df,ls):
#evaluates the quasi space value
allqs=[]
#at the moment, the most easy method(everything else failed...)
#get the mean absolute value of the complete spectrum. works best
for i in range(len(ns)):
# xvalues=py.fftfreq(len(ns[i]),df)*c/4
QSr=py.fft(ns[i].real-py.mean(ns[i].real))
QSi=py.fft(ns[i].imag-py.mean(ns[i].real))
#naive cut:
ix=range(3,int(len(QSr)/2-1))
QSr=QSr[ix]
QSi=QSi[ix]
allqs.append(py.mean(abs(QSr))+py.mean(abs(QSi)))
return allqs
def main(nbits):
"""Main Program"""
import pylab
nbits = int(nbits)
m = mls(nbits)
fm = pylab.fft(m)
fig, ax = pylab.subplots(2, sharex=True)
ax[0].plot(20 * pylab.log10(numpy.abs(fm)))
ax[1].plot(numpy.angle(fm))
def windowAndFFT(self,window,fftlength):
if ( self.Length != len(window) ):
raise Exception, 'size mismatch, length of data != that of window'
if ( fftlength < self.Length ):
raise ValueError, 'fftlength < length of data'
z = numpy.zeros(fftlength - self.Length)
Abar = pylab.fft( numpy.array( list(self.A*window) + list(z) ) )
Ebar = pylab.fft( numpy.array( list(self.E*window) + list(z) ) )
Tbar = pylab.fft( numpy.array( list(self.T*window) + list(z) ) )
Abar = Abar[ 0:numpy.floor( fftlength/2+1 ) ]*self.Cadence
Ebar = Ebar[ 0:numpy.floor( fftlength/2+1 ) ]*self.Cadence
Tbar = Tbar[ 0:numpy.floor( fftlength/2+1 ) ]*self.Cadence
self.winfftA = Abar
self.winfftE = Ebar
self.winfftT = Tbar
self.FreqOffset = 0
self.FreqCadence = 1./(self.Cadence*fftlength)
self.FreqLength = len(Abar)
return
def PlotFftPhase(signal, dt):
N=numpy.size(signal)
fund = 1/(N*dt) # freq fundamental, as outras sao multiplas dessa
f=range(N)
for k in range(N-1):
f[k+1]=fund*k #multiplas da fundamental
z=pylab.fft(signal)
pylab.plot(f, numpy.angle(z, deg=True))
pylab.ylabel('Phase (degrees)')
pylab.xlabel('Frequency (Hz)')
def windowAndFFT(self, window, fftlength):
if (self.Length != len(window)):
raise Exception, 'size mismatch, length of data != that of window'
if (fftlength < self.Length):
raise ValueError, 'fftlength < length of data'
z = numpy.zeros(fftlength - self.Length)
Abar = pylab.fft(numpy.array(list(self.A * window) + list(z)))
Ebar = pylab.fft(numpy.array(list(self.E * window) + list(z)))
Tbar = pylab.fft(numpy.array(list(self.T * window) + list(z)))
Abar = Abar[0:numpy.floor(fftlength / 2 + 1)] * self.Cadence
Ebar = Ebar[0:numpy.floor(fftlength / 2 + 1)] * self.Cadence
Tbar = Tbar[0:numpy.floor(fftlength / 2 + 1)] * self.Cadence
self.winfftA = Abar
self.winfftE = Ebar
self.winfftT = Tbar
self.FreqOffset = 0
self.FreqCadence = 1. / (self.Cadence * fftlength)
self.FreqLength = len(Abar)
return
def PlotFftPhase(signal, dt):
N = numpy.size(signal)
fund = 1 / (N * dt) # freq fundamental, as outras sao multiplas dessa
f = range(N)
for k in range(N - 1):
f[k + 1] = fund * k #multiplas da fundamental
z = pylab.fft(signal)
pylab.plot(f, numpy.angle(z, deg=True))
pylab.ylabel('Phase (degrees)')
pylab.xlabel('Frequency (Hz)')
def my_transform(x,dir):
# my_transform - perform either FFT with energy conservation.
# Works on array of size (w,w,a,b) on the 2 first dimensions.
w = np.shape(x)[0]
if dir == 1 :
y = np.transpose(pyl.fft(np.transpose(x)))/np.sqrt(w)
else :
y = np.transpose(pyl.ifft(np.transpose(x)*np.sqrt(w)))
return y
def zMeasures(current, v, delay, sampr, f1, bwinsz=1):
## zero padding
current = current[int(delay * sampr - 0.5 * sampr +
1):-int(delay * sampr - 0.5 * sampr)]
current = np.hstack((np.repeat(current[0], int(delay * sampr)), current,
np.repeat(current[-1], int(delay * sampr))))
current = current - np.mean(current)
v = v[int(delay * sampr - 0.5 * sampr) +
1:-int(delay * sampr - 0.5 * sampr)]
v = np.hstack((np.repeat(v[0], int(delay * sampr)), v,
np.repeat(v[-1], int(delay * sampr))))
v = v - np.mean(v)
## input and transfer impedance
f_current = (fft(current) / len(current))[0:int(len(current) / 2)]
f_cis = (fft(v) / len(v))[0:int(len(v) / 2)]
z = f_cis / f_current
## impedance measures
Freq = np.linspace(0.0, sampr / 2.0, len(z))
zAmp = abs(z)
zPhase = np.arctan2(np.imag(z), np.real(z))
zRes = np.real(z)
zReact = np.imag(z)
## smoothing
fblur = np.array([1.0 / bwinsz for i in range(bwinsz)])
zAmp = convolve(zAmp, fblur, 'same')
zPhase = convolve(zPhase, fblur, 'same')
## trim
mask = (Freq >= 0.5) & (Freq <= f1)
Freq, zAmp, zPhase, zRes, zReact = Freq[mask], zAmp[mask], zPhase[
mask], zRes[mask], zReact[mask]
## resonance
zResAmp = np.max(zAmp)
zResFreq = Freq[np.argmax(zAmp)]
Qfactor = zResAmp / zAmp[0]
fVar = np.std(zAmp) / np.mean(zAmp)
return Freq, zAmp, zPhase, zRes, zReact, zResAmp, zResFreq, Qfactor, fVar
def PlotFftAbs(signal, dt):
N = numpy.size(signal)
fund = 1 / (N * dt) # freq fundamental, as outras sao multiplas dessa
f = range(N)
for k in range(N - 1):
f[k + 1] = fund * k #multiplas da fundamental
z = pylab.fft(signal)
z_real = abs(z)
pylab.plot(f, z_real)
pylab.ylabel('Amplitude')
pylab.xlabel('Frequency (Hz)')
def fourierPower(signal):
"Calculates the Fourier power spectrum of the signal"
A = pl.fft(signal)
n = (len(signal) + 1) / 2
power = pl.zeros(n) #+1)
#power[0] = A[0].real**2
#for i in range(1,n+1):
for i in range(n):
power[i] = 4 * (A[i + 1].real**2 + A[i + 1].imag**2)
return power
def FourierDerivative(f):
"""
this derivatie just works for periodic 2*pi multiple series
have to figure out how to make that work for any function
"""
N = np.size(f)
n = np.arange(0, N)
# df discrete differential operator
df = np.complex(0, 1) * py.fftshift(n - N / 2)
dfdt = py.ifft(df * py.fft(f))
return py.real(dfdt)
def PlotFftAbs(signal, dt):
N=numpy.size(signal)
fund = 1/(N*dt) # freq fundamental, as outras sao multiplas dessa
f=range(N)
for k in range(N-1):
f[k+1]=fund*k #multiplas da fundamental
z=pylab.fft(signal)
z_real=abs(z)
pylab.plot(f, z_real)
pylab.ylabel('Amplitude')
pylab.xlabel('Frequency (Hz)')
def FourierDerivative(f):
"""
this derivatie just works for periodic 2*pi multiple series
have to figure out how to make that work for any function
"""
N = np.size(f)
n = np.arange(0,N)
# df discrete differential operator
df = np.complex(0,1)*py.fftshift(n-N/2)
dfdt = py.ifft( df*py.fft(f) )
return py.real(dfdt)
def getDR(self):
#this function should return the dynamic range
#this should be the noiselevel of the fft
noiselevel=py.sqrt(py.mean(abs(py.fft(self._tdData.getAllPrecNoise()[0]))**2))
#apply a moving average filter on log
window_size=5
window=py.ones(int(window_size))/float(window_size)
hlog=py.convolve(20*py.log10(self.getFAbs()), window, 'valid')
one=py.ones((2,))
hlog=py.concatenate((hlog[0]*one,hlog,hlog[-1]*one))
return hlog-20*py.log10(noiselevel)
def freqVect(self,f): # FFT of cos(f) by doing FFT(cos(f)) x = N.zeros(self.winSize,N.double) for i in range(self.winSize): x[i] = N.cos( float(f)*(i+0.5)*N.pi*2.0 / self.freqT / (self.winSize)) X=P.fft(x) return X
def _calculate_spectra_fft(sx, sy, Q_x, Q_y, Q_s, n_lines):
n_turns, n_files = sx.shape
# Allocate memory for output.
oxx, axx = plt.zeros((n_lines, n_files)), plt.zeros((n_lines, n_files))
oyy, ayy = plt.zeros((n_lines, n_files)), plt.zeros((n_lines, n_files))
for file_i in xrange(n_files):
t = plt.linspace(0, 1, n_turns)
ax = plt.absolute(plt.fft(sx[:, file_i]))
ay = plt.absolute(plt.fft(sy[:, file_i]))
# Amplitude normalisation
ax /= plt.amax(ax, axis=0)
ay /= plt.amax(ay, axis=0)
# Tunes
if file_i==0:
tunexfft = t[plt.argmax(ax[:n_turns/2], axis=0)]
tuneyfft = t[plt.argmax(ay[:n_turns/2], axis=0)]
print "\n*** Tunes from FFT"
print " tunex:", tunexfft, ", tuney:", tuneyfft, "\n"
# Tune normalisation
ox = (t - (Q_x[file_i] % 1)) / Q_s[file_i]
oy = (t - (Q_y[file_i] % 1)) / Q_s[file_i]
# Sort
CX = plt.rec.fromarrays([ox, ax], names='ox, ax')
CX.sort(order='ax')
CY = plt.rec.fromarrays([oy, ay], names='oy, ay')
CY.sort(order='ay')
ox, ax, oy, ay = CX.ox[-n_lines:], CX.ax[-n_lines:], CY.oy[-n_lines:], CY.ay[-n_lines:]
oxx[:,file_i], axx[:,file_i], oyy[:,file_i], ayy[:,file_i] = ox, ax, oy, ay
spectra = {}
spectra['horizontal'] = (oxx, axx)
spectra['vertical'] = (oyy, ayy)
return spectra
def main():
pylab.ion()
ind = [
0,
]
ldft = [
0,
]
lfft = [
0,
]
lpfft = [
0,
]
# plot a graph Dft vs Fft, lists just support size until 2**9
for i in range(1, 9, 1):
t_before = time.clock()
dsprocessing.dspDft(rand(2**i).tolist())
dt = time.clock() - t_before
ldft.append(dt)
print("dft ", 2**i, dt)
#pylab.plot([2**i,], [time.clock()-t_before,]);
t_before = time.clock()
dsprocessing.dspFft(rand(2**i).tolist())
dt = time.clock() - t_before
print("fft ", 2**i, dt)
lfft.append(dt)
#pylab.plot([2**i,], [time.clock()-t_before,]);
ind.append(2**i)
# python fft just to compare
t_before = time.clock()
pylab.fft(rand(2**i).tolist())
dt = time.clock() - t_before
lpfft.append(dt)
pylab.plot(ind, ldft)
pylab.plot(ind, lfft)
pylab.plot(ind, lpfft)
pylab.show()
return [ind, ldft, lfft, lpfft]
def freqVectH(self,f,rads): # FFT of sin(wt).hann(t) by explicit FFT(sin(wt).hann(t)) x = N.zeros(self.winSize,N.double) for i in range(self.winSize): x[i] = N.sin(rads+float(f)*(i+0.5)*N.pi*2.0 / self.freqT / (self.winSize)) X=P.fft(x*self.hann) return X
def fourier(ts, xs):
def safe_dB(x):
abs_x = abs(x)
if abs_x < 1e-10:
return -200
return 20 * math.log10(abs_x)
n = len(ts)
dt = ts[1] - ts[0]
fs = [float(k + 1) / (dt * n) for k in range(n / 2)]
Xs = [safe_dB(X) for X in pylab.fft(xs)[:n / 2]]
return fs, Xs
def fft_based(input_signal, filter_coefficients, boundary=0):
"""applied fft if the signal is too short to be splitted in windows
Params :
input_signal : the audio signal
filter_coefficients : coefficients of the chirplet bank
boundary : manage the bounds of the signal
Returns :
audio signal with application of fast Fourier transform
"""
num_coeffs = filter_coefficients.size
half_size = num_coeffs//2
if boundary == 0:#ZERO PADDING
input_signal = np.lib.pad(input_signal, (half_size, half_size), 'constant', constant_values=0)
filter_coefficients = np.lib.pad(filter_coefficients, (0, input_signal.size-num_coeffs), 'constant', constant_values=0)
newx = ifft(fft(input_signal)*fft(filter_coefficients))
return newx[num_coeffs-1:-1]
elif boundary == 1:#symmetric
input_signal = concatenate([flipud(input_signal[:half_size]), input_signal, flipud(input_signal[half_size:])])
filter_coefficients = np.lib.pad(filter_coefficients, (0, input_signal.size-num_coeffs), 'constant', constant_values=0)
newx = ifft(fft(input_signal)*fft(filter_coefficients))
return newx[num_coeffs-1:-1]
else:#periodic
return roll(ifft(fft(input_signal)*fft(filter_coefficients, input_signal.size)), -half_size).real
def plot2piFft(self, func, Fs, L):
''' Fs is the sampling freq.
L is length of signal list.
This plot is for a func that has period of 2pi.
If you found the time domain wave is not very accurate,
that is because you set too small Fs, which leads to
to big step Ts.
'''
base_freq = 1.0/(2*np.pi) #频域横坐标除以基频，即以基频为单位，此处的基频为 2*pi rad/s
Ts = 1.0/Fs
t = [el*Ts for el in range(0,L)]
x = [func(el) for el in t]
# https://www.ritchievink.com/blog/2017/04/23/understanding-the-fourier-transform-by-example/
# 小明给的代码：
# sampleF = Fs
# print('小明：')
# for f, Y in zip(
# np.arange(0, len(x)*sampleF,1) * 1/len(x) * sampleF,
# np.log10(np.abs(np.fft.fft(x) / len(x)))
# ):
# print('\t', f, Y)
L_4pi = int(4*np.pi / Ts) +1 # 画前两个周期的
self.fig_plot2piFft = plt.figure(7)
plt.subplot(211)
plt.plot(t[:L_4pi], x[:L_4pi])
#title('Signal in Time Domain')
#xlabel('Time / s')
#ylabel('x(t)')
plt.title('Winding Function')
plt.xlabel('Angular location along air gap [mech. rad.]')
plt.ylabel('Current Linkage by unit current [Ampere]')
NFFT = 2**nextpow2(L)
print('NFFT =', NFFT, '= 2^%g' % (nextpow2(L)), '>= L =', L)
y = fft(x,NFFT) # y is a COMPLEX defined in numpy
Y = [2 * el.__abs__() / L for el in y] # /L for spectrum aplitude consistent with actual signal. 2* for single-sided. abs for amplitude.
f = Fs/2.0/base_freq*linspace(0,1,int(NFFT/2+1)) # unit is base_freq Hz
#f = Fs/2.0*linspace(0,1,NFFT/2+1) # unit is Hz
plt.subplot(212)
plt.plot(f, Y[0:int(NFFT/2+1)])
plt.title('Single-Sided Amplitude Spectrum of x(t)')
plt.xlabel('Frequency divided by base_freq [base freq * Hz]')
#plt.ylabel('|Y(f)|')
plt.ylabel('Amplitude [1]')
plt.xlim([0,50])
def visualize ():
sample_rate, snd = load_sample(".\\hh-closed\\dh9.WAV")
print snd.dtype
data = normalize(snd)
print data.shape
n = data.shape[0]
length = float(n)
print length / sample_rate, "s"
timeArray = arange(0, length, 1)
timeArray = timeArray / sample_rate
timeArray = timeArray * 1000 #scale to milliseconds
ion()
if False:
plot(timeArray, data, color='k')
ylabel('Amplitude')
xlabel('Time (ms)')
raw_input("press enter")
exit()
p = fft(data) # take the fourier transform
nUniquePts = ceil((n+1)/2.0)
print nUniquePts
p = p[0:nUniquePts]
p = abs(p)
p = p / float(n) # scale by the number of points so that
# the magnitude does not depend on the length
# of the signal or on its sampling frequency
p = p**2 # square it to get the power
# multiply by two (see technical document for details)
# odd nfft excludes Nyquist point
if n % 2 > 0: # we've got odd number of points fft
p[1:len(p)] = p[1:len(p)] * 2
else:
p[1:len(p) -1] = p[1:len(p) - 1] * 2 # we've got even number of points fft
print p
freqArray = arange(0, nUniquePts, 1.0) * (sample_rate / n);
plot(freqArray/1000, 10*log10(p), color='k')
xlabel('Frequency (kHz)')
ylabel('Power (dB)')
raw_input("press enter")
m = average(freqArray, weights = p)
v = average((freqArray - m)**2, weights= p)
r = sqrt(mean(data**2))
s = var(data**2)
print "mean freq", m #TODO: IMPORTANT: this is currently the mean *power*, not the mean freq. What we want is mean freq weighted by power
print "var freq", v
print "rms", r
print "squared variance", s
def detrend(data,detrend_Kernel_Length = 10,sampling_Frequency = 100000,channels = 8):
from pylab import fft, ifft, sin , cos,log,plot,show,conj,legend
import random
n=len(data[0])
detrend_fft_Length = (2**((log(detrend_Kernel_Length * sampling_Frequency)/log(2))))
ma = [1.0]*sampling_Frequency
ma.extend([0.0]*(detrend_fft_Length - sampling_Frequency))
mafft = fft(ma)
trend = [0.0]*n
for nch in range(channels):
count = 0
while count + detrend_fft_Length <= len(data[nch]):
temp = data[nch][count:count+int(detrend_fft_Length)]
y = fft(temp)
z = ifft( conj(mafft)*y)
for cc in xrange(count,count+(int(detrend_fft_Length)-sampling_Frequency)):
trend[cc] = z[cc-count].real / sampling_Frequency
count = count+(int(detrend_fft_Length)-sampling_Frequency)
for cc in xrange(len(trend)):
data[nch][cc] = data[nch][cc] - trend[cc]
def amplitude_spectrum(self, use_fft=True):
N = len(self.values)
if use_fft:
transform = fft(self.values)
else:
transform = zeros(N, dtype=complex)
for m in range(self.fs):
sum = 0
for n in range(N):
inner = -(2 * pi * n * m) / N
sum += self.values[n] * (cos(inner) + 1j * sin(inner))
transform[m] = sum
transform_abs = [abs(x) for x in transform]
values = [transform_abs[f] / N for f in range(int(self.fs / 2) + 1)]
return self.__class__(values)
def getFFT(s1, n):
p = fft(s1) # take the fourier transform
nUniquePts = ceil((n + 1) / 2.0)
p = p[0:nUniquePts]
p = abs(p)
p = p / n # scale by the number of points so that
# the magnitude does not depend on the length
# of the signal or on its sampling frequency
p = p ** 2 # square it to get the power
# multiply by two (see technical document for details)
# odd nfft excludes Nyquist point
if n % 2 > 0: # we've got odd number of points fft
p[1:len(p)] = p[1:len(p)] * 2
else:
p[1:len(p) - 1] = p[1:len(p) - 1] * 2 # we've got even number of points fft
return nUniquePts, p
def PlotFftRPhase(y, dt=0.1): # plot the grah of fft (fase) of y with command plot
"""
plot also just half spectrum (just positive frequencies),
and from frequencys 1:N/2
"""
N = numpy.size(y)
fund = 1/(N*dt) # freq fundamental, as outras sao multiplas dessa
f=range(N)
for k in range(N-1):
f[k+1]=fund*k #multiplas da fundamental
yn = pylab.fft(y)
angyn = numpy.angle(yn, deg=True)
pylab.plot(f[0:numpy.size(f)/2], angyn[0:numpy.size(f)/2])
def fft(self): self.sortx(); xmin=min(self['x']); xmax=max(self['x']); dx=numpy.fabs(numpy.mean(diff(self['x']))); Nx=(xmax-xmin)/dx; xnew=numpy.linspace(xmin,xmax,Nx+1); spect=self.copyxy(); spect['x']=xnew; spect['y']=xnew; spect1=self.copyxy(); spect1.mapx(spect); newy=pylab.fft(spect1['y']); newx=arange(len(spect1['x']))/float(len(spect1['x']))/dx; spect1['x']=newx; spect1['y']=newy.real; return spect1;
def freqz(sosmat, nsamples=44100, sample_rate=44100, plot=True):
"""Plots Frequency response of sosmat."""
from pylab import np, plt, fft, fftfreq
x = np.zeros(nsamples)
x[nsamples/2] = 0.999
y, states = sosfilter_double_c(x, sosmat)
Y = fft(y)
f = fftfreq(len(x), 1.0/sample_rate)
if plot:
plt.grid(True)
plt.axis([0, sample_rate / 2, -100, 5])
L = 20*np.log10(np.abs(Y[:len(x)/2]) + 1e-17)
plt.semilogx(f[:len(x)/2], L, lw=0.5)
plt.hold(True)
plt.title('freqz sos filter')
plt.xlabel('Frequency / Hz')
plt.ylabel('Damping /dB(FS)')
plt.xlim((10, sample_rate/2))
plt.hold(False)
return x, y, f, Y
def process_window(sample_rate, data):
# print "processing window"
# print data.dtype
# print data.shape
n = data.shape[0]
length = float(n)
# print length / sample_rate, "s"
p = fft(data) # take the fourier transform
nUniquePts = ceil((n+1)/2.0)
p = p[0:nUniquePts]
p = abs(p)
p = p / float(n) # scale by the number of points so that
# the magnitude does not depend on the length
# of the signal or on its sampling frequency
p = p**2 # square it to get the power
# multiply by two (see technical document for details)
# odd nfft excludes Nyquist point
if n % 2 > 0: # we've got odd number of points fft
p[1:len(p)] = p[1:len(p)] * 2
else:
p[1:len(p) -1] = p[1:len(p) - 1] * 2 # we've got even number of points fft
freqArray = arange(0, nUniquePts, 1.0) * (sample_rate / n);
if sum(p) == 0:
raise Silence
m = average(freqArray, weights = p)
v = sqrt(average((freqArray - m)**2, weights= p))
r = sqrt(mean(data**2))
s = var(data**2)
print "mean freq", m #TODO: IMPORTANT: this is currently the mean *power*, not the mean freq. What we want is mean freq weighted by power
# print freqArray
# print (freqArray - m)
# print p
print "var freq", v
print "rms", r
print "squared variance", s
return [m, v, r, s]
def __gauss(sacobj, Tn, alpha):
"""
Return envelope and gaussian filtered data
"""
import pylab as pl
data = pl.array(sacobj.data)
delta = sacobj.delta
Wn = 1 / float(Tn)
Nyq = 1 / (2 * delta)
old_size = data.size
pad_size = 2**(int(pl.log2(old_size))+1)
data.resize(pad_size)
spec = pl.fft(data)
spec.resize(pad_size)
W = pl.array(pl.linspace(0, Nyq, pad_size))
Hn = spec * pl.exp(-1 * alpha * ((W-Wn)/Wn)**2)
Qn = complex(0,1) * Hn.real - Hn.imag
hn = pl.ifft(Hn).real
qn = pl.ifft(Qn).real
an = pl.sqrt(hn**2 + qn**2)
an.resize(old_size)
hn = hn[0:old_size]
return(an, hn)
def PlotFftNAbs(y, dt=0.1): # plot the grah of fft (amplitude) of y with command plot
"""
Normalized version plot of FFT
equal abs(FFT) * 2/y.length
plot also just half spectrum (just positive frequencies),
and from frequencys 1:N/2
the mean (DC component) is removed before fft
"""
N = numpy.size(y)
#remove the DC component just the mean (could use a detrend function)
y = y - numpy.mean(y)
fund = 1/(N*dt) # freq fundamental, as outras sao multiplas dessa
f=range(N)
for k in range(N-1):
f[k+1]=fund*k #multiplas da fundamental
yn = pylab.fft(y)
# perfeito. se vpce colocar um seno de 30 e outro de 60, com aplitude 1 e 2 no espectro
# de amplitude voce vai encontrar exatamente 1 e 2
absyn=2*numpy.abs(yn)/numpy.size(yn);
pylab.plot(f[0:numpy.size(f)/2], absyn[0:numpy.size(f)/2])
return [f, absyn]
def _calculatefdData(self,tdData):
#no need to copy it before (access member variable is ugly but shorter)
#calculate the fft of the X channel
fd=py.fft(tdData.getEX())
#calculate absolute and phase values
fdabs=abs(fd)
fdph=abs(py.unwrap(py.angle(fd)))
#calculate frequency axis
dfreq=py.fftfreq(tdData.num_points,tdData.dt)
#extrapolate phase to 0 and 0 frequency
fdph=self.removePhaseOffset(dfreq,fdph)
#set up the fdData array
t=py.column_stack((dfreq,fd.real,fd.imag,fdabs,fdph))
#this is so not nice!
self.fdData=t
unc=self.calculateFDunc()
t=self.getcroppedData(t,0,unc[-1,0])
#interpolate uncertainty
intpunc=interp1d(unc[:,0],unc[:,1:],axis=0)
unc=intpunc(t[:,0])
return py.column_stack((t,unc))
def main():
# get all files, sorted by temp
os.chdir('./data')
files = sorted(glob.glob("*obdm*") and glob.glob("*dat*"))[::-1]
if files == []:
print 'no files located'
sys.exit()
os.chdir('..')
BECfrac = pl.array([])
Temps = pl.array([])
# compute zero frequency FFT term for each file
for fileName in files:
# get header line
estFile = open('./data/'+fileName,'r')
estLines = estFile.readlines();
pimcid = estLines[0]
headers = estLines[1].split()
estFile.close()
headers.pop(0)
headers = pl.array(headers).astype(float)
rawData = pl.loadtxt('./data/'+fileName)
avgs = pl.array([])
Ncol = 1.0*rawData.size/rawData[0].size
for j in range(int(Ncol)):
for i in range(rawData[j].size):
if j == 0:
avgs = pl.append(avgs, rawData[j][i])
else:
avgs[i] += rawData[j][i]
avgs /= Ncol
# compute FFT
f = pl.fft(avgs)
BECfrac = pl.append(BECfrac, f[0].real)
Temps = pl.append(Temps, (fileName[9:15]))
print '=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-='
print 'Zero Frequency FFT term: ',f[0].real
print '=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-='
# plotting
#pl.plot(headers, avgs, marker='o', linewidth=0, markerfacecolor='None',
# markeredgecolor='Indigo')
#pl.ylabel(r'$\langle \hat{\Psi}^{\dagger}(r)\ \hat{\Psi}(r\') \rangle$',
# fontsize=20)
pl.plot(Temps, BECfrac, marker='o', linewidth=0, markerfacecolor='None',
markeredgecolor='Indigo')
pl.ylabel('Condensate Fraction',
fontsize=20)
pl.xlabel(r'$T\ [K]$', fontsize=20)
pl.title('BEC Fraction')
pl.grid(True)
pl.show()
