Added FFT function (uses np.fft.fft) to utilities module, to avoid repeating same code in several modules.

这个提交包含在:
Craig Warren
2017-12-20 15:01:12 +00:00
父节点 4e7aecb8bb
当前提交 91a8a6a14f
共有 4 个文件被更改,包括 65 次插入80 次删除

查看文件

@@ -30,6 +30,7 @@ from gprMax.constants import floattype
from gprMax.constants import complextype
from gprMax.materials import Material
from gprMax.pml import PML
from gprMax.utilities import fft_power
from gprMax.utilities import round_value
@@ -234,45 +235,32 @@ def dispersion_analysis(G):
else:
# User-defined waveform
if waveform.type == 'user':
waveformvalues = waveform.uservalues
iterations = G.iterations
# Built-in waveform
else:
# Time to analyse waveform - 4*pulse_width as using entire
# time window can result in demanding FFT
# Time to analyse waveform - 4*pulse_width as using entire
# time window can result in demanding FFT
waveform.calculate_coefficients()
time = np.arange(0, 4 * waveform.chi, G.dt)
waveformvalues = np.zeros(len(time))
timeiter = np.nditer(time, flags=['c_index'])
iterations = round_value(4 * waveform.chi / G.dt)
if iterations > G.iterations:
iterations = G.iterations
while not timeiter.finished:
waveformvalues[timeiter.index] = waveform.calculate_value(timeiter[0], G.dt)
timeiter.iternext()
waveformvalues = np.zeros(G.iterations)
for iteration in range(G.iterations):
waveformvalues[iteration] = waveform.calculate_value(iteration * G.dt, G.dt)
# Ensure source waveform is not being overly truncated before attempting any FFT
if np.abs(waveformvalues[-1]) < np.abs(np.amax(waveformvalues)) / 100:
# Calculate magnitude of frequency spectra of waveform
mag = np.abs(np.fft.fft(waveformvalues))**2
# Calculate power (ignore warning from taking a log of any zero values)
with np.errstate(divide='ignore'):
power = 10 * np.log10(mag)
# Replace any NaNs or Infs from zero division
power[np.invert(np.isfinite(power))] = 0
# Frequency bins
freqs = np.fft.fftfreq(power.size, d=G.dt)
# Shift powers so that frequency with maximum power is at zero decibels
power -= np.amax(power)
# FFT
freqs, power = fft_power(waveformvalues, G.dt)
# Get frequency for max power
freqmaxpower = np.where(np.isclose(power[1::], np.amax(power[1::])))[0][0]
freqmaxpower = np.where(power == 0)[0][0]
# Set maximum frequency to a threshold drop from maximum power, ignoring DC value
try:
freq = np.where((np.amax(power[freqmaxpower::]) - power[freqmaxpower::]) > G.highestfreqthres)[0][0] + 1
results['maxfreq'].append(freqs[freq])
freqthres = np.where(power[freqmaxpower:] < -G.highestfreqthres)[0][0]
results['maxfreq'].append(freqs[freqthres])
except:
results['error'] = 'unable to calculate maximum power from waveform, most likely due to undersampling.'

查看文件

@@ -138,6 +138,35 @@ def round32(value):
return int(32 * np.ceil(float(value) / 32))
def fft_power(waveform, dt):
"""Calculate a FFT of the given waveform of amplitude values;
converted to decibels and shifted so that maximum power is 0dB
Args:
waveform (ndarray): time domain waveform
dt (float): time step
Returns:
freqs (ndarray): frequency bins
power (ndarray): power
"""
# Calculate magnitude of frequency spectra of waveform (ignore warning from taking a log of any zero values)
with np.errstate(divide='ignore'): #
power = 10 * np.log10(np.abs(np.fft.fft(waveform))**2)
# Replace any NaNs or Infs from zero division
power[np.invert(np.isfinite(power))] = 0
# Frequency bins
freqs = np.fft.fftfreq(power.size, d=dt)
# Shift powers so that frequency with maximum power is at zero decibels
power -= np.amax(power)
return freqs, power
def human_size(size, a_kilobyte_is_1024_bytes=False):
"""Convert a file size to human-readable form.

查看文件

@@ -27,6 +27,7 @@ import matplotlib.gridspec as gridspec
from gprMax.exceptions import CmdInputError
from gprMax.receivers import Rx
from gprMax.utilities import fft_power
def mpl_plot(filename, outputs=Rx.defaultoutputs, fft=False):
@@ -83,22 +84,17 @@ def mpl_plot(filename, outputs=Rx.defaultoutputs, fft=False):
# Plotting if FFT required
if fft:
# Calculate magnitude of frequency spectra of waveform (ignore warning from taking a log of any zero values)
with np.errstate(divide='ignore'):
power = 10 * np.log10(np.abs(np.fft.fft(outputdata))**2)
# Replace any NaNs or Infs from zero division
power[np.invert(np.isfinite(power))] = 0
# FFT
freqs, power = fft_power(outputdata, dt)
freqmaxpower = np.where(power == 0)[0][0]
# Frequency bins
freqs = np.fft.fftfreq(power.size, d=dt)
# Set plotting range to -60dB from maximum power or 4 times
# frequency at maximum power
try:
pltrange = np.where(power[freqmaxpower:] < -60)[0][0] + freqmaxpower + 1
except:
pltrange = freqmaxpower * 4
# Shift powers so that frequency with maximum power is at zero decibels
power -= np.amax(power)
# Set plotting range to -60dB from maximum power
pltrange = np.where((np.amax(power[1::]) - power[1::]) > 60)[0][0] + 1
# To a maximum frequency
# pltrange = np.where(freqs > 2e9)[0][0]
pltrange = np.s_[0:pltrange]
# Plot time history of output component

查看文件

@@ -24,6 +24,7 @@ import numpy as np
import matplotlib.pyplot as plt
from gprMax.exceptions import CmdInputError
from gprMax.utilities import fft_power
from gprMax.utilities import round_value
from gprMax.waveforms import Waveform
@@ -83,12 +84,7 @@ def mpl_plot(w, timewindow, dt, iterations, fft=False):
print('Waveform characteristics...')
print('Type: {}'.format(w.type))
if w.type == 'user':
waveform = w.uservalues
w.amp = np.max(np.abs(waveform))
print('Maximum amplitude: {:g}'.format(w.amp))
print('Maximum (absolute) amplitude: {:g}'.format(np.max(np.abs(waveform))))
if w.freq and not w.type == 'gaussian':
print('Centre frequency: {:g} Hz'.format(w.freq))
@@ -100,44 +96,20 @@ def mpl_plot(w, timewindow, dt, iterations, fft=False):
delay = np.sqrt(2) / w.freq
print('Time to centre of pulse: {:g} s'.format(delay))
# Calculate pulse width for gaussian
if w.type == 'gaussian':
powerdrop = -3 # dB
with np.errstate(divide='ignore'): # Ignore warning from taking a log of any zero values
startpower = 10 * np.log10(waveform / np.amax(waveform))
stopower = 10 * np.log10(waveform[start:] / np.amax(waveform))
# Replace any NaNs or Infs from zero division
startpower[np.invert(np.isfinite(startpower))] = 0
stoppower[np.invert(np.isfinite(stoppower))] = 0
start = np.where(startpower > powerdrop)[0][0]
stop = np.where(stoppower < powerdrop)[0][0] + start
print('Pulse width at {:d}dB, i.e. full width at half maximum (FWHM): {:g} s'.format(powerdrop, time[stop] - time[start]))
print('Time window: {:g} s ({} iterations)'.format(timewindow, iterations))
print('Time step: {:g} s'.format(dt))
if fft:
# Calculate magnitude of frequency spectra of waveform (ignore warning from taking a log of any zero values)
with np.errstate(divide='ignore'): #
power = 10 * np.log10(np.abs(np.fft.fft(waveform))**2)
# FFT
freqs, power = fft_power(waveform, dt)
# Replace any NaNs or Infs from zero division
power[np.invert(np.isfinite(power))] = 0
# Frequency bins
freqs = np.fft.fftfreq(power.size, d=dt)
# Shift powers so that frequency with maximum power is at zero decibels
power -= np.amax(power)
if w.type == 'user':
freqmaxpower = np.where(np.isclose(power[1::], np.amax(power[1::])))[0][0]
w.freq = freqs[freqmaxpower]
# Set plotting range to 4 times centre frequency of waveform
pltrange = np.where(freqs > 4 * w.freq)[0][0]
# Set plotting range to 4 times frequency at max power of waveform or
# 4 times the centre frequency
freqmaxpower = np.where(power == 0)[0][0]
if freqs[freqmaxpower] > w.freq:
pltrange = np.where(freqs > 4 * freqs[freqmaxpower])[0][0]
else:
pltrange = np.where(freqs > 4 * w.freq)[0][0]
pltrange = np.s_[0:pltrange]
fig, (ax1, ax2) = plt.subplots(nrows=1, ncols=2, num=w.type, figsize=(20, 10), facecolor='w', edgecolor='w')