FFT
Perform a Discrete Fourier Transform on the input vector. Through the FFT algorithm, the input vector will be transformed from a time domain into a frequency domain, which will be an ordered pair of arrays.
Inputs
------
default : OrderedPair
The data to apply FFT to. Params: window : select the node will apply a window to the signal to avoid spectral leakage real_signal : boolean whether the input signal is real or complex sample_rate : int the sample rate of the signal, defaults to 1 display : boolean whether the output would be graphed, set to false for pure data and true for data that is more suitable to be graphed Returns: out : OrderedPair if display is true x: frequency
y: spectrum of the signal out : DataFrame if display is false time: time domain
frequency: frequency domain
real: real section of the signal
imag: imaginary section of the signal
Python Code
from scipy import signal, fft
from numpy import abs
from flojoy import flojoy, OrderedPair, DataFrame
from typing import Literal
from pandas import DataFrame as df
@flojoy
def FFT(
default: OrderedPair,
window: Literal[
"none",
"boxcar",
"triang",
"blackman",
"hamming",
"hann",
"bartlett",
"flattop",
"parzen",
"bohman",
"blackmanharris",
"nuttall",
"barthann",
"cosine",
"exponential",
"tukey",
"taylor",
"lanczos",
] = "hann",
real_signal: bool = True,
sample_rate: int = 1,
display: bool = True,
) -> OrderedPair | DataFrame:
"""Perform a Discrete Fourier Transform on the input vector.
Through the FFT algorithm, the input vector will be transformed from a time domain into a frequency domain, which will be an ordered pair of arrays.
Inputs
------
default : OrderedPair
The data to apply FFT to.
Parameters
----------
window : select
the node will apply a window to the signal to avoid spectral leakage
real_signal : boolean
whether the input signal is real or complex
sample_rate : int
the sample rate of the signal, defaults to 1
display : boolean
whether the output would be graphed, set to false for pure data and true for data that is more suitable to be graphed
Returns
-------
OrderedPair if display is true
x: frequency
y: spectrum of the signal
DataFrame if display is false
time: time domain
frequency: frequency domain
real: real section of the signal
imag: imaginary section of the signal
"""
if sample_rate <= 0:
raise ValueError("Sample rate must be greater than 0")
signal_value = default.y
x = default.x
sample_spacing = 1.0 / sample_rate
# x-axis
frequency = (
fft.rfftfreq(x.shape[-1], sample_spacing)
if real_signal and display
else fft.fftfreq(x.shape[-1], sample_spacing)
)
frequency = fft.fftshift(frequency)
if display:
# y-axis
if window == "none":
fourier = fft.rfft(signal_value) if real_signal else fft.fft(signal_value)
else:
window = signal.get_window(window, len(signal_value))
fourier = (
fft.rfft(signal_value * window)
if real_signal
else fft.fft(signal_value * window)
)
fourier = fft.fftshift(fourier)
fourier = abs(fourier)
return OrderedPair(x=frequency, y=fourier)
# for processing
fourier = fft.fft(signal_value)
d = {"x": x, "frequency": frequency, "real": fourier.real, "imag": fourier.imag}
return DataFrame(df=df(data=d))
Example
Having problems with this example app? Join our Discord community and we will help you out!
In this example, LINSPACE generates an array of 600 samples, which would be the value of step.
The sample rate in this case is 800, so the end parameter is samples/sample_rate = 0.75.
The array is then passed down to two SINE nodes, with one generating a sine wave of 30hz with 1 amplitude,
and the other generating a sine wave of 100hz with 0.5 amplitude.
Finally, the signal is passed down to FFT which performs the fast fourier transform algorithm,
transforming the input from the time domain into the frequency domain. Since we want to display the data,
the display option is set to true. The input signal is only consist of real numbers, so real is also true.
Note that a hann window is used here to reduce spectral leakage, but any other window type or no window also works.