11/14/21, 6:17 AM Python Fft Filters
([Link]
Python/v3
(/python/v3) > Signal Suggest an ([Link]
Analysis (/python/v3/#signal- edit to this docs/tree/master/_posts/python-v3/signal-
analysis) > FFT Filters page analysis/fft-filters/)
FFT Filters
in Python/v3
Learn how filter out the frequencies of a signal by using low-pass, high-pass and band-pass FFT
filtering.
Note: this page is part of the documentation for version 3 of
[Link], which is not the most recent version.
See our Version 4 Migration Guide ([Link]
migration/) for information about how to upgrade.
New to Plotly?
Plotly's Python library is free and open source! Get started ([Link]
started/) by downloading the client and reading the primer
([Link]
You can set up Plotly to work in online ([Link]
started/#initialization-for-online-plotting) or offline ([Link]
started/#initialization-for-offline-plotting) mode, or in jupyter notebooks
([Link]
We also have a quick-reference cheatsheet ([Link]
documentation/images/python_cheat_sheet.pdf) (new!) to help you get started!
Imports
The tutorial below imports NumPy ([Link] Pandas
([Link] SciPy ([Link] and Plotly
([Link]
[Link] 1/12
11/14/21, 6:17 AM Python Fft Filters
import [Link] as py
import plotly.graph_objs as go
import plotly.figure_factory as ff
import numpy as np
import pandas as pd
import scipy
from scipy import signal
Import Data
An FFT Filter is a process that involves mapping a time signal from time-space to frequency-
space in which frequency becomes an axis. By mapping to this space, we can get a better
picture for how much of which frequency is in the original time signal and we can ultimately
cut some of these frequencies out to remap back into time-space. Such filter types include
low-pass, where lower frequencies are allowed to pass and higher ones get cut off -, high-
pass, where higher frequencies pass, and band-pass, which selects only a narrow range or
"band" of frequencies to pass through.
Let us import some stock data to apply FFT Filtering:
[Link] 2/12
11/14/21, 6:17 AM Python Fft Filters
data = pd.read_csv('[Link]
peed_laurel_nebraska.csv')
df = data[0:10]
table = ff.create_table(df)
[Link](table, filename='wind-data-sample')
10 Min Std Dev Time 10 M
2.73 2001-06-11 11:00 22.3
1.98 2001-06-11 11:10 23.0
1.87 2001-06-11 11:20 23.3
2.03 2001-06-11 11:30 22.0
3.1 2001-06-11 11:40 20.5
2.3 2001-06-11 11:50 25.2
2.46 2001-06-11 12:00 24.8
1.87 2001-06-11 12:10 24.0
1.71 2001-06-11 12:20 22.9
1.76 2001-06-11 12:30 EDIT CHART
17.9
Plot the Data
Let's look at our data in its raw form before doing any filtering.
[Link] 3/12
11/14/21, 6:17 AM Python Fft Filters
trace1 = [Link](
x=list(range(len(list(data['10 Min Std Dev'])))),
y=list(data['10 Min Std Dev']),
mode='lines',
name='Wind Data'
layout = [Link](
showlegend=True
trace_data = [trace1]
fig = [Link](data=trace_data, layout=layout)
[Link](fig, filename='wind-raw-data-plot')
14
12
10
0 50 100 150
EDIT CHART
[Link] 4/12
11/14/21, 6:17 AM Python Fft Filters
Low-Pass Filter
A Low-Pass Filter is used to remove the higher frequencies in a signal of data.
fc is the cutoff frequency as a fraction of the sampling rate, and b is the transition band also
as a function of the sampling rate. N must be an odd number in our calculation as well.
[Link] 5/12
11/14/21, 6:17 AM Python Fft Filters
fc = 0.1
b = 0.08
N = int([Link]((4 / b)))
if not N % 2: N += 1
n = [Link](N)
sinc_func = [Link](2 * fc * (n - (N - 1) / 2.))
window = 0.42 - 0.5 * [Link](2 * [Link] * n / (N - 1)) + 0.08 * [Link](4 * [Link] * n
/ (N - 1))
sinc_func = sinc_func * window
sinc_func = sinc_func / [Link](sinc_func)
s = list(data['10 Min Std Dev'])
new_signal = [Link](s, sinc_func)
trace1 = [Link](
x=list(range(len(new_signal))),
y=new_signal,
mode='lines',
name='Low-Pass Filter',
marker=dict(
color='#C54C82'
layout = [Link](
title='Low-Pass Filter',
showlegend=True
trace_data = [trace1]
fig = [Link](data=trace_data, layout=layout)
[Link](fig, filename='fft-low-pass-filter')
[Link] 6/12
11/14/21, 6:17 AM Python Fft Filters
Low-Pass Filter
0 50 100 150 200
EDIT CHART
High-Pass Filter
Similarly a High-Pass Filter will remove the lower frequencies from a signal of data.
Again, fc is the cutoff frequency as a fraction of the sampling rate, and b is the transition
band also as a function of the sampling rate. N must be an odd number.
Only by performing a spectral inversion afterwards after setting up our Low-Pass Filter will
we get the High-Pass Filter.
[Link] 7/12
11/14/21, 6:17 AM Python Fft Filters
fc = 0.1
b = 0.08
N = int([Link]((4 / b)))
if not N % 2: N += 1
n = [Link](N)
sinc_func = [Link](2 * fc * (n - (N - 1) / 2.))
window = [Link](N)
sinc_func = sinc_func * window
sinc_func = sinc_func / [Link](sinc_func)
# reverse function
sinc_func = -sinc_func
sinc_func[int((N - 1) / 2)] += 1
s = list(data['10 Min Std Dev'])
new_signal = [Link](s, sinc_func)
trace1 = [Link](
x=list(range(len(new_signal))),
y=new_signal,
mode='lines',
name='High-Pass Filter',
marker=dict(
color='#424242'
layout = [Link](
title='High-Pass Filter',
showlegend=True
trace_data = [trace1]
fig = [Link](data=trace_data, layout=layout)
[Link](fig, filename='fft-high-pass-filter')
[Link] 8/12
11/14/21, 6:17 AM Python Fft Filters
High-Pass Filter
−2
0 50 100 150 200
EDIT CHART
Band-Pass Filter
The Band-Pass Filter will allow you to reduce the frequencies outside of a defined range of
frequencies. We can think of it as low-passing and high-passing at the same time.
In the example below, fL and fH are the low and high cutoff frequencies respectively as a
fraction of the sampling rate.
[Link] 9/12
11/14/21, 6:17 AM Python Fft Filters
fL = 0.1
fH = 0.3
b = 0.08
N = int([Link]((4 / b)))
if not N % 2: N += 1 # Make sure that N is odd.
n = [Link](N)
# low-pass filter
hlpf = [Link](2 * fH * (n - (N - 1) / 2.))
hlpf *= [Link](N)
hlpf = hlpf / [Link](hlpf)
# high-pass filter
hhpf = [Link](2 * fL * (n - (N - 1) / 2.))
hhpf *= [Link](N)
hhpf = hhpf / [Link](hhpf)
hhpf = -hhpf
hhpf[int((N - 1) / 2)] += 1
h = [Link](hlpf, hhpf)
s = list(data['10 Min Std Dev'])
new_signal = [Link](s, h)
trace1 = [Link](
x=list(range(len(new_signal))),
y=new_signal,
mode='lines',
name='Band-Pass Filter',
marker=dict(
color='#BB47BE'
layout = [Link](
title='Band-Pass Filter',
showlegend=True
trace_data = [trace1]
fig = [Link](data=trace_data, layout=layout)
[Link](fig, filename='fft-band-pass-filter')
[Link] 10/12
11/14/21, 6:17 AM Python Fft Filters
Band-Pass Filter
−1
−2
−3
0 50 100 150 200 250
EDIT CHART
([Link]
Products Pricing About Us
Dash Enterprise Pricing Careers
([Link] ([Link] ([Link]
Consulting and Training pricing/) Resources
([Link] ([Link]
and-oem/) Blog
([Link]
Support JOIN OUR MAILING LIST
[Link] 11/12
11/14/21, 6:17 AM Python Fft Filters
Community Support Sign up to stay in the loop with all
([Link] things Plotly — from Dash Club to
Documentation product
updates, webinars, and more!
([Link] S U B S C RI B E
libraries) ( H TTPS :/ / G [Link] / S U B S C RI PTI ON)
Copyright © 2021 Plotly. All rights Terms of Service ([Link] Privacy Policy
reserved. service/) ([Link]
[Link] 12/12