FIR
Python Code
from scipy import signal
from flojoy import flojoy, OrderedPair
from typing import Literal
@flojoy
def FIR(
default: OrderedPair,
sample_rate: int = 100,
filter_type: Literal["lowpass", "highpass", "bandpass", "bandstop"] = "lowpass",
window: Literal[
"boxcar",
"triang",
"blackman",
"hamming",
"hann",
"bartlett",
"flattop",
"parzen",
"bohman",
"blackmanharris",
"nuttall",
"barthann",
"cosine",
"exponential",
"tukey",
"taylor",
"lanczos",
] = "hann",
cutoff_low: float = 10.0,
cutoff_high: float = 15.0,
taps: int = 200,
) -> OrderedPair:
"""Apply a low-pass FIR filter to an input vector. The filter is designed with the window method.
This filter takes a few inputs: the sample_rate (will be passed as a parameter if the target node is not connected), the window type of the filter, the cutoff frequency, and the number of taps (or length) of the filter.
Inputs
------
default : OrderedPair
The data to apply a FIR filter to.
Parameters
----------
sample_rate : int
the amount of samples within a second
filter_type : select
how the filter behaves
window : select
the window function used in the FIR
cutoff_low : float
the frequency cutoff to filter out the lower frequencies
cutoff_high : float
the frequency cutoff to filter out the upper frequencies
taps : int
the length of the filter
Returns
-------
OrderedPair
x: time domain
y: filtered signal
"""
sample_rate: int = sample_rate # Hz
filter_type: str = filter_type
window_type: str = window
cutoff_low: float = cutoff_low
cutoff_high: float = cutoff_high
n_taps: int = taps
times = default.x
input_signal = default.y
if input_signal.size < n_taps * 3:
raise ValueError("length of the data should be three times longer than taps")
elif (
n_taps % 2 == 0
): # in the case where the passband contains the Nyquist frequency
n_taps = n_taps + 1
# create the filter with the parameter inputs
if filter_type == "bandpass" or filter_type == "bandstop":
fil = signal.firwin(
numtaps=n_taps,
cutoff=[cutoff_low, cutoff_high],
fs=sample_rate,
pass_zero=filter_type,
window=window_type,
)
elif filter_type == "lowpass":
fil = signal.firwin(
numtaps=n_taps,
cutoff=cutoff_high,
fs=sample_rate,
pass_zero=filter_type,
window=window_type,
)
else:
fil = signal.firwin(
numtaps=n_taps,
cutoff=cutoff_low,
fs=sample_rate,
pass_zero=filter_type,
window=window_type,
)
# ... and then apply it to the signal
filtered_x = signal.filtfilt(fil, 1.0, input_signal)
return OrderedPair(x=times, y=filtered_x)
Example App
In this example, five BASIC_OSCILLATOR nodes generates an array of 400 samples with a sample rate of 100.
Each with different amplitude and a unique frequency between 10-35hz.
They are then all summed together into one signal using the ADD node.
To then get one of the initial sine waves, which is the one generated by OSC_3 in this case, the FIR node
is used to filter out all other frequencies that’s not 25hz.
For the filter type, it will be bandpass as there are unwanted frequencies both above and below 25hz.
The high cutoff would be 28hz and the low cutoff would be 23hz. For the window, barthann is used for
a sharper cutoff since we don’t want any other frequencies in this example.
For taps, there are a couple of ways of calculating depending on how much attenuation or how much ripple is in the signal. Here are some formulas, the top-rated answer would be the one used in this example. With the transition width being 20, the number of taps would be 80 after calculation.
Finally, the output of FIR node will be displayed with the LINE node, which gives us a very similar graph to OSC_3.
Also, two FFT nodes are included to show the frequencies before and after the FIR node.