kopia lustrzana https://github.com/projecthorus/horus-gui
112 wiersze
3.3 KiB
Python
112 wiersze
3.3 KiB
Python
# FFT
|
|
import logging
|
|
import time
|
|
import numpy as np
|
|
from queue import Queue
|
|
from threading import Thread
|
|
|
|
|
|
class FFTProcess(object):
|
|
""" Process an incoming stream of samples, and calculate FFTs """
|
|
|
|
def __init__(
|
|
self,
|
|
nfft=8192,
|
|
stride=4096,
|
|
update_decimation=1,
|
|
fs=48000,
|
|
sample_width=2,
|
|
range=[100, 4000],
|
|
callback=None,
|
|
):
|
|
self.nfft = nfft
|
|
self.stride = stride
|
|
self.update_decimation = update_decimation
|
|
self.update_counter = 0
|
|
self.fs = fs
|
|
self.sample_width = sample_width
|
|
self.range = range
|
|
|
|
self.callback = callback
|
|
|
|
self.sample_buffer = bytearray(b"")
|
|
|
|
self.input_queue = Queue(512)
|
|
|
|
self.init_window()
|
|
|
|
self.processing_thread_running = True
|
|
|
|
self.t = Thread(target=self.processing_thread)
|
|
self.t.start()
|
|
|
|
def init_window(self):
|
|
""" Initialise Window functions and FFT scales. """
|
|
self.window = np.hanning(self.nfft)
|
|
self.fft_scale = np.fft.fftshift(np.fft.fftfreq(self.nfft)) * self.fs
|
|
self.mask = (self.fft_scale > self.range[0]) & (self.fft_scale < self.range[1])
|
|
|
|
def perform_fft(self):
|
|
""" Perform a FFT on the first NFFT samples in the sample buffer, then shift the buffer along """
|
|
|
|
# Convert raw data to floats.
|
|
raw_data = np.fromstring(
|
|
bytes(self.sample_buffer[: self.nfft * self.sample_width]), dtype=np.int16
|
|
)
|
|
raw_data = raw_data.astype(np.float64) / (2 ** 15)
|
|
|
|
# Advance sample buffer
|
|
self.sample_buffer = self.sample_buffer[self.stride * self.sample_width :]
|
|
|
|
# Calculate Maximum value
|
|
_raw_max = raw_data.max()
|
|
if(_raw_max>0):
|
|
# Calculate FFT
|
|
_fft = 20 * np.log10(
|
|
np.abs(np.fft.fftshift(np.fft.fft(raw_data * self.window)))
|
|
) - 20 * np.log10(self.nfft)
|
|
|
|
# Calculate dBFS value.
|
|
_dbfs = 20*np.log10(_raw_max)
|
|
else:
|
|
_fft = np.zeros(self.nfft)*np.nan
|
|
_dbfs = -99.0
|
|
|
|
if self.callback != None:
|
|
if self.update_counter % self.update_decimation == 0:
|
|
self.callback({"fft": _fft[self.mask], "scale": self.fft_scale[self.mask], 'dbfs': _dbfs})
|
|
|
|
self.update_counter += 1
|
|
|
|
def process_block(self, samples):
|
|
""" Add a block of samples to the input buffer. Calculate and process FFTs if the buffer is big enough """
|
|
|
|
self.sample_buffer.extend(samples)
|
|
|
|
while len(self.sample_buffer) > self.nfft * self.sample_width:
|
|
self.perform_fft()
|
|
|
|
def processing_thread(self):
|
|
|
|
while self.processing_thread_running:
|
|
if self.input_queue.qsize() > 0:
|
|
data = self.input_queue.get()
|
|
self.process_block(data)
|
|
else:
|
|
time.sleep(0.01)
|
|
|
|
def add_samples(self, samples):
|
|
""" Add a block of samples to the input queue """
|
|
try:
|
|
self.input_queue.put_nowait(samples)
|
|
except:
|
|
logging.error("Input overrun!")
|
|
|
|
def flush(self):
|
|
""" Clear the sample buffer """
|
|
self.sample_buffer = bytearray(b"")
|
|
|
|
def stop(self):
|
|
""" Halt processing """
|
|
self.processing_thread_running = False
|