[Discuss-gnuradio] problem with audio extraction in NTSC

[Achilleas Anastasopoulos]

Dear all,

I am working on extracting the audio FM signal from the NTSC
signal, and experimenting with the data file

ntsc-short-complex-baseband-8MS.dat

The idea seems simple: recenter the audio carrier to 0 frequency,
LPF and decimate and then do standard fm demod.
This is done in the attached file.

Unfortunately, although I hear something like

"winter storm will still going to stick around..."

the signal seems highly distorted as if the audio
carrier is very unstable.
I tried to fix this by tracking the video carrier,
but the problem gets worse.

One possible reason for this is that the local oscilator used to 
generate the IQ signal is not stable, which unfortunately cannot be fixed...

Can someone verify this problem and/or suggest any solution.

Thanks
Achilleas

#!/usr/bin/env python

from gnuradio import gr
from gnuradio import audio_oss as audio
from gnuradio.wxgui import stdgui
from gnuradio.wxgui import fftsink, oldfftsink, oldfftsink_linear, scopesink
import wx
import math

                                                                                
class test_app_flow_graph (stdgui.gui_flow_graph):
    def __init__(self, frame, panel, vbox, argv):
        stdgui.gui_flow_graph.__init__ (self, frame, panel, vbox, argv)
                                                                                
        # build our flow graph
        input_rate = 8.00e6
        fm_decim = 25 
        fm_rate = input_rate / fm_decim          # 320 kHz
        audio_decimation = 10 
        audio_rate = fm_rate / audio_decimation  # 32 kHz

        filename = "ntsc-short-complex-baseband-8MS.dat"
                                                                                
        src = gr.file_source (gr.sizeof_short, filename,True)
        conv = gr.interleaved_short_to_complex ()

# Pilot extraction:
        #Ap= 100.0
        #p_width = 100e3
        #pilot_coefs = gr.firdes.band_pass (2/Ap, input_rate, 1.75e6-p_width/2, 
1.75e6+p_width/2, p_width, gr.firdes.WIN_HAMMING)
        #pilot_filter = gr.fir_filter_ccf(1, pilot_coefs)
        #real = gr.complex_to_real()
        #imag = gr.complex_to_imag()
        #neg = gr.multiply_const_ff(-1.0)
        #conj = gr.float_to_complex()


# FM signal extraction
        #correction = gr.multiply_cc()
        fm_coeffs = 
gr.firdes.low_pass(1.0,input_rate,100e3,50e3,gr.firdes.WIN_HAMMING)
        fmc = gr.freq_xlating_fir_filter_ccf (fm_decim,fm_coeffs,-2.75e6, 
input_rate)
        peak_amp = 1.0
        fm_demod_gain = 1.0 / (2 * math.pi * 75e3 / peak_amp / fm_rate)
        volume = 0.4
        fm_demod = gr.quadrature_demod_cf (volume*fm_demod_gain)
# L+R processing: LPF + decimation
# compute FIR filter taps for audio filter
        width = 1e3
        audio_coefs = gr.firdes.low_pass (1.0, fm_rate, 15e3, width, 
gr.firdes.WIN_HAMMING)
        audio_filter_lpr = gr.fir_filter_fff (audio_decimation, audio_coefs)
        audio_sink = audio.sink (int (audio_rate))


        block1, fft_win1 = oldfftsink.make_fft_sink_c (self, panel, "Data", 
1024, input_rate)
        block2, fft_win2 = oldfftsink.make_fft_sink_c (self, panel, "Data", 
512, fm_rate)
        block3, fft_win3 = oldfftsink.make_fft_sink_f (self, panel, "Data", 
256, audio_rate)
        
        self.connect (src, conv)
        #self.connect (conv, pilot_filter)
        #self.connect (pilot_filter,real)
        #self.connect (pilot_filter,imag)
        #self.connect (imag,neg)
        #self.connect (real,(conj,0))
        #self.connect (neg,(conj,1))
        #self.connect (conv,(correction,0))
        #self.connect (conj,(correction,1))
        #self.connect (correction, fmc)
        self.connect (conv, fmc)
        self.connect (fmc,fm_demod)
        self.connect (fm_demod,audio_filter_lpr)
        self.connect (audio_filter_lpr,audio_sink)

        self.connect (conv, block1)
        self.connect (fmc, block2)
        self.connect (audio_filter_lpr, block3)
        vbox.Add (fft_win1, 1, wx.EXPAND)
        vbox.Add (fft_win2, 1, wx.EXPAND)
        vbox.Add (fft_win3, 1, wx.EXPAND)
                                                                                
                                                                                
def main ():
    app = stdgui.stdapp (test_app_flow_graph, "FFT Sink Test App")
    app.MainLoop ()
                                                                                
if __name__ == '__main__':
    main ()