def test_ccf_000(self): N = 1000 # number of samples to use fs = 1000 # baseband sampling rate rrate = 1.123 # resampling rate nfilts = 32 freq = 100 signal = gr.sig_source_c(fs, gr.GR_SIN_WAVE, freq, 1) head = gr.head(gr.sizeof_gr_complex, N) pfb = filter.pfb_arb_resampler_ccf(rrate, taps) snk = gr.vector_sink_c() self.tb.connect(signal, head, pfb, snk) self.tb.run() Ntest = 50 L = len(snk.data()) t = map(lambda x: float(x)/(fs*rrate), xrange(L)) phase = 0.53013 expected_data = map(lambda x: math.cos(2.*math.pi*freq*x+phase) + \ 1j*math.sin(2.*math.pi*freq*x+phase), t) dst_data = snk.data() self.assertComplexTuplesAlmostEqual(expected_data[-Ntest:], dst_data[-Ntest:], 3)
def __init__(self, rate, taps=None, flt_size=32, atten=100): gr.hier_block2.__init__(self, "pfb_arb_resampler_ccf", gr.io_signature(1, 1, gr.sizeof_gr_complex), # Input signature gr.io_signature(1, 1, gr.sizeof_gr_complex)) # Output signature self._rate = rate self._size = flt_size if (taps is not None) and (len(taps) > 0): self._taps = taps else: # Create a filter that covers the full bandwidth of the input signal bw = 0.4 tb = 0.2 ripple = 0.1 #self._taps = filter.firdes.low_pass_2(self._size, self._size, bw, tb, atten) made = False while not made: try: self._taps = optfir.low_pass(self._size, self._size, bw, bw+tb, ripple, atten) made = True except RuntimeError: ripple += 0.01 made = False print("Warning: set ripple to %.4f dB. If this is a problem, adjust the attenuation or create your own filter taps." % (ripple)) # Build in an exit strategy; if we've come this far, it ain't working. if(ripple >= 1.0): raise RuntimeError("optfir could not generate an appropriate filter.") self.pfb = filter.pfb_arb_resampler_ccf(self._rate, self._taps, self._size) #print "PFB has %d taps\n" % (len(self._taps),) self.connect(self, self.pfb) self.connect(self.pfb, self)
def test_ccf_000(self): N = 1000 # number of samples to use fs = 1000 # baseband sampling rate rrate = 1.123 # resampling rate nfilts = 32 taps = filter.firdes.low_pass_2( nfilts, nfilts * fs, fs / 2, fs / 10, attenuation_dB=80, window=filter.firdes.WIN_BLACKMAN_hARRIS) freq = 100 data = sig_source_c(fs, freq, 1, N) signal = blocks.vector_source_c(data) pfb = filter.pfb_arb_resampler_ccf(rrate, taps) snk = blocks.vector_sink_c() self.tb.connect(signal, pfb, snk) self.tb.run() Ntest = 50 L = len(snk.data()) t = map(lambda x: float(x) / (fs * rrate), xrange(L)) phase = 0.53013 expected_data = map(lambda x: math.cos(2.*math.pi*freq*x+phase) + \ 1j*math.sin(2.*math.pi*freq*x+phase), t) dst_data = snk.data() self.assertComplexTuplesAlmostEqual(expected_data[-Ntest:], dst_data[-Ntest:], 3)
def test_ccf_000(self): N = 1000 # number of samples to use fs = 1000 # baseband sampling rate rrate = 1.123 # resampling rate nfilts = 32 freq = 100 signal = gr.sig_source_c(fs, gr.GR_SIN_WAVE, freq, 1) head = gr.head(gr.sizeof_gr_complex, N) pfb = filter.pfb_arb_resampler_ccf(rrate, taps) snk = gr.vector_sink_c() self.tb.connect(signal, head, pfb, snk) self.tb.run() Ntest = 50 L = len(snk.data()) t = map(lambda x: float(x) / (fs * rrate), xrange(L)) phase = 0.53013 expected_data = map(lambda x: math.cos(2.*math.pi*freq*x+phase) + \ 1j*math.sin(2.*math.pi*freq*x+phase), t) dst_data = snk.data() self.assertComplexTuplesAlmostEqual(expected_data[-Ntest:], dst_data[-Ntest:], 3)
def test02(self): # Test QPSK sync M = 4 theta = 0 loop_bw = cmath.pi/100.0 fmin = -0.5 fmax = 0.5 mu = 0.5 gain_mu = 0.01 omega = 2 gain_omega = 0.001 omega_rel = 0.001 self.test = digital.mpsk_receiver_cc(M, theta, loop_bw, fmin, fmax, mu, gain_mu, omega, gain_omega, omega_rel) data = 10000*[complex( 0.707, 0.707), complex(-0.707, 0.707), complex(-0.707, -0.707), complex( 0.707, -0.707)] data = [0.5*d for d in data] self.src = gr.vector_source_c(data, False) self.snk = gr.vector_sink_c() # pulse shaping interpolation filter nfilts = 32 excess_bw = 0.35 ntaps = 11 * int(omega*nfilts) rrc_taps0 = filter.firdes.root_raised_cosine( nfilts, nfilts, 1.0, excess_bw, ntaps) rrc_taps1 = filter.firdes.root_raised_cosine( 1, omega, 1.0, excess_bw, 11*omega) self.rrc0 = filter.pfb_arb_resampler_ccf(omega, rrc_taps0) self.rrc1 = filter.fir_filter_ccf(1, rrc_taps1) self.tb.connect(self.src, self.rrc0, self.rrc1, self.test, self.snk) self.tb.run() expected_result = 10000*[complex(-0.5, +0.0), complex(+0.0, -0.5), complex(+0.5, +0.0), complex(+0.0, +0.5)] # get data after a settling period dst_data = self.snk.data()[200:] # Only compare last Ncmp samples Ncmp = 1000 len_e = len(expected_result) len_d = len(dst_data) expected_result = expected_result[len_e - Ncmp - 1:-1] dst_data = dst_data[len_d - Ncmp:] #for e,d in zip(expected_result, dst_data): # print "{0:+.02f} {1:+.02f}".format(e, d) self.assertComplexTuplesAlmostEqual(expected_result, dst_data, 1)
def __init__(self, rate, taps=None, flt_size=32, atten=100): gr.hier_block2.__init__(self, "pfb_arb_resampler_ccf", gr.io_signature(1, 1, gr.sizeof_gr_complex), # Input signature gr.io_signature(1, 1, gr.sizeof_gr_complex)) # Output signature self._rate = rate self._size = flt_size if (taps is not None) and (len(taps) > 0): self._taps = taps else: # Create a filter that covers the full bandwidth of the output signal # If rate >= 1, we need to prevent images in the output, # so we have to filter it to less than half the channel # width of 0.5. If rate < 1, we need to filter to less # than half the output signal's bw to avoid aliasing, so # the half-band here is 0.5*rate. percent = 0.80 if(self._rate < 1): halfband = 0.5*self._rate bw = percent*halfband tb = (percent/2.0)*halfband ripple = 0.1 # As we drop the bw factor, the optfir filter has a harder time converging; # using the firdes method here for better results. self._taps = filter.firdes.low_pass_2(self._size, self._size, bw, tb, atten, filter.firdes.WIN_BLACKMAN_HARRIS) else: halfband = 0.5 bw = percent*halfband tb = (percent/2.0)*halfband ripple = 0.1 made = False while not made: try: self._taps = optfir.low_pass(self._size, self._size, bw, bw+tb, ripple, atten) made = True except RuntimeError: ripple += 0.01 made = False print("Warning: set ripple to %.4f dB. If this is a problem, adjust the attenuation or create your own filter taps." % (ripple)) # Build in an exit strategy; if we've come this far, it ain't working. if(ripple >= 1.0): raise RuntimeError("optfir could not generate an appropriate filter.") self.pfb = filter.pfb_arb_resampler_ccf(self._rate, self._taps, self._size) #print "PFB has %d taps\n" % (len(self._taps),) self.connect(self, self.pfb) self.connect(self.pfb, self)
def test01(self): # Test BPSK sync excess_bw = 0.35 sps = 4 loop_bw = cmath.pi/100.0 nfilts = 32 init_phase = nfilts/2 max_rate_deviation = 1.5 osps = 1 ntaps = 11 * int(sps*nfilts) taps = filter.firdes.root_raised_cosine(nfilts, nfilts*sps, 1.0, excess_bw, ntaps) self.test = digital.pfb_clock_sync_ccf(sps, loop_bw, taps, nfilts, init_phase, max_rate_deviation, osps) data = 10000*[complex(1,0), complex(-1,0)] self.src = blocks.vector_source_c(data, False) # pulse shaping interpolation filter rrc_taps = filter.firdes.root_raised_cosine( nfilts, # gain nfilts, # sampling rate based on 32 filters in resampler 1.0, # symbol rate excess_bw, # excess bandwidth (roll-off factor) ntaps) self.rrc_filter = filter.pfb_arb_resampler_ccf(sps, rrc_taps) self.snk = blocks.vector_sink_c() self.tb.connect(self.src, self.rrc_filter, self.test, self.snk) self.tb.run() expected_result = 10000*[complex(-1,0), complex(1,0)] dst_data = self.snk.data() # Only compare last Ncmp samples Ncmp = 1000 len_e = len(expected_result) len_d = len(dst_data) expected_result = expected_result[len_e - Ncmp:] dst_data = dst_data[len_d - Ncmp:] #for e,d in zip(expected_result, dst_data): # print e, d self.assertComplexTuplesAlmostEqual(expected_result, dst_data, 1)
def test01(self): # Test BPSK sync excess_bw = 0.35 sps = 4 loop_bw = cmath.pi / 100.0 nfilts = 32 init_phase = nfilts / 2 max_rate_deviation = 1.5 osps = 1 ntaps = 11 * int(sps * nfilts) taps = filter.firdes.root_raised_cosine(nfilts, nfilts * sps, 1.0, excess_bw, ntaps) self.test = digital.pfb_clock_sync_ccf(sps, loop_bw, taps, nfilts, init_phase, max_rate_deviation, osps) data = 10000 * [complex(1, 0), complex(-1, 0)] self.src = blocks.vector_source_c(data, False) # pulse shaping interpolation filter rrc_taps = filter.firdes.root_raised_cosine( nfilts, # gain nfilts, # sampling rate based on 32 filters in resampler 1.0, # symbol rate excess_bw, # excess bandwidth (roll-off factor) ntaps) self.rrc_filter = filter.pfb_arb_resampler_ccf(sps, rrc_taps) self.snk = blocks.vector_sink_c() self.tb.connect(self.src, self.rrc_filter, self.test, self.snk) self.tb.run() expected_result = 10000 * [complex(-1, 0), complex(1, 0)] dst_data = self.snk.data() # Only compare last Ncmp samples Ncmp = 1000 len_e = len(expected_result) len_d = len(dst_data) expected_result = expected_result[len_e - Ncmp:] dst_data = dst_data[len_d - Ncmp:] #for e,d in zip(expected_result, dst_data): # print e, d self.assertComplexTuplesAlmostEqual(expected_result, dst_data, 1)
def test01(self): # Test BPSK sync M = 2 theta = 0 loop_bw = cmath.pi / 100.0 fmin = -0.5 fmax = 0.5 mu = 0.5 gain_mu = 0.01 omega = 2 gain_omega = 0.001 omega_rel = 0.001 self.test = digital.mpsk_receiver_cc(M, theta, loop_bw, fmin, fmax, mu, gain_mu, omega, gain_omega, omega_rel) data = 10000 * [complex(1, 0), complex(-1, 0)] #data = [2*random.randint(0,1)-1 for x in xrange(10000)] self.src = blocks.vector_source_c(data, False) self.snk = blocks.vector_sink_c() # pulse shaping interpolation filter nfilts = 32 excess_bw = 0.35 ntaps = 11 * int(omega * nfilts) rrc_taps0 = filter.firdes.root_raised_cosine(nfilts, nfilts, 1.0, excess_bw, ntaps) rrc_taps1 = filter.firdes.root_raised_cosine(1, omega, 1.0, excess_bw, 11 * omega) self.rrc0 = filter.pfb_arb_resampler_ccf(omega, rrc_taps0) self.rrc1 = filter.fir_filter_ccf(1, rrc_taps1) self.tb.connect(self.src, self.rrc0, self.rrc1, self.test, self.snk) self.tb.run() expected_result = [0.5 * d for d in data] dst_data = self.snk.data() # Only compare last Ncmp samples Ncmp = 1000 len_e = len(expected_result) len_d = len(dst_data) expected_result = expected_result[len_e - Ncmp - 1:-1] dst_data = dst_data[len_d - Ncmp:] #for e,d in zip(expected_result, dst_data): # print "{0:+.02f} {1:+.02f}".format(e, d) self.assertComplexTuplesAlmostEqual(expected_result, dst_data, 1)
def __init__(self, rate, taps=None, flt_size=32, atten=100): gr.hier_block2.__init__( self, "pfb_arb_resampler_ccf", gr.io_signature(1, 1, gr.sizeof_gr_complex), # Input signature gr.io_signature(1, 1, gr.sizeof_gr_complex)) # Output signature self._rate = rate self._size = flt_size if (taps is not None) and (len(taps) > 0): self._taps = taps else: # Create a filter that covers the full bandwidth of the output signal # If rate >= 1, we need to prevent images in the output, # so we have to filter it to less than half the channel # width of 0.5. If rate < 1, we need to filter to less # than half the output signal's bw to avoid aliasing, so # the half-band here is 0.5*rate. percent = 0.80 if (self._rate < 1): halfband = 0.5 * self._rate bw = percent * halfband tb = (percent / 2.0) * halfband ripple = 0.1 # As we drop the bw factor, the optfir filter has a harder time converging; # using the firdes method here for better results. self._taps = filter.firdes.low_pass_2( self._size, self._size, bw, tb, atten, filter.firdes.WIN_BLACKMAN_HARRIS) else: halfband = 0.5 bw = percent * halfband tb = (percent / 2.0) * halfband ripple = 0.1 made = False while not made: try: self._taps = optfir.low_pass(self._size, self._size, bw, bw + tb, ripple, atten) made = True except RuntimeError: ripple += 0.01 made = False print( "Warning: set ripple to %.4f dB. If this is a problem, adjust the attenuation or create your own filter taps." % (ripple)) # Build in an exit strategy; if we've come this far, it ain't working. if (ripple >= 1.0): raise RuntimeError( "optfir could not generate an appropriate filter." ) self.pfb = filter.pfb_arb_resampler_ccf(self._rate, self._taps, self._size) #print "PFB has %d taps\n" % (len(self._taps),) self.connect(self, self.pfb) self.connect(self.pfb, self)