def test_estimate_channel_multiple_rx(self):
        Nsc = 300                            # 300 subcarriers
        size = Nsc // 2
        Nzc = 139

        user1_seq = SrsUeSequence(
            RootSequence(root_index=25, size=size, Nzc=Nzc), 1)
        user2_seq = SrsUeSequence(
            RootSequence(root_index=25, size=size, Nzc=Nzc), 4)

        ue1_channel_estimator = CazacBasedChannelEstimator(user1_seq)

        speed_terminal = 3/3.6               # Speed in m/s
        fcDbl = 2.6e9                        # Central carrier frequency (in Hz)
        subcarrier_bandwidth = 15e3          # Subcarrier bandwidth (in Hz)
        wave_length = 3e8/fcDbl              # Carrier wave length
        Fd = speed_terminal / wave_length    # Doppler Frequency
        Ts = 1./(Nsc * subcarrier_bandwidth) # Sampling interval
        L = 16                               # Number of jakes taps

        # Create the fading generators and set multiple receive antennas
        jakes1 = JakesSampleGenerator(Fd, Ts, L, shape=(3, 1))
        jakes2 = JakesSampleGenerator(Fd, Ts, L, shape=(3, 1))

        # Create a TDL channel object for each user
        tdlchannel1 = TdlChannel(jakes1, channel_profile=COST259_TUx)
        tdlchannel2 = TdlChannel(jakes2, channel_profile=COST259_TUx)

        # Generate channel that would corrupt the transmit signal.
        tdlchannel1._generate_impulse_response(1)
        tdlchannel2._generate_impulse_response(1)

        # Get the generated impulse response
        impulse_response1 = tdlchannel1.get_last_impulse_response()
        impulse_response2 = tdlchannel2.get_last_impulse_response()

        # Get the corresponding frequency response
        freq_resp_1 = impulse_response1.get_freq_response(Nsc)
        H1 = freq_resp_1[:, :, 0, 0]
        freq_resp_2 = impulse_response2.get_freq_response(Nsc)
        H2 = freq_resp_2[:, :, 0, 0]

        # Sequence of the users
        r1 = user1_seq.seq_array()
        r2 = user2_seq.seq_array()

        # Received signal (in frequency domain) of user 1
        comb_indexes = np.arange(0, Nsc, 2)
        Y1 = H1[comb_indexes, :] * r1[:, np.newaxis]
        Y2 = H2[comb_indexes, :] * r2[:, np.newaxis]
        Y = Y1 + Y2

        # Calculate expected estimated channel for user 1
        y1 = np.fft.ifft(np.conj(r1[:, np.newaxis]) * Y, size, axis=0)
        tilde_h1_espected = y1[0:16]
        tilde_H1_espected = np.fft.fft(tilde_h1_espected, Nsc, axis=0)

        # Test the CazacBasedChannelEstimator estimation
        H1_estimated = ue1_channel_estimator.estimate_channel_freq_domain(Y.T, 15)
        np.testing.assert_array_almost_equal(
            H1_estimated, tilde_H1_espected.T)

        # Test if true channel and estimated channel are similar. Since the
        # channel estimation error is higher at the first and last
        # subcarriers we will test only the inner 200 subcarriers
        error = np.abs(H1[50:-50, :] - tilde_H1_espected[50:-50, :])
        np.testing.assert_almost_equal(error/2., np.zeros(error.shape), decimal=2)
    def test_estimate_channel_with_dmrs(self):
        Nsc = 24
        size = Nsc

        user1_seq = DmrsUeSequence(
            RootSequence(root_index=17, size=size), 1)
        user2_seq = DmrsUeSequence(
            RootSequence(root_index=13, size=size), 4)

        ue1_channel_estimator = CazacBasedChannelEstimator(user1_seq,
                                                           size_multiplier=1)

        speed_terminal = 3/3.6               # Speed in m/s
        fcDbl = 2.6e9                        # Central carrier frequency (in Hz)
        subcarrier_bandwidth = 15e3          # Subcarrier bandwidth (in Hz)
        wave_length = 3e8/fcDbl              # Carrier wave length
        Fd = speed_terminal / wave_length    # Doppler Frequency
        Ts = 1./(Nsc * subcarrier_bandwidth) # Sampling interval
        L = 16                               # Number of jakes taps

        jakes1 = JakesSampleGenerator(Fd, Ts, L)
        jakes2 = JakesSampleGenerator(Fd, Ts, L)

        # Create a TDL channel object for each user
        tdlchannel1 = TdlChannel(jakes1, channel_profile=COST259_TUx)
        tdlchannel2 = TdlChannel(jakes2, channel_profile=COST259_TUx)

        # Generate channel that would corrupt the transmit signal.
        tdlchannel1._generate_impulse_response(1)
        tdlchannel2._generate_impulse_response(1)

        # Get the generated impulse response
        impulse_response1 = tdlchannel1.get_last_impulse_response()
        impulse_response2 = tdlchannel2.get_last_impulse_response()

        # Get the corresponding frequency response
        freq_resp_1 = impulse_response1.get_freq_response(Nsc)
        H1 = freq_resp_1[:, 0]
        freq_resp_2 = impulse_response2.get_freq_response(Nsc)
        H2 = freq_resp_2[:, 0]

        # Sequence of the users
        r1 = user1_seq.seq_array()
        r2 = user2_seq.seq_array()

        # Received signal (in frequency domain) of user 1
        Y1 = H1 * r1
        Y2 = H2 * r2
        Y = Y1 + Y2

        # Calculate expected estimated channel for user 1
        y1 = np.fft.ifft(np.conj(r1) * Y, size)
        tilde_h1 = y1[0:4]
        tilde_H1 = np.fft.fft(tilde_h1, Nsc)

        # xxxxxx DEBUG xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
        a = y1[0:4]
        A = np.fft.fft(a, Nsc)
        import matplotlib.pyplot as plt
        # xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

        # Test the CazacBasedChannelEstimator estimation
        np.testing.assert_array_almost_equal(
            ue1_channel_estimator.estimate_channel_freq_domain(Y, 3),
            tilde_H1)

        # Test if true channel and estimated channel are similar. Since the
        # channel estimation error is higher at the first and last
        # subcarriers we will test only the inner 200 subcarriers
        error = np.abs(H1[5:-5] - tilde_H1[5:-5])
        np.testing.assert_almost_equal(
            error/2., np.zeros(error.size), decimal=2)