ii) (2%) As long as the sampling frequency is larger than 2 times the signal bandwidth, one can easily reconstruct the original signal from samples.

iii) (2%) Split phase line coding can provide self-synchronization.

iv) (2%) For sending a same signal, linear modulation typically requires a larger bandwidth than angle modulation.

2. Consider sending a message $m(t)$ whose power spectral density (PSD) is $S_m(f) = f^4$ for $|f| \le W$ and $S_m(f) = 0$ otherwise. The transmitted signal is given by $x(t) = m(t) \cos 2\pi f_c t$. The signal is corrupted by an additive white Gaussian noise with PSD $N_0/2$. i) (5%) Pass the received signal through an ideal bandpass filter centered at \( f_{c} \) with bandwidth \( B \geq 2W \). The noise then becomes \( n_{c}(t) \cos 2\pi f_{c}t - n_{s}(t) \sin 2\pi f_{c}t \) narrowband noise. Draw \( S_{n_c}(f) \) the PSD of \( n_{c}(t) \).

ii) (8%) The output signal now is \( y(t) = x(t) + n_c(t) \cos 2\pi f_c t - n_s(t) \sin 2\pi f_c t \). Use a coherent demodulator followed by a low-pass filter with bandwidth \( W \) to demodulate it. Derive the demodulation signal and the SNR at the demodulator output.

iii) (8%) Suppose the signal is further corrupted by an interference \( k(t) = \text{sinc}(wt) \) and the received signal after the bandpass filter becomes \( y_2(t) = (t) + k(t) \cos 2\pi(f_c + 2W)t + n_c(t) \cos 2\pi f_c t - n_s(t) \sin 2\pi f_c t \). Design a low-pass filter for the demodulator in part ii) such that the signal-to-interference-plus-noise ratio (signal power divided by the sum of interference and noise power) is maximized.

3. Let $\{a_k\}$ be a sequence of data which is multiplied with a pulse shaping function $h(t)$ and sent every $T_s$ seconds. The signal is given by $x(t) = \sum_{k=-\infty}^{\infty} a_k h(t - kT_s)$. We sample this signal every $T_s$ seconds to get the discrete signal $x[n] = \sum_{k=-\infty}^{\infty} a_k h[n - k]$ for every integer $n$. i) (5%) Derive the condition on \( h(t) \), in time domain, such that there will be no inter-symbol interference?

ii) (8%) Now, derive from part i) the condition on \( H(f) \), in frequency domain, which guarantees no inter-symbol interference.

iii) (8%) Provide an example of \( H(f) \) satisfying the condition in part ii). Show in time domain that it indeed will not result in any inter-symbol interference.

4. (10%) In wireless communication systems, Additive White Gaussian Noise (AWGN) channel is often assumed. What do "Additive", "White", and "Gaussian" mean?

5. (10%) Plot the 16QAM and 8PSK constellations with normalized power 1. How many bits can one 16QAM symbol and one 8PSK symbol carry, respectively?

114年 - 114 國立臺北大學_碩士班一般入學考試試題_民俗藝術與文化資產研究所：語文論述能力(民俗與文化)#147240

113年 - 113 國立臺北大學_碩士班一般入學考試試題_民俗藝術與文化資產研究所：語文論述能力(民俗與文化)#147238

112年 - 112 國立臺北大學_碩士班一般入學考試試題_民俗藝術與文化資產研究所：語文論述能力(民俗與文化)#147239

111年 - 111 國立臺北大學_碩士班一般入學考試試題_資訊工程學系：線性代數與離散數學#147257

111年 - 111 國立臺北大學_碩士班一般入學考試試題_民俗藝術與文化資產研究所：語文論述能力(民俗與文化)#147245

111年 - [無官方正解]111 國立臺北大學_碩士班一般入學考試試題_金融與合作經營學系：合作經濟學#147235

110年 - 110 國立臺北大學_碩士班一般入學考試試題_民俗藝術與文化資產研究所：語文論述能力(民俗與文化)#147241

110年 - [無官方正解]110 國立臺北大學_碩士班一般入學考試試題_金融與合作經營學系：合作經濟學#147234

109年 - 109 國立臺北大學_碩士班一般入學考試試題_資訊工程學系：線性代數與離散數學#147254

109年 - 109 國立臺北大學_碩士班一般入學考試試題_通訊工程學系：通訊原理#147253