Problem 1
A causal discrete-time linear time-invariant (LTI) digital filter has the pole-zero constellation plotted in the complex z-plane in the diagram. The system has: - Two complex-conjugate poles located at $p_{1,2} = 0.80 e^{\pm j\pi/4}$ - Two zeros located at $z_1 = 0$ and $z_2 = -1$ Based on the pole-zero geometry and the system transfer function $H(z) = \frac{z(z + 1)}{z^2 - 1.1314z + 0.640}$, what is the DC magnitude gain $|H(e^{j\omega})|$ at zero frequency ($\\omega = 0$), and is the filter Bounded-Input Bounded-Output (BIBO) stable?
- A1.85 (and the filter is marginally stable)
- B3.93 (and the filter is BIBO stable)correct
- C5.56 (and the filter is BIBO unstable)
- D0.51 (and the filter is BIBO stable)
Show the solution
**Step 1: Formulate the System Function H(z)** From the FE Reference Handbook (Signal Processing / Z-Transforms): The transfer function of a discrete-time system with zeros at $z_1 = 0, z_2 = -1$ and poles at $p_{1,2} = 0.80 e^{\pm j\pi/4}$: $$N(z) = (z - 0)(z - (-1)) = z(z + 1) = z^2 + z$$ The denominator polynomial is: $$D(z) = (z - 0.80 e^{j\pi/4})(z - 0.80 e^{-j\pi/4}) = z^2 - 2(0.80)\cos(\pi/4)z + (0.80)^2$$ Since $\cos(\pi/4) = \sqrt{2}/2 = 0.7071$: $$2(0.80)(0.7071) = 1.1314$$ $$D(z) = z^2 - 1.1314z + 0.640$$ $$H(z) = \frac{z(z + 1)}{z^2 - 1.1314z + 0.640}$$ **Step 2: Assess BIBO Stability** For a causal discrete-time LTI system, BIBO stability requires that all poles lie strictly inside the unit circle: $$|p_{1,2}| = 0.80 0.80$). Thus, the filter is **strictly BIBO stable**. **Step 3: Calculate DC Frequency Response at ω = 0 (z = e^(j0) = 1)** Evaluating $H(z)$ on the unit circle at $z = 1$: $$H(1) = \frac{1(1 + 1)}{1^2 - 1.1314(1) + 0.640} = \frac{2.0}{1 - 1.1314 + 0.640} = \frac{2.0}{0.5086}$$ $$|H(e^{j0})| = \frac{2.0}{0.5086} = 3.9324 \approx 3.93$$ *(For comparison, at the Nyquist frequency $\omega = \pi$, $z = -1$, yielding zero gain $H(-1) = 0$ due to the zero at $z = -1$).* Conclusion: The DC gain is **3.93** and the system is **BIBO stable**. Distractor Analysis: - Option A: and the filter is marginally stable - Option B: and the filter is BIBO stable (Correct) - Option C: and the filter is BIBO unstable - Option D: and the filter is BIBO stable
Problem 2
A digital signal processing block computes the frequency spectrum of a real-time sensor array using an $N$-point transform, where $N = 4{,}096 = 2^{12}$: - Direct calculation of the Discrete Fourier Transform (DFT) requires $N^2$ complex multiplications. - The Decimation-in-Time (DIT) Radix-2 Fast Fourier Transform (FFT) algorithm reduces this complexity to $\frac{N}{2} \log_2(N)$ complex multiplications. What is the theoretical computational speedup factor $\frac{\text{Ops}_{DFT}}{\text{Ops}_{FFT}}$ achieved by utilizing the Radix-2 FFT?
- A341
- B512
- C683correct
- D854
Show the solution
**Step 1: Quantify Direct DFT Multiplications** From the FE Reference Handbook (Signal Processing / Discrete Fourier Transform & FFT): Direct matrix multiplication of an $N$-point DFT requires: $$M_{DFT} = N^2 = (4{,}096)^2 = 16{,}777{,}216\text{ complex multiplications}$$ **Step 2: Quantify Radix-2 FFT Multiplications** For a Radix-2 Cooley-Tukey algorithm with $N = 2^m = 2^{12}$: $$M_{FFT} = \frac{N}{2} \log_2(N) = \frac{4{,}096}{2} \times 12 = 2{,}048 \times 12 = 24{,}576\text{ complex multiplications}$$ **Step 3: Calculate the Speedup Ratio** $$\text{Speedup} = \frac{M_{DFT}}{M_{FFT}} = \frac{N^2}{\frac{N}{2} \log_2(N)} = \frac{2N}{\log_2(N)} = \frac{2(4{,}096)}{12} = \frac{8{,}192}{12} = 682.67 \approx 683$$ Conclusion: The computational speedup factor is **683**. Distractor Analysis: - Option A: Compares FFT directly with DFT without real-time sampling rate. - Option B: Uses log10 instead of base-2 log2 in Cooley-Tukey complexity. - Option C: Correct. Determines computational speedup ratio N^2 / (N * log2(N)) = 683. - Option D: Overestimates speedup by omitting complex multiplication scaling.
Problem 3
A digital IIR low-pass filter is designed from an analog continuous-time prototype using the Bilinear Transform: $$s = \frac{2}{T} \frac{1 - z^{-1}}{1 + z^{-1}}$$ Operating parameters: - Digital sampling frequency: $f_s = 20.0\text{ kHz}$ (sampling interval $T = 50.0\ \mu\text{s}$) - Desired discrete-time cutoff frequency: $f_{c,dig} = 4.00\text{ kHz}$ To compensate for the non-linear tangent frequency compression inherent to the Bilinear Transform, what is the prewarped analog cutoff frequency $\omega_a$ (in radians per second) required for the analog prototype filter?
- A25,133 rad/s (unwarped linear frequency 4.00 kHz)
- B34,500 rad/s (excessive prewarping factor)
- C29,062 rad/s (equivalent to f_a = 4.63 kHz, prewarping to compensate tangent compression)correct
- D18,850 rad/s (reciprocal tangent error)
Show the solution
**Step 1: Compute Normalized Digital Angular Frequency $\Omega$** From the FE Reference Handbook (Signal Processing / Digital Filters / Bilinear Transform & Frequency Warping): $$\Omega = \frac{2\pi f_{c,dig}}{f_s} = \frac{2\pi (4{,}000\text{ Hz})}{20{,}000\text{ Hz}} = 0.400\pi\text{ rad/sample} = 1.256637\text{ rad/sample}$$ **Step 2: Apply Tangent Prewarping Formula** The mapping between analog continuous angular frequency $\omega_a$ and digital frequency $\Omega$ is: $$\omega_a = \frac{2}{T} \tan\left(\frac{\Omega}{2}\right)$$ Given $T = \frac{1}{20{,}000\text{ s}^{-1}} = 5.00 \times 10^{-5}\text{ s}$: $$\frac{2}{T} = 2 \times 20{,}000 = 40{,}000\text{ rad/s}$$ $$\frac{\Omega}{2} = \frac{0.400\pi}{2} = 0.200\pi\text{ radians} = 36.0^\circ$$ $$\tan(0.200\pi) = \tan(36.0^\circ) \approx 0.7265425$$ **Step 3: Calculate Prewarped Frequency $\omega_a$** $$\omega_a = 40{,}000 \times 0.7265425 = 29{,}061.7\text{ rad/s} \approx 29{,}062\text{ rad/s}$$ In Hertz: $$f_a = \frac{\omega_a}{2\pi} = \frac{29{,}061.7}{2\pi} = 4{,}625.3\text{ Hz} \approx 4.63\text{ kHz}$$ Conclusion: The required prewarped analog prototype cutoff is **$29,062\text{ rad/s}$** ($4.63\text{ kHz}$). Distractor Analysis: - Option A: unwarped linear frequency 4.00 kHz - Option B: excessive prewarping factor - Option C: equivalent to f_a = 4.63 kHz, prewarping to compensate tangent compression (Correct) - Option D: reciprocal tangent error