1.07 Supplementary Signal Functions, Hold Circuits & DAC Reconstruction
Related Concepts: 1.01 Signals, Systems & Singularity Functions.md | 1.05 The Sampling Theorem Bridge & Multi-Rate Processing.md | 1.06 Typical Signal Processing Operations.md
Overview
This note plugs critical curriculum gaps in Chapter 1. Part I expands on advanced continuous-time and discrete-time signals—Real Exponentials, Complex Exponentials, and the Sinc Function. Part II details physical signal reconstruction in the Digital-to-Analog (DAC) pipeline, focusing on the mathematical transfer functions of Zero-Order Hold (ZOH) and First-Order Hold (FOH) circuits.
Part I: Advanced Elementary Signals
1. Real Exponential Function
A real exponential signal changes its amplitude monotonically {unidirectionally, either always growing or always shrinking} with time. It is governed by a real constant exponent.
A. Continuous-Time Real Exponential
Mathematically represented as: Where:
- is the amplitude scaling factor at .
- is the exponential parameter {damping factor or growth rate}.
Depending on the sign of , the signal exhibits two distinct physical behaviors:
- Exponential Decay (): The signal’s amplitude decays asymptotically toward zero as . This models physical systems returning to equilibrium, such as charge dissipating across a resistor.
- Exponential Growth (): The signal’s amplitude grows boundlessly as . This represents unstable positive-feedback systems.
[GRAPH: Continuous Exponential Decay (y = e^{-0.5t}) vs. Growth (y = e^{0.5t}) — source: Textbook Fig 1.16]
B. Discrete-Time Real Exponential
For discrete index , the real exponential is defined as: Where is the base of the exponential.
- Decaying Sequence (): If is positive (), the sequence decays monotonically toward zero. If is negative (), the sequence oscillates back and forth between positive and negative values while its envelope decays toward zero.
- Growing Sequence (): The sequence grows boundlessly toward infinity.
[GRAPH: Discrete stems showing monotonic decay for a = 0.8 and oscillating decay for a = -0.8 — source: Textbook Ch 1]
2. Complex Exponential Function
The complex exponential function is the mathematical bridge that unifies real exponential dampening and sinusoidal oscillations.
A. Continuous-Time Complex Exponential
Mathematically represented as: Where is a complex variable of frequency:
- represents the real part {the damping coefficient, measured in Nepers/sec}.
- represents the imaginary part {the angular oscillation frequency, measured in radians/sec}.
B. Euler’s Expansion
Substituting into the complex exponential equation: Applying Euler’s Identity ():
ight] = A e^{\sigma t}\cos(\Omega t) + j A e^{\sigma t}\sin(\Omega t)$$ ##### C. Physical Interpretations * **Case 1: $\sigma = 0 \implies x(t) = A e^{j\Omega t}$ (Undamped / Pure Sinusoid):** Represents a pure complex sinusoidal wave of constant amplitude. * **Case 2: $\sigma < 0 \implies x(t) = A e^{\sigma t} e^{j\Omega t}$ (Damped Sinusoid):** Represents a sinusoidal oscillation enclosed inside an exponentially decaying envelope. * **Case 3: $\sigma > 0 \implies x(t) = A e^{\sigma t} e^{j\Omega t}$ (Growing Sinusoid):** Represents a sinusoidal oscillation whose amplitude grows exponentially. [GRAPH: Exponentially damped sinusoid 5*sin(2*pi*1*t).*exp(-0.4*t) showing sinusoidal peaks bounded by an exponential envelope — source: Rabiul Sir Notes, Fig 1.20] --- #### 3. Sinc Function The **Sinc Function** (abbreviation of *sine cardinal*) is highly prominent in signal processing and Fourier analysis because it represents the time-domain impulse response of an ideal brick-wall low-pass filter. ##### A. Definition There are two common definitions of the sinc function. ECE 2107 utilizes the **Normalized Sinc Function** (DSP standard): $$ ext{sinc}(t) = rac{\sin(\pi t)}{\pi t}$$ > [!tip] **Sinc Notation Alert** > > In pure mathematics, the **Unnormalized Sinc Function** (often written as $ ext{Sa}(t)$, sine-arg) is defined as: > $$ ext{Sa}(t) = rac{\sin(t)}{t}$$ > To transition from unnormalized to normalized: $ ext{sinc}(t) = ext{Sa}(\pi t)$. Always read the axis labels in exam sheets! ##### B. Mathematical Evaluation at the Origin ($t = 0$) Direct substitution of $t = 0$ yields an indeterminate form ($0/0$). Applying **L'Hôpital's Rule** *{differentiating the numerator and denominator separately with respect to t}*: $$\lim_{t o 0} ext{sinc}(t) = \lim_{t o 0} rac{ rac{d}{dt}[\sin(\pi t)]}{ rac{d}{dt}[\pi t]} = \lim_{t o 0} rac{\pi \cos(\pi t)}{\pi} = \cos(0) = 1$$ Therefore, the function has a peak amplitude of **1** at $t = 0$. ##### C. Zero Crossings The sinc function crosses the horizontal axis at values of $t$ where the numerator is zero but the denominator is non-zero: $$\sin(\pi t) = 0 \implies \pi t = k\pi \implies t = k \quad ext{for } k \in \mathbb{Z}, \ k eq 0$$ Thus, zero crossings occur at all non-zero integer values: $t = \pm 1, \pm 2, \pm 3, \dots$ [GRAPH: Sinc(t) = sin(\pi t)/(\pi t) showing a central lobe of height 1 at t=0, crossing zero at all non-zero integers, and decaying side lobes — source: Senior Notes, Fig 1.19] --- ### Part II: Practical Signal Reconstruction (Hold Circuits) #### 1. The Digital-to-Analog (DAC) Reconstruction Pipeline In **Note 1.05**, we explored sampling—discretizing a continuous signal $x_a(t)$ into impulses. The reverse process, converting a discrete sequence $x[n]$ back into a continuous-time signal $x_a(t)$, takes place in a DAC. ``` +------------+ +--------------+ +---------------------+ +---------+ | Discrete | ---> | Impulse | ---> | Hold Circuit | ---> | Smoothing| ---> Analog | Sequence | | Generator | | (ZOH or FOH LPF) | | Filter | Signal | x[n] | | x_p(t) | | "Staircase/Polygon" | | (Ideal) | x_a(t) +------------+ +--------------+ +---------------------+ +---------+ ``` --- #### 2. Zero-Order Hold (ZOH) Reconstruction The **Zero-Order Hold** is the simplest and most common hardware reconstruction filter. ##### A. Operating Concept The ZOH takes a discrete sample value at $t = nT_s$ and holds its amplitude completely constant over the entire sampling interval until the next sample arrives at $t = (n+1)T_s$. This produces a jagged, staircase-like continuous waveform. ##### B. Impulse Response The impulse response $h_0(t)$ of a ZOH is a single rectangular gating pulse of width $T_s$ starting at the origin: $$h_0(t) = u(t) - u(t - T_s)$$ [GRAPH: Rectangular pulse of height 1 spanning from t = 0 to t = T_s — source: Textbook Ch 6] ##### C. Derivation of the ZOH s-Domain Transfer Function To find the transfer function $H_0(s)$, we take the Laplace Transform of its impulse response $h_0(t)$: $$H_0(s) = \mathcal{L}\{u(t) - u(t - T_s)\} = \mathcal{L}\{u(t)\} - \mathcal{L}\{u(t - T_s)\}$$ Using the Laplace time-shifting property: $$H_0(s) = rac{1}{s} - rac{1}{s} e^{-s T_s} = rac{1 - e^{-s T_s}}{s}$$ ##### D. Derivation of the ZOH Frequency Response To analyze its filtering characteristics, substitute $s = j\Omega$: $$H_0(j\Omega) = rac{1 - e^{-j\Omega T_s}}{j\Omega}$$ Using Euler's identity to factor out a phase delay of half a sampling period ($e^{-j\Omega T_s/2}$): $$H_0(j\Omega) = rac{e^{-j\Omega T_s/2} \left( e^{j\Omega T_s/2} - e^{-j\Omega T_s/2} ight)}{j\Omega}$$ We know that $e^{j heta} - e^{-j heta} = 2j\sin heta$. Substituting this in: $$H_0(j\Omega) = e^{-j\Omega T_s/2} \left[ rac{2j \sin\left( rac{\Omega T_s}{2} ight)}{j\Omega} ight] = e^{-j\Omega T_s/2} \cdot 2 rac{\sin\left( rac{\Omega T_s}{2} ight)}{\Omega}$$ Multiply numerator and denominator by $T_s$: $$H_0(j\Omega) = T_s \left[ rac{\sin\left( rac{\Omega T_s}{2} ight)}{ rac{\Omega T_s}{2}} ight] e^{-j\Omega T_s/2}$$ Using the unnormalized sinc notation ($ ext{Sa}( heta) = rac{\sin heta}{ heta}$): $$H_0(j\Omega) = T_s \, ext{Sa}\left( rac{\Omega T_s}{2} ight) e^{-j\Omega T_s/2} \quad \iff \quad T_s \, ext{sinc}\left( rac{\Omega T_s}{2\pi} ight) e^{-j\Omega T_s/2}$$ ##### E. Physical Implications of the Frequency Response 1. **Aperture Distortion (Sinc Roll-off):** The magnitude spectrum $|H_0(j\Omega)| = T_s \left| ext{sinc}\left( rac{\Omega T_s}{2\pi} ight) ight|$ acts as a low-pass filter, but its main lobe rolls off significantly before reaching the Nyquist frequency ($\Omega = \pi/T_s$). This dampens high-frequency signal components (aperture effect). 2. **Phase Delay:** The phase term $e^{-j\Omega T_s/2}$ introduces a constant linear phase delay of exactly half a sampling interval ($T_s/2$), lagging the reconstructed output behind the actual digital values. 3. **Spectral Leakage:** Sinc side lobes exist outside the Nyquist band. These lobes leak unwanted high-frequency staircase-harmonics into the analog output, requiring an active low-pass smoothing filter to suppress. --- #### 3. First-Order Hold (FOH) Reconstruction The **First-Order Hold** is a more sophisticated reconstruction filter that utilizes linear interpolation. ##### A. Operating Concept Instead of holding a constant value, the FOH reconstructs the analog signal by joining the sample points with straight lines (linear interpolation). It estimates the slope between consecutive samples, producing a continuous polygonal line waveform with no sharp staircase discontinuities. ##### B. Impulse Response (Symmetric Model) To perform linear interpolation, the impulse response $h_1(t)$ must be a symmetric triangular pulse centered at $t = 0$ with a total base width of $2T_s$: $$h_1(t) = \left( 1 - rac{|t|}{T_s} ight) \left[ u(t + T_s) - u(t - T_s) ight]$$ [GRAPH: Triangular pulse centered at origin of height 1 spanning from -T_s to +T_s — source: Textbook Ch 6] ##### C. Frequency Response of FOH The Fourier transform of a symmetrical triangular pulse is the squared sinc function: $$H_1(j\Omega) = T_s \left( ext{sinc}\left( rac{\Omega T_s}{2\pi} ight) ight)^2$$ ##### D. Comparison: ZOH vs. FOH * **Harmonic Suppression:** The FOH magnitude response drops off as the square of the sinc function, causing its side lobes to be significantly smaller than ZOH. This dramatically reduces high-frequency staircase harmonic leakage and simplifies the design of the post-DAC smoothing filter. * **Causality & Delay:** The symmetric FOH is non-causal *{requires future samples to calculate the connecting slope}*. A causal physical implementation of FOH requires delaying the output by exactly one sampling interval ($T_s$), whereas the ZOH only introduces a delay of $T_s/2$. --- ### Common Mistakes That Cost Marks > [!warning] **Exam Pitfall Checklist** > > * **The Sinc Peak Value:** Never state $ ext{sinc}(0) = 0$ or undefined. Differentiate using L'Hôpital's Rule to prove that $ ext{sinc}(0) = 1$. > * **Complex vs. Real Exponentials:** Do not mix up the real damping factor $\sigma$ with the complex frequency $s$. Remember that $e^{\sigma t}$ represents simple decay/growth, while $e^{j\Omega t}$ represents oscillatory movement. > * **ZOH Phase Delay:** In derivation questions, do not omit the phase shift term $e^{-j\Omega T_s/2}$ when stating the transfer function. This phase shift is a physical delay of $T_s/2$ and is highly critical for stability analysis. > * **ZOH vs. FOH Spectral Roll-off:** Remember that ZOH magnitude decays as $ ext{sinc}(f)$, while FOH magnitude decays much faster as $ ext{sinc}^2(f)$. --- ### Potential Exam Questions (PYQ Prep) > [!question] **Unilateral vs. Bilateral Exponentials** > > **Question:** Define a real exponential signal. Sketch the continuous-time signal $x(t) = e^{-3t}u(t)$ and explain why multiplying by $u(t)$ is necessary in practical systems. > **Answer Plan:** > 1. Write the general definition of a real exponential $x(t) = Ae^{\sigma t}$. > 2. Detail that $\sigma = -3 < 0 \implies$ decay. > 3. Explain that multiplying by $u(t)$ forces the signal to be causal ($x(t) = 0$ for $t < 0$). This matches physically realizable systems that are turned on at a starting time $t = 0$. > [!question] **Sinc Function Peak and Zero Crossings** > > **Question:** Mathematically define the normalized sinc function. Prove its value at the origin $t = 0$ and find all of its zero-crossing points. > **Answer Plan:** > 1. State the formula: $ ext{sinc}(t) = rac{\sin(\pi t)}{\pi t}$. > 2. Use L'Hôpital's rule to derive the limit as $t o 0$, showing the peak is $1$. > 3. Solve $\sin(\pi t) = 0$ for $t eq 0$ to prove that zero crossings occur at all non-zero integers $t = \pm 1, \pm 2, \dots$ > [!question] **ZOH Transfer Function Derivation** > > **Question:** Draw the impulse response of a Zero-Order Hold (ZOH) circuit with a sampling period $T_s$. Derive its transfer function in both the s-domain and the frequency domain. > **Answer Plan:** > 1. Draw a rectangular pulse of height 1 from $t = 0$ to $T_s$. > 2. Write $h_0(t) = u(t) - u(t - T_s)$ and take the Laplace transform to find $H_0(s) = rac{1 - e^{-s T_s}}{s}$. > 3. Substitute $s = j\Omega$ and factor out $e^{-j\Omega T_s/2}$ to arrive at the final sinc form: $H_0(j\Omega) = T_s ext{sinc}\left( rac{\Omega T_s}{2\pi} ight) e^{-j\Omega T_s/2}$. --- ### Self-Check Before Moving On - [ ] Can you write Euler's identity and expand a complex exponential into its real and imaginary parts? - [ ] Do you understand why the sinc function represents the impulse response of an ideal low-pass filter? - [ ] Can you derive the s-domain transfer function of a ZOH from its impulse response step-by-step? - [ ] Do you know the physical trade-off between ZOH and FOH in terms of causal delay versus harmonic leakage? --- ### Source Citations * *K. Deergha Rao, Signals and Systems, Chapter 1 (Section 1.5 - Singularity Functions).* * *Senior Lecture Notes, "1 intro to Signal & systems.pdf" (Sampling Process & Hold Circuits).* * *Syllabus ECE 2107, Weeks 10-11 (DAC & Hold Circuits).*