Chapter 4 - The Sampling Theorem & Multi-Rate Processing

04 Chapter Map - The Sampling Theorem & Multi-Rate Processing

Chapter 4 Overview & Map of Content (MOC)

Nyquist boundaries, aperture effect sinc droops, zero-order holds, and multi-rate decimation/interpolation arrays.

---

📚 Study Notes Index

Read in order — each note assumes the previous one.

#	Note	What it covers
4.00	4.00 Sampling and Multi-Rate Processing Compact Review	Sampling Compact Review, Sampling Cheat Sheet
4.01	4.01 The Shannon Nyquist Sampling Theorem	Sampling Theorem, Nyquist Rate, Sampling Proof
4.02	4.02 Spectral Replications and Aliasing Dynamics	Aliasing Dynamics, Spectral Replications, Frequency Folding, Anti-Aliasing Filtering
4.03	4.03 Ideal Natural and Flat-Top Sampling Techniques	Sampling Techniques, Flat-Top Sampling, Natural Sampling, Aperture Effect, Equalizer Filter
4.04	4.04 Multi-Rate Signal Processing Decimation and Interpolati	Decimation, Interpolation, Up-sampling, Down-sampling, Multi-rate Processing

---

🎯 Exam Weight

ECE 2107 Exam Relevance

Master the core derivations, mathematical definitions, and problem-solving techniques. Refer to ECE 2107 - Signals and Systems for syllabus boundaries and past year questions.

---

🔗 Related Resources

• Course Teaching Plan: ECE 2107 - Signals and Systems
• Previous chapter: 03 Chapter Map - State-Space Representation of CT Systems
• Next chapter: 05 Chapter Map - Analog Filter Design

---

Chapter 4: Sampling & Multi-Rate Processing — Compact Review

1. The Shannon-Nyquist Sampling Theorem

1.1 Theorem Statement and Fundamental Bounds

*(Target: Theory Descriptive / Numerical Solving)* [1.05, 4.01]

• Shannon-Nyquist Theorem: A continuous-time, band-limited analog signal $x_a(t)$ containing no frequency components higher than $f_{\text{max}}$ Hz can be uniquely and perfectly reconstructed from its discrete samples $x_a(nT)$ if and only if the sampling frequency $F_s$ (samples/second) satisfies [1.05, 6.1.2]: $F_s\ge 2f_{\text{max}}$
• Nyquist Rate ($F_N$): The absolute minimum theoretical sampling rate to avoid overlap (aliasing) [1.05, 6.1.2]: $F_N=2f_{\text{max}}$
• Nyquist Interval ($T_N$): The maximum allowable time duration between consecutive samples [1.05]: $T_N=\frac{1}{F_N}=\frac{1}{2f_{\text{max}}}$
• Band-Pass Sampling Theorem: For a band-pass signal with non-zero energy restricted to the frequency band $f_L<|f|<f_H$, the minimum sampling rate to prevent aliasing is [6.1.2]: $F_s\ge 2\Delta f=2(f_H-f_L)$

---

1.2 Frequency-Domain Representation of Sampling (Spectral Replication)

*(Target: Mathematical Proof / 10-Mark Derivation)* [4.01]

• Mathematical Modeling: The sampled signal $x_s(t)$ is modeled as the multiplication of the analog signal $x_a(t)$ by an infinite periodic Dirac impulse train $p(t)={\sum }_{n=-\infty }^{\infty }\delta (t-n{T}_s)$ with period $T_s$ [6.1.1]: $x_s(t)=x_a(t)\cdot p(t)={\sum }_{n=-\infty }^{\infty }{x}_a(n{T}_s)\delta (t-n{T}_s)\text{(derivation)}$
• Frequency-Domain Spectrum: Multiplying in the time domain corresponds to convolving their respective Fourier transforms in the frequency domain, replicating the analog spectrum $X_a(j\Omega )$ at integer multiples of the sampling angular frequency ${\Omega }_s=\frac{2\pi }{T_s}$ [4.01, 7.4]: $X_s(j\Omega )=\frac{1}{T_s}{\sum }_{k=-\infty }^{\infty }{X}_a\left(j\left(\Omega -k{\Omega }_s\right)\right)\text{(derivation)}$

---

2. The Three Sampling Conditions

*(Target: Theory Descriptive / Spectral Diagrams)* [1.05, 4.02]

Condition	Mathematical Criteria	Spectral Dynamics	Reconstruction Filter Requirement
Over-sampling	$F_s>2f_{\text{max}}$	Adjacent spectral replications are separated by a guard band ($F_s-2f_{\text{max}}$) [1.05].	Practical, cheap low-pass filter with a gradual transition band [1.05].
Nyquist-rate sampling	$F_s=2f_{\text{max}}$	Adjacent spectral replications touch exactly at their edges without overlapping [1.05].	Ideal “brick-wall” low-pass filter with an infinite roll-off slope [1.05].
Under-sampling	$F_s<2f_{\text{max}}$	Adjacent spectral replications overlap, causing aliasing (frequency folding) [1.05].	Non-recoverable: Original signal cannot be reconstructed by any filter [1.05].

• [GRAPH: Replicated frequency spectra showing: (1) Over-sampling (spectral copies separated by guard bands), (2) Nyquist sampling (copies exactly touching at f_max), and (3) Under-sampling (copies overlapping, highlighting shaded aliasing fold-over regions) — source: Ch 4.02 / Ch 7 Fig 7.12]

---

3. Aliasing Dynamics & Anti-Aliasing Filters

3.1 Aliasing Effect

*(Target: Theory Descriptive)* [1.05, 4.02]

• Definition: A destructive, irreversible frequency-folding distortion that occurs when an analog signal is under-sampled ($F_s<2f_{\text{max}}$) [1.05, 6.1.2].
• Mechanism: High-frequency spectral components above the folding frequency (Nyquist frequency, $F_s/2$) fold back into the baseband spectrum and masquerade as false low-frequency components [1.05, 6.1.2].

---

3.2 Anti-Aliasing Filter

*(Target: Theory Descriptive)* [4.02]

• Definition: A high-order analog low-pass filter placed before the sampler (A/D converter) [1.05, 4.02].
• Function: It band-limits the input analog signal by stripping away all noise and frequency components higher than $F_s/2$ {the folding limit} before sampling occurs, ensuring the Nyquist criterion is strictly satisfied [1.05, 4.02].

---

4. Classification of Sampling Techniques

*(Target: Theory Descriptive / Structural Comparison)* [1.05, 4.03]

The physical switching implementation of sampling is divided into three distinct methodologies:

Parameter / Feature	Ideal (Impulse) Sampling	Natural Sampling (Chopping)	Flat-Top Sampling (Sample & Hold)
Mathematical Multiplier	Infinite train of zero-width Dirac delta impulses: ${\delta }_{T_s}(t)$ [1.05, 4.03].	Periodic train of finite-width rectangular pulses of duration $\tau$ [1.05, 4.03].	Sampled value held strictly constant over duration $\tau$ via capacitor [1.05, 4.03].
Waveform Top Profile	Zero width, infinite height impulses [1.05].	Tops of pulses follow the exact analog curve of $x_a(t)$ [1.05, 141].	Tops of pulses are perfectly flat, producing a staircase profile [1.05, 141].
Spectral Envelope	Flat amplitude scaling factor $\frac{1}{T_s}$ across all frequencies [7.4].	Monotonically decaying sinc envelope: scaled by Fourier coefficients $C_n$ [4.03].	Decaying sinc function envelope: $H_{\text{sh}}(j\Omega )=\tau \cdot \text{sinc}\left(\frac{\Omega \tau }{2}\right)e^{-j\Omega \tau /2}$ [4.03].
High-Frequency Distortion	None.	None (within the baseband) [4.03].	Aperture Effect: Sinc envelope attenuates high frequencies within the baseband [4.03].
Equalizer Required?	No.	No [4.03].	Yes: Requires a ${\text{sinc}}^{-1}$ equalizer filter to restore flat response [4.03].

• [GRAPH: Graphical wave representation comparing: (1) Analog wave, (2) Natural sampled wave (curving tops), and (3) Flat-top sampled wave (flat horizontal tops) — source: Ch 4.03 / Ch 6 Fig 6.3]

---

4.1 The Aperture Effect

*(Target: Theory Descriptive)* [4.03]

• Aperture Effect: The high-frequency amplitude distortion (droop) introduced during flat-top sampling because the pulse width $\tau$ is finite (not an ideal impulse) [4.03].
• Mathematical Envelope: The baseband spectrum is multiplied by a sinc envelope [4.03]: $H_{\text{sh}}(j\Omega )=\tau \left(\frac{\sin (\Omega \tau /2)}{\Omega \tau /2}\right)e^{-j\Omega \tau /2}$
• Remedy: An active equalizer circuit with frequency response proportional to $H_{\text{eq}}(j\Omega )\approx \frac{1}{\text{sinc}(\Omega \tau /2)}$ is placed in the reconstruction stage to compensate for the high-frequency attenuation [4.03].

---

5. Reconstruction & Zero-Order Hold (ZOH)

5.1 Ideal Reconstruction (Whittaker-Shannon Interpolation)

*(Target: Mathematical Proof / 10-Mark Derivation)* [7.5]

• Mechanism: Ideal reconstruction is achieved by passing the sampled impulse train $x_s(t)$ through an ideal “brick-wall” low-pass filter $H_{\text{LP}}(j\Omega )$ with cutoff frequency ${\Omega }_c={\Omega }_s/2=\pi /T_s$ and gain $T_s$ [7.5]: $H_{\text{LP}}(j\Omega )=\{\begin{array}{ll}{T}_s,& |\Omega |\le {\Omega }_c\\ 0,& |\Omega |>{\Omega }_c\end{array}\leftrightarrow h_{\text{LP}}(t)=\frac{\sin (\pi t/T_s)}{\pi t/T_s}=\text{sinc}\left(\frac{t}{T_s}\right)\text{(derivation)}$
• Reconstruction Formula: In the time domain, convolving $x_s(t)$ with $h_{\text{LP}}(t)$ yields the continuous-time signal $x_a(t)$ via sinc interpolation [7.5]: $x_a(t)={\sum }_{n=-\infty }^{\infty }{x}_a(n{T}_s)\cdot \text{sinc}\left(\frac{t-n{T}_s}{T_s}\right)\text{(derivation)}$

---

5.2 Quantization Mechanics & SQNR

*(Target: Theory Descriptive / Numerical Solving)* [6.1.2, 6.1.3]

• Quantization Step Size ($\Delta$): For a $b$-bit linear quantizer with a full-scale voltage range $A$, the step size is [6.1.3]: $\Delta =\frac{A}{2^b}$
• Quantization Noise Power ($P_{\text{qn}}$ or ${\sigma }_e^2$): Assuming the quantization error $e(n)$ is uniformly distributed in the range $[-\Delta /2,\Delta /2]$, the mean-square error (noise power) is [6.1.3]: $P_{\text{qn}}={\sigma }_e^2=\frac{1}{\Delta }{\int }_{-\Delta /2}^{\Delta /2}{e}^{2}de=\frac{{\Delta }^2}{12}\text{(derivation)}$
• Signal-to-Quantization Noise Ratio (SQNR): Expressed as a function of the number of bits $b$, the maximum theoretical SQNR for a full-scale sinusoidal input is [6.1.3]: ${\text{SQNR}}_{\text{dB}}\approx 6.02b+1.76\text{ dB}$ {Note: Every additional bit added to the ADC resolution improves the SQNR by exactly 6.02 dB}.

---

6. Multi-Rate Signal Processing

*(Target: Theory Descriptive / Numerical Solving)* [4.04]

6.1 Discrete-Time Down-sampler (Decimator)

*(Target: Numerical Solving / Array Manipulation)* [1.05, 4.04]

• Definition: Reduces the sampling rate of a discrete sequence by an integer factor $M$ [1.05, 4.04].
• Operation: Keeps every $M$-th sample and discards the intermediate $M-1$ samples [1.05, 4.04]: $x_d[n]=x[nM]$
• Anti-Aliasing Filter Requirement: To prevent digital aliasing before discarding samples, the discrete signal must be band-limited to $|\omega |<\pi /M$ using a digital low-pass decimation filter [4.04].

---

6.2 Discrete-Time Up-sampler (Interpolator)

*(Target: Numerical Solving / Array Manipulation)* [1.05, 4.04]

• Definition: Artificially increases the sampling rate of a discrete sequence by an integer factor $L$ [1.05, 4.04].
• Operation: Inserts $L-1$ zero-valued samples between consecutive original samples [1.05, 4.04]: $x_e[n]=\{\begin{array}{ll}x[n/L],& n=0,\pm L,\pm 2L,\dots \\ 0,& \text{otherwise}\end{array}$
• Anti-Imaging Filter Requirement: To remove the high-frequency spectral images created by the inserted zeros, the up-sampled signal must be passed through a digital low-pass interpolation filter with cutoff ${\omega }_c=\pi /L$ and gain factor $L$ [4.04].

---

6.3 Down-sampling vs. Up-sampling Comparison

*(Target: Theory Descriptive)* [4.04]

Feature / Property	Down-sampler (Compressor)	Up-sampler (Expander)
Mathematical Identity	$x_d[n]=x[nM]$ [1.05, 4.04]	$x_e[n]=x[n/L]$ for multiples of $L$ [1.05, 4.04]
Effect on Array Length	Slashes length by a factor of $M$.	Expands length by a factor of $L$.
Time-Domain Operation	Discards $M-1$ in-between samples [1.05, 4.04].	Inserts $L-1$ zeros in-between samples [1.05, 4.04].
Frequency Spectrum Effect	Stretches the spectrum in frequency.	Compresses the spectrum, creating images.
Filtering Required	Low-pass decimation filter before down-sampling [4.04].	Low-pass interpolation filter after up-sampling [4.04].
Data Loss Profile	Lossy (irreversible unless over-sampled) [1.05].	Non-lossy (information is fully preserved).

---

7. Common Mistakes That Cost Marks

Critical Exam Pitfalls

1. Bit-Reversal / Shift Origin Mismatch: In down-sampling numericals, forgetting to track the origin index ($n=0$, underlined or marked with an arrow). Down-sampling keeps samples at $nM$ measured strictly relative to the original origin sample, not just from the left-most element of the array.
2. Omitting Zero Insertion in Up-sampling: Writing interpolations by directly duplicating adjacent samples (staircase interpolation) instead of inserting actual zeros first before passing through the low-pass interpolation filter.
3. Forgetting the Equalizer in Flat-Top Sampling: Stating that flat-top sampling preserves baseband signal spectrum shape without distortion. You must specify that it introduces aperture effect droop and requires a ${\text{sinc}}^{-1}$ equalizer [4.03].
4. Nyquist Interval Variable Swap: Confusing the Nyquist Rate (frequency, $2f_{\text{max}}$) with the Nyquist Interval (time, $1/(2f_{\text{max}})$).

---

8. PYQ Bank — Verbatim Questions & Answer Plans

Q1: Down-sampling and Up-sampling Array Solving [KUET 2024, 2022]

• Question: What is discrete-time up-sampler and down-sampler? Find the up-sampling sequence and down-sampling sequence for a given signal array by factor 3.
• Answer Plan:
  1. Define down-sampler and write the compressor identity: $x_d[n]=x[nM]$ [1.05].
  2. Define up-sampler and write the expander identity: $x_e[n]=x[n/L]$ [1.05].
  3. Identify the index $n=0$ of the given array.
  4. Extract indices at multiples of $3$ to formulate the down-sampled array.
  5. Insert exactly 2 zeros between every adjacent sample of the original array to formulate the up-sampled array.

Q2: Shannon-Nyquist Theorem Proof [KUET 2019, 2018, 2017]

• Question: State and prove the Shannon-Nyquist sampling theorem.
• Answer Plan:
  1. State the theorem: $F_s\ge 2f_{\text{max}}$ [1.05].
  2. Model the impulse-train sampler mathematically: $x_s(t)=x_a(t)\cdot \sum \delta (t-n{T}_s)$ [6.1.1].
  3. (derivation): Apply continuous-time Fourier transform to show $X_s(j\Omega )=\frac{1}{T_s}\sum X_a(j(\Omega -k{\Omega }_s))$ [4.01].
  4. Discuss how overlapping spectra (aliasing) occurs if ${\Omega }_s<2{\Omega }_m$, whereas guard bands are maintained if ${\Omega }_s>2{\Omega }_m$ [1.05, 4.02].

Q3: Distinguish Between Sampling Techniques [KUET 2018, 2017]

• Question: Classify sampling techniques and distinguish between Ideal, Natural, and Flat-top sampling.
• Answer Plan:
  1. List the three sampling classifications [1.05].
  2. Insert comparison matrix detailing: Mathematical multiplier, Waveform top profile, Spectrum envelope, and Equalizer requirements [4.03].
  3. Explicitly define the aperture effect droop and the role of the recovery filter [4.03].

---

Chapter 9 Self-Check Revision Checklist

• Can you write down the frequency-domain representation of uniform sampling from memory?
• Do you know how the aperture effect shifts flat-top sampling spectrum profiles compared to ideal impulse sampling?
• Can you solve discrete down-sampling array mappings relative to a shifted index origin?
• Do you know the exact SQNR improvement obtained per additional ADC bit?

---

Related Notes: 4.02 Spectral Replications & Aliasing Dynamics | 4.03 Ideal, Natural & Flat-Top Sampling Techniques | 4.04 Multi-Rate Signal Processing: Decimation & Interpolation

---

4.01 The Shannon-Nyquist Sampling Theorem

Core Idea

The Shannon-Nyquist Sampling Theorem is the mathematical bridge that allows us to convert a continuous-time analog signal into a discrete-time sequence without any loss of information. It asserts that if a band-limited continuous-time signal has no frequency components higher than $f_{\text{max}}$ Hz, it can be uniquely and perfectly reconstructed if the sampling frequency $f_s$ satisfies $f_s\ge 2f_{\text{max}}$.

---

1. Mathematical Definitions and Concepts

To understand sampling, we must mathematically define the transition from continuous-time $t$ to discrete-time indices $n$ under a uniform sampling period $T_s$ {the time interval between successive samples}.

1.1 The Sampling Equations

Let $x_a(t)$ represent a continuous-time analog input signal. Uniform sampling is modeled as taking snapshots of the signal every $T_s$ seconds: $x[n]=x_a(n{T}_s)\text{for }n\in \mathbb{Z}$

Here:

• $T_s$ = Sampling Period [seconds]
• $f_s=\frac{1}{T_s}$ = Sampling Frequency [Hz or samples/second]
• ${\Omega }_s=\frac{2\pi }{T_s}=2\pi f_s$ = Sampling Angular Frequency [rad/s]

1.2 The Nyquist Conditions

1. Nyquist Rate ($f_{\text{Nyquist}}$): The absolute minimum theoretical sampling rate that avoids overlapping spectral replicas. $f_{\text{Nyquist}}=2f_{\text{max}}$
2. Nyquist Interval ($T_{\text{Nyquist}}$): The maximum allowable time spacing between adjacent samples: $T_{\text{Nyquist}}=\frac{1}{f_{\text{Nyquist}}}=\frac{1}{2f_{\text{max}}}$

---

2. Rigorous Proof of the Sampling Theorem [PYQ 2019, 2017]

The proof of the sampling theorem relies on modeling the physical sampling switch as a continuous-time multiplication of the analog signal $x_a(t)$ by a periodic impulse train $p(t)$:

2.1 Time-Domain Formulation

Let $p(t)$ be an impulse train of unit area spaced by $T_s$: $p(t)={\sum }_{n=-\infty }^{\infty }\delta (t-n{T}_s)$

The sampled signal $x_s(t)$ is the product: $x_s(t)=x_a(t)\cdot p(t)=x_a(t){\sum }_{n=-\infty }^{\infty }\delta (t-n{T}_s)$

Using the sampling property of the impulse ($x(t)\delta (t-t_0)=x(t_0)\delta (t-t_0)$): $x_s(t)={\sum }_{n=-\infty }^{\infty }{x}_a(n{T}_s)\delta (t-n{T}_s)$

2.2 Frequency-Domain Formulation

Since multiplication in the time domain corresponds to convolution in the frequency domain (scaled by $1/2\pi$): $X_s(j\Omega )=\frac{1}{2\pi }[X_a(j\Omega )\ast P(j\Omega )]$

Because $p(t)$ is periodic with period $T_s$, we can express it as a complex exponential Fourier Series: $p(t)={\sum }_{k=-\infty }^{\infty }{c}_k{e}^{jk{\Omega }_{s}t}\text{where }{\Omega }_s=\frac{2\pi }{T_s}$

The Fourier coefficients $c_k$ are: $c_k=\frac{1}{T_s}{\int }_{-T_s/2}^{T_s/2}p(t)e^{-jk{\Omega }_{s}t}dt=\frac{1}{T_s}{\int }_{-T_s/2}^{T_s/2}\delta (t)e^{-jk{\Omega }_{s}t}dt=\frac{1}{T_s}$

Thus: $p(t)=\frac{1}{T_s}{\sum }_{k=-\infty }^{\infty }{e}^{jk{\Omega }_{s}t}$

Taking the Fourier transform of this exponential series yields a train of impulses in the frequency domain: $P(j\Omega )=\frac{2\pi }{T_s}{\sum }_{k=-\infty }^{\infty }\delta (\Omega -k{\Omega }_s)$

Now, convolve $X_a(j\Omega )$ with $P(j\Omega )$: $X_s(j\Omega )=\frac{1}{2\pi }{\int }_{-\infty }^{\infty }{X}_a(j\theta )P(j(\Omega -\theta ))d\theta$ $X_s(j\Omega )=\frac{1}{2\pi }\cdot \frac{2\pi }{T_s}{\sum }_{k=-\infty }^{\infty }{X}_a(j\Omega )\ast \delta (\Omega -k{\Omega }_s)$

Since convolving any function with a shifted impulse simply shifts the function itself: $X_s(j\Omega )=\frac{1}{T_s}{\sum }_{k=-\infty }^{\infty }{X}_a(j(\Omega -k{\Omega }_s))$

Physical Meaning of the Result

This equation proves that the spectrum of the sampled signal $X_s(j\Omega )$ consists of an infinite series of replicated copies of the original analog spectrum $X_a(j\Omega )$, spaced at integer multiples of the sampling frequency ${\Omega }_s$ and scaled in amplitude by $\frac{1}{T_s}$.

---

3. High-Yield Solved Numericals

3.1 Example 1: Multi-Tone Composite Waveform [PYQ 2016, 2015]

Question: Calculate the Nyquist Rate and Nyquist Interval for the analog signal: $x(t)=5\cos (200\pi t)+10\sin (500\pi t)$

Step-by-Step Solution:

1. Extract individual frequencies:

   • For Term 1 ($5\cos (200\pi t)$): ${\omega }_1=200\pi \text{ rad/s}⟹f_1=\frac{{\omega }_1}{2\pi }=100\text{ Hz}$
   • For Term 2 ($10\sin (500\pi t)$): ${\omega }_2=500\pi \text{ rad/s}⟹f_2=\frac{{\omega }_2}{2\pi }=250\text{ Hz}$

2. Identify maximum frequency ($f_{\text{max}}$): $f_{\text{max}}=\max (f_1,f_2)=250\text{ Hz}$

3. Evaluate Nyquist parameters:

   • Nyquist Rate: $f_{\text{Nyquist}}=2f_{\text{max}}=2\times 250\text{ Hz}=500\text{ Hz}$
   • Nyquist Interval: $T_{\text{Nyquist}}=\frac{1}{f_{\text{Nyquist}}}=\frac{1}{500}=0.002\text{ seconds}(2\text{ ms})$

---

3.2 Example 2: Sinc and Sinc-Squared Functions

Question: Calculate the Nyquist Rate for the signal: $x(t)={\text{sinc}}^2(100t)$

Step-by-Step Solution:

To solve this, we must check both standard engineering definitions of the Sinc function since different instructors use different standards.

• Interpretation A: Normalized Sinc (Rao Standard: $\text{sinc}(u)=\frac{\sin (\pi u)}{\pi u}$)

  1. Express the base term: $\text{sinc}(100t)=\frac{\sin (100\pi t)}{100\pi t}$
  2. Find the base maximum frequency: ${\omega }_{\text{base}}=100\pi \text{ rad/s}⟹f_{\text{base}}=50\text{ Hz}$
  3. Because squaring in time corresponds to frequency-domain convolution, the band limit doubles: $f_{\text{max}}=2\times f_{\text{base}}=2\times 50=100\text{ Hz}$
  4. Calculate the Nyquist Rate: $f_{\text{Nyquist}}=2f_{\text{max}}=2\times 100\text{ Hz}=200\text{ Hz}$

• Interpretation B: Unnormalized Sinc ($\text{sinc}(u)=\frac{\sin (u)}{u}$)

  1. Express the base term: $\text{sinc}(100t)=\frac{\sin (100t)}{100t}$
  2. Find base maximum frequency: ${\omega }_{\text{base}}=100\text{ rad/s}⟹f_{\text{base}}=\frac{50}{\pi }\text{ Hz}$
  3. Doubling the band limit due to multiplication: ${\omega }_{\text{max}}=2\times 100=200\text{ rad/s}⟹f_{\text{max}}=\frac{100}{\pi }\text{ Hz}$
  4. Calculate the Nyquist Rate: $f_{\text{Nyquist}}=2f_{\text{max}}=\frac{200}{\pi }\approx 63.66\text{ Hz}$

---

4. Common Mistakes That Cost Marks

The Sinc frequency factor error

When squaring a signal (like $x(t)=g^2(t)$ or convolving in frequency), the highest frequency component doubles ($f_{\text{max}}=2f_{\text{base}}$). Many students forget to double the base frequency and evaluate the Nyquist rate incorrectly, losing major marks.

The Angular vs. Cyclic frequency trap ($\Omega$ vs. $f$)

Always verify if the question asks for the Nyquist rate in Hz or rad/s.

• Nyquist rate in Hz: $f_s=2f_{\text{max}}$
• Nyquist rate in rad/s: ${\Omega }_s=2{\Omega }_{\text{max}}$ Confusing the two factors in the proof or numericals leads to scaling errors by $2\pi$.

---

5. PYQ Bank — Verbatim Questions & Answer Plans

5.1 PYQ 2019/2017 [10 Marks]

Question: State and prove the Shannon-Nyquist sampling theorem.

• Answer Plan:
  1. State the theorem formally (Section 1).
  2. Model uniform impulse train sampling in the time domain (Section 2.1).
  3. Compute the Fourier series representation of the impulse train $p(t)$ (Section 2.2).
  4. Carry out the frequency-domain convolution integration to derive $X_s(j\Omega )=\frac{1}{T_s}\sum X_a(j(\Omega -k{\Omega }_s))$ (Section 2.2).
  5. Explain the reconstruction criteria (over-sampling, Nyquist, under-sampling).

5.2 PYQ 2016/2015 [5 Marks]

Question: Determine the Nyquist Rate and Nyquist Interval of the signal $x(t)=5\cos (200\pi t)+10\sin (500\pi t)$.

• Answer Plan: Follow the step-by-step extraction and numerical solution laid out in Section 3.1 to secure all 5 marks.

---

6. Self-Check Before Moving On

• Can you state the Shannon-Nyquist theorem verbatim without skipping the “band-limited” constraint?
• Can you derive the Fourier Transform of an impulse train $P(j\Omega )=\frac{2\pi }{T_s}\sum \delta (\Omega -k{\Omega }_s)$?
• Can you prove that multiplication by $p(t)$ in time replicates the spectrum in frequency?
• Do you know how to calculate the maximum frequency for composite signals and sinc-squared signals?

---

Source: (k.Deergha Rao) signals and systems.pdf (Chapter 6 & 7), 1.05 The Sampling Theorem Bridge & Multi-Rate Processing.md.

---

4.01 The Shannon-Nyquist Sampling Theorem | 4.03 Ideal, Natural & Flat-Top Sampling Techniques

---

4.02 Spectral Replications & Aliasing Dynamics

Core Idea

Uniform time-domain sampling mathematically translates to a periodic replication of the signal’s continuous spectrum in the frequency domain at integer multiples of the sampling frequency ${\Omega }_s$. The relationship between the sampling rate ${\Omega }_s$ and the maximum signal frequency ${\Omega }_m$ determines whether these replicated spectral bands remain isolated or overlap. When under-sampled, overlapping bands cause Aliasing, a destructive, irreversible frequency-folding distortion that can only be prevented prior to sampling using an analog Anti-Aliasing Filter.

---

1. The Physics of Spectral Replications

To understand what happens in the frequency domain when we sample an analog signal, we recall the continuous-time impulse-train sampled signal $x_s(t)$ from Note 4.01: $x_s(t)=x(t)\cdot p(t)={\sum }_{n=-\infty }^{\infty }x(n{T}_s)\delta (t-n{T}_s)$

Since the sampling function $p(t)=\sum \delta (t-n{T}_s)$ is periodic with period $T_s$, its Fourier transform $P(j\Omega )$ is a discrete impulse train in the frequency domain spaced by the angular sampling frequency ${\Omega }_s=\frac{2\pi }{T_s}$: $P(j\Omega )=\frac{2\pi }{T_s}{\sum }_{k=-\infty }^{\infty }\delta (\Omega -k{\Omega }_s)$

Applying the frequency convolution property {multiplication in time is convolution in frequency}, we derive the spectrum of the sampled signal $X_s(j\Omega )$: $X_s(j\Omega )=\mathcal{F}\{x(t)\cdot p(t)\}=\frac{1}{2\pi }\left[X(j\Omega )\ast P(j\Omega )\right]$

$X_s(j\Omega )=\frac{1}{2\pi }\left[X(j\Omega )\ast \left(\frac{2\pi }{T_s}{\sum }_{k=-\infty }^{\infty }\delta (\Omega -k{\Omega }_s)\right)\right]$

$X_s(j\Omega )=\frac{1}{T_s}{\sum }_{k=-\infty }^{\infty }X\left(j\left(\Omega -k{\Omega }_s\right)\right)$

This elegant formula reveals that the spectrum of a sampled signal consists of the original spectrum $X(j\Omega )$ scaled by $\frac{1}{T_s}$ and repeated infinitely at every integer multiple of the sampling frequency ${\Omega }_s$.

---

2. The Three Sampling Conditions

The boundary condition that determines the separation of these replicated spectral bands is established by comparing the angular sampling frequency ${\Omega }_s$ with twice the maximum frequency component of the analog signal ${\Omega }_m$.

2.1 Over-Sampling (${\Omega }_s>2{\Omega }_m$)

When the sampling frequency is strictly greater than twice the highest frequency component, the replicated spectral copies are spaced far apart.

• Guard Band: A clear frequency gap {buffer band} exists between adjacent replicas, spanning the interval: $\Delta \Omega ={\Omega }_s-2{\Omega }_m>0$
• Reconstruction Filter: Because of this gap, a practical, low-cost analog low-pass reconstruction filter with a gradual transition band can easily isolate the baseband spectrum $X(j\Omega )$ without introducing phase or amplitude distortion.

[GRAPH: Replicated frequency spectra under the over-sampling condition (\Omega_s > 2\Omega_m). Plot shows a series of triangular-shaped spectral bands of height 1/T_s centered at \Omega = 0, \pm\Omega_s, \pm2\Omega_s. The baseband triangle is bounded by -\Omega_m and \Omega_m. Clear gaps of width \Omega_s - 2\Omega_m representing guard bands are highlighted between adjacent triangles - source: K. Deergha Rao Fig 7.14]

---

2.2 Nyquist-Rate Sampling (${\Omega }_s=2{\Omega }_m$)

This is the boundary condition where the sampling rate is exactly equal to twice the maximum frequency of the signal.

• Edge-Touching: The adjacent spectral replicas touch exactly at their edges with no frequency gap: ${\Omega }_s-2{\Omega }_m=0⟹{\Omega }_s-{\Omega }_m={\Omega }_m$
• The “Brick-Wall” Necessity: To reconstruct the original signal perfectly, we require an ideal low-pass filter with an infinitely sharp, vertical cutoff at ${\Omega }_c={\Omega }_m$: $H_{LP}(j\Omega )=\{\begin{array}{ll}{T}_s,& |\Omega |\le {\Omega }_m\\ 0,& |\Omega |>{\Omega }_m\end{array}$ Physical Constraint: Such a filter is mathematically causal but physically non-realizable because its impulse response is a double-sided sinc function extending from $t=-\infty$ to $t=\infty$, requiring infinite time delay (anticipatory memory) to implement.

[GRAPH: Replicated frequency spectra under the Nyquist-rate sampling condition (\Omega_s = 2\Omega_m). Plot shows triangular spectral bands of height 1/T_s centered at 0 and \pm\Omega_s touching exactly at the boundary frequency \Omega = \Omega_m, showing zero gap between the adjacent bands - source: K. Deergha Rao Fig 7.12]

---

2.3 Under-Sampling (${\Omega }_s<2{\Omega }_m$)

When the sampling frequency is less than twice the maximum frequency component, the adjacent spectral replicas overlap in the frequency domain.

• Fold-Over Region: The high-frequency tails of one replica overlap with the low-frequency portion of the adjacent replica.
• Destructive Interference: Within this overlapping band, the spectra sum together algebraically. The original high frequencies fold back into the lower frequencies, making it physically impossible to reconstruct the original waveform. This spectral overlap region is called the fold-over region or aliasing zone.

[GRAPH: Replicated frequency spectra under the under-sampling condition (\Omega_s < 2\Omega_m) demonstrating aliasing. Plot shows overlapping triangular spectral bands centered at 0, \pm\Omega_s, \pm2\Omega_s. The overlapping regions between adjacent triangles are heavily shaded and labeled as 'Fold-over / Aliasing region' to show the corruption of low-frequency data by high-frequency replicas - source: K. Deergha Rao Fig 7.13]

---

3. Aliasing Distortion & The Frequency-Folding Mechanism

3.1 Mathematical Definition of Aliasing

Aliasing is the frequency-folding distortion that occurs when a continuous-time signal is sampled at a rate below its Nyquist limit. High-frequency components “fold” around the folding frequency $f_{\text{fold}}=\frac{f_s}{2}$ {also known as the folding boundary or half-sampling limit} and reappear as false, lower-frequency components in the reconstructed baseband.

3.2 The Symmetrical Folding Formula

If a continuous-time sinusoidal signal has a true frequency $f_0$, and it is sampled at a rate $f_s$, the apparent reconstructed frequency $f_{\text{alias}}$ in the baseband $|f|\le \frac{f_s}{2}$ can be calculated using the symmetrical folding formula: $f_{\text{alias}}=\left|f_0-k{f}_s\right|\text{where }k\in \mathbb{Z}\text{ such that }{f}_{\text{alias}}\le \frac{f_s}{2}$

---

3.3 High-Yield Algebraic Aliasing Proof

To see this folding mechanism algebraically, let us sample a high-frequency cosine wave $x(t)=\cos (2\pi f_{0}t)$ at a sampling rate $f_s=\frac{1}{T_s}$ where $f_s<2f_0$.

The discrete-time sequence of samples is: $x[n]=x(n{T}_s)=\cos \left(2\pi f_{0}n{T}_s\right)=\cos \left(2\pi \frac{f_0}{f_s}n\right)$

Let the true frequency be represented as an integer multiple of the sampling frequency plus an offset: $f_0=k{f}_s\pm f_{\text{alias}}⟹\frac{f_0}{f_s}=k\pm \frac{f_{\text{alias}}}{f_s}$

Substitute this ratio back into our discrete-time expression: $x[n]=\cos \left(2\pi \left(k\pm \frac{f_{\text{alias}}}{f_s}\right)n\right)$ $x[n]=\cos \left(2\pi kn\pm 2\pi \frac{f_{\text{alias}}}{f_s}n\right)$

Using the cosine angle addition identity $\cos (\theta \pm 2\pi kn)=\cos (\theta )$ for any integer $k$ and $n$: $x[n]=\cos \left(2\pi \frac{f_{\text{alias}}}{f_s}n\right)=\cos \left(2\pi f_{\text{alias}}n{T}_s\right)$

This algebraic identity proves that the discrete sample values of the high-frequency sinusoid $f_0$ are mathematically identical to the sample values of a lower-frequency sinusoid $f_{\text{alias}}$. During reconstruction, the DAC cannot distinguish between these two signals and will always output the lower frequency $f_{\text{alias}}$, resulting in irreversible data loss.

---

3.4 Grounded Numerical Example

Problem: A continuous-time signal $x(t)=\cos (200\pi t)$ is sampled at a rate $f_s=150\text{ Hz}$. Find the true frequency of the signal, verify if aliasing occurs, and determine the apparent frequency of the reconstructed signal.

Step 1: Extract the True Frequency

From the expression $x(t)=\cos (2\pi f_{0}t)=\cos (200\pi t)$: $2\pi f_0=200\pi ⟹f_0=100\text{ Hz}$

Step 2: Test the Nyquist Criteria

The minimum theoretical sampling rate to avoid aliasing is the Nyquist Rate: $f_{\text{Nyquist}}=2f_0=2(100)=200\text{ Hz}$

Since our actual sampling frequency $f_s=150\text{ Hz}<200\text{ Hz}$, aliasing will definitely occur.

Step 3: Apply the Folding Formula

We seek an integer $k$ such that the aliased frequency falls below the half-sampling folding limit $f_{\text{fold}}=\frac{150}{2}=75\text{ Hz}$: For $k=1$: $f_{\text{alias}}=|f_0-f_s|=|100-150|=50\text{ Hz}$

Since $50\text{ Hz}\le 75\text{ Hz}$, the apparent reconstructed frequency is: $f_{\text{alias}}=50\text{ Hz}$

Step 4: Verify Graphically and Algebraically

$x[n]=\cos \left(2\pi \frac{100}{150}n\right)=\cos \left(\frac{4\pi }{3}n\right)$ $\cos \left(\frac{4\pi }{3}n\right)=\cos \left(2\pi n-\frac{2\pi }{3}n\right)=\cos \left(-\frac{2\pi }{3}n\right)=\cos \left(\frac{2\pi }{3}n\right)$ $\cos \left(\frac{2\pi }{3}n\right)=\cos \left(2\pi \frac{50}{150}n\right)=\cos \left(2\pi f_{\text{alias}}n{T}_s\right)$

The $100\text{ Hz}$ wave has folded around the $75\text{ Hz}$ folding frequency and is reconstructed as a false $50\text{ Hz}$ sine wave.

---

4. The Anti-Aliasing Filter (AAF)

4.1 The Fundamental Post-Sampling Rule

The Post-Sampling Filtering Fallacy

Once a signal is sampled and aliasing occurs, the overlapping spectral components merge completely. At this point, the original low frequencies and folded-back high frequencies are mathematically summed into a single amplitude value at each frequency index. No digital filter or software algorithm can ever separate them. Aliasing is an irreversible physical process.

4.2 The Solution: Analog Pre-Filtering

To prevent aliasing, we must remove all frequency components above the half-sampling limit before the signal reaches the sampler or Analog-to-Digital Converter (ADC). This is achieved using an analog Anti-Aliasing Filter (AAF).

                 +-----------------------+      +---------+
Analog Input --->| Anti-Aliasing Filter  |----->| Sampler |---> Sampled Output
   x(t)          |  (Analog Low-Pass)    |      |  (ADC)  |        x[n]
                 +-----------------------+      +---------+
                         Cutoff:
                    f_c <= f_s / 2

4.3 Structural Design Specifications of an AAF

1. Filter Classification: Active or passive analog low-pass filter (LPF).
2. Cutoff Frequency ($f_c$): Designed to match the folding frequency boundary: $f_c\le \frac{f_s}{2}$
3. The Over-Sampling Relaxer: In practical engineering, designing a sharp analog filter with a very narrow transition band is incredibly difficult and expensive. Therefore, we deliberately over-sample the signal ($f_s\gg 2f_c$) to create a wide guard band. This wide guard band allows us to use a cheap, low-order analog anti-aliasing filter while still completely preventing aliasing.

---

5. Common Mistakes That Cost Marks

The Post-ADC Digital Filtering Trap

In exam papers, examiners often ask if a digital filter implemented in a DSP can eliminate aliasing. Always answer NO. Explain that once the analog signal passes through the ADC sampler, the overlapping spectral copies are already summed. A digital filter cannot distinguish between a true $50\text{ Hz}$ signal and a $100\text{ Hz}$ signal aliased to $50\text{ Hz}$.

Incorrect Folding Calculations

When calculating folding frequencies, do not just subtract the frequency as $f_{\text{alias}}=f_s-f_0$. Always take the absolute value $|f_0-k{f}_s|$. Forgetting the absolute value yields a negative frequency, which is mathematically correct under exponential notation but physically incorrect for real-valued signals without a corresponding phase-sign inversion.

---

6. PYQ Bank — Verbatim Questions & Answer Plans

6.1 PYQ 2021 Question 1c / 2019 Question 2b [06 Marks]

Question: Explain the aliasing effect and describe the means/filters used to avoid it.

• Answer Plan:
  1. Define Aliasing mathematically and conceptually (Section 3.1).
  2. State the condition for under-sampling ($f_s<2f_{\text{max}}$) which leads to spectral overlap (Section 2.3).
  3. Draw the overlapping triangular spectrum diagram (Section 2.3).
  4. Explain the solution: Placing an analog Anti-Aliasing Filter before the sampler with $f_c\le f_s/2$ (Section 4.2 & 4.3).

6.2 PYQ 2018 Question 2b [06 Marks]

Question: Draw the criteria and frequency spectrum in case of over-sampling, under-sampling & Nyquist rate sampling.

• Answer Plan:
  1. Divide your answer into three distinct sub-sections matching Section 2.
  2. For Over-sampling, state ${\Omega }_s>2{\Omega }_m$ and draw the spaced-out triangular replicas with highlighted guard bands.
  3. For Nyquist-rate sampling, state ${\Omega }_s=2{\Omega }_m$ and draw the replicas touching exactly.
  4. For Under-sampling, state ${\Omega }_s<2{\Omega }_m$ and draw overlapping replicas with shaded fold-over aliasing regions.

---

7. Self-Check Before Moving On

• Can you derive the periodic replicated spectrum equation $X_s(j\Omega )$ using time-convolution?
• Do you know the exact frequency criteria for over-sampling, under-sampling, and Nyquist rate sampling?
• Can you calculate the apparent folding frequency of any under-sampled sinusoid using $f_{\text{alias}}=|f_0-k{f}_s|$?
• Can you explain why a digital filter cannot remove aliasing after a signal has been sampled?
• Do you understand the role, physical placement, and cutoff constraints of an analog Anti-Aliasing Filter?

---

Source: (k.Deergha Rao) signals and systems.pdf (Chapter 7.4), ECE 2107 - Signals and Systems.md, 1.05 The Sampling Theorem Bridge & Multi-Rate Processing.md, Azmat Sir Class Slides.

---

4.02 Spectral Replications & Aliasing Dynamics | 4.04 Multi-Rate Signal Processing: Decimation & Interpolation

---

4.03 Ideal, Natural & Flat-Top Sampling Techniques

Core Idea

Continuous-time analog signals cannot be directly processed by digital systems without conversion. The physical interface is the Analog-to-Digital Converter (ADC), where the continuous time axis is discretized. Depending on the physical implementation of the sampling switch, this discretization process is classified into Ideal (Impulse), Natural (Chopping), and Flat-Top (Sample-and-Hold) sampling. While ideal and natural sampling preserve the exact, undistorted frequency spectrum of the analog signal within each replica, practical flat-top sampling introduces high-frequency roll-off distortion known as the Aperture Effect, requiring post-filtering equalization.

---

1. The Physical Analog-to-Digital Conversion (ADC) Pipeline

Before looking at the mathematics of different sampling techniques, we must understand where sampling fits in the physical hardware pipeline of a modern receiver.

                                  THE ADC PIPELINE
                                  
   Analog        +------------+   Filtered    +-------------+   Flat-top    +-----------+   Quantized   +---------+   Digital
   Signal        |    Anti-   |    Analog     |  Sample &   |   Sampled     |           |    Samples    |         |    Word
  x_a(t) ------->|  Aliasing  |-------------->|    Hold     |-------------->| Quantizer |-------------->| Encoder |------------->
                 |  LPF (f_c) |   x_filt(t)   |  (Flat-top) |   x_ft(t)     |           |    x_q(n)     |         |   b-bits
                 +------------+               +-------------+               +-----------+               +---------+
                                                     ^
                                                     | Trigger Clock
                                                     | f_s >= 2 f_max

[DIAGRAM: Complete Block Diagram of the Analog-to-Digital Conversion (ADC) Process, showing the cascade of anti-aliasing filtering, sample-and-hold discretization, quantization, and binary encoding - source: Textbook Fig 6.2]

1.1 Role of Individual Blocks

1. Anti-Aliasing Filter: An analog low-pass filter with cutoff frequency $f_c\le f_s/2$. It strips away high-frequency noise and out-of-band components that would otherwise fold back into the baseband and cause irreversible aliasing distortion during the sampling stage.
2. Sample-and-Hold (S&H) Circuit: The physical hardware executing flat-top sampling. It tracks the analog input voltage during a very short window (sample) and holds it strictly constant (hold) across the remainder of the sampling period to give the quantizer stable voltage levels to convert.
3. Quantizer: Converts the continuously variable amplitude of the flat-topped staircase samples into discrete voltage levels chosen from a finite pool of $2^b$ levels, where $b$ is the ADC bit resolution.
4. Encoder: Maps each quantized voltage level to a unique $b$-bit binary word, emitting a stream of digital data ready for processor manipulation.

---

2. Ideal (Impulse Train) Sampling

Ideal Sampling—also called Impulse Sampling—represents the theoretical mathematical baseline where the analog signal $x_a(t)$ is multiplied by an infinite train of zero-width Dirac delta impulses.

                     IDEAL SAMPLING WAVEFORM CHARACTERISTICS
                     
        x_a(t) (Continuous Waveform)            p(t) (Periodic Impulse Train)
             
            _--""--_                                 |   |   |   |   |   |   |
          _-"        "-_                             |   |   |   |   |   |   |
         /              \                            |   |   |   |   |   |   |
     ------------------------- t                 ------------------------- t
                                                     0   T_s 2T_s
                                                     
                                 x_s(t) (Ideally Sampled Impulses)
                                             |
                                             |   |
                                             |   |   |
                                                 |   |   |
                                 ------------------------- t

[GRAPH: Waveform representation of Ideal Sampling: Multiplying a smooth continuous-time analog wave x_a(t) by an infinite train of zero-width Dirac impulses p(t) yields an impulse train x_s(t) whose impulse areas match the sample values - source: Textbook Fig 6.1]

2.1 Mathematical Formulation

In the time domain, we multiply $x_a(t)$ by the periodic impulse train $p(t)$: x_s(t) = x_a(t) \cdot p(t) \tag{1}

The periodic impulse train is defined as: p(t) = \sum_{n=-\infty}^{\infty} \delta(t - nT_s) \tag{2}

Substituting Eq. (2) into Eq. (1) and applying the impulse sifting property ($w(t)\delta (t-t_0)=w(t_0)\delta (t-t_0)$) yields: x_s(t) = \sum_{n=-\infty}^{\infty} x_a(nT_s) \delta(t - nT_s) \tag{3}

2.2 Frequency-Domain Representation

Taking the Fourier transform of Eq. (3) maps this multiplication to convolution in the frequency domain, yielding replicated copies of the analog spectrum spaced at integer multiples of the sampling frequency ${\Omega }_s=2\pi /T_s$: X_s(j\Omega) = \frac{1}{T_s} \sum_{k=-\infty}^{\infty} X_a(j(\Omega - k\Omega_s)) \tag{4}

Conceptual Checkpoint

Because each spectral replica in $X_s(j\Omega )$ is scaled by the constant factor $1/T_s$, the shape of the original spectrum $X_a(j\Omega )$ is completely undistorted inside each replicated frequency band. However, because Dirac impulses have infinite height and zero width, this technique is a mathematical abstraction that is physically impossible to realize in hardware.

---

3. Natural Sampling (Chopping)

Natural Sampling is a physically realizable technique where the continuous-time signal $x_a(t)$ is electronically switched (or “chopped”) by a train of narrow, rectangular pulses of finite width $\tau$ and period $T_s$.

                    NATURAL SAMPLING WAVEFORM CHARACTERISTICS
                    
        p(t) (Periodic Rectangular Pulses)        x_ns(t) (Naturally Sampled Wave)
             ___     ___     ___                       _--""--_ 
            |   |   |   |   |   |                    _|   |  | |  |_  
            |   |   |   |   |   |                   | |   |  | |  | | 
            |   |   |   |   |   |                   | |   |  | |  | | 
        ------------------------- t                 ------------------------- t
            <--> tau                                  (Tops curve along wave contour)
            <-------> T_s

[GRAPH: Waveform representation of Natural Sampling: Multiplying an analog wave by a periodic rectangular pulse train of pulse-width tau and period T_s yields a pulsed wave x_ns(t) where the top of each rectangular slice preserves the natural curve of the original analog wave - source: Textbook Fig 6.1]

3.1 Mathematical Formulation

The naturally sampled signal is: x_{ns}(t) = x_a(t) \cdot p_{rect}(t) \tag{5}

where $p_{rect}(t)$ is a periodic rectangular pulse train of amplitude $1$, period $T_s$, and pulse width $\tau$: p_{rect}(t) = \sum_{n=-\infty}^{\infty} \text{rect}\left(\frac{t - nT_s}{\tau}\right) \tag{6}

Because $p_{rect}(t)$ is a periodic function of time, it can be represented by its exponential Fourier series: p_{rect}(t) = \sum_{n=-\infty}^{\infty} C_n e^{j n \Omega_s t} \tag{7}

where the Fourier coefficients $C_n$ are evaluated over one period $T_s$: C_n = \frac{1}{T_s} \int_{-T_s/2}^{T_s/2} p_{rect}(t) e^{-j n \Omega_s t} \, dt = \frac{1}{T_s} \int_{-\tau/2}^{\tau/2} (1) e^{-j n \Omega_s t} \, dt \tag{8} C_n = \frac{1}{T_s} \left[ \frac{e^{-j n \Omega_s t}}{-j n \Omega_s} \right]_{-\tau/2}^{\tau/2} = \frac{e^{j n \Omega_s \tau/2} - e^{-j n \Omega_s \tau/2}}{j n \Omega_s T_s} \tag{9} C_n = \frac{2 \sin(n \Omega_s \tau / 2)}{n \Omega_s T_s} = \frac{\tau}{T_s} \frac{\sin(n \Omega_s \tau / 2)}{n \Omega_s \tau / 2} = \frac{\tau}{T_s} \text{sinc}\left(\frac{n \Omega_s \tau}{2\pi}\right) \tag{10}

3.2 Frequency-Domain Representation

Substituting the Fourier series representation of Eq. (7) back into Eq. (5) yields: x_{ns}(t) = x_a(t) \cdot \sum_{n=-\infty}^{\infty} C_n e^{j n \Omega_s t} = \sum_{n=-\infty}^{\infty} C_n x_a(t) e^{j n \Omega_s t} \tag{11}

Applying the Frequency Shifting (Modulation) Property of the Fourier transform ($\mathcal{F}\{f(t)e^{j{\omega }_{0}t}\}=F(j(\omega -{\omega }_0))$) to Eq. (11) gives: X_{ns}(j\Omega) = \sum_{n=-\infty}^{\infty} C_n X_a(j(\Omega - n\Omega_s)) \tag{12}

Substituting the explicit expression for $C_n$ from Eq. (10) into Eq. (12) yields the final spectral representation of naturally sampled signals: X_{ns}(j\Omega) = \frac{\tau}{T_s} \sum_{n=-\infty}^{\infty} \text{sinc}\left(\frac{n \Omega_s \tau}{2\pi}\right) X_a(j(\Omega - n\Omega_s)) \tag{13}

                       SPECTRUM OF NATURALLY SAMPLED SIGNAL
                       
                     Sinc Envelope [ sinc( \Omega \tau / 2\pi ) ]
                . . . . - - - - - - . . . .
              .-                           -.
             /                               \
            |        |           |            |
            |     |  |  |     |  |  |     |  |  |
         ---|-----|--|--|-----|--|--|-----|--|--|--- \Omega
               -\Omega_s   0   \Omega_s   2\Omega_s

[GRAPH: Frequency spectrum of Natural Sampling: The analog spectrum repeats at integer multiples of \Omega_s. Each replica is scaled as a whole by the constant coefficient C_n. This results in no internal frequency distortion within individual replicas - source: Textbook Ch 6]

3.3 Physical Interpretation

1. The spectrum consists of replicated copies of the original spectrum $X_a(j\Omega )$ centered at integer multiples of the sampling frequency ${\Omega }_s$.
2. Each individual spectral replica is scaled as a whole by the constant value $C_n=\frac{\tau }{T_s}\text{sinc}(\frac{n{\Omega }_s\tau }{2\pi })$.
3. Because the scaling factor $C_n$ is constant for a given harmonic index $n$, the baseband spectrum ($n=0$) and all individual replicas are completely undistorted in shape. They are merely scaled in overall amplitude.

---

4. Flat-Top Sampling (Sample-and-Hold)

In modern microelectronics, natural sampling is difficult to execute because quantizers require a completely stable, static voltage level during conversion. Instead, Flat-Top Sampling is used, where the analog signal’s amplitude is sampled at $t=n{T}_s$ and held constant for the duration $\tau$ of the sampling pulse, producing a staircase-like waveform.

                   FLAT-TOP SAMPLING WAVEFORM CHARACTERISTICS
                    
        x_a(t) (Continuous Waveform)             x_ft(t) (Flat-Top Staircase Wave)
             _--""--_                                 _----_  
           _-"        "-_                            |      | _----_ 
          /              \                           |      ||      |
     ------------------------- t                 ------------------------- t
                                                     <--> tau

[GRAPH: Waveform representation of Flat-Top Sampling: The analog amplitude is sampled at discrete instants and held strictly flat for duration tau, producing a staircase waveform - source: Textbook Fig 6.2]

4.1 Mathematical Modeling as a Convolution Cascade

To model flat-top sampling mathematically, we can think of it as a two-stage cascade:

1. First, the analog signal $x_a(t)$ is ideally sampled with a zero-width impulse train to generate $x_s(t)$.
2. Second, this impulse train $x_s(t)$ is convolved with a single rectangular pulse $h(t)$ of width $\tau$ and amplitude $1$ to “stretch” each impulse into a flat-topped block of width $\tau$.

               MATHEMATICAL DECOMPOSITION OF FLAT-TOP SAMPLING
               
    Analog       +-----------------+  Ideal Impulses  +-----------------+  Flat-top
    Signal ----->| Ideal Sampler   |----------------->| Rectangular     |---> Staircase
    x_a(t)       |  Clock: T_s     |     x_s(t)       | Filter h(t)     |     x_ft(t)
                 +-----------------+                  +-----------------+

Let the single rectangular stretching pulse $h(t)$ be: h(t) = \text{rect}\left(\frac{t - \tau/2}{\tau}\right) = \begin{cases} 1, & 0 \le t < \tau \\ 0, & \text{otherwise} \end{cases} \tag{14}

Using this pulse shape, the flat-top sampled signal $x_{ft}(t)$ is defined as: x_{ft}(t) = x_s(t) * h(t) \tag{15}

Substituting Eq. (3) for $x_s(t)$ into Eq. (15) yields: x_{ft}(t) = \left[ \sum_{n=-\infty}^{\infty} x_a(nT_s) \delta(t - nT_s) \right] * h(t) \tag{16}

Since convolution is a linear operator, we distribute it inside the summation: x_{ft}(t) = \sum_{n=-\infty}^{\infty} x_a(nT_s) \left[ \delta(t - nT_s) * h(t) \right] \tag{17}

By the impulse shifting property of convolution ($f(t)\ast \delta (t-t_0)=f(t-t_0)$), we obtain the final time-domain expression for a flat-top sampled signal: x_{ft}(t) = \sum_{n=-\infty}^{\infty} x_a(nT_s) h(t - nT_s) \tag{18}

4.2 Frequency-Domain Representation

Applying the convolution theorem ($\mathcal{F}\{f(t)\ast g(t)\}=F(j\Omega )\cdot G(j\Omega )$) to Eq. (15) gives: X_{ft}(j\Omega) = X_s(j\Omega) \cdot H(j\Omega) \tag{19}

We first evaluate the Fourier transform of the rectangular stretching pulse $h(t)$ defined in Eq. (14) using the time-shifting property: H(j\Omega) = \mathcal{F}\left\{\text{rect}\left(\frac{t - \tau/2}{\tau}\right)\right\} = \left[ \tau \text{sinc}\left(\frac{\Omega \tau}{2\pi}\right) \right] e^{-j \Omega \tau / 2} \tag{20}

Now, substituting $X_s(j\Omega )$ from Eq. (4) and $H(j\Omega )$ from Eq. (20) into Eq. (19) yields: X_{ft}(j\Omega) = \left[ \frac{1}{T_s} \sum_{k=-\infty}^{\infty} X_a(j(\Omega - k\Omega_s)) \right] \cdot \tau \text{sinc}\left(\frac{\Omega \tau}{2\pi}\right) e^{-j \Omega \tau / 2} \tag{21}

Distributing the terms yields the final frequency-domain expression for flat-top sampled signals: X_{ft}(j\Omega) = \frac{\tau}{T_s} \text{sinc}\left(\frac{\Omega \tau}{2\pi}\right) e^{-j \Omega \tau / 2} \sum_{k=-\infty}^{\infty} X_a(j(\Omega - k\Omega_s)) \tag{22}

---

5. The Aperture Effect & Frequency Distortion

Looking closely at the spectrum of the flat-top sampled signal in Eq. (22) reveals a critical engineering problem that distinguishes it from both ideal and natural sampling.

5.1 The Mathematical Proof of Baseband Distortion

In natural sampling (Eq. 13), each spectral replica is scaled as a whole by a constant factor $C_n$. In flat-top sampling (Eq. 22), the entire replicated spectrum is multiplied by a continuous, frequency-dependent envelope $H(j\Omega )=\tau \text{sinc}(\frac{\Omega \tau }{2\pi })e^{-j\Omega \tau /2}$.

Let us look at the reconstructed baseband signal ($k=0$) after passing the flat-top sampled signal through an ideal low-pass reconstruction filter: X_{ft,\text{baseband}}(j\Omega) = \left[ \frac{\tau}{T_s} \text{sinc}\left(\frac{\Omega \tau}{2\pi}\right) e^{-j \Omega \tau / 2} \right] \cdot X_a(j\Omega) \tag{23}

                      THE FLAT-TOP APERTURE DISTORTION EFFECT
                      
         Ideal Reconstructed Spectrum               Flat-top Reconstructed Spectrum
                |H(j\Omega)|                                  |H(j\Omega)|
               +-------------+                               +-\          /-+
               |             |                               |  \        /  |
               |             |                               |   \      /   |
               |             |                               |    \    /    |
         ------|-------------|------ \Omega            ------|-----\--/-----|------ \Omega
              -f_m          f_m                             -f_m     0      f_m
              (Flat Passband)                         (High-Frequency Roll-off / Droop)

[GRAPH: Comparison of ideal reconstructed passband (perfectly flat) vs. flat-top reconstructed passband: The continuous Sinc envelope of flat-top sampling attenuates higher frequencies within the baseband, creating severe high-frequency attenuation - source: Textbook Ch 6]

5.2 Physical Consequences of the Aperture Effect

1. High-Frequency Roll-Off: Because the $\text{sinc}(\frac{\Omega \tau }{2\pi })$ envelope is a continuously decreasing function of frequency $\Omega$ over the baseband, it attenuates higher frequencies within the desired signal $X_a(j\Omega )$.
2. Aperture Distortion: This non-uniform attenuation across the passband distorts the signal.
3. The Pulse-Width Trade-off:
   • If we make the pulse width $\tau$ very small relative to $T_s$ ($\tau \ll T_s$), the Sinc envelope becomes very wide and flat over the baseband, minimizing distortion. However, a very narrow pulse contains very little signal energy, causing a severe drop in the signal-to-noise ratio (SNR).
   • If we increase $\tau$ to maximize signal energy, we increase the Sinc roll-off, resulting in severe aperture distortion.

5.3 Hardware Correction: The Equalizer Filter

To resolve this trade-off, engineers use a larger pulse width $\tau$ to preserve signal energy and place an analog Equalizer Filter (or Aperture Corrector) with frequency response $E(j\Omega )$ immediately after the low-pass reconstruction filter.

                     APERTURE EQUALIZATION SIGNAL CASCADE
                     
      Staircase   +---------------+  Attenuated   +---------------+  Restored flat
       Signal --->|  Reconstruction|--> Baseband  |  Equalizer    |--> Analog Out
       x_ft(t)    |  LPF (f_s/2)  |   x_LPF(t)    |  Filter E(j\Omega)|   x_out(t)
                  +---------------+               +---------------+

The equalizer’s frequency response must be the mathematical inverse of the aperture distortion envelope to restore a flat frequency response across the baseband: E(j\Omega) = \frac{1}{H(j\Omega)} = \frac{1}{\tau \text{sinc}\left(\frac{\Omega \tau}{2\pi}\right) e^{-j \Omega \tau / 2}} \tag{24}

The magnitude response of the equalizer filter over the passband $[-{\Omega }_m,{\Omega }_m]$ is: |E(j\Omega)| = \frac{1}{\tau \text{sinc}\left(\frac{\Omega \tau}{2\pi}\right)} = \frac{\frac{\Omega\tau}{2}}{\tau \sin\left(\frac{\Omega\tau}{2}\right)} = \frac{\Omega}{2 \sin\left(\frac{\Omega\tau}{2}\right)} \tag{25}

This rising frequency response compensates for the Sinc droop, restoring a flat baseband frequency response.

---

6. Unified Comparison of Sampling Techniques

This comparison table summarizes the key properties of the three sampling techniques:

Feature	Ideal (Impulse) Sampling	Natural (Chopping) Sampling	Flat-Top (S&H) Sampling
Physical Realizability	Mathematically ideal, physically impossible.	Physically realizable using high-speed switches.	Standard practical interface in modern ADCs.
Pulsed Waveform Shape	Train of zero-width Dirac delta impulses.	Train of narrow pulses; tops curve along wave contour.	Staircase-like waveform; tops are flat.
Time-Domain Equation	$x_s(t)=\sum x_a(n{T}_s)\delta (t-n{T}_s)$	$x_{ns}(t)=x_a(t)\sum \text{rect}\left(\frac{t-n{T}_s}{\tau }\right)$	$x_{ft}(t)=\sum x_a(n{T}_s)\text{rect}\left(\frac{t-n{T}_s}{\tau }\right)$
Frequency-Domain Equation	$X_s(j\Omega )=\frac{1}{T_s}\sum X_a(j(\Omega -k{\Omega }_s))$	$X_{ns}(j\Omega )=\sum C_n{X}_a(j(\Omega -n{\Omega }_s))$	$X_{ft}(j\Omega )=H(j\Omega )X_s(j\Omega )$
Baseband Spectral Shape	Undistorted: Scaled evenly by constant factor $1/T_s$.	Undistorted: Scaled evenly by constant factor $C_0=\tau /T_s$.	Distorted: Scaled unevenly by Sinc envelope (Aperture Effect).
Replication Envelope	Flat envelope (infinite bandwidth).	Attenuated at high replica indexes by Sinc envelope.	Replicas and baseband are both filtered by continuous Sinc envelope.
Equalization Needed	No.	No.	Yes: Requires ${\text{sinc}}^{-1}$ equalizer filter.

---

7. Common Mistakes That Cost Marks

The Flat-Top Equalizer Oversight

Students often assume that since flat-top sampling uses a low-pass reconstruction filter, the analog signal is perfectly recovered. On exams, you must state that flat-top sampling requires an equalizer filter ($E(j\Omega )={\text{sinc}}^{-1}(f\tau )$) to correct for the aperture effect. Neglecting this detail can cost you several marks on theoretical questions.

Natural vs. Flat-Top Spectral Envelope Confusions

Do not confuse the spectral envelopes of natural and flat-top sampling:

• In natural sampling, the Sinc function $\text{sinc}(n{f}_s\tau )$ acts on the discrete harmonic index $n$, meaning each spectral replica is scaled by a single constant value. No distortion is introduced inside any replica.
• In flat-top sampling, the Sinc function $\text{sinc}(f\tau )$ is a continuous function of operating frequency $f$, meaning the scaling factor varies across each replica. This introduces severe distortion inside every replica, including the baseband.

---

8. PYQ Bank — Verbatim Questions & Answer Plans

8.1 PYQ 2023 Question 4b [5 Marks]

Question: Draw the block diagram of the analog-to-digital (ADC) conversion process.

• Answer Plan:
  1. Draw the functional hardware block diagram shown in Section 1.
  2. Label all intermediate signals clearly: Analog Input $x_a(t)$, Anti-Aliased Output $x_{\text{filt}}(t)$, Flat-topped Samples $x_{ft}(t)$, Quantized Samples $x_q(n)$, and Digital $b$-bit Output.
  3. Briefly summarize the role of the 4 key blocks (Anti-Aliasing Filter, Sample and Hold, Quantizer, Encoder) in 1 sentence each to secure full marks.

8.2 PYQ 2017 Question 4b [7 Marks]

Question: Distinguish between Ideal, Natural, and Flat-top sampling techniques.

• Answer Plan:
  1. Define each of the three techniques clearly.
  2. Write down the time-domain multiplication or convolution expressions for each technique.
  3. Reproduce the Unified Comparison Table from Section 6 to highlight differences in physical realizability, wave shape, baseband distortion, and the need for equalization.

8.3 PYQ 2018 Question 4b [4 Marks]

Question: Classify the sampling techniques.

• Answer Plan:
  1. State that physical sampling techniques are classified into three types based on the hardware switch model: Ideal (Impulse) sampling, Natural (Chopping) sampling, and Flat-top (Sample-and-Hold) sampling.
  2. Provide a 1-sentence definition of each along with their corresponding time-domain mathematical expressions.

---

9. Self-Check Before Moving On

• Can you reproduce the complete block diagram of the ADC process from memory?
• Do you understand why natural sampling preserves an undistorted baseband spectrum while flat-top sampling does not?
• Can you derive the flat-top frequency-domain expression $X_{ft}(j\Omega )=X_s(j\Omega )\cdot H(j\Omega )$ using convolution?
• Can you define the Aperture Effect and write down the magnitude response of the equalizer filter needed to correct it?

---

Source: (k.Deergha Rao) signals and systems.pdf (Chapters 6 & 7), 1.05 The Sampling Theorem Bridge & Multi-Rate Processing.md, Rabiul sir class note.pdf.

---

4.03 Ideal, Natural & Flat-Top Sampling Techniques | Course Roadmap

---

4.04 Multi-Rate Signal Processing: Decimation & Interpolation

Core Idea

Multi-rate signal processing is the branch of digital signal processing that deals with systems whose sampling rates change at different stages. Because a discrete-time sequence index $n$ must strictly be an integer, discrete scaling is non-trivial. It is achieved through Down-sampling (Decimation), which decreases the sampling rate by discarding samples, and Up-sampling (Interpolation), which increases the sampling rate by inserting zeros.

---

1. The Core Philosophy of Discrete-Time Scaling

In continuous-time systems, time scaling is mathematically straightforward: we replace the independent variable $t$ with $at$. For $a>1$, the signal is compressed; for $0<a<1$, the signal is expanded.

However, in discrete-time systems, this operation is highly restricted. The independent variable is an integer sample index $n$. We cannot define a value at a fractional index like $x[1.5]$. Thus:

• A direct compression like $y[n]=x[2n]$ is possible but causes permanent data loss because we must throw away all odd-indexed samples ($x[1],x[3],\dots$).
• A direct expansion like $y[n]=x[n/2]$ is undefined for odd values of $n$ (such as $y[1]=x[0.5]$).

To resolve these physical constraints, multi-rate digital systems use decimation and interpolation to scale discrete arrays without losing mathematical consistency.

---

2. Discrete-Time Down-Sampler (Decimation by $M$)

The down-sampler (also known as a sampling rate compressor) reduces the sampling rate of a discrete signal by an integer factor $M$.

                +-------+
    x[n] ------>|  | M  |------> x_d[n] = x[nM]
                +-------+

2.1 Mathematical Formulation

A down-sampler by factor $M$ keeps only every $M$-th sample of the input sequence and discards the intermediate $M-1$ samples: $x_d[n]=x[nM]\text{for }n\in \mathbb{Z}$

If the sampling period of the input signal $x[n]$ is $T$ (sampling frequency $F_s=1/T$), the sampling period of the down-sampled signal $x_d[n]$ becomes: $T^{\prime }=MT⟹F_s^{\prime }=\frac{F_s}{M}$

2.2 Spectral Effects & Digital Aliasing

In the frequency domain, shrinking a signal in the time domain expands its spectrum. Specifically, down-sampling by factor $M$ stretches the DTFT spectrum by a factor of $M$: $X_d(e^{j\omega })=\frac{1}{M}{\sum }_{i=0}^{M-1}X\left(e^{j\frac{\omega -2\pi i}{M}}\right)$

This expansion can cause the shifted copies of the spectrum to overlap, resulting in digital aliasing.

The Decimation Filter Rule

To prevent destructive aliasing during decimation, we must place a digital low-pass Anti-Aliasing Filter (or Decimation Filter) before the down-sampling block. The filter must have a normalized cutoff frequency of: ${\omega }_c=\frac{\pi }{M}$ This strips away any high-frequency components that would otherwise fold back into the baseband.

---

3. Discrete-Time Up-Sampler (Interpolation by $L$)

The up-sampler (also known as a sampling rate expander) increases the sampling rate of a discrete signal by an integer factor $L$.

                +-------+
    x[n] ------>|  ^ L  |------> x_e[n]
                +-------+

3.1 Mathematical Formulation

An up-sampler by factor $L$ inserts exactly $L-1$ zero-valued samples between each consecutive pair of original samples: $x_e[n]=\{\begin{array}{ll}x\left[\frac{n}{L}\right],& n=0,\pm L,\pm 2L,\dots \\ 0,& \text{otherwise}\end{array}$

If the original sampling period is $T$, the up-sampled sequence has a sampling period of: $T^{\prime }=\frac{T}{L}⟹F_s^{\prime }=L\cdot F_s$

3.2 Spectral Effects & Imaging Distortion

In the frequency domain, expanding a signal in the time domain compresses its spectrum. Up-sampling compresses the baseband spectrum $[-\pi ,\pi ]$ by a factor of $L$: $X_e(e^{j\omega })=X(e^{j\omega L})$

Because of this compression, $L-1$ unwanted duplicates of the original spectrum (called spectral images) appear within the new interval $[-\pi ,\pi ]$.

The Interpolation Filter Rule

The zero-inserted sequence $x_e[n]$ has a rough, staircase-like shape. To eliminate the high-frequency spectral images and interpolate the actual intermediate signal values, we must pass $x_e[n]$ through an Interpolation Filter (a digital low-pass filter) with:

1. Cutoff Frequency: ${\omega }_c=\frac{\pi }{L}$
2. Passband Gain: $L$ (to compensate for the amplitude drop caused by zero-insertion)

---

4. Comprehensive Multi-Rate Comparison Table

The physical and mathematical characteristics of the two multi-rate operators are summarized below:

Feature	Down-sampling ($\downarrow M$)	Up-sampling ($\uparrow L$)
Mathematical Equation	$x_d[n]=x[nM]$	$x_e[n]=x[n/L]$ (for $n=kL$)
Time Domain Effect	Discards $M-1$ samples	Inserts $L-1$ zeros
Information Volume	Data loss occurs	Zero-padding (no new info yet)
New Sampling Period	$T^{\prime }=MT$ (slower)	$T^{\prime }=T/L$ (faster)
Spectral Effect	Spectrum expands by $M$	Spectrum compresses by $L$
Associated Distortion	Digital Aliasing	Spectral Imaging
Associated Filter	Anti-Aliasing Filter (before $\downarrow M$)	Anti-Imaging Filter (after $\uparrow L$)
Filter Cutoff & Gain	${\omega }_c=\frac{\pi }{M}$, $\text{Gain}=1$	${\omega }_c=\frac{\pi }{L}$, $\text{Gain}=L$

---

5. Comprehensive Worked Examples (The Exam Killers)

5.1 Example 1: The 10-Mark 2024 KUET Exam Classic [PYQ 2024]

Question: Find out the up-sampling sequence $x_u[n]$ and down-sampling sequence $x_d[n]$ from the input signal sequence $x[n]=[1,6,11,14]$ for a factor of 3.

Step-by-Step Solution:

Scenario A: Default Origin Assumption (Leftmost Element is $n=0$)

Under the standard convention, if the origin is not explicitly marked with an arrow ($\uparrow$) or underline ($\underline{}$), the first element is located at the origin $n=0$: $x[0]=1,x[1]=6,x[2]=11,x[3]=14$ All other indices are zero: $x[n]=0$ for $n<0$ or $n\ge 4$.

1. Down-sampling by Factor $M=3$: The formula is: $x_d[n]=x[3n]$ Let’s map individual indices $n$:

• $n=0⟹x_d[0]=x[0]=1$
• $n=1⟹x_d[1]=x[3]=14$
• $n=2⟹x_d[2]=x[6]=0$
• $n=-1⟹x_d[-1]=x[-3]=0$

Thus, the down-sampled sequence is: ${\mathbf{x}}_{\mathbf{d}}[\mathbf{n}]=[\underline{1},14]$

2. Up-sampling by Factor $L=3$: The up-sampler inserts $L-1=2$ zeros between each consecutive sample of the original sequence:

• $x_u[0]=x[0]=1$
• $x_u[1]=0$
• $x_u[2]=0$
• $x_u[3]=x[1]=6$
• $x_u[4]=0$
• $x_u[5]=0$
• $x_u[6]=x[2]=11$
• $x_u[7]=0$
• $x_u[8]=0$
• $x_u[9]=x[3]=14$

Thus, the up-sampled sequence is: ${\mathbf{x}}_{\mathbf{u}}[\mathbf{n}]=[\underline{1},0,0,6,0,0,11,0,0,14]$

---

Scenario B: Alternative Origin Assumption (Origin at $n=0$ is marked, e.g., at value $11$)

If the examiner defines a different origin during the exam, say at the third sample ($11$): $x[-2]=1,x[-1]=6,x[\underline{0}]=11,x[1]=14$

1. Down-sampling by Factor $M=3$: $x_d[n]=x[3n]$

• $n=0⟹x_d[0]=x[0]=11$
• $n=1⟹x_d[1]=x[3]=0$
• $n=-1⟹x_d[-1]=x[-3]=0$

Thus, the down-sampled sequence collapses to: ${\mathbf{x}}_{\mathbf{d}}[\mathbf{n}]=[\underline{11}]$

2. Up-sampling by Factor $L=3$: Insert 2 zeros between each consecutive element, maintaining the origin at value $11$: ${\mathbf{x}}_{\mathbf{u}}[\mathbf{n}]=[1,0,0,6,0,0,\underline{11},0,0,14]$

---

5.2 Example 2: The 2022 Exam Sequence [PYQ 2022]

Question: If $x[n]=[2,6,\underline{0},3,1,2]$, find the compressed (down-sampled) and expanded (up-sampled) sequences for a factor of 3.

Step-by-Step Solution:

1. Extract the Index Map relative to the underlined origin: $x[-2]=2,x[-1]=6,x[\underline{0}]=0,x[1]=3,x[2]=1,x[3]=2$

2. Calculate Down-sampling ($M=3$): $x_d[n]=x[3n]$

   • $n=0⟹x_d[0]=x[0]=0$
   • $n=1⟹x_d[1]=x[3]=2$
   • $n=-1⟹x_d[-1]=x[-3]=0$ (out of bounds)

   Therefore, the down-sampled sequence is: ${\mathbf{x}}_{\mathbf{d}}[\mathbf{n}]=[\underline{0},2]$

3. Calculate Up-sampling ($L=3$): Insert $L-1=2$ zeros between each consecutive pair of elements, keeping the origin aligned at the original sample $x[0]=0$: ${\mathbf{x}}_{\mathbf{e}}[\mathbf{n}]=[2,0,0,6,0,0,\underline{0},0,0,3,0,0,1,0,0,2]$

---

6. Common Mistakes That Cost Marks

The Origin Alignment Trap

Do not always assume that the first element in a bracket is the origin! Always check the array for an arrow ($\uparrow$) or underline ($\underline{}$). If you down-sample without mapping the indices relative to $n=0$ first, your entire output array will shift, resulting in a zero-mark evaluation on the numerical.

The Anti-Imaging Filter Gain Oversight

When asked to draw the block diagram of an interpolator, students often forget that up-sampling scales down the signal amplitude. The matching digital low-pass reconstruction filter must contain a scaling gain factor of $L$ to restore the signal’s original physical voltage range.

---

7. PYQ Bank — Verbatim Questions & Answer Plans

7.1 PYQ 2024 Question 7c [10 Marks]

Question: Find out the up-sampling sequence $x_u(n)$ and down-sampling sequence $x_d(n)$ from the input signal sequence $x(n)=[1,6,11,14]$ for the discrete-time system under a factor of 3.

• Answer Plan:
  1. State the mathematical down-sampling and up-sampling equations.
  2. Write out the step-by-step coordinate substitutions showing how index values map under both Scenario A (default leftmost origin) and Scenario B (alternative marked origin).
  3. Provide the final bolded output arrays clearly labeling the $n=0$ coordinate.

7.2 PYQ 2021 [5 Marks]

Question: Define (i) Discrete time up-sampler; (ii) Discrete time down-sampler.

• Answer Plan:
  1. Provide the formal time-domain equations for both operators.
  2. Draw the standard block diagrams ($\uparrow L$ and $\downarrow M$).
  3. Explain the physical effect of both blocks in the time domain (zero-insertion vs. sample compression).

---

8. Self-Check Before Moving On

• Can you define down-sampling and up-sampling mathematically and graphically?
• Do you know why an anti-aliasing filter is required before a down-sampler, and what its cutoff frequency must be?
• Can you explain why up-sampling creates spectral images, and how the interpolation filter removes them?
• Can you reliably execute index mappings on discrete arrays with non-zero time origins?

---

Source: (k.Deergha Rao) signals and systems.pdf (Section 6.6), 1.05 The Sampling Theorem Bridge & Multi-Rate Processing.md, pyq catagorised (Table 7).