Sampling and reconstruction in communication in Signal Processing - Mini Project: Build & Apply

Sampling and Reconstruction in Communication

📖 Scenario: Imagine you are working on a communication system that sends sound signals. To send the sound digitally, you need to sample the continuous sound wave at certain points in time and then reconstruct it back to a smooth wave at the receiver.This project will help you understand how sampling and reconstruction work using simple data science techniques.

🎯 Goal: You will create a simple signal, sample it at a fixed rate, reconstruct the signal using interpolation, and then display the original and reconstructed signals to see how well the process works.

📋 What You'll Learn

Create a continuous signal using a mathematical function

Define a sampling rate to pick points from the signal

Use interpolation to reconstruct the signal from samples

Plot the original and reconstructed signals for comparison

💡 Why This Matters

🌍 Real World

Sampling and reconstruction are key steps in converting real-world signals like sound and images into digital form and back, used in phones, radios, and streaming.

💼 Career

Understanding these concepts is important for roles in signal processing, telecommunications, audio engineering, and data science involving time series data.

1

Create the continuous signal

Create a variable called time_continuous that holds 1000 points evenly spaced between 0 and 1 seconds using np.linspace(0, 1, 1000). Then create a variable called signal_continuous that is a sine wave with frequency 5 Hz using np.sin(2 * np.pi * 5 * time_continuous).

Signal Processing

import numpy as np
# Create time_continuous and signal_continuous variables
# Your code here

Hint

Use np.linspace to create the time points and np.sin with the formula for a sine wave.

2

Define the sampling rate and sample the signal

Create a variable called sampling_rate and set it to 20. Then create a variable called time_samples that holds points from 0 to 1 seconds spaced by 1 / sampling_rate. Finally, create signal_samples by sampling signal_continuous at the indices corresponding to time_samples.

Signal Processing

import numpy as np

time_continuous = np.linspace(0, 1, 1000)
signal_continuous = np.sin(2 * np.pi * 5 * time_continuous)
# Define sampling_rate, time_samples, and signal_samples
# Your code here

Hint

Use np.arange to create sample times and sample the sine wave at those times.

3

Reconstruct the signal using interpolation

Import interp1d from scipy.interpolate. Create an interpolation function called reconstruct_func using interp1d(time_samples, signal_samples, kind='linear'). Then create signal_reconstructed by applying reconstruct_func to time_continuous.

Signal Processing

import numpy as np
from scipy.interpolate import interp1d

time_continuous = np.linspace(0, 1, 1000)
signal_continuous = np.sin(2 * np.pi * 5 * time_continuous)

sampling_rate = 20
time_samples = np.arange(0, 1 + 1/sampling_rate, 1 / sampling_rate)
signal_samples = np.sin(2 * np.pi * 5 * time_samples)
# Create reconstruct_func and signal_reconstructed
# Your code here

Hint

Use linear interpolation to reconstruct the signal from samples.

4

Plot the original and reconstructed signals

Import matplotlib.pyplot as plt. Plot time_continuous vs signal_continuous with label 'Original Signal'. Plot time_continuous vs signal_reconstructed with label 'Reconstructed Signal' using a dashed line. Add a legend and show the plot.

Signal Processing

import numpy as np
from scipy.interpolate import interp1d
import matplotlib.pyplot as plt

time_continuous = np.linspace(0, 1, 1000)
signal_continuous = np.sin(2 * np.pi * 5 * time_continuous)

sampling_rate = 20
time_samples = np.arange(0, 1 + 1/sampling_rate, 1 / sampling_rate)
signal_samples = np.sin(2 * np.pi * 5 * time_samples)

reconstruct_func = interp1d(time_samples, signal_samples, kind='linear')
signal_reconstructed = reconstruct_func(time_continuous)
# Plot the signals
# Your code here

Hint

Use plt.plot to draw both signals and plt.legend() to show labels.

Practice

1. What is the main purpose of sampling in communication systems?

easy

A. To convert a continuous signal into discrete values

B. To amplify the signal strength

C. To filter out noise from the signal

D. To encrypt the signal for security

Solution

1. Step 1: Understand the role of sampling

   Sampling takes a continuous-time signal and measures its amplitude at discrete time intervals.

2. Step 2: Identify the purpose in communication

   This process allows the signal to be represented digitally for processing and transmission.

3. Final Answer:

   To convert a continuous signal into discrete values -> Option A

4. Quick Check:

   Sampling = Continuous to discrete conversion [OK]

Hint: Sampling means turning smooth signals into numbers [OK]

Common Mistakes:

• Confusing sampling with filtering
• Thinking sampling increases signal strength
• Mixing sampling with encryption

2. Which of the following is the correct Nyquist sampling rate formula for a signal with maximum frequency f_max?

easy

A. f_s > f_max / 2

B. f_s > 2 * f_max

C. f_s = f_max

D. f_s < f_max / 2

Solution

1. Step 1: Recall Nyquist theorem

   The Nyquist rate states the sampling frequency must be greater than twice the highest frequency in the signal.

2. Step 2: Apply to formula

   So, f_s > 2 * f_max ensures no loss of information during sampling.

3. Final Answer:

   f_s > 2 * f_max -> Option B

4. Quick Check:

   Nyquist rate = 2 * max frequency [OK]

Hint: Sample faster than twice the highest frequency [OK]

Common Mistakes:

• Using sampling rate less than max frequency
• Confusing less than and greater than signs
• Ignoring the factor of 2 in Nyquist rate

3. Given a continuous signal with maximum frequency 3 kHz, what is the minimum sampling frequency to avoid aliasing? Choose the correct Python code output.

f_max = 3000
f_s = 2 * f_max
print(f_s)

medium

A. 6000

B. 1500

C. 3000

D. 9000

Solution

1. Step 1: Calculate minimum sampling frequency

   Using Nyquist, minimum sampling frequency is twice the max frequency: 2 * 3000 = 6000 Hz.

2. Step 2: Check Python code output

   The code multiplies 3000 by 2 and prints 6000, matching the calculation.

3. Final Answer:

   6000 -> Option A

4. Quick Check:

   2 * 3000 = 6000 [OK]

Hint: Double the max frequency to get sampling rate [OK]

Common Mistakes:

• Printing max frequency instead of double
• Dividing instead of multiplying
• Confusing units (Hz vs kHz)

4. Identify the error in the following Python code snippet for reconstructing a signal using sinc interpolation:

import numpy as np

def reconstruct_signal(samples, t, T):
    x_recon = 0
    for n in range(len(samples)):
        x_recon += samples[n] * np.sinc((t - n*T))
    return x_recon

# samples = [1, 2, 3], t = 1.5, T = 1
print(reconstruct_signal([1, 2, 3], 1.5, 1))

medium

A. np.sinc argument should be normalized by pi

B. Loop should iterate over t values, not samples

C. x_recon should be initialized as an array, not zero

D. The function should return samples instead of x_recon

Solution

1. Step 1: Understand np.sinc function

   In numpy, np.sinc(x) = sin(pi*x)/(pi*x), so input is already normalized by pi.

2. Step 2: Check argument passed to np.sinc

   The code passes (t - n*T) directly, which is correct because np.sinc expects normalized input.

3. Step 3: Identify error

   Actually, the argument is correct; the error is that x_recon is initialized as 0 (a scalar), but the function expects to handle array inputs for t for vectorized reconstruction.

4. Step 4: Correct error

   x_recon should be initialized as np.zeros_like(t) if t is an array, but since t is scalar here, initialization is fine.

5. Step 5: Reconsider options

   np.sinc argument should be normalized by pi says argument should be normalized by pi, but numpy's sinc already does that, so no need to normalize.

6. Step 6: Final conclusion

   No error in normalization; the code is correct for scalar t. The main issue is if t is array, initialization fails.

7. Final Answer:

   x_recon should be initialized as an array, not zero -> Option C

8. Quick Check:

   Accumulator initialization must match output type [OK]

Hint: Remember numpy sinc is normalized by pi already [OK]

Common Mistakes:

• Assuming np.sinc needs manual pi normalization
• Initializing accumulator incorrectly for arrays
• Confusing loop variable roles

5. You have a signal sampled at 8 kHz with maximum frequency 5 kHz. What problem will occur during reconstruction, and how can it be fixed?

hard

A. Signal will be noisy; fix by applying a low-pass filter after sampling

B. Signal will be delayed; fix by adjusting reconstruction timing

C. No problem; sampling rate is sufficient for reconstruction

D. Aliasing will occur; fix by increasing sampling rate above 10 kHz

Solution

1. Step 1: Compare sampling rate and Nyquist rate

   Nyquist rate = 2 * max frequency = 2 * 5 kHz = 10 kHz. Sampling rate is 8 kHz, which is less than Nyquist rate.

2. Step 2: Identify consequence of insufficient sampling

   Sampling below Nyquist rate causes aliasing, where higher frequencies appear as lower frequencies, distorting the signal.

3. Step 3: Determine solution

   To avoid aliasing, increase sampling rate above 10 kHz.

4. Final Answer:

   Aliasing will occur; fix by increasing sampling rate above 10 kHz -> Option D

5. Quick Check:

   Sampling rate < Nyquist rate causes aliasing [OK]

Hint: Sample above twice max frequency to avoid aliasing [OK]

Common Mistakes:

• Assuming 8 kHz is enough for 5 kHz max frequency
• Confusing noise with aliasing
• Ignoring Nyquist sampling theorem