Compute the circular convolution x 1 (n) N ⃝ x 2 (n) for N = 4, 7, and 8.
1. Compute the circular convolution x 1 (n) N ⃝ x 2 (n) for N = 4, 7, and 8.
2. Compute the linear convolution x 1 (n) ∗ x 2 (n).
3. Using results of calculations, determine the minimum value of N necessary so that linear and circular convolutions are same on the N -point interval. 4. Without performing the actual convolutions, explain how you could have obtained the result of P5.33.3.
P5.34 Let
x(n) =
A cos (2πℓn/N ), 0 ≤ n ≤ N − 1 = A cos (2πℓn/N ) R
N (n)
elsewhere
Problems 209
where ℓ is an integer. Notice that x(n) contains exactly ℓ periods (or cycles) of the cosine waveform in N samples. This is a windowed cosine sequence containing no leakage.
1. Show that the DFT X(k) is a real sequence given by
X(k) = AN δ (k − ℓ) + AN δ (k − N + ℓ);
0 ≤ k ≤ (N − 1), 0<ℓ<N
2. Show that if ℓ = 0, then the DFT X(k) is given by
X(k) = AN δ(k);
0 ≤ k ≤ (N − 1)
3. Explain clearly how these results should be modified if ℓ < 0 or ℓ > N . 4. Verify the results of parts 1, 2, and 3 using the following sequences. Plot the real parts of
the DFT sequences using the stem function.
(a) x 1 (n) = 3 cos (0.04πn) R 200 (n) (b) x 2 (n) = 5R 50 (n) (c) x 3 (n) = [1 + 2 cos (0.5πn) + cos (πn)] R 100 (n) (d) x 4 (n) = cos (25πn/16) R 64 (n) (e) x 5 (n) = [4 cos (0.1πn) − 3 cos (1.9πn)] R 40 (n)
P5.35 Let x(n) = A cos (ω 0 n) R N (n), where ω 0 is a real number.
1. Using the properties of the DFT, show that the real and the imaginary parts of X(k) are given by
X(k) = X R (k) + jX I (k)
4 π (N −1)
5 sin [π (k − f 0 N )]
X R (k) = (A/2) cos
(k − f 0 N)
sin [π(k − f 0 N )/N ]
4 π (N −1)
5 sin [π (k − N + f 0 N )]
+ (A/2) cos
(k + f 0 N)
sin [π(k − N + f 0 N )/N ]
I (k) = − (A/2) sin (N −1)
5 sin [π (k − f 0 N )]
(k − f 0 N)
sin [π(k − f 0 N )/N ]
− 5 (N −1) sin [π (k − N + f 0 N )]
(A/2) sin
(k + f 0 N)
sin [π(k − N + f 0 N )/N ] 2. This result implies that the original frequency ω 0 of the cosine waveform has leaked into
other frequencies that form the harmonics of the time-limited sequence, and hence it is called the leakage property of cosines. It is a natural result due to the fact that bandlimited periodic cosines are sampled over noninteger periods. Explain this result using the periodic extension ˜ x(n) of x(n) and the result in Problem P5.34.1.
3. Verify the leakage property using x(n) = cos (5πn/99) R 200 (n). Plot the real and the imaginary parts of X(k) using the stem function.
P5.36 Let
x(n) =
A sin (2πℓn/N ) , 0 ≤ n ≤ N − 1 = A sin (2πℓn/N ) R
N (n)
Elsewhere
Chapter 5
THE DISCRETE FOURIER TRANSFORM
where ℓ is an integer. Notice that x(n) contains exactly ℓ periods (or cycles) of the sine waveform in N samples. This is a windowed sine sequence containing no leakage.
1. Show that the DFT X(k) is a purely imaginary sequence given by
AN
AN
X(k) =
(k − ℓ) −
δ (k − N + ℓ);
0 ≤ k ≤ (N − 1), 0<ℓ<N
2j
2j
2. Show that if ℓ = 0, then the DFT X(k) is given by
X(k) = 0;
0 ≤ k ≤ (N − 1)
3. Explain clearly how these results should be modified if ℓ < 0 or ℓ > N . 4. Verify the results of parts 1, 2, and 3 using the following sequences. Plot the imaginary
parts of the DFT sequences using the stem function.
(a) x 1 (n) = 3 sin (0.04πn) R 200 (n) (b) x 2 (n) = 5 sin 10πnR 50 (n) (c) x 3 (n) = [2 sin (0.5πn) + sin (πn)] R 100 (n) (d) x 4 (n) = sin (25πn/16) R 64 (n) (e) x 5 (n) = [4 sin (0.1πn) − 3 sin (1.9πn)] R 20 (n)
P5.37 Let x(n) = A sin (ω 0 n) R N (n), where ω 0 is a real number.
1. Using the properties of the DFT, show that the real and the imaginary parts of X(k) are given by
X(k) = X R (k) + jX I (k)
4 π (N −1)
5 sin [π (k − f 0 N )]
X R (k) = − (A/2) sin
(k − f 0 N)
sin [π(k − f 0 N )/N ]
4 + (A/2) sin π (N −1)
(k + f 5 N) sin [π (k − N + f 0 N )]
0 sin [π(k − N + f
0 N )/N ]
I (k) = − (A/2) cos
(N −1)
(k − f 0 N)
sin [π (k − f 0 N )] sin [π(k − f 0 N )/N ]
+ (A/2) cos
4 π (N −1)
5 sin [π (k − N + f
0 N )] sin [π(k − N + f 0 N )/N ]
(k + f 0 N)
2. This result is the leakage property of sines. Explain it using the periodic extension ˜ x(n) of x(n) and the result in Problem P5.36.1. 3. Verify the leakage property using x(n) = sin (5πn/99) R 100 (n). Plot the real and the imaginary parts of X(k) using the stem function.
P5.38 An analog signal x a (t) = 2 sin (4πt) + 5 cos (8πt) is sampled at t = 0.01n for n = 0, 1, . . . , N − 1 to obtain an N -point sequence x(n). An N -point DFT is used to obtain
an estimate of the magnitude spectrum of x a (t).
1. From the following values of N , choose the one that will provide the accurate estimate of the spectrum of x a (t). Plot the real and imaginary parts of the DFT spectrum X(k). (a) N = 40,
(b) N = 50,
(c) N = 60.
Problems 211
2. From the following values of N , choose the one that will provide the least amount of leakage in the spectrum of x a (t). Plot the real and imaginary parts of the DFT spectrum X(k). (a) N = 90,
(b) N = 95,
(c) N = 99.
P5.39 Using (5.49), determine and draw the signal flow graph for the N = 8 point, radix-2 decimation-in-frequency FFT algorithm. Using this flow graph, determine the DFT of the
sequence
x(n) = cos (πn/2) , 0≤n≤7
P5.40 Using (5.49), determine and draw the signal flow graph for the N = 16 point, radix-4
decimation-in-time FFT algorithm. Using this flow graph, determine the DFT of the sequence
x(n) = cos (πn/2) ,
0 ≤ n ≤ 15
P5.41 Let x(n) be a uniformly distributed random number between [−1, 1] for 0 ≤ n ≤ 10 6 . Let
h(n) = sin(0.4πn),
0 ≤ n ≤ 100
1. Using the conv function, determine the output sequence y(n) = x(n) ∗ h(n). 2. Consider the overlap-and-save method of block convolution along with the FFT
algorithm to implement high-speed block convolution. Using this approach, determine y(n) with FFT sizes of 1024, 2048, and 4096.
3. Compare these approaches in terms of the convolution results and their execution times.