Assignment 1 Solutions

Chap1 (Review/Introduction) and Chap2 (Sampling, A/D conversion)

1. Basedband signals g1 (t) and g2 (t) with bandwidth 100kHz and 150kHz respectively. Determine
the Nyquist interval and the sampling rate for signals g1 (t), g2 (t), g12 (t), g23 (t), and g1 (t)g2 (t).

Solution: The bandwidth of g1 (t) and g2 (t) are 100 kHz and 150 kHz, respectively. Therefore,

• the Nyquist sampling rates for g1 (t) is 200 kHz, sampling interval Ts = 1/200k = 5µs

• the Nyquist sampling rates for g2 (t) is 300 kHz, sampling interval Ts = 1/300k = 3.33µs.

• the bandwidth of g12 (t) is 200 kHz, fN yq = 400 kHz, fN yq = 1/400k = 0.25µs.

• the bandwidth of g23 (t) is 450 kHz, fN yq = 900 kHz, fN yq = 1/900k = 1.11µs.

• the bandwidth of g1 (t) · g2 (t) is 250 kHz, fN yq = 500 kHz, fN yq = 1/500k = 2µs.

2. Determine the Nyquist sampling rate and the Nyquist sampling interval for the signals:

(a) sinc(100πt);

(b) sinc2 (100πt);

(c) sinc(100πt) + sinc(50πt);

(d) sinc(100πt) + 3sinc2 (60πt);

(e) sinc(50πt) sinc(100πt).

Solution:

(a) since ( )
1 f
sinc(2Bπt) → rect
2B 2B
here 2B = 100, the bandwidth B of this signal is 50 Hz. The Nyquist rate is 100 Hz
(samples/sec).
(b)
( )
1 f
sinc (Bπt) → ∆
2
B 2B

here B = 100, the bandwidth of this signal B is 100 Hz. The Nyquist rate is 200 Hz
(samples/sec).

1
 (c)
( ) ( )
f f
sinc(100πt) + sinc(50πt) → 0.01rect + 0.02rect
100 50

the bandwidth of the first term on the right-hand side is 50 Hz and the second term is 25
Hz. Clearly the bandwidth of the composite signal is the higher of the two, that is, 50 Hz.
The Nyquist rate is 100 Hz (samples/sec).
(d)
( ) ( )
f f
sinc(100πt) + 3sinc (60πt) → 0.01rect
2
+ 0.05∆
100 120

the bandwidth of the first term is 50 Hz and that of the second term is 60 Hz. The
bandwidth of the sum is the higher of the two, that is, 60 Hz. The Nyquist sampling rate
is 120 Hz.
(e)
( ) ( )
f f
sinc(50πt) → 0.02rect sinc(100πt) → 0.01rect
50 100

The two signals have BW 25 Hz and 50 Hz respectively. The spectrum of the product of
two signals is 1/(2π) times the convolution of their spectra. From width property of the
convolution, the width of the convoluted signals is the sum of the widths of the signals
convolved. Therefore, the BW of the product is 25+50=75 Hz. The Nyquist rate is 150
Hz.

3. A signal g(t) = sinc2 (5πt) is sampled (using uniformly spaced impulses) at a rate of: (i) 5 Hz;
(ii) 10 Hz; (iii) 20 Hz. For each of the three case:

(a) Sketch the sampled signal;

(b) Sketch the spectrum of the sampled signal;

(c) Explain whether you can recover the signal g(t) from the sampled signal;

(d) If the sampled signal is passed through an ideal low-pass filter of bandwidth 5 Hz, sketch
the spectrum of the output signal.

Solution: ( )
f
2
g(t) = sinc (5πt) → G(f ) = 0.2∆
10
The BW of g(t) is 5 Hz. Therefore, the Nyquist rate is 10 Hz, and the Nyquist interval is
T = 1/10 = 0.1s.

1
• When fs = 5Hz, the spectrum T G(f ) repeats every 5 Hz. The successive spectra overlaps,
and the spectrum G(f ) is not recoverable from Gs (f ), that is, g(t) cannot be recovered

2
 from its samples. If the sampled signal is passed through an ideal lowpass filter of BW 5
Hz, the output spectrum is rect(f /10), and the output signal is 10sinc(20πt), which is not
the desired signal sinc2 (5πt).

• When fs = 10Hz, the spectrum Gs (f ) consists of back-to-back, nonoverlapping repetition

1
of T G(f ) repeating every 10 Hz. Hence, G(f ) can be recovered from Gs (f ) using an ideal
lowpass filter of BW 5 Hz, and the output is 10sinc2 (5πt).

• in the last case of oversampling (fs = 20 Hz), with empty band between successive cycles.
Hence, G(f ) can be recovered from Gs (f ) using an ideal lowpass filter or even a practical
lowpass filter. The output is 20sinc2 (5πt).

4. Prove that a signal cannot be simultaneously time-limited and band-limited. Hint: show that
contrary assumption leads to contradiction. Assume a signal simultaneously time-limited and
band-limited so that G(f ) = 0 for |f | > B. In this case G(f ) = G(f )rect(f /2B ′ ) for B ′ > B.
This means that g(t) is equal to g(t) ∗ 2B ′ sinc(2πB ′ t). Show that the latter cannot be time-
limited.

Solution: Assuming a signal g(t) that is simultaneously time-limited and bandlimited. Let
g(f ) = 0 for |f | > B. Therefore,
( )
f
g(f )rect = g(f ) f or B ′ > B.
2B ′
Therefore, from the time-convolution property

g(t) = g(t) ∗ [2B ′ sinc(2πB ′ t)] = 2B ′ g(t) ∗ sinc(2πB ′ t).

Because g(t) is time-limited, g(t) = 0 for |t| > T . But g(t) is equal to convolution of g(t) with
sinc(2πB ′ t) which is not time-limited. It is impossible to obtain a time-limited signal from the
convolution of a time-limited signal with a non-timelimited signal.

5. Solution: (a): the bandwidth is 15 kHz. The Nyquist rate is 30 kHz.

(b): 65536 = 216 , so that 16 binary digits are needed to encode each sample.
(c): 30, 000 × 16 = 480, 000 bits/s.
(d): 44, 100 × 16 = 705, 600 bits/s.

6. Solution: (a): The Nyquist rate is 2 × 4.5 × 106 = 9 MHz. The actual sampling rate =1.2 × 9 =
10.8 MHz.
(b): 1024 = 210 , so that 10 bits or binary pulses are needed to encode each sample.
(c): 10.8 × 106 × 10 = 108 × 106 or 108 Mbits/s.

7. Solution: If mp is the peak sample amplitude, then

mp
quantization error ≤ 0.2% × mp =
500

3
 Because the maximum quantization error is

q 2mp
= 0.5 ×
2 L

• it follows that L ≥ 500. Since L should be a power of 2, we choose L = 512 = 29 . This

requires 9-bit binary code per sample. The Nyquist rate is 2 × 1000 = 2000 Hz. 20% above
this rate is 2 × 1.2 = 2.4 kHz. Thus, each signal has 2400 samples/second, and each sample
is encoded by 9 bits. Therefore, each signal uses 9 × 2.4 = 21.6 kbits/second.

• Five such signals are multiplexed. Hence, we need a total of 5 × 21.6 = 108 Kbits/second
data bits.

• Framing and synchronization requires additional 0.5% bits, that is 108, 000 × 0.005 = 540
bits, yielding a total of 108,540 bits/second.

• The minimum transmission bandwidth is 108.54/2 = 54.27 kHz.

8. Solution: Let mp denote the peak amplitude of the sinusoid signal, signal power is E[m2 (t)] =
m2p /2. Let L denote the number of steps, then the stepsize is q = 2mp /L. The noise power is
σ 2 = q 2 /2 = m2p /(3L2 ). The required SNR is 47 dB=50119, which is

E[m2 (t)] 3L2

= ≥ 50119
σ2 2

which implies L ≥ 182.8. Because L is a power of 2, we select L = 256 = 28 . The SNR for this
value L is
E[m2 (t)] 3L2
= = 3(256)2 (0.5) = 98304 = 49.43dB
σ2 2

9. Use the sampling property of the unit impulse function to evaluate the following integrals.
∫∞
(a) −∞
cos 6t δ(t − 3) dt
∫∞
(b) −∞
10δ(t) (1 + t)−1 dt
∫∞
(c) −∞
10δ(t + 4) (t2 + 6t + 1) dt
∫∞
(d) −∞
exp(−t2 ) δ(t − 2) dt

Solution:
(a)
∫ ∞
cos 6t δ(t − 3) dt = cos(6 · 3) = cos(18)
−∞

(b)
∫ ∞
10δ(t) (1 + t)−1 dt = 10(1 + 0)−1 = 10
−∞

4
 (c)
∫ ∞
10δ(t + 4) (t2 + 6t + 1) dt = 10(16 − 24 + 1) = 90
−∞

(d)
∫ ∞
exp(−t2 ) δ(t − 2) dt = exp(−4)
−∞

10. Solution: (a) Let L denote the number of quantization levels. The peak signal power to
quantization noise power is ( )
S
= 3L2 .
N peak

We have 10 log10 (3L2 ) ≥ 30 (dB), and L can be solved as

L = ⌈18.26⌉ = 19 levels

The number of bits per sample is

N = ⌈log2 L⌉ = ⌈log2 19⌉ = 5 bits/sample

(b) Let Tb denote the time duration of a bit. Since the sample rate is 8000 samples/s, each
sample is represented by 5 bits. Therefore, there are 8000 × 5 bits each second and

1
Tb = = 25 µs
8000 × 5

the required bandwidth W is

1
W = = 40 kHz
Tb

11. Solution:
(a) The maximum frequency is fm = 1000Hz. Therefore, sampling rate should be

fs ≥ 2fm = 2 × 1000 = 2000 samples/s

The sampling interval should satisfy:

1
Tm = ≤ 0.5 ms
fs

(b) 2000 samples/s × 3600 s = 7.2 × 106 samples.

12. Solution: (a) Binary case:

R = 8000 samples/s × 6 bits/sample = 48, 000 bits/s

1
W = = R = 48, 000Hz
Tb

5
 ( )
S
= 3L2 = 3(64)2 = 12, 288 ≈ 41dB
N q

(b) Four-level case:

48, 000 bits/s
Rs = = 24, 000 symbols/s
2 bits/symbol
1
W = = Rs = 24, 000Hz
T
( )
S
= the same as in the binary case ≈ 41dB
N q

13. Solution: (a) Assuming that the L quantization levels are equally spaced and symmetrical
about zero. Then, the maximum possible quantization noise voltage equals 1/2 the q volt
interval between any two neighboring levels. Also, the peak quantization noise power, Nq , can
be expressed as (q/2)2 .

The peak signal power, S, can be designated (Vpp /2)2 , where Vpp = Vp − (−Vp ) is the peak-to-
peak signal voltage, and Vp is the peak voltage.

since there are L quantization levels and (L − 1) intervals (each interval corresponding to q
volts), we can write:
( )
S (Vpp /2)2 [q(L − 1)/2]2 q 2 L2 /4
= = ≈ = L2
Nq peak (q/2)2 (q/2)2 q 2 /4
Thus, we need to compute how many levels, L, will yield a (S/Nq )peak = 96 dB. We therefore
write:
96dB = 10 log10 (S/Nq )peak = 10 log10 L2 = 20 log10 L

L = 1096/20 = 63096 levels

(b) The number of bits that corresponding to 63096 levels is

l = ⌈log2 L⌉ = ⌈log2 63096⌉ = 16 bits/sample

(c) R=16 bits/sample × 44.1 k samples/s = 705,600 bits/s

14. Solution: see course notes.

15. Solution: (a) For each sampling interval, there are 24+1=25 pulses. Therefore, each pulse
occupies:
Ts 1
= T = = 5µs
25 8000 × 25
Therefore, the spacing between successive pulses of the multiplexed signal is 5-1=4 µs.

(b) With Nyquist sampling, each pulse occupies:

Ts 1
T = = = 6.25µs
25 6400 × 25
Therefore, the spacing between successive pulses of the multiplexed signal is 6.25-1=5.25 µs.