Solved Problems On Shannon's Capacity

b) The bandwidth of the signal is the range of frequencies, that is, the difference
between the highest and lowest frequency. The highest frequency is 16, the
lowest is 3, so the answer is: 16 – 3 = 13 Hz.
c) As discussed in class, a digital signal is represented as a square wave which
requires an infinite number of frequencies to represent. Since there are a finite
number of frequencies in this signal, is must be analog.

3. “CAT-5” twisted-pair cable has a bandwidth of roughly 100MHz. We would like to

transmit information at a bit rate of 500Mbps. Is a signal-to-noise ratio of 30dB enough
to reliably transmit this much information? Why or why not?

Answer: This question requires a simple application of the Shannon Theorem:

C = B log 2 (1 + SNR)
In this case we are given C = 500Mbps, B = 100MHz, and SNR = 30dB. In order to
answer the question of whether or not the SNR is sufficient, we need to see if the capacity
provided by the bandwidth and SNR is >= 500Mbps. As a first step, we need to convert
the SNR from decibels to the “unitless” form:
SNRdB = 10 log10 SNR
so
SNR = 10 SNRdB / 10
= 10 30 / 10
= 1000
We can now apply the Shannon Theorem:
C = B log 2 (1 + SNR )
= 100 MHz log 2 (1 + 1000)
≈ 100 MHz * 9.967 * (1Mbps / MHz )
= 996.7 Mbps
Note the use of “Mbps/MHz” to underscore the fact that we are explicitly converting
from “Hz” to “bps”.
Thus the Shannon Theorem tells us that the theoretical limit for transferring data over this
medium at this SNR is approximately 1000Mbps (or 1Gbps). Thus we can reliably
transmit 500Mbps over this connection.

4. What is the maximum capacity of a medium with a bandwidth of 750KHz and a

signal-to-noise ratio of 30dB?

Answer: This is a straight-forward application of the Shannon Theorem, except that we

must first convert the SNR from decibels to its “unitless” form:
SNRdB = 10 log10 SNR
so
SNR = 10 SNRdB / 10
= 10 30 / 10
= 1000
So, applying the Shannon Theorem:
C = B log 2 (1 + SNR )
= 750 KHz log 2 (1 + 1000) * (1Kbps / KHz )
= 750 Kbps * 9.967
≈ 7475.25 Kbps
≈ 7.475Mbps
So the maximum capacity of a medium with a bandwidth of 750KHz and a signal-to-
noise ratio of 30dB is 7.475Mbps.

5. What is the minimum signal-to-noise ratio, in decibels, that must be maintained in

order to transmit a 600Kbps signal over a medium with a bandwidth of 20,000Hz?

Answer: Let’s work around the Shannon Theorem to solve for the SNR:
C = B log 2 (1 + SNR )
C
= log 2 (1 + SNR )
B
2 C / B = 1 + SNR
SNR = 2 C / B − 1
We can now solve this for the given capacity and bandwidth. Note that our capacity is in
Kbps, but our bandwidth is in Hz. Since 20,000Hz = 20Khz, we’ll do the calculations in
Kbps/Khz:
SNR = 2 ( 600 / 20 ) − 1
SNR = 2 30 − 1
SNR = 1073741823
The question however, asks for the SNR in decibels, so we need to convert:
SNRdB = 10 log10 SNR
SNRdB = 10 log10 1073741823
SNRdB ≈ 90.31
So we would need at least 90.31dB as a signal-to-noise ratio in order to transmit the
desire capacity.
6. We are given a medium that will reliably transmit frequencies between 0 and
25,000Hz. Is it possible to transmit 200Kbps of information along this line? If so, then
describe a method and any conditions that must be satisfied. If not, explain why.

Answer:
We could apply the Shannon Theorem here but that would just give us a limit on the
capacity that could be transmitted by the medium. It would not describe how to do so. If
we apply the Nyquist Theorem for two-level signals:
C = 2B
C = 2 * (25000 − 0)
C = 50 Kbps
we see that with a two-level digital signal, we can only transmit at most 50Kbps. We can
apply the reformulated Nyquist Theorem to examine a possible multi-level signal:
C = 2 B log 2 M
200 Kbps = 2(25 KHz ) log 2 M
log 2 M = 200 / 50
log 2 M = 4
M = 2 4 = 16
So we can transmit 200Kbps of information along this medium provided we use use a
digital signal with at least 16 levels.
However this does place some restrictions – the Nyquist Theorem assumes that there is
no noise and that transmission is error-free.

7. A triangle wave has the following shape:

This wave can be generally represented by the Fourier series

∞
1 π
∑ (2k + 1)
k =0
2
sin( 2π (2k + 1) f1t +
2
)

where f1 is the base frequency of the wave.

a) What is the bandwidth of this signal?