3.

4 / CHANNEL CAPACITY 91
Data
transmitted: 1 0 1 0 0 1 1 0 0 1 1 0 1 0 1

Signal:

Noise:

Signal plus
noise:

Sampling
times:

Data received: 1 0 1 0 0 1 0 0 0 1 1 0 1 1 1

Original data: 1 0 1 0 0 1 1 0 0 1 1 0 1 0 1
Bits in error
Figure 3.16 Effect of Noise on a Digital Signal

communication. For example, a sharp spike of energy of 0.01 s duration would not
destroy any voice data but would wash out about 560 bits of digital data being trans-
mitted at 56 kbps. Figure 3.16 is an example of the effect of noise on a digital signal.
Here the noise consists of a relatively modest level of thermal noise plus occasional
spikes of impulse noise. The digital data can be recovered from the signal by sam-
pling the received waveform once per bit time. As can be seen, the noise is occasion-
ally sufficient to change a 1 to a 0 or a 0 to a 1.

3.4 CHANNEL CAPACITY

We have seen that there are a variety of impairments that distort or corrupt a signal.
For digital data, the question that then arises is to what extent these impairments
limit the data rate that can be achieved. The maximum rate at which data can be
transmitted over a given communication path, or channel, under given conditions, is
referred to as the channel capacity.
There are four concepts here that we are trying to relate to one another.
• Data rate: The rate, in bits per second (bps), at which data can be com-
municated
92 CHAPTER 3 / DATA TRANSMISSION

• Bandwidth: The bandwidth of the transmitted signal as constrained by the

transmitter and the nature of the transmission medium, expressed in cycles per
second, or Hertz
• Noise: The average level of noise over the communications path
• Error rate: The rate at which errors occur, where an error is the reception of a
1 when a 0 was transmitted or the reception of a 0 when a 1 was transmitted
The problem we are addressing is this: Communications facilities are
expensive and, in general, the greater the bandwidth of a facility, the greater the
cost. Furthermore, all transmission channels of any practical interest are of
limited bandwidth. The limitations arise from the physical properties of the
transmission medium or from deliberate limitations at the transmitter on the
bandwidth to prevent interference from other sources. Accordingly, we would
like to make as efficient use as possible of a given bandwidth. For digital data,
this means that we would like to get as high a data rate as possible at a particu-
lar limit of error rate for a given bandwidth. The main constraint on achieving
this efficiency is noise.

Nyquist Bandwidth
To begin, let us consider the case of a channel that is noise free. In this environ-
ment, the limitation on data rate is simply the bandwidth of the signal. A formu-
lation of this limitation, due to Nyquist, states that if the rate of signal
transmission is 2B, then a signal with frequencies no greater than B is sufficient
to carry the signal rate. The converse is also true: Given a bandwidth of B, the
highest signal rate that can be carried is 2B. This limitation is due to the effect of
intersymbol interference, such as is produced by delay distortion. The result is
useful in the development of digital-to-analog encoding schemes and is, in
essence, based on the same derivation as that of the sampling theorem, described
in Appendix F.
Note that in the preceding paragraph, we referred to signal rate. If the signals
to be transmitted are binary (two voltage levels), then the data rate that can be sup-
ported by B Hz is 2B bps. However, as we shall see in Chapter 5, signals with more
than two levels can be used; that is, each signal element can represent more than one
bit. For example, if four possible voltage levels are used as signals, then each signal
element can represent two bits. With multilevel signaling, the Nyquist formulation
becomes

C = 2B log2 M

where M is the number of discrete signal or voltage levels.

So, for a given bandwidth, the data rate can be increased by increasing the
number of different signal elements. However, this places an increased burden
on the receiver: Instead of distinguishing one of two possible signal elements
during each signal time, it must distinguish one of M possible signal elements.
Noise and other impairments on the transmission line will limit the practical
value of M.
 3.4 / CHANNEL CAPACITY 93

EXAMPLE 3.3 Consider a voice channel being used, via modem, to transmit
digital data. Assume a bandwidth of 3100 Hz. Then the Nyquist capacity, C, of the
channel is 2B = 6200 bps. For M = 8, a value used with some modems, C
becomes 18,600 bps for a bandwidth of 3100 Hz.

Shannon Capacity Formula

Nyquist’s formula indicates that, all other things being equal, doubling the band-
width doubles the data rate. Now consider the relationship among data rate, noise,
and error rate. The presence of noise can corrupt one or more bits. If the data rate is
increased, then the bits become “shorter” so that more bits are affected by a given
pattern of noise.
Figure 3.16 illustrates this relationship. If the data rate is increased, then more
bits will occur during the interval of a noise spike, and hence more errors will occur.
All of these concepts can be tied together neatly in a formula developed by the
mathematician Claude Shannon. As we have just illustrated, the higher the data rate,
the more damage that unwanted noise can do. For a given level of noise, we would
expect that a greater signal strength would improve the ability to receive data cor-
rectly in the presence of noise. The key parameter involved in this reasoning is the
signal-to-noise ratio (SNR, or S/N),10 which is the ratio of the power in a signal to the
power contained in the noise that is present at a particular point in the transmission.
Typically, this ratio is measured at a receiver, because it is at this point that an attempt
is made to process the signal and recover the data. For convenience, this ratio is often
reported in decibels:

signal power
SNR dB = 10 log10
noise power

This expresses the amount, in decibels, that the intended signal exceeds the noise
level. A high SNR will mean a high-quality signal and a low number of required
intermediate repeaters.
The signal-to-noise ratio is important in the transmission of digital data
because it sets the upper bound on the achievable data rate. Shannon’s result is that
the maximum channel capacity, in bits per second, obeys the equation
C = B log 211 + SNR2 (3.1)
where C is the capacity of the channel in bits per second and B is the bandwidth of
the channel in Hertz. The Shannon formula represents the theoretical maximum
that can be achieved. In practice, however, only much lower rates are achieved. One
reason for this is that the formula assumes white noise (thermal noise). Impulse
noise is not accounted for, nor are attenuation distortion or delay distortion. Even in

10
Some of the literature uses SNR; others use S/N. Also, in some cases the dimensionless quantity is
referred to as SNR or S/N and the quantity in decibels is referred to as SNR db or 1S/N2db . Others use just
SNR or S/N to mean the dB quantity. This text uses SNR and SNR db .
94 CHAPTER 3 / DATA TRANSMISSION

an ideal white noise environment, present technology still cannot achieve Shannon
capacity due to encoding issues, such as coding length and complexity.
The capacity indicated in the preceding equation is referred to as the error-free
capacity. Shannon proved that if the actual information rate on a channel is less than
the error-free capacity, then it is theoretically possible to use a suitable signal code to
achieve error-free transmission through the channel. Shannon’s theorem unfortu-
nately does not suggest a means for finding such codes, but it does provide a yardstick
by which the performance of practical communication schemes may be measured.
Several other observations concerning the preceding equation may be instructive.
For a given level of noise, it would appear that the data rate could be increased by
increasing either signal strength or bandwidth. However, as the signal strength increases,
so do the effects of nonlinearities in the system, leading to an increase in intermodula-
tion noise. Note also that, because noise is assumed to be white, the wider the band-
width, the more noise is admitted to the system.Thus, as B increases, SNR decreases.

EXAMPLE 3.4 Let us consider an example that relates the Nyquist and Shan-
non formulations. Suppose that the spectrum of a channel is between 3 MHz and
4 MHz and SNR dB = 24 dB. Then

B = 4 MHz - 3 MHz = 1 MHz

SNR dB = 24 dB = 10 log101SNR2
SNR = 251
Using Shannon’s formula,
C = 106 * log211 + 2512 L 106 * 8 = 8 Mbps
This is a theoretical limit and, as we have said, is unlikely to be reached. But
assume we can achieve the limit. Based on Nyquist’s formula, how many signal-
ing levels are required? We have
C = 2B log 2 M
8 * 106 = 2 * 11062 * log 2 M
4 = log 2 M
M = 16

The Expression Eb/N0

Finally, we mention a parameter related to SNR that is more convenient for determining
digital data rates and error rates and that is the standard quality measure for digital com-
munication system performance. The parameter is the ratio of signal energy per bit to
noise power density per Hertz, Eb/N0 . Consider a signal, digital or analog, that contains
binary digital data transmitted at a certain bit rate R. Recalling that 1 Watt = 1 J/s, the
energy per bit in a signal is given by Eb = STb , where S is the signal power and Tb is the
time required to send one bit.The data rate R is just R = 1/Tb . Thus
 3.4 / CHANNEL CAPACITY 95
Eb S>R S
= =
N0 N0 kTR
or, in decibel notation,

a b = SdBW - 10 log R - 10 log k - 10 log T

Eb
N0 dB
= SdBW - 10 log R + 228.6 dBW - 10 log T
The ratio Eb/N0 is important because the bit error rate for digital data is a (decreas-
ing) function of this ratio. Given a value of Eb/N0 needed to achieve a desired error
rate, the parameters in the preceding formula may be selected. Note that as the bit
rate R increases, the transmitted signal power, relative to noise, must increase to
maintain the required Eb/N0 .
Let us try to grasp this result intuitively by considering again Figure 3.16. The
signal here is digital, but the reasoning would be the same for an analog signal. In
several instances, the noise is sufficient to alter the value of a bit. If the data rate
were doubled, the bits would be more tightly packed together, and the same passage
of noise might destroy two bits. Thus, for constant signal to noise ratio, an increase in
data rate increases the error rate.
The advantage of Eb/N0 over SNR is that the latter quantity depends on the
bandwidth.

EXAMPLE 3.5 For binary phase-shift keying (defined in Chapter 5), Eb/N0 =
8.4 dB is required for a bit error rate of 10-4 (one bit error out of every 10,000). If
the effective noise temperature is 290°K (room temperature) and the data rate is
2400 bps, what received signal level is required?
We have
8.4 = S1dBW2 - 10 log 2400 + 228.6 dBW - 10 log 290
= S1dBW2 - 110213.382 + 228.6 - 110212.462
S = - 161.8 dBW

We can relate Eb/N0 to SNR as follows. We have

Eb S
=
N0 N0R
The parameter N0 is the noise power density in Watts/Hertz. Hence, the noise in a
signal with bandwidth B is N = N0B. Substituting, we have
Eb S BT
= (3.2)
N0 N R
Another formulation of interest relates Eb/N0 to spectral efficiency. Shannon’s
result (Equation 3.1) can be rewritten as:
96 CHAPTER 3 / DATA TRANSMISSION

S
= 2 C>B - 1
N
Using Equation (3.2), and equating R with C, we have

= 12 C>B - 12
Eb B
N0 C
This is a useful formula that relates the achievable spectral efficiency C/B to Eb/N0 .

EXAMPLE 3.6 Suppose we want to find the minimum Eb/N0 required to

achieve a spectral efficiency of 6 bps/Hz. Then
Eb/N0 = 11/6212 6 - 12 = 10.5 = 10.21 dB.

3.5 RECOMMENDED READING AND WEB SITE

There are many books that cover the fundamentals of analog and digital transmission.
[COUC01] is quite thorough. Other good reference works are [FREE05], which includes
some of the examples used in this chapter, and [HAYK01].

COUC01 Couch, L. Digital and Analog Communication Systems. Upper Saddle River, NJ:
Prentice Hall, 2001.
FREE05 Freeman, R. Fundamentals of Telecommunications. New York: Wiley, 2005.
HAYK01 Haykin, S. Communication Systems. New York: Wiley, 2001.

Recommended Web site:

• Fourier series synthesis: An excellent visualization tool for Fourier series

3.6 KEY TERMS, REVIEW QUESTIONS, AND PROBLEMS

Key Terms

absolute bandwidth attenuation distortion data

analog data audio dc component
analog signal bandwidth decibel (dB)
analog transmission center frequency delay distortion
aperiodic channel capacity digital data
attenuation crosstalk digital signal