EE 421: Communications I

Dr. Mohammed Hawa

Digital Modulation Techniques

For more information: read Chapters 7 and 10 in your textbook.

There are four major modulation techniques used by communication systems

nowadays to transport baseband digital data onto a carrier. These modulation
techniques are:
• Amplitude-Shift Keying (ASK)
• Frequency-Shift Keying (FSK)
• Phase-Shift Keying (PSK)
• Quadrature Amplitude Modulation (QAM)

Amplitude-Shift Keying (ASK):

example the transmitter could send the carrier 2 cos

ASK represents digital data as variations in the amplitude of a carrier signal. For

while using the carrier cos

to represent a logic 1,
to represent a logic 0. This is shown in the diagram
below. The receiver detects the amplitude of the carrier to recover the original bit
stream.

A special case of ASK is when a logic 1 is represented by cos (i.e., the presence
of a carrier) and a logic 0 is represented by a zero voltage (i.e., the absence of a
carrier). This special case is called On-Off Keying (OOK) and is shown below.

Notice that you can visualize ASK as the process of Amplitude Modulation (AM)
using a “Polar NRZ” digital baseband message signal. In other words, we say that
ASK is the result of multiplying a binary Polar NRZ signal (with appropriate
DC shift) times a sinusoidal carrier. This is shown in the diagram below:

Page 1 of 18
The above diagram shows that a general ASK signal is simply an AM signal with a
modulation index m < 1, while an OOK is an AM signal with a modulation index
m = 1. Hence, an envelope detector can be used at the receiver to demodulate the
ASK signal. Actually, all types of modulators and demodulators we explained
earlier for AM are now applicable to ASK provided that the message signal m(t)
used is a digital baseband Polar NRZ signal.

In addition, since ASK is a special case of AM modulation, the bandwidth of ASK is

2B centered around the carrier frequency , where B is the bandwidth of the Polar
NRZ signal. Since the bandwidth of Polar NRZ is equal to the data bit rate (fo) of the
bit stream to be sent, the bandwidth of ASK is 2fo (Hz).

The following is a sketch of the PSD for an ASK signal. It consists of two replicas of
the PSD for a Polar NRZ signal with additional carrier impulses. You can see that
the bandwidth of this ASK signal is approximately 2fo (Hz).

The advantages and disadvantages of ASK are summarized below:

Advantages Disadvantages
• ASK is the simplest kind of • It can be used only when the signal-to-
modulation to generate and noise ratio (SNR) is very high.
detect. • Its bandwidth is too big (equals 2fo).

Frequency-shift keying (FSK):

In FSK the instantaneous frequency of the carrier signal is shifted between two
possible frequency values termed the mark frequency (representing a logic 1) and
the space frequency (representing a logic 0). This is shown in the diagram below.

Page 2 of 18
Notice that FSK can be thought of as Frequency Modulation (FM) using a “Polar
NRZ” digital baseband signal as the message, and hence FSK can be seen as a subset
of FM modulation.

≈ ∆ +
Since FSK is a special case of FM modulation, the bandwidth of FSK is given by
Carson’s rule which says that , where B is the bandwidth of the
Polar NRZ signal (equal to fo (the bit rate)). Hence, the bandwidth of FSK is
2∆f + 2fo (Hz). In addition, all modulator and demodulator circuits for FM are still
applicable for FSK.

FSK has several advantages over ASK due to the fact that the carrier has a constant
amplitude. These are the same advantages present in FM which include: immunity
to non-linearities, immunity to rapid fading, immunity to adjacent channel
interference, and the ability to exchange SNR for bandwidth. FSK was used in early
slow dial-up modems.

Phase-shift keying (PSK):

In PSK, the data is conveyed by changing the phase of the carrier wave. One
possible representation (called Binary Phase-Shift Keying or BPSK) is to send logic 1
as a cosine signal with zero phase shift and a logic 0 as a cosine signal but with a
180° phase shift. We say in this case that the BPSK signal can assume one of two
possible symbols: 0° and 180°. This case is shown in the following Figure.

Page 3 of 18
BPSK can be thought of as a special case of Phase Modulation (PM) using a “Polar

deviation to be ∆ = /2 (i.e., 2∆ = − = ). This value maximizes

NRZ” digital baseband message1. In the case of BPSK, we select the peak phase

immunity to phase noise.

Since BPSK is a special case of PM, the bandwidth of PSK is 2B + 2∆f, where B is
the bandwidth for the polar NRZ signal and ∆f = 0 since the sinusoidal carrier signal
does not change its frequency. Hence, the bandwidth of BPSK is 2fo (Hz).

A convenient way to represent PSK modulation is using a constellation diagram. A

constellation diagram consists of a group of points representing the different
symbols the carrier in a PSK modulated signal can assume. For example, for BPSK,
in which each bit is represented by one symbol (i.e., either A cos(ωct) or
A cos(ωct – 180°)), the constellation diagram consists of two points (see Figure
below). These two points have the same amplitude A, but they are 180° apart. This
means that a logic 1 corresponds to A cos(ωct), while a logic 0 corresponds to
A cos(ωct – 180°).

Another common example of PSK is Quadrature (or

Quaternary) Phase-Shift Keying (QPSK). QPSK uses
four possible phases for the carrier (45°, 135°, 225°, 315°)
but with the same carrier amplitude, as shown in the
constellation diagram to the right. With four phases,
QPSK can encode two bits per one symbol (see Figure
below).

Data 10 11 00 01

QPSK(t)

Tsymb = 2T0 Symbol Time

1Notice that BPSK can also be thought of as a special case of DSB-SC in which the Polar NRZ signal
DSB-SC modulates a sinusoidal carrier. This is because multiplying a carrier by positive and negative
values switches its phase by 180°.

Page 4 of 18
You can imagine QPSK as a special case of Phase Modulation (PM) in which the
baseband message signal m(t) is a digital M-ary signal (with M = 4). In this case, the
bandwidth of the M-ary baseband signal is B = Baud Rate = fo / 2, which means that
the bandwidth of the QPSK signal is 2B + 2∆f = fo instead of 2 fo for BPSK. Hence,
QPSK can be used to double the data rate compared to a BPSK system while
maintaining the same bandwidth of the modulated signal.

Notice that any number of phases may be used to

construct a PSK constellation. Usually, 8-PSK is the
highest order PSK constellation deployed in
practice (see the figure to the right). In this case,
each carrier symbol represents three bits.

With more than eight phases, the error-rate

becomes too high and there are better, though more
complex, modulation schemes available (such as
QAM).

Notice that in PSK, the constellation points are usually positioned with uniform
angular spacing around a circle. This gives maximum phase-separation between
adjacent points and thus the best immunity to noise. Points are positioned on a
circle so that all the different phases can be transmitted with the same carrier
amplitude.

The axes in a constellation diagram are called the in-phase (I) and quadrature (Q)
axes, respectively, due to their 90° separation. The nice thing about a constellation
diagram is that it lends itself to straightforward and simple implementation of PSK
modulation in hardware. This is because the PSK modulated signal can be
generated by individually DSB-SC modulating both a sine wave and a cosine wave
and then adding the resulting modulated carriers to each other. In such case, the
constellation diagram is extremely helpful since the amplitude of each point along
the in-phase axis is the one used to modulate the cosine wave and the amplitude
along the quadrature axis is the one used to modulate the sine wave. This procedure
will be much more obvious when we discuss QAM modulation in the next section.

It is worth mentioning that BPSK and QPSK can be regarded special cases of the
more general QAM modulation, where the amplitude of the modulating signal is
constant (see next section).

Example: Find the bandwidth of an 8-PSK modulated signal if the data bit rate is

Solution: For 8-PSK, Bandwidth = 2B = 2×Baud Rate = 2 &'( * = 2 + # ,%/%- # =

!!"#$% !!"#$%
100 kbit/s.

)
66.67 kHz.

Page 5 of 18
Quadrature Amplitude Modulation (QAM):
QAM is a modulation scheme which conveys data by modulating the amplitude of
two carrier waves. These two waves (a cosine and a sine) are out of phase with each
other by 90° and are thus called quadrature carriers — hence the name of the
scheme.

Both analog and digital QAM are possible. Analog QAM was used in NTSC and
PAL television systems, where the I- and Q-signals carry the components of
chrominance (color) information.

Let us start by remembering analog QAM, which allowed us to transmit two

message signals using two orthogonal carriers of the same frequency. The following
Figure shows this scheme. Notice that both modulated signals will occupy the same
frequency band around ωc.

The two baseband signals can be separated at the receiver by synchronous

considering the multiplier output 1

detection using two local carriers in phase quadrature. This can be shown by
of the top branch (see Figure above):

1 = 2345 cos =6 cos + 7 sin : cos

1 1 1
= + cos 2 + sin 2
2 2 2 7

/2. Thus, in QAM two signals can be transmitted simultaneously over

The last two terms are suppressed by the lowpass filter (LPF), yielding the desired
output
a bandwidth of 2B, and still get separated at the receiver.

Digital QAM, on the other hand, is constructed using two M-ary baseband signals
(called i(t) and q(t)) modulating the two quadrature carriers. For example, in 16-

them can assume one of four possibilities. This results in 4 4 = 16 possible carrier
QAM both i(t) and q(t) are 4-ary digital baseband signals, which means each one of

symbols as shown in the constellation diagram below. Hence, 16-QAM uses 16

symbols, with each symbol representing a specific four-bit pattern.

Page 6 of 18
For example, to send the bit sequence 100101110000 using 16-QAM, the bit stream is
split into 4-bit groups, with each 4-bit pattern affecting i(t) and q(t) as shown in the
figure below.
Data 1001 0111 0000

i(t) 3A
4-ary A
signal A
3A
180 0
i(t) cos( ct)
t

-A cos( ct) Data 1001 0111 0000

3A cos( ct) -3A cos( ct)
QAM(t)
Data 1001 0111 0000
q(t) 3A
t
4-ary A
signal A A 2 cos( ct - 225 )
3A
A 10 cos( ct - 18.4 ) A 18 cos( ct - 135 )
180 180
q(t) sin( ct)
t

A sin( ct) -A sin( ct)

3A sin( ct)

Notice that the baud rate (i.e., the symbol rate) of the resulting 16-QAM signal is one
fourth that of the data bit rate. This is why the bandwidth of 16-QAM is 2×Baud Rate
= 2 fo/4 = fo/2. You can see that this is correct because the bandwidth of each one of
the 4-ary signals is B = fo / 4 (one symbol per four bits). Performing DSB-SC
modulation for each one of these signals (i.e., QAM) results in a total bandwidth of
2B = 2 (fo/4) = fo/2.

Example: Find the bandwidth of an 16-QAM modulated signal if the data bit rate is

Solution: For 16-QAM, Bandwidth = 2×Baud Rate = 2 &'( = = 2 > # ,%/%- #

*5#$% *5#$%
8 Mbit/s.

)
= 4 MHz.

Page 7 of 18
In QAM, the constellation points are usually arranged in a square grid with equal
vertical and horizontal spacing called rectangular QAM (see the above constellation
diagram). The number of points in the grid is usually a power of 2 (2, 4, 8...). The
most common forms of QAM are 16-QAM, 64-QAM, 128-QAM and 256-QAM. By
moving to higher-order constellations, it is possible to transmit more bits per
symbol, which reduces bandwidth. However, if the mean energy of the constellation
is to remain the same, the points must be closer together and are thus more
susceptible to noise; this results in a higher bit error rate (BER) and, hence, higher-
order QAM can deliver more data less reliably than lower-order QAM unless, of
course, the SNR is increased.

Rectangular QAM constellations are, in general, sub-optimal in the sense that they
do not maximally space the constellation points for a given energy. However, they
have the considerable advantage that they are easier to generate and demodulated
using simple hardware. Non-square constellations
achieve marginally better performance but are
harder to modulate and demodulate.

For example, the diagram of circular 16-QAM

constellation is shown to the right. The
constellation diagram shown below is the one used
in the V.32bis dial-up modem. This modem
provides 14.4 kbit/s using only 2400 baud rate.
Can you calculate the number of constellation
points from these numbers?2

Note: It is worth mentioning that in practical systems, M-ary signals are shaped

case, the bandwidth of QAM (or PSK) becomes 2 ?@AB 1 + C /2 instead of just
using a raised-cosine pulse before modulating the two quadrature carriers. In such

2 ?@AB.

2Remember that baud rate = symbol rate = bit rate / bits-per-symbol. This means that the number of

above constellation diagram, you will notice 128 symbols corresponding to D = log 7 128 = 7 bits.
data bits-per-symbol is k = 14400 / 2400 = 6 bits. However, if you count the number of symbols in the

This seeming contradiction is a result of using one of the seven bits as a parity bit not as a data bit.
Using parity bits combined with high-order modulation is called Trellis Coded Modulation (TCM).

Page 8 of 18
Performance of digital modulation techniques in presence of Noise:
Remember that we measured the performance for analog modulation techniques in
terms of signal quality, which was related to output signal-to-noise ratio (SNRout).
For digital modulation techniques, the performance is measured in terms of output
bit error rate (BER), which represents the number of erroneous bits that the receiver
expects per second. For example, a BER = 10-4 means that we expect on average 1 bit

performance if the ?HI ≤ 10L= .

error out of every 10,000 transmitted bits. We say the system exhibits good

Remember that we are using the Additive White Gaussian Noise (AWGN)
mathematical model to describe the noise on a communication channel. Hence, the

variance M 7 . The variance of the noise M 7 is its average power.

noise n(t) is considered as a Gaussian random process with zero average and a

Recall that for a standard Gaussian random variable X with zero-mean and unity-
variance , the probability density function (pdf) is:

1 L )
N 1 = P 7
√2

For the purpose of our performance analysis, we will define the Quantile function
Q(x) as the complement of the cumulative distribution function F(x) of the standard
Gaussian random variable, i.e.,

U
1 U LV)
Q 1 = 1 − R 1 = 1 − S N T BT = S N T BT = S P 7 BT
LU √2

The diagram below gives a visual representation for Q(x) which represents the
shaded area under the standard Gaussian density curve:

Gaussian pdf, fn(n)

F(x)
Q(x)

x n

Usually we use a table (similar to the one shown below) to lookup Q(x) values for
specific x arguments since the above integral has no closed form solution.

Page 9 of 18
 x Q (x ) x Q (x ) x Q (x ) x Q (x )

2.00 2.28E-02 3.00 1.35E-03 4.00 3.17E-05 5.00 2.87E-07

2.05 2.02E-02 3.05 1.14E-03 4.05 2.56E-05 5.05 2.21E-07
2.10 1.79E-02 3.10 9.68E-04 4.10 2.07E-05 5.10 1.70E-07
2.15 1.58E-02 3.15 8.16E-04 4.15 1.66E-05 5.15 1.30E-07
2.20 1.39E-02 3.20 6.87E-04 4.20 1.33E-05 5.20 9.96E-08
2.25 1.22E-02 3.25 5.77E-04 4.25 1.07E-05 5.25 7.60E-08
2.30 1.07E-02 3.30 4.83E-04 4.30 8.54E-06 5.30 5.79E-08
2.35 9.39E-03 3.35 4.04E-04 4.35 6.81E-06 5.35 4.40E-08
2.40 8.20E-03 3.40 3.37E-04 4.40 5.41E-06 5.40 3.33E-08
2.45 7.14E-03 3.45 2.80E-04 4.45 4.29E-06 5.45 2.52E-08
2.50 6.21E-03 3.50 2.33E-04 4.50 3.40E-06 5.50 1.90E-08
2.55 5.39E-03 3.55 1.93E-04 4.55 2.68E-06 5.55 1.43E-08
2.60 4.66E-03 3.60 1.59E-04 4.60 2.11E-06 5.60 1.07E-08
2.65 4.02E-03 3.65 1.31E-04 4.65 1.66E-06 5.65 8.02E-09
2.70 3.47E-03 3.70 1.08E-04 4.70 1.30E-06 5.70 5.99E-09
2.75 2.98E-03 3.75 8.84E-05 4.75 1.02E-06 5.75 4.46E-09
2.80 2.56E-03 3.80 7.23E-05 4.80 7.93E-07 5.80 3.32E-09
2.85 2.19E-03 3.85 5.91E-05 4.85 6.17E-07 5.85 2.46E-09
2.90 1.87E-03 3.90 4.81E-05 4.90 4.79E-07 5.90 1.82E-09
2.95 1.59E-03 3.95 3.91E-05 4.95 3.71E-07 5.95 1.34E-09

This is not part of the exam material …

The analysis presented here is quite general and applies for both digital baseband
and digital modulated signals. We will start with the special case of sending a pulse
p(t) when transmitting a 1 and sending the negative of the same pulse, i.e., –p(t)
when transmitting a 0. Notice that this analysis is equally applicable to Polar NRZ
signals (where p(t) = rect(t)) and to BPSK signals (where p(t) = A cos(ωct)). See the
following figure.
Data 1 0 0 1 1 0
p(t)

Generalized
Polar p(t)

–p(t)

p(t) = 5 V
5V
Polar
NRZ
-5 V
p(t) = A cos( ct)

A
BPSK t
–A
–A cos( ct) =
A cos( ct – 180 )

Page 10 of 18
We start by assuming that our receiver will sample the received pulse p(t) in the
middle of the pulse (also called symbol) period to read its amplitude. For example,
if the transmitted signal is Polar NRZ, as shown above, the detected samples are
supposed to be A = +5V or –A = –5V, which decides whether the received bit is a
logic 1 or a logic 0.

However, because of the channel noise the detected samples will be ±A + n, where
n is a zero-mean normally (Gaussian) distributed random noise, as shown below. In
threshold detection, and because of the symmetry of the situation, the preferred
detection threshold is zero; that is, the received pulse is detected as a 1 or 0
depending on whether the sample value is positive or negative.

Notice that because of added noise, occasionally a transmitted 0 will be sampled as

a positive voltage, causing the received 0 to be read as 1. This means that an error
can occur with a probability equal to the area of the Gaussian function

above figure). Hence, for Gaussian noise with zero-mean and variance M 7 the
corresponding to a transmitted 0 but falling above the zero-voltage threshold (see

probability of making an error in reading a 0 at the receiver is given by3:

Pr6PYYZY|\: = Pr6] > : = Q _ `

M
And similarly,
Pr6PYYZY|a: = Pr6] < − : = Q _ `
M

3 Notice that for a Gaussian random variable that has zero mean and variance M 7 , the probability that
f) f)
the random variable exceeds a value x is c N T BT = c P BT = chij P BT = Q h j.
L L
)g) )
∞
d√7e
∞ ∞
√7e g d

Page 11 of 18
Hence, using conditional probabilities, we have:

lP@D m@ lnP o@nAP

Pr6PYYZY: = k Q_ ` Pr6qrstuv:
Y m ZN ]ZpmP
w&& xyz{'&x
1 1
= Pr6PYYZY|a: Pr6a: + Pr6PYYZY|\: Pr6\: = Q _ ` +Q_ `
M 2 M 2
= Q_ `
M

It is important to notice, however, that for AWGN noise the above threshold
detection is not optimal. For example, a better method is to sample the pulse p(t)
three times; if at least two out of three samples are positive, the received bit is
considered 1. It turns out that the optimal scenario is to sample the pulse infinitely
many times, which is equivalent to calculating the energy of the received symbol
instead of just using one sample value to make the final decision.

To achieve this optimal behavior, receivers use a matched filter before sampling the
output of the filter at the end of each symbol time (see the Figure below). The

the energy of the incoming symbol. The matched filter is represented by |

matched filter is simply a convolution of the received symbol with itself, resulting in
in the
Figure below.

sampling the received symbol l + ] , the receiver now samples the output of
To derive the mathematical equations at the receiver, we notice that instead of

the matched filter, denoted by l} + ]} . Of course, we know that in the

frequency domain:

~} =~ |

1 U
and,
l} = ℱ L €~} • = ℱ L €~ | •= S ~ | P ‚ƒ, B
2 LU

where ~ is the Fourier transform of the incoming symbol l . Since the output
of the matched filter is sampled at = „%- # , the peak sample value is now:

1 U
= l} …„%- #† = S ~ | P ‚ƒ‡ˆ‰Š‹ B
}
2 LU

But it is also important to realize that the AWGN noise signal ]

through the matched filter which means we get an output noise signal ]}
will also pass
with a
power spectral density of:

Page 12 of 18
 Œ • =Œ || |7
Hence, the rms of output noise is:

1 U 1 U
M} = ŽM 7• = Ž•••••••
]}7 =• S Œ B =• S Œ || |7 B
2 LU •
2 LU

‘} U
= • S || |7 B
4 LU

We explained earlier that the probability of error is given by the following equation:

lP@D m@ lnP o@nAP

Pr6PYYZY: = k Q_ ` Pr6qrstuv:
Y m ZN ]ZpmP
w&& xyz{'&x

And since the quantile function Q 1 decreases monotonically with 1, our objective
is to design the matched filter | such that the ratio } /M} is maximum . It can be
shown that (see Chapter 10 in your textbook for the analysis), that the matched filter
that maximizes the above ratio is:

| =~ − P L‚ƒ‡ˆ‰Š‹

If we use such a matched filter at the receiver (i.e., we use optimal threshold
detection), then irrespective of the symbol type or shape, we will have:

1 U
lP@D m@ lnP o@nAP = = S |~ |7 B = H%’ = P]PY“” p] ℎP p ,– m” —Zn
}
2 LU

‘! U ‘! H%’
Y m ZN ]ZpmP = M} = • S |~ |7 B =•
4 LU 2
Hence,

lP@D m@ lnP o@nAP ZN p ,– m” —Zn

BER = Pr6PYYZY: = k Q› œ Pr•žŸ qrstuv¡
Y m ZN ]ZpmP
w&& xyz{'&x

H%’ 2H%
= k Q£ ¥ Pr•žŸ qrstuv¡ = k Q ¦ • ’ § Pr•žŸ qrstuv¡
‘! H%’ ‘!
w&& xyz{'&x Ž w&& xyz{'&x
¢ 2 ¤
2 ∑ H%’ Pr6žŸ qrstuv:
= Q¦• §
‘!

If we use H% to represent the average energy-per-transmitted-symbol (over all

possible symbols in the modulated signal), we can write the above equation as
follows:

Page 13 of 18
 2H% 2H#
BER = Q ¦ • § = Q¦• §
‘! ‘!

The above analysis applies for other types of digital modulated signals, not just
polar signals, but with slight variations. A summary of the BER equations for the
different modulation techniques we discussed in this document is given below.

For the rest of this document, we will use the following notation:
• M = Number of possible symbols that the modulated signal can assume.
• k = the number of bits sent per transmitted symbol = log2 (M).
• Es = Average energy-per-transmitted-symbol in the modulated signal (Joule).

• Œ = \ = Double-sided noise power spectral density (in W/Hz = Joule).

©
• Eb = Average energy-per-transmitted-bit in the modulated signal (Joule) = Es/k.

• To = Bit duration.
• Tsymb = Symbol duration = k To
• BER = Probability of bit-error = bit error rate.

Modulation with AWGN Error Probability

H#
?HI = Q ¦ • §
ASK
‘!

H#
?HI = Q ¦ • §
FSK
‘!

2H#
?HI = Q ¦ • §
BPSK
‘!

2H#
?HI = Q ¦ • §
QPSK
‘!

2 H#
?HI ≅ Q ¦ •2D sin h j §
PSK (order M)
D ‘! «

4 1 3D H#
~# = _1 − `Q¦ • §
QAM (order M)
D √« « − 1 ‘!
(Rectangular QAM)

?HI = 1 − 1 − ~# 7

Page 14 of 18
Example:

amplitude of the carrier is = 0.5 V, data rate is 2 bps, and ‘! = 2 10L7 W/Hz.
Find the BER for BPSK if we use an optimal detector (a matched filter). Assume the

Solution:
In BPSK there is one symbol per bit (i.e., a total of two symbols that the modulated
signal can assume). The two symbols can be written as:

m = cos m7 = − cos = cos −

The energy-per-symbol here is the same as the energy-per-bit and is equal for both
possible symbols. Hence, its average is:

7 7 7 7 7
1
H# = H% = › „%- # œ Pr6a: + › „%- # œ Pr6\: = „%- = „} =
2 2 2 #
2 2 N!
Hence,
2H# 7 0.57
BER = Q ¦ • § = Q¦• § = Q¦• § = Q…√6.25† = Q 2.5
‘! ‘! N! 2 10L7 2
= 6.21 10L+

Example:

detector (a matched filter). Assume the data rate is 4 bps, and ‘! = 5 10L7 W/Hz.
Find the BER for the 16-QAM constellation shown below if we use an optimal

2
2.236
1

1.414
I
−2 −1 1 2
−1

2.828
−2

Solution:
In this system there are 16 possible symbols, which we assume to be equally
probable, i.e., each occurs with a probability of 1/16. Hence, the energy-per-symbol

1.4147 4 2.2367 8 2.8287 4

is:

H% = › „%- # œ _ ` + › „%- # œ _ ` + › „%- # œ _ `

2 16 2 16 2 16

H% = 60.25 + 1.25 + 1:„%- # = 2.5…„%- #†

H% „%- # 2.5 2.5

H# = = 2.5 _ ` = 2.5 „! = = = 0.625 J
D D N! 4

Page 15 of 18
 4 1 3D H# 4 1 3 4 0.625 3
~# = _1 − `Q¦ • § = _1 − `Q¦ • § = Q… √10†
D √« « − 1 ‘! 4 √16 16 − 1 0.05 4
3 3
= Q 3.162 = 8 10L> = 6 10L>
4 4

?HI = 1 − 1 − ~# 7
= 1− 1−6 10L> 7
= 1.2 10L+

Comparison of Digital Modulation Schemes

We show below the BER curves for the different digital modulation schemes:
0
10
ASK, FSK
BPSK, QPSK
-1 8-PSK
10
16-PSK
16-QAM
-2 64-QAM
10
256-QAM

-3
10
BER

-4
10

-5
10

-6
10

3 dB 4.5 dB
-7
10
0 5 10 15 20 25
E /N (dB)
b 0

Comparing BPSK and QPSK with ASK and FSK, we notice that BPSK and QPSK
provide smaller bit error rate for the same Eb/No. In other words, for the same bit
error rate, we need less signal-to-noise ratio (Eb/No) to send BPSK and QPSK. This
means that BPSK and QPSK have better immunity to noise than ASK and FSK.

Notice also that the performance of BPSK is the same as that for QPSK, while the
performance of 8-PSK and 16-PSK are worse (i.e., they require more signal-to-noise
ratio to achieve the same bit error rate). This is an expected result because 8-PSK and
16-PSK have more constellation diagram points (which are now closer and closer to
each other).
Also notice how 16-QAM has a superior performance compared to 16-PSK, which is
to be expected because the constellation points are further apart in 16-QAM
compared to 16-PSK.

Page 16 of 18
noise ratio (Eb/No) to achieve near error free transmission (this is ?HI ≈ 10L= ).
The following table shows the bandwidth requirements and the necessary signal-to-

Notice that for higher order modulation techniques, we require less bandwidth but
we need more signal-to-noise ratio (Eb/No) to maintain small bit error rate (i.e., to
maintain good performance).

2N}
Modulation Bandwidth Error free Eb/No (i.e., BER < 10-6)

2∆N + 2? = 2N} C + 1
ASK 13.5 dB

2 ?@AB = 2N}
FSK 13.5 dB

2 ?@AB = N}
BPSK 10.5 dB

2 ?@AB = 2N} /3
QPSK 10.5 dB

2 ?@AB = N} /2
8-PSK 14 dB

2 ?@AB = N} /2
16-PSK 18 dB

2 ?@AB = N} /3
16-QAM 14.5 dB

2 ?@AB = N} /4
64-QAM 18.5 dB
256-QAM 23.4 dB

Remember this for digital modulation:

The available bandwidth of the channel decides the baud rate (symbols per second)

The available signal-to-noise ratio (H# /‘! ) decides the level of modulation you can
you can send.

use while still maintaining a small bit error rate (good quality). In other words, it
decides the number of bits you can send per symbol.
Hence, the two factors together (bandwidth and SNR) decide the total bit rate you
can achieve over any single channel. Now compare this to Shannon’s Limit!!

Applications of digital modulation techniques:

The following are some current-day communication systems that use digital
modulation:

IEEE 802.11 (Wi-Fi): A very important Wireless Local Area Networking technology.
Since Wi-Fi has many variants, it uses different modulation techniques such as:
BPSK, QPSK, 16-QAM, 64-QAM and CCK (Complementary Code Keying) (CCK is
an extension of QPSK).

IEEE 802.16 (Wi-MAX): A very important Wireless Metropolitan Area Network,

and currently competes with ADSL for Internet delivery. Wi-MAX switches
dynamically between different modulation schemes such as: BPSK, QPSK, 16-QAM,
and 64-QAM. It uses these modulation schemes in combination with OFDM
(Orthogonal Frequency division multiplexing) (OFDM is an extension of FDM).

DVB (Digital Video Broadcasting): This is the European standard for digital
television broadcasting. There are many variants within the standard: DVB-S (for
satellite broadcasting) uses QPSK or 8-PSK; DVB-C (for cable) uses 16-QAM,
32-QAM, 64-QAM, 128-QAM or 256-QAM; and DVB-T (for terrestrial television
broadcasting) uses 16-QAM or 64-QAM.

Page 17 of 18
DAB (Digital Audio Broadcasting): Future European standard for digital radio
broadcasting, which should replace AM and FM radio broadcasting. DAB use
DQPSK (Differential QPSK) (DQPSK is a variation of QPSK).

ADSL: Currently one of the main choices for connecting to the Internet. Uses
adaptive QAM in a scheme called DMT (Discrete Multi-Tone modulation).

Shannon's Limit:
You see by now that the two main resources in communication systems are the
channel bandwidth and the transmitted power (or SNR). In a given communication
channel, one resource may be more valuable than the other, and the communication
scheme should be designed accordingly.

The limitation imposed on communication by the channel bandwidth and the SNR
is dramatically highlighted by Shannon's noisy channel theorem, which applies for
channels contaminated with Additive White Gaussian Noise (AWGN) noise.
Shannon's equation states that:

C = B log2 (1 + SNR) bits/second

where C is the rate of information transmission in bits per second, B is the channel
bandwidth (in Hz), and SNR is the signal-to-noise ratio (unitless) on the channel.

This rate C (known as the channel capacity) is the maximum number of binary bits
that can be transmitted per second with a probability of error arbitrarily close to
zero. In other words, it is impossible to transmit at a rate higher than C without
incurring errors. Shannon's equation clearly brings out the limitation on the rate of
communication imposed by B and SNR. This is why bandwidth and SNR are the
main quantities that we study in communication systems. Notice that if noise was
zero (i.e., SNR = ∞), or the bandwidth was infinite (B = ∞), then we can transmit
infinity information and communication would cease to be a problem.

It should be remembered that Shannon's result represents the upper limit on the rate
of communication over a channel and can be achieved only with a systems of great
complexity and a time delay in reception approaching infinity. Practical systems
operate at rates below the Shannon rate.

Page 18 of 18