EEE 437 Wireless Communication

Performance of Digital Modulation

Dr. Md. Forkan Uddin

Professor
Department of EEE, BUET

Curtesy: Dr. Md. Farhad Hossain

 A Communication System
Transmitter

Receiver

2
 Introduction
 Digital modulation consists of mapping a bit stream of finite length into an analog
signal for transmission over the channel.
 Detection consists of determining the original bit sequence based on the signal
received over the channel.
 The main considerations in choosing a particular digital modulation technique are
 high data rate
 high spectral efficiency (minimum bandwidth occupancy)
 high power efficiency (minimum required transmit power)
 robustness to channel impairments (minimum probability of bit error)
 low power/cost implementation
 Often these are conflicting requirements, and the choice of modulation is based on
finding the technique that achieves the best tradeoff between these requirements.
 Once the modulation technique is determined, the constellation size must be chosen.
Modulations with large constellations have higher data rates for a given signal bandwidth,
but are more susceptible to noise, fading, and hardware imperfections. 3
 Signal Space Analysis
 Digital modulation encodes a bit stream of finite length into one of several
possible transmitted signals.
 The receiver minimizes the probability of detection error by decoding the
received signal as the signal in the set of possible transmitted signals that is
“closest” to the one received.
 Determining the distance between the transmitted and received signals requires
a metric for the distance between signals.
 By representing signals as a vector in a vector space, we can have the metric for
the distance between signals.

QPSK

4
 Signal Space Analysis

 Let the source has M={m1, . . .,mM} set of all possible messages, where ith message
sequence mi = {b1, . . . , bK} ∈ M is a bit sequence of length K. There are M = 2K
possible sequences of K bits and thus K = log2M.
 The message mi has a probability pi of being selected for transmission, where sum of
all pi is equal to 1.
 Every T seconds, the system sends K = log2M bits of information through the channel
and thus, data rate R = K/T bps.
 Suppose message mi is to be transmitted over the channel during the time interval [0,
T). Since the channel is analog, the message must be embedded into an analog signal
for channel transmission. Thus, each message mi ∈ M is mapped to a unique analog
signal si(t) ∈ S = {s1(t), . . . , sM(t)}, where si(t) is defined on the time interval [0, T) and
has energy
5
 Signal Space Analysis
 Since each message represents a bit sequence, each signal si(t) ∈ S also represents
a bit sequence, and detection of the transmitted signal si(t) at the receiver is
equivalent to the detection of the transmitted bit sequence.
 When messages are sent sequentially, the transmitted signal becomes a sequence
of the corresponding analog signals and s(t) thus becomes

where si(t) is the analog signal corresponding to the message mi designated for the
transmission interval [kT, (k+1)T).
 An example of the transmitted signal s(t) = s1(t) + s2(t − T) + s1(t − 2T) + s1(t − 3T)
corresponding to the string of messages m1, m2, m1, m1 with message mi mapped to
signal si(t) is shown below.

6
 Signal Space Analysis
 In the AWGN model, transmitted signal is sent through an AWGN channel, where a
white Gaussian noise process n(t) of power spectral density N0/2 is added to form the
received signal r(t) = s(t) + n(t).
 Given r(t), the receiver must determine the best estimate of which si(t) ∈ S was
transmitted during each transmission interval [kT, (k + 1)T).
 This best estimate for si(t) is mapped to a best estimate of the message mi(t)∈ M and
the receiver then outputs this best estimate of the transmitted
bit sequence.
 The goal of the receiver design in estimating the transmitted message is to minimize
the average probability of message error:

over each time interval [kT, (k + 1)T).

 By representing the signals {si(t), i = 1, . . ., M} geometrically, we can solve for the
optimal receiver design in AWGN based on a minimum distance criterion. 7
 Geometric Representation of Signals
Any set of M real energy signals S = {s1(t), . . . , sM(t)} defined on [0, T) can be
represented as a linear combination of N ≤ M real orthonormal basis functions {φ1(t), . .
. , φN(t)}.
Thus, we can write each si(t) ∈S in terms of its basis function representation as

If all the signals {si(t)} are linearly independent, then N = M, otherwise N <M.
Example

Using Gram–Schmidt
orthogonalization process 8
 Geometric Representation of Signals
 We denote the coefficients {sij} as a vector si = (si1, . . . , siN ) ∈ RN which is
called the signal constellation point corresponding to the signal si(t). The
signal constellation consists of all constellation points {s1, . . . , sM}.
 si(t) can be obtained from si and si can be obtained from si(t). Thus, it is
equivalent to characterize the transmitted signal by si(t) or si.
 The representation of si(t) in terms of its constellation point si ∈ R N is called
its signal space representation and the vector space containing the
constellation is called the signal space.

One-dimensional signal space A two-dimensional signal space

9
 Geometric Representation of Signals

 The length of a vector in R N:

 The distance between two signal constellation points si and sk:

 Inner product of two signals:

10
 Q function

Some useful bounds for the Q function for x > 0:

11
 Performance in Wireless Channel

 We now consider the performance of the digital modulation techniques when used
over AWGN channels and channels with flat fading.
 There are two performance criteria of interest: the probability of error, defined
relative to either symbol or bit errors, and the outage probability, defined as the
probability that the instantaneous signal-to-noise ratio falls below a given threshold.
 Flat fading can cause a dramatic increase in either the average bit-error-rate or
the signal outage probability.
 Wireless channels may also exhibit frequency selective fading and Doppler shift.
 Frequency-selective fading gives rise to intersymbol interference (ISI), which
causes an irreducible error floor in the received signal.
 Doppler causes spectral broadening, which leads to adjacent channel interference
(typically small at reasonable user velocities).
12
 Performance in AWGN Channel
 In an AWGN channel, the modulated signal s(t) = R{u(t)ej2πfct} has noise n(t) added
to it prior to reception, where u(t) is the complex envelope of s(t).
 The noise n(t) is a white Gaussian random process with mean zero and power
spectral density N0/2. The received signal is thus, r(t) = s(t) + n(t).
 Received signal-to-noise power ratio (SNR): Ratio of the received signal power
Pr to the power of the noise within the bandwidth of the transmitted signal s(t).
 The received power Pr is determined by the transmitted power, and the path-loss,
shadowing, and multipath fading. AWGN channel models ignore the last three factors.
 The noise power is determined by the bandwidth of the transmitted signal and the
spectral properties of n(t). Specifically, if the bandwidth of the complex envelope u(t) of
s(t) is B, then the bandwidth of the transmitted signal s(t) is 2B. Since the noise n(t) has
uniform power spectral density N0/2, the total noise power within the bandwidth 2B is
PN = N0/2 × 2B = N0B. Eb: Signal energy per bit
 So the received SNR is given by Es: Energy per symbol
Ts: Symbol time
Tb: Bit duration/period
(Binary modulation:
13
Ts = Tb and Es = Eb)
 Performance in AWGN Channel
 The quantities γs = Es/N0 and γb = Eb/N0 are sometimes called the SNR per
symbol and the SNR per bit, respectively.
 For performance specification, we are interested in the bit error probability Pb as a
function of γb.
 However, for M-ary signaling (e.g., MPAM and MPSK), the bit error probability Pb
depends on both the symbol error probability and the mapping of bits to symbols. Thus,
we typically compute the symbol error probability Ps as a function of γs based on
the signal space concepts and then obtain Pb as a function of γb using an exact or
approximate conversion.
 The approximate conversion typically assumes that the symbol energy is divided
equally among all bits, and that Gray encoding is used so that at reasonable SNRs, one
symbol error corresponds to exactly one bit error. These assumptions for M-ary
signaling lead to the approximations:

14
 Error Probability: BPSK
With binary modulation, each symbol corresponds to one bit, so the symbol and bit
error rates are the same.

The transmitted signal s(t):

s1(t) = Ag(t) cos(2πfct) = Aφ1(t) to sent a 1 bit &
s2(t) = - Ag(t) cos(2πfct) = - Aφ1(t) to send a 0 bit.

dmin: Minimum distance in among constellation points

15
Error Probability: BPSK

16
 Error Probability: BPSK
Received signal Gaussian distributed:

or

Probability of error if s1 is transmitted:

17
 Error Probability: BPSK

Due to symmetry:

Average symbol probability of error:

(Symbols are considered equiprobable)

18
 Error Probability: QPSK
QPSK modulation consists of BPSK modulation on both the in-phase and
quadrature components of the signal. With perfect phase and carrier recovery, the
received signal components corresponding to each of these branches are orthogonal.
Therefore, the bit error probability on each branch is the same as for BPSK.
1. Exact symbol error probability equals the probability that either branch has a bit
error:

Since the symbol energy is split between the in-phase and

quadrature branches, we have γs = 2γb.

2. Union bound:

Writing this in terms of γs = 2γb = Es/N0 = A2/N0 yields

19
Error Probability: QPSK

20
 Error Probability: QPSK
3. Nearest-neighbor bound:
Minimum distance between constellation points:

With Gray encoding, we can approximate Pb from Ps:

Pb ≈ Ps /2
(since we have 2 bits per symbol)
Decision regions

The reflected binary code or Gray code is an ordering

of the binary numeral system such that two successive
values differ in only one bit (binary digit). 21
 Error Probability: QPSK
Example 6.1:
Find the bit error probability Pb and symbol error probability Ps of QPSK
assuming γb = 7dB. Compare the exact Pb with the approximation Pb = Ps/2
based on the assumption of Gray coding. Finally, compute Ps based on the
nearest-neighbor bound using γs = 2γb, and compare with the exact Ps.

22
 Error Probability: MPSK
For MPSK all of the information is encoded in the phase of the
transmitted signal.
Thus, the transmitted signal over one symbol time is given by

Comparing above si(t) with the generalized below form, we can get constellation points.

Here, g(t) is pulse shaping filter, which can be rectangular or any suitable pulse. Also, φ1(t) = g(t)
cos(2πfct) and φ2(t) = g(t) sin(2πfct) are the orthonormal basis functions with unit energy.
Constellation points or symbols (si1, si2):

Es = A2 & γs = A2/N0 23
 Error Probability: MPSK

Decision regions

An error occurs if the ith signal constellation point is transmitted and

The joint distribution of r and θ can be obtained through a bivariate transformation of

the noise n1 and n2 on the in-phase and quadrature branches

24
 Error Probability: MPSK
Since the error probability depends only on the distribution of θ, we can integrate out
the dependence on r, yielding

By symmetry, the probability of error is the same for each constellation point.
Thus, we can obtain Ps from the probability of error assuming the constellation point s1
= (A, 0) is transmitted, which is

A closed-form solution to this integral does not exist for M > 4, and hence the
exact value of Ps must be computed numerically.

25
 Error Probability: MPSK

Each point in the MPSK constellation has two nearest neighbors at distance dmin
= 2A sin(π/M). Thus, the nearest neighbor approximation to Ps is given by

26
Error Probability: MPSK

27
 Error Probability: MPSK
Example 6.2
Compare the probability of bit error for 8PSK and 16PSK assuming γb = 15 dB
and using the approximated Ps.

Note that Pb is much larger for 16PSK than for 8PSK for the same γb. This result is
expected, since 16PSK packs more bits per symbol into a given constellation, so for a fixed
energy-per-bit, the minimum distance between constellation points will be smaller. 28
 Error Probability: MPAM
Constellation points are:

Decision regions
for 8-PAM

Demodulations of MPAM:

Only a single basis function as MPAM is one-dimensional:

29
 Error Probability: MPAM
As MPAM is one-dimensional, it has no quadrature component (si2 = 0).
For MPAM all of the information is encoded into the signal amplitude Ai = (2i-1-M)d
Transmitted signal over one symbol time is given by

The amplitude of the transmitted signal takes on M different values, which implies
that each pulse conveys log2M = K bits per symbol time Ts.
Each of the (M-2) inner constellation points have two nearest neighbors at distance
2d. The outermost 2 points have only one neighbor at a distance 2d.
Minimum distance between the constellation points: dmin = mini,j |Ai − Aj | = 2d
Energy of ith symbol:

Average energy per symbol:

30
 Error Probability: MPAM
For inner constellation points:

For outermost constellation points:

Probability of symbol error Ps:

Probability of symbol error:

31
 Error Probability: MQAM

Decision Regions for 16-QAM

For MQAM, the information bits are encoded in both the amplitude and phase of
the transmitted signal.
Thus, both MPAM and MPSK have one degree of freedom in which to encode the
information bits (amplitude or phase), MQAM has two degrees of freedom.
As a result, MQAM is more spectrally-efficient than MPAM and MPSK, in that it can
encode the most number of bits per symbol for a given average energy. 32
 Error Probability: MQAM
Transmitted signal:

Energy of transmitted signal:

For square signal constellations, si1 and si2 take values:

Ai = (2i − 1 − L)d, i = 1, 2, . . ., L = 2l
l is the bit per dimension and bits/symbol = 2l.
MQAM Constellations size: M = L2
Minimum distance between signal points dmin = 2d, the same as for MPAM.
In fact, MQAM with square constellations of size M = L2 is equivalent to
MPAM modulation with constellations of size L on each of the in-phase and
quadrature signal components, each with half the energy of the original
33
MQAM system.
 Error Probability: MQAM
The symbol error probability for each branch of the MQAM system is same as that
of MPAM) with M replaced by L: and equal to the average
energy per symbol in the MQAM constellation.

(for each branch)

Exact probability of symbol error for the MQAM system:

The nearest neighbor approximation to probability of symbol error depends on

whether the constellation point is an inner or outer point. If we average the nearest
neighbor approximation over all inner and outer points, we obtain the MQAM
probability of error:

34
 Error Probability: MQAM
For nonrectangular constellations, it is relatively straightforward to show that the
probability of symbol error is upper bounded as

The nearest neighbor approximation for nonrectangular

constellations:

Mdmin: Largest number of nearest neighbors for any constellation point in the
constellation
dmin: Minimum distance in the constellation

35
 Approximate Error Probability: Coherent Detections
General form (approximated):

αM : Number of nearest neighbors to a constellation at the minimum distance

βM : A constant that relates minimum distance to average symbol energy

36
 Error Probability: Fading Channels
 In AWGN channel, the probability of symbol error Ps depends on the received
SNR or, equivalently, on γs.
 In a fading environment, the received signal power varies randomly over
distance or time due to shadowing and/or multipath fading.
 Thus, in fading γs is a random variables with distribution pγs(γ), and therefore
Ps(γs) is also random.
 The performance metric when γs is random depends on the rate of change of the
fading.
 Three different performance criteria that can be used to characterize the random
variable Ps:
1. Outage probability, Pout : It is the probability that γs falls below a given value
corresponding to the maximum allowable symbol error Ps.
2. Average error probability : It is Ps averaged over the distribution of γs.
3. Combined average error probability and outage: It is defined as the
average error probability that can be achieved at some percentage of time or
some percentage of spatial locations.
37
 Error Probability: Fading Channels
1. Outage probability Pout: If the signal power is changing slowly (Ts << Tc), then a
deep fade will affect many simultaneous symbols. Thus, fading may lead to large
error bursts, which cannot be corrected with coding of reasonable complexity.
Therefore, these error bursts can seriously degrade end-to-end performance. In this
case, acceptable performance cannot be guaranteed over all time or,
equivalently, throughout a cell, without drastically increasing transmit power.
Under these circumstances, an outage probability is specified so that the channel
is deemed unusable for some fraction of time or space.
2. Average probability of symbol error applies when the signal fading is on the
order of a symbol time (Ts ≈ Tc), so that the signal fade level is constant over
roughly one symbol time. Since many error correction coding techniques can
recover from a few bit errors, and end-to-end performance is typically not seriously
degraded by a few simultaneous bit errors, the average error probability is a
reasonably good figure of merit for the channel quality under these conditions.
3. Outage and average error probability are often combined when the channel is
modeled as a combination of large-scale and small-scale fading, e.g., log-
normal shadowing with fast Rayleigh fading. 38
Error Probability: Fading Channels

Deep fade (Ts ≈ Tc)

Deep fade (Ts << Tc)

39
 1. Outage Probability: Fading Channels
 The outage probability relative to γ0 (minimum SNR required for acceptable
performance) is defined as:

In Rayleigh fading, the outage probability becomes:

Inverting this formula shows that for a given outage probability Pout, the required
average SNR is

In dB, this means that 10log10(γs) must exceed the target 10log10(γ0) by
Fd = −10 log10[−ln(1 − Pout)] to maintain acceptable performance more than (1−Pout)
*100% of the time. The quantity Fd is typically called the dB fade margin. 40
Outage Probability: Fading Channels

41
 2. Average Probability of Error: Fading Channels
 The average probability of error is used as a performance metric when Ts ≈ Tc. Thus,
we can assume that γs is roughly constant over a symbol time.
 Then the averaged probability of error is computed by integrating the error
probability in AWGN over the fading distribution:

where Ps(γ) is the probability of symbol error in AWGN with SNR γ.

For a given distribution of the fading amplitude r (i.e., Rayleigh, Rician, log-normal,
etc.), we compute by making the change of variable as below
This follows from the fact that the probability contained in a
differential area must be invariant under change of variables.

In Rayleigh fading, the received signal amplitude r has the Rayleigh distribution,

and the signal power is exponentially distributed with mean 2σ2. 42

 Average Probability of Error: Fading Channels

The SNR per symbol Es/N0 for a given amplitude r is

Where is the PSD of the noise in the in-phase and quadrature branches.

 Differentiating both sides of this expression yields

Then, we get,

For binary signaling, this (pdf of SNR per symbol γs) becomes,

43
 Average Probability of Error: Fading Channels

For BPSK in AWGN channel,

Integrating this BPSK Pb over the above pdf of γb yields,

If we use the general approximation , then the

average probability of symbol error Ps in Rayleigh fading can be
approximated as

44
BPSK in Rayleigh Fading

BPSK

45
MQAM in Rayleigh Fading

MQAM

46
3. Combined Outage and Average Error Probability
When the fading environment is a superposition of both large-scale fading and
small-scale fading (i.e., log-normal shadowing and small-scale fading), a common
performance metric is combined outage and average error probability, where
outage occurs when the large-scale fading falls below some target value and the
average performance in non-outage is obtained by averaging over the small-scale
fading.
We use the following notation:
Let, γs denotes the random SNR for fixed path-loss, random shadowing and random
small-scale fading.
Let, denotes the (random) SNR per symbol for a fixed path-loss and random
shadowing, but averaged over small-scale fading. Its average value is . So, is
random due to shadow fading.
Let, denotes the average SNR per symbol for a fixed path-loss with averaging
over small-scale fading and shadowing.

An outage is declared when the received SNR per symbol due to shadowing and
path-loss alone, , falls below a given target value . 47
 Combined Outage and Average Error Probability
An outage is declared when the received SNR per symbol due to shadowing and
path-loss alone, , falls below a given target value , i.e., < .
When not in outage , the average probability of error is obtained by
averaging over the distribution of the small-scale fading conditioned on the mean
SNR:

The criterion used to determine the outage target is typically based on a

given maximum average probability of error, i.e., , where the target
must then satisfy

Clearly whenever , the average error probability will be below the target
value.
48
Q-function Table

49
The END

50
 Additional Information (David Tse)
We make the standard assumption that w(t) is zero-mean additive white Gaussian noise (AWGN) with
power spectral density N0/2, i.e., E[w(0)w(t)] = N0/2 δ(t).

Hence the discrete-time noise process w(m) is white, i.e., independent over time; moreover, the real and
imaginary components are i.i.d. Gaussians with variances N0/2.

Complex noise, w ∼ CN(0, N0)

The noise energy per complex symbol time is N0

The assumption of AWGN essentially means that we are assuming that the primary source of the noise is at
the receiver or is radiation impinging on the receiver that is independent of the paths over which the signal is
being received. This is normally a very good assumption for most communication situations.
The orthogonal modulation scheme considered here uses only real symbols and hence transmits only on the
I channel. Hence it may seem more natural to define the SNR in terms of noise energy per real symbol, i.e.,
N0/2. However, later we will consider modulation schemes that use complex symbols and hence transmit on
both the I and Q channels. In order to be consistent throughout, we choose to define SNR this way.

Channel tap gains: For Rayleigh fading, all the taps are modeled as: h ∼ CN(0, σ2). Then the magnitude |h|
of each tap is Rayleigh distributed and |h|2 is exponentially distributed. For flat fading, only one tap in the
channel model.
Usually, for Rayleigh fading, h ∼ CN(0, 1), where we normalize the variance to be 1.
Recall that we normalized the channel gain such that E[h2] = 1. 51