Chapter 5:

Signaling on ISI Channels and

Equalization

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE DHT, HCMUT
 References

[1] J. G. Proakis, Digital Communications, 5th Edition, McGraw-

Hill, 2008.

[2] Athanasios Papoulis, Probability, Random Variables, and Stochastic

Processes, McGraw-Hill, 1991 (3rd Ed.), 2001 (4th Ed.).

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 2 DHT, HCMUT
 5. Goals (1)
 In this chapter, we consider the problem of signal design when the
channel is bandlimited to some specified bandwidth of W Hz. Under
this condition, the channel may be modeled as a linear filter having an
equivalent lowpass frequency response C( f ) that is zero for | f | > W.
 The first topic that is treated is the design of the signal pulse g(t) in a
linearly modulated signal, represented as

that efficiently utilizes the total available channel bandwidth W. We

shall see that when the channel is ideal for | f| ≤ W, a signal pulse can
be designed that allows us to transmit at symbol rates comparable to or
exceeding the channel bandwidth W. On the other hand, when the
channel is not ideal, signal transmission at a symbol rate equal to or
exceeding W results in intersymbol interference (ISI) among a
number of adjacent symbols.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 3 DHT, HCMUT
 5. Goals (2)
 The second topic that we consider is the design of the receiver in the
presence of intersymbol interference and AWGN. The solution to the
ISI problem is to design a receiver that employs a means for
compensating or reducing the ISI in the received signal. The
compensator for the ISI is called an equalizer.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 4 DHT, HCMUT
 5. Characterization of Band-limited Channel (1)
 A bandlimited channel such as a telephone channel will be
characterized as a linear filter having an equivalent lowpass frequency-
response characteristic C( f ). Its equivalent lowpass impulse response
is denoted by c(t). Then, if a signal of the form:

is transmitted over a bandpass telephone channel, the equivalent low-

pass received signal is

where the integral represents the convolution of c(t) with v(t), and z(t)
denotes the additive noise. Within the bandwidth of the channel, we
may express the frequency response C( f ) as

where |C( f )| is the amplitude-response characteristic and θ( f ) is the

phase-response characteristic.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 5 DHT, HCMUT
 5. Characterization of Band-limited Channel (2)
 Furthermore, the envelope delay characteristic is defined as

A channel is said to be nondistorting or ideal if the amplitude response

|C(f)| is constant for all | f | ≤ W and θ(f) is a linear function of
frequency, i.e., τ (f) is a constant for all | f | ≤ W. On the other hand, if
|C(f)| is not constant for all | f | ≤ W, we say that the channel distorts
the transmitted signal V(f) in amplitude, and, if τ(f) is not constant
for all | f | ≤ W, we say that the channel distorts the signal V(f) in delay.

As a result of the amplitude and delay distortion caused by the

nonideal channel frequency-response characteristic C(f), a succession
of pulses transmitted through the channel at rates comparable to the
bandwidth W are smeared to the point that they are no longer
distinguishable as well-defined pulses at the receiving terminal.
Instead, they overlap, and, thus, we have intersymbol interference.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 6 DHT, HCMUT
5. Characterization of Band-limited Channel (3)
Example: Effect of channel distortion: (a) channel input; (b) channel
output; (c) equalizer output.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 7 DHT, HCMUT
 5. Digital Transmissions
Digital transmission techniques are becoming more popular in all areas of
telecommunications.

Most of the new telecommunication systems are based on digital

technology.

Baseband digital transmission:

 Digital PAM
 Line coding
 Nyquist pulse shaping

Digital modulation methods: QAM, PSK, FSK, MSK digital modulations

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 8 DHT, HCMUT
 5. Bits and Symbols
The idea of digital transmission is to transmit bit sequences or, more generally,
multilevel symbol sequences using PAM-modulation (pulse amplitude
modulation).

A multilevel symbol is obtained when several bits are combined into a

symbol. E.g., if 4 bits are combined into 1 symbol, then the number of bit
combinations is 24 = 16. In general, B bits can be represented as M = 2B
levels.

The number of levels or bit combinations depends on the application and on

channel requirements, so that levels can be distinct in noisy channel.

If we combine several bits into one symbol, the symbol rate (or baud rate) is
reduced. This affects the transmitted signal bandwidth.

Transmitting bandwidth sets an upper limit to the symbol rate and noise causes
errors. Thus, bit/symbol rates and error probability play important roles in
digital transmission (similar to bandwidth and S/N in analog transmission)
Dept. of Telecomm. Eng. Comm II 2013
Faculty of EEE 9 DHT, HCMUT
 5. Digital PAM Signals (1)
Digital message representation at baseband takes a form of an amplitude
modulated pulse (PAM) train. Digital PAM-signal is transmitted over
the continuous-time channel as the following waveform:

Here p(t) is a basic pulse waveform whose amplitude is scaled by the

transmitted symbol ak .

It’s important that adjacent pulses do not interfere with each other in the
reception. Ideally, this happens when the following condition holds:

This condition ensures that we can recover the message by sampling x(t)
periodically at t = KD, where K = ± 1, ± 2, … since:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 10 DHT, HCMUT
 5. Digital PAM Signals (2)
The rectangular pulse satisfies the condition for p(t) if τ ≤ D. D may
not need to be pulse duration but rather pulse-to-pulse interval or the time
allotted to one symbol. Thus, the signaling rate is:

measured in symbols per second (or baud). In the case of binary signaling
(M = 2), we write D = Tb for the bit duration and the bit rate is:

measured in bits per second (bps or b/s).

For the condition of p(t), this is possible to implement in two different ways:
 Using short pulses not overlapping in time domain (i.e. line coding)
⇒ the bandwidth is not the smallest possible, but easy to implement.
 Using pulses overlapping in time domain (i.e., Nyquist pulse
shaping) ⇒ signal bandwidth can be minimized, more complicated.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 11 DHT, HCMUT
 5. Digital PAM Signals (3)
Spectrum of Digital PAM Signal:
The power spectrum of the transmitted digital baseband signal can be
obtained from the Fourier transform of digital PAM signal x(t):

where P(f) is the Fourier transform of the pulse waveform p(t) and Ra(n) is
the autocorrelation function of ak.

The transmitted signal spectrum must be matched with the channel

properties. In baseband systems, e.g., in cables, the degradation is not
constant within the used frequency band. Typically, the channel degradation
is increased in high frequencies.

Therefore, a higher signal power should be put into lower frequencies,

where the cable degradation is the smallest. This reduces cross talks and
radio distortions.
Dept. of Telecomm. Eng. Comm II 2013
Faculty of EEE 12 DHT, HCMUT
 5. Line Coding vs. Nyquist-Pulse Shaping (1)
Two different approaches to shape spectrum:
(1) Line coding
 The pulse waveform is a square pulse.
 The spectrum is a sinc-type wide spectrum.
 The DC-component can be removed by constructing the signal
properly.
 Usually, the symbol train is generated to have some correlation, in
order to modify (or “shape”) the transmitted spectrum.
 Mostly used for binary signaling.
 In a pure line coding, the bandwidth consumption is not a limiting
factor.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 13 DHT, HCMUT
 5. Line Coding vs. Nyquist-Pulse Shaping (2)
(2) Nyquist-pulse shaping
 Used when we want to reduce the bandwidth consumption. It is
assumed that transmitted symbols are uncorrelated ⇒ the
transmitted spectrum has the shape of Fourier transform of the
pulse waveform.
 The pulse waveform is optimized so that the needed bandwidth is
small ⇒ adjacent pulses are overlapped in time domain.
 The methods can also be combined. In practical systems, one of
them is chosen.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 14 DHT, HCMUT
 5. Line Coding (1)
The goal of line coding:
 Spectrum management and spectrum shaping.
 To remove the variation of DC-component in AC-coupled systems.
 To avoid synchronization problems when the transmitted symbol
train consists of long sequences with constant 0 or 1.
 System monitoring during the normal operation is possible by
using proper line codes.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 15 DHT, HCMUT
 5. Line Coding (2)

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 16 DHT, HCMUT
 5. Line Coding (3)

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 17 DHT, HCMUT
 5. Transmission Limitations (1)
Digital baseband transmission model:

The signal-plus-noise-and-interference waveform:

where td is transmission delay and stands for pulse shape with

transmission distortion (See a possibility waveform of y(t) in next slide).

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 18 DHT, HCMUT
 5. Transmission Limitations (2)

The task of the regenerator is to recover the digital message from y(t).
The synchronization signal may help the regenerator by identifying the
optimum sampling times:

If then

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 19 DHT, HCMUT
 5. Transmission Limitations (3)
When rectangular pulses are passed through a bandlimited channel, the
pulses will spread in time and the pulse for each symbol will smear into the
time intervals of succeeding symbols. This leads to an increased
probability of the receiver making an error in detecting a symbol ⇒
intersymbol interference – ISI.

The combined effects of noise and ISI may result in errors in the
regenerated message.

If n(t) is white noise, then the noise power can be reduced by reducing the
bandwidth of the LPF at receiver. However, the low pass filtering causes
pulses to spread out which would increase the ISI. Consequently, the
fundamental limitations of digital transmission is the relationship
between ISI, bandwidth and signaling rate.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 20 DHT, HCMUT
 5. Transmission Limitations (4)
The Nyquist statement:
Given an ideal low-pass channel with bandwidth B, it is possible to
transmit independent symbols at a rate r ≤ 2B baud without ISI. It is not
possible to transmit independent symbols at rate r > 2B.

Signaling at the maximum rate r = 2B requires a special pulse shape, that

is sinc pulse:

having the bandlimited spectrum:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 21 DHT, HCMUT
 5. Transmission Limitations (5)
Eye diagram:
An experimental display to know the channel characteristics, it further
clarifies digital transmission limitations.

Distorted polar binary signal and eye diagram:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 22 DHT, HCMUT
 5. Transmission Limitations (6)
General binary eye diagram:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 23 DHT, HCMUT
 5. Bandlimited Digital PAM Systems (1)
Consider digital baseband transmission with bandlimited channel.
Consequently, the rectangular signaling pulses would be severely distorted
(resulting in intersymbol interference - ISI). Instead, we must use
bandlimited pulses specially shaped to avoid ISI.

 Nyquist- Pulse Shaping:

Assumed that noise is absent, the signal at the input of the regenerator is:

As before, the condition for p(t) is:

which eliminates ISI, but now we impose additional requirement that the
pulse spectrum be bandlimited:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 24 DHT, HCMUT
 5. Bandlimited Digital PAM Systems (2)
where

This means that the signaling rate is:

in which B may be considered as the minimum required transmission

bandwidth, so that BT ≥ B.

The Nyquist theorem states that the above bandlimited spectrum is

satisfied if the p(t) has the form:

With a cosine rolloff spectrum:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 25 DHT, HCMUT
 5. Bandlimited Digital PAM Systems (3)
Then, the spectrum of p(t) is:

and the corresponding pulse shape is:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 26 DHT, HCMUT
 5. Bandlimited Digital PAM Systems (4)

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 27 DHT, HCMUT
 5. Bandlimited Digital PAM Systems (5)
When β = r/2 (100% rolloff), the pulse spectrum has the raised cosine
shape:

and

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 28 DHT, HCMUT
 5. Noise and Errors (1)
We assumed that the channel is distortionless so the receiver signal is free
of ISI. Assumed that, the additive white noise with zero mean, independent
of the signal.

Binary Error Probability:

Baseband binary receiver:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 29 DHT, HCMUT
 5. Noise and Errors (2)
A sample-and-hold (S/H) extracts from y(t) the sample values:

These sample values are compared with a fixed threshold level V:

If y(tk) > V, the output of the comparator gets high level (bit 1). If y(tk) < V,
the comparator goes low level (bit 0).

Considering x(t) to be unipolar signal (ak = 1 for bit 1, and ak = 0 for bit 0).
Let variable Y represents y(tk) at an arbitrary sampling time, and n represents
n(tk).
If H0 denotes hypothesis that ak = 0 and Y = n, then the pdf:

where pN(n) is the pdf of noise alone. Similar, H1 denotes hypothesis that
ak = A and Y = A + n, then:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 30 DHT, HCMUT
 5. Noise and Errors (3)

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 31 DHT, HCMUT
 5. Noise and Errors (4)
The comparator implements the decision rule:
 Choose H0 (ak = 0), if Y < V
 Choose H1 (ak = A), if Y > V

The corresponding regeneration error probabilities are then given by:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 32 DHT, HCMUT
 5. Noise and Errors (5)
The threshold value is adjusted to minimize the average error probability:

where

Normally,

Then, for optimum threshold Vopt, we have:

Assumed that the noise is with Gaussian distribution with zero mean and
variance σ2, so:

Then, we obtain:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 33 DHT, HCMUT
 5. Noise and Errors (6)

Since pN(n) is even function and Vopt = A/2, then:

For the polar signal, ak = ± A/2, we have Vopt = 0.

From [1], we can write:

The Q function is then obtained from the Table.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 34 DHT, HCMUT
 5. Noise and Errors (7)
Regenerative Repeater:
Long-haul transmission requires repeaters. For analog repeaters, we obtain:

where (S/N)1 is signal to noise ratio after one hop and m is number of hops.
The transmitted power per repeater must be increased linearly with m. The
contaminating noise progressively builds up from repeater to repeater.

In contrast, a digital repeater is a regenerator, regenerating new digital

signal to next repeater. For m is not too large, we obtain:

It requires much smaller transmitted power per repeater than analog repeater.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 35 DHT, HCMUT
 5. Noise and Errors (8)
Matched Filtering:
Every baseband digital receiver (including repeaters) should include a
LPF designed to remove excess noise without introducing ISI. The
optimum LPF for timelimited pulses in white noise is a matched filter.

Let the received signal with duration τ (τ ≤ D) as:

The matched filter is designed to maximize the signal to noise ratio (that
means minimizing the error probability) at time:

From [1], the impulse response of the matched filter is:

with

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 36 DHT, HCMUT
 5. Noise and Errors (9)

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 37 DHT, HCMUT
 5. Noise and Errors (10)
When the x(t) is in white noise, the output noise power from the matched
filter is:

Considering binary transmission systems with bit rate rb, average received
power SR and noise density N0. We can characterize this system in terms
of two parameters:

where Eb corresponds to average energy per bit, while γb represents the

ratio of bit energy to noise density.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 38 DHT, HCMUT
 5. Noise and Errors (11)
If the signal consists of timelimited pulse p(t) with amplitude sequence ak,
then:

where for unipolar signal and for polar signal.

Thus, we obtain:

Therefore,

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 39 DHT, HCMUT
 5. Noise and Errors (12)
M-ary Error Probability:
The bit error probability (or bit error rate - BER) is:

in which:

where r is the M-ary signaling rate (symbol rate), rb is the bit rate, and SR
is average received power.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 40 DHT, HCMUT
 5. Linear Equalizer (1)
 Equivalent discrete-time model of ISI channel with AWGN:

where {Ik} is information sequence at the input of the channel, the

output sequence {vk} that can be expressed as

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 41 DHT, HCMUT
 5. Linear Equalizer (2)

where {ηk} is a white Gaussian noise sequence and {fk} is a set of tap
coefficients of an equivalent discrete-time transversal filter having a
transfer function F(z).

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 42 DHT, HCMUT
 5. Linear Equalizer (3)
 In this and the following sections, we describe suboptimum channel
equalization approaches to compensate for the ISI using a linear
transversal filter. The linear filter most often used for equalization is
the transversal filter shown as

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 43 DHT, HCMUT
 5. Linear Equalizer (4)
where its input is the sequence {vk} is given as

and its output in the estimate of the information sequence {Ik}. The
estimate of the kth symbol may be expressed as

 Considerable research has been performed on the criterion for

optimizing the filter coefficients {ck}. Since the most meaningful
measure of performance for a digital communication system is the
average probability of error, it is desirable to choose the coefficients to
minimize the probability of error. Two criteria have found widespread
use in optimizing the equalizer coefficients: the peak distortion
criterion and the mean-square-error criterion.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 44 DHT, HCMUT
 5. Linear Equalizer (5)
 Peak distortion criterion: The peak distortion is simply defined as the
worst-case intersymbol interference at the output of the equalizer. The
minimization of this performance index is called the peak distortion
criterion.
We observe that the cascade of the discrete-time linear filter model
having an impulse response {fn} and an equalizer having an impulse
response {cn} can be represented by a single equivalent filter having
the impulse response:

The equalizer is assumed to have an infinite number of taps. Its output

at the kth sampling instant can be expressed in the form:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 45 DHT, HCMUT
 5. Linear Equalizer (6)
The first term in above equation represents a scaled version of the
desired symbol. For convenience, we normalize q0 to unity. The
second term is the intersymbol interference. The peak value of this
interference, which is called the peak distortion, is

With an equalizer having an infinite number of taps, it is possible to

select the tap weights so that D(c) = 0, i.e., qn = 0 for all n except n = 0.
That is, the ISI can be completely eliminated. The values of the tap
weights for accomplishing this goal are determined from the condition:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 46 DHT, HCMUT
 5. Linear Equalizer (7)
By taking the z transform of above equation, we obtain:

or

where C(z) denotes the z transform of the {cj}. Note that the equalizer,
with transfer function C(z), is simply the inverse filter to the linear
filter model F(z). In other words, complete elimination of the ISI
requires the use of an inverse filter to F(z).We call such a filter a zero-
forcing filter.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 47 DHT, HCMUT
 5. Linear Equalizer (8)
Note that, in the case of finite-length equalizer, it is generally
impossible to completely eliminate the ISI at the output of the
equalizer.

 Mean-Square-Error (MSE) criterion:

In the MSE criterion, the tap weight coefficients {cj} of the equalizer
are adjusted to minimize the mean square value of the error:

From [1], the transfer function of the equalizer based on the MSE
criterion is:

where N0 noise spectral density factor.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 48 DHT, HCMUT
 5. Decision-Feedback Equalizer
 In this section, we consider a nonlinear type of channel equalizer for
mitigating the ISI, which is also suboptimum, but whose performance
is generally better than that of the linear equalizer. The nonlinear
equalizer consists of two filters, a feedforward filter and a feedback
filter, arranged as shown in below figure, and it is called a decision-
feedback equalizer (DFE).

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 49 DHT, HCMUT
 5. Adaptive Filtering (1)
 Conventional frequency-selective digital filters with fixed coefficients
are designed to have a given frequency response chosen to alter the
spectrum of the input signal in a desired manner. Their key features are
as follows:
 The filters are linear and time-invariant.
 The design procedure uses the desired passband, transition bands,
passband ripple, and stopband attenuation. We do not need to
know the sample values of the signals to be processed.
 Since the filters are frequency-selective, they work best when the
various components of the input signal occupy non-overlapping
frequency bands. For example, it is easy to separate a signal and
additive noise when their spectra do not overlap.
 The filter coefficients are chosen during the design phase and are
held constant during the normal operation of the filter.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 50 DHT, HCMUT
 5. Adaptive Filtering (2)
 However, there are many practical application problems that cannot be
successfully solved by using fixed digital filters because either we do
not have sufficient information to design a digital filter with fixed
coefficients or the design criteria change during the normal operation
of the filter.

Most of these applications can be successfully solved by using special

“smart” filters known collectively as adaptive filters. The
distinguishing feature of adaptive filters is that they can modify their
response to improve performance during operation without any
intervention from the user.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 51 DHT, HCMUT
 5. Adaptive Filtering (3)
 Basis elements of a general adaptive filters:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 52 DHT, HCMUT
 5. Adaptive Filtering (4)
 Filtering structure. This module forms the output of the filter
using measurements of the input signal(s). The filtering structure is
linear if the output is obtained as a linear combination of the input
measurements; otherwise, it is said to be nonlinear. The structure is
fixed by the designer, and its parameters are adjusted by the
adaptive algorithm.
 Criterion of performance (COP). The output of the adaptive
filter and the desired response (when available) are processed by
the COP module to assess its quality with respect to the
requirements of the particular application.
 Adaptive algorithm. The adaptive algorithm uses the value of the
criterion of performance, or some function of it, and the
measurements of the input and desired response (when available)
to decide how to modify the parameters of the filter to improve its
performance.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 53 DHT, HCMUT
 5. Adaptive Filtering (5)
 Typical application problems that can be effectively solved by using an
adaptive filter:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 54 DHT, HCMUT
 5. Adaptive Filtering (6)
System identification [5]:

System inversion [5]:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 55 DHT, HCMUT
 5. Adaptive Filtering (7)
Signal prediction [5]:

Interference cancellation [5]:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 56 DHT, HCMUT
 5. Adaptive Filtering (8)
Simple model of a digital communications system with channel
equalization:

Pulse trains: (a) without intersymbol

interference (ISI) and (b) with ISI.
ISI distortion: The tails of adjacent
pulses interfere the current pulse and
can lead to an incorrect decision.
The equalizer can compensate for
the ISI distortion.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 57 DHT, HCMUT
 5. Adaptive Filtering (9)
Block diagram of the basic components of an active noise control
system:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 58 DHT, HCMUT
 Chapter 6:
Multiple-Antenna Systems

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE DHT, HCMUT
 References

[1] J. G. Proakis, Digital Communications, 5th Edition, McGraw-

Hill, 2008.

[2] Athanasios Papoulis, Probability, Random Variables, and Stochastic

Processes, McGraw-Hill, 1991 (3rd Ed.), 2001 (4th Ed.).

[3] Frank Gross, Smart Antennas for Wireless Communication with

Matlab, McGraw-Hill, 2005.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 60 DHT, HCMUT
 6. Array Processing (1)
 Array processing deals with techniques for the analysis and
processing of signals collected by a group of sensors (sensor array).

The collection of sensors makes up the array, and the manner in which
the signals from the sensors are combined and handled constitutes the
processing.

The type of processing is dictated by the needs of the particular

application. Array processing has found widespread application in a
large number of areas, including radar, sonar, communications,
seismology, geophysical prospecting for oil and natural gas, diagnostic
ultrasound, and multi-channel audio systems.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 61 DHT, HCMUT
 6. Array Processing (2)
Spatial filtering (Beamforming) [5]:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 62 DHT, HCMUT
 6. Array Processing (3)

Beamforming based on
Beamforming based on Adaptive algorithms (e.g. at Rx)
DoA estimation (e.g. at Tx)

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 63 DHT, HCMUT
 6. Array Processing (4)
Example of (adaptive) beamforming with an airborne for interference
mitigation [5]:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 64 DHT, HCMUT
 6. Array Processing (5)
(Adaptive) Sidelobe canceler:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 65 DHT, HCMUT
 6. Array Fundamentals (1)
 See sections 4.1, 4.2 of [3]

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 66 DHT, HCMUT
 6. Array Fundamentals (2)
 Two-element linear array

Total electric field:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 67 DHT, HCMUT
 6. Array Fundamentals (3)
or

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 68 DHT, HCMUT
 6. Array Fundamentals (4)
 Uniform N-element linear array

In this case, the array factor is

or

where δ is the phase shift from element to element, and

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 69 DHT, HCMUT
 6. Array Fundamentals (5)
 Array steering vector (array response vector, array manifold vector)

Therefore, the array factor can alternatively be expressed as the sum of

the elements of the array steering vector as

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 70 DHT, HCMUT
 6. Array Fundamentals (6)
 See Section 4.1 for determinng the array factor nulls, maxima, and the
mainlobe beamwidth.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 71 DHT, HCMUT
 6. Array Fundamentals (7)
 Broadside linear array: The most common mode of operation for a
linear array is in the broadside mode. This is the case where δ = 0 such
that all element currents are in phase.
This array is called a broadside array because the maximum radiation
is broadside to the array geometry.
As the array element spacing increases, the array physically is longer,
thereby decreasing the mainlobe width. The general rule for array
radiation is that the mainlobe width is inversely proportional to the
array length.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 72 DHT, HCMUT
 6. Array Fundamentals (8)
Example broadside linear array:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 73 DHT, HCMUT
 6. Array Fundamentals (9)
 End-fire linear array: The name end-fire indicates that this array’s
maximum radiation is along the axis containing the array elements.
Thus, maximum radiation is “out the end” of the array. This case is
achieved when δ = −kd.
It should be noted that the mainlobe width for the ordinary end-fire
case is much greater than the mainlobe width for the broadside case.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 74 DHT, HCMUT
 6. Array Fundamentals (10)
Example end-fire linear array:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 75 DHT, HCMUT
 6. Array Fundamentals (11)
 Beamsteered linear array: A beamsteered linear array is an array
where the phase shift δ is a variable thus allowing the mainlobe to be
directed toward any direction of interest. The broadside and end-fire
conditions are special cases of the more generalized beamsteered array.
The beamsteering conditions can be satisfied by defining the phase
shift δ = −kd sin θ0. We may rewrite the array factor in terms of
beamsteering such that

The beamwidth of the beamsteered array can be determined by

The beamsteered array beamwidth is now given as:

Dept. of Telecomm. Eng. Comm II 2013
Faculty of EEE 76 DHT, HCMUT
 6. Array Fundamentals (12)
Example of beamsteered linear array:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 77 DHT, HCMUT
 6. Array Fundamentals (13)
 Array weighting: The previous derivation for the array factor
assumed that all of the isotropic elements had unity amplitude. It
results in that large sidelobes appear. The presence of sidelobes means
that the array is radiating energy in untended directions. Additionally,
due to reciprocity, the array is receiving energy from unintended
directions. In a multipath environment, the sidelobes can receive the
same signal from multiple angles (fading problem). The sidelobes can
be suppressed by weighting, shading, or windowing the array
elements. The array factor for weighted array can be expressed as

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 78 DHT, HCMUT
 6. Array Fundamentals (14)
Example of Blackman weights are defined by

See Section 4.2, [3] for more examples of array weighting.

Dept. of Telecomm. Eng. Comm II 2013
Faculty of EEE 79 DHT, HCMUT
 6. Array Fundamentals (15)
 Beamsteered and weighted arrays: The nonuniformly weighted
array can also be modified in order to steer the beam to any direction
desired and with suppressed sidelobe levels.
In general, any array can be steered to any direction by either using
phase shifters in the hardware or by digitally phase shifting the data at
the back end of the receiver. If the received signal is digitized and
processed, this signal processing is often called digital beamforming
(DBF). Current technologies are making it more feasible to perform
DBF and therefore allow the array designer to bypass the need for
hardware phase shifters. The DBF performed can be used to steer the
antenna beam according to any criteria specified by the user.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 80 DHT, HCMUT
 6. Array Fundamentals (16)

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 81 DHT, HCMUT
 6. Array Fundamentals (17)
 Fixed Beam Arrays: Fixed beam arrays are designed such that the
array pattern consists of several simultaneous spot beams transmitting
in fixed angular directions.
 Normally these directions are in equal angular increments so as to
insure a relatively uniform coverage of a region in space.
 These fixed beams can be also used in satellite communications to
create spot beams toward fixed earth-based locations.
 Fixed beams can also be used for mobile communication base
stations in order to provide space division multiple access (SDMA)
capabilities.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 82 DHT, HCMUT
 6. Array Fundamentals (18)
One method for easily creating fixed beams is through the use of
Butler matrices. The Butler matrix is an analog means of producing
several simultaneous fixed beams through the use of phase shifters. As
an example, let us assume a linear array of N elements. If N = 2n
elements, the array factor can be given as

with

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 83 DHT, HCMUT
 6. Array Fundamentals (19)
Example: Butler matrix with N = 4 element array with d = λ/2.
Since N = 4, then l = −3/2, −1/2, 1/2, 3/2. Substituting these values
into the equation for sin θ, we get the polar plot as

These beams can be created using

fixed phase shifters by noting that
βl = lπ = ±π/2 , ±3π/2 .

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 84 DHT, HCMUT
 6. Array Fundamentals (20)
 Fixed sidelobe canceling: The basic goal of a fixed sidelobe
canceller (SLC) is to choose array weights such that a null is placed in
the direction of interference while the mainlobe maximum is in the
direction of interest.

Example:
Considering a fixed sidelobe
canceling for one fixed known
desired source and two fixed
undesired interferers. All signals
are assumed to operate at the same
carrier frequency. Let us assume
a 3-element array with the desired
signal and interferers as shown.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 85 DHT, HCMUT
 6. Array Fundamentals (21)
The array vector is given by

The array weights are given by

Therefore, the total array output from the summer is given as

The array output for the desired signal will be designated by SD

whereas the array output for the interfering signals will be designated
by S1 and S2. Since there are three unknown weights, there must be
three conditions satisfied.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 86 DHT, HCMUT
 6. Array Fundamentals (22)
or in matrix form:

One can invert the matrix to find the required complex weights w1, w2,
and w3. As an example, if the desired signal is arriving from θD = 0◦
while θ1 = −45◦ and θ2 = 60◦, the necessary weights can be calculated
to be

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 87 DHT, HCMUT
 6. Array Fundamentals (23)
The array factor is then plotted as

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 88 DHT, HCMUT
 6. Array Fundamentals (24)
There are some limitations to this scheme. The number of nulls cannot
exceed the number of array elements. In addition, the array maximum
cannot be closer to a null than the array resolution allowed. The array
resolution is inversely proportional to the array length.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 89 DHT, HCMUT
 6. Smart Antennas (1)
 See Chapter 8 of [3]

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 90 DHT, HCMUT
 6. Smart Antennas (2)
 Traditional array antennas, where the main beam is steered to directions
of interest, are called phased arrays, beamsteered arrays, or scanned
arrays. The beam is steered via phase shifters oftenly implemented at
RF frequencies. This general approach to phase shifting has been
referred to as electronic beamsteering because of the attempt to change
the phase of the current directly at each antenna element.

Modern beamsteered array antennas, where the pattern is shaped

according to certain optimum criteria, are called smart antennas.
Smart antennas have alternatively been called digital beamformed
(DBF) arrays or adaptive arrays (when adaptive algorithms are
employed). The term smart implies the use of digital signal processing
in order to shape the beam pattern according to certain conditions.

Smart antennas can be applied for improved radar systems, improved

system capacities with mobile wireless, and improved wireless
communications through the implementation of space division
multiple access (SDMA).
Dept. of Telecomm. Eng. Comm II 2013
Faculty of EEE 91 DHT, HCMUT
 6. Smart Antennas (3)
Smart antenna patterns are controlled via algorithms based upon certain
criteria. These criteria could be maximizing the signal-tointerference
ratio (SIR), minimizing the variance, minimizing the meansquare
error (MSE), steering toward a signal of interest, or nulling the
interfering signals.

The implementation of these algorithms can be performed electronically

through analog devices but it is generally more easily performed using
digital signal processing. This requires that the array outputs be digitized
through the use of an A/D converter. This digitization can be performed
at either IF or baseband frequencies. Since an antenna pattern (or beam)
is formed by digital signal processing, this process is often referred to as
digital beamforming. When the algorithms used are adaptive algorithms,
this process is referred to as adaptive beamforming. Adaptive
beamforming is generally the more useful and effective beamforming
solution because it is dynamically optimizes the array pattern according
to the changing electromagnetic environment.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 92 DHT, HCMUT
 6. Smart Antennas (4)

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 93 DHT, HCMUT
 6. Smart Antennas (5)
The main advantage of digital beamforming is that phase shifting and
array weighting can be performed on the digitized data rather than by
being implemented in hardware. If the parameters of operation are
changed or the detection criteria are modified, the beamforming can be
changed by simply changing an algorithm rather than by replacing
hardware.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 94 DHT, HCMUT
 6. Smart Antennas (6)
 Fixed weight beamforming basics
 Maximum signal-to-interference ratio: One criterion which can
be applied to enhancing the received signal and minimizing the
interfering signals is based upon maximizing the SIR.
Example: 3-element array with one fixed known desired source
and two fixed undesired interferers. All signals are assumed to
operate at the same carrier frequency.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 95 DHT, HCMUT
 6. Smart Antennas (7)
The required complex weights w1, w2, and w3 can be determined
as

The steering vector for each source is given by

and σn2 is the noise variace (noise power).

As an example, if the desired signal is arriving from θ0 = 0◦

while θ1 = −45◦ and θ2 = 60◦, the necessary weights can be
calculated to be

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 96 DHT, HCMUT
 6. Smart Antennas (8)
The array factor is plotted as

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 97 DHT, HCMUT
 6. Smart Antennas (9)
See example 8.1, [3] with Matlab function.

Array pattern with approximate nulls at −15◦ and 25◦.

Dept. of Telecomm. Eng. Comm II 2013
Faculty of EEE 98 DHT, HCMUT
 6. Smart Antennas (10)
The general case for max SIR:
It shows one desired signal arriving from the angle θ0 and N
interferers arriving from angles θ1, . . . , θN. The signal and the
interferers are received by an array of M elements with M
potential weights. Each received signal at element m also
includes additive Gaussian noise. Time is represented by the kth
time sample.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 99 DHT, HCMUT
 6. Smart Antennas (11)
The array output y can be given in the following form:

where

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 100 DHT, HCMUT
 6. Smart Antennas (12)
Therefore, the array output can be re-writen as

where

The (SIR) is defined as the ratio of the desired signal power

divided by the undesired signal power

where

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 101 DHT, HCMUT
 6. Smart Antennas (13)
The final optimized weight for max SIR is

where

See example 8.2, [3]:

The M = 3-element array with spacing d = .5λ has a noise
variance σ2n = .001, a desired received signal arriving at θ = 30 ,
0 ◦

and two interferers arriving at angles θ1 = −30◦ and θ2 = 45◦.

Assume that the signal and interferer amplitudes are constant.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 102 DHT, HCMUT
 6. Smart Antennas (14)

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 103 DHT, HCMUT
 6. Smart Antennas (15)
 Minimum variance (or minimum variance distortionless
response): The goal of the minimum variance (MV) method is to
minimize the array output noise variance. The weighted array
output is given by

In order to ensure a distortionless response, we must also add the

constraint that

Finally, the minimum variance optimum weights can be obtained

as

where

is the correlation matrix of unwanted signals and noise.

Dept. of Telecomm. Eng. Comm II 2013
Faculty of EEE 104 DHT, HCMUT
 6. Smart Antennas (16)
See example 8.5, [3].

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 105 DHT, HCMUT
 6. Smart Antennas (17)
 Adaptive Beamforming
If the desired arrival angles change with time, it is necessary to devise
an optimization scheme that operates on-the-fly so as to keep
recalculating the optimum array weights. The receiver signal
processing algorithm then must allow for the continuous adaptation to
an ever-changing electromagnetic environment.
The adaptive algorithm takes the fixed beamforming process one step
further and allows for the calculation of continuously updated weights.

 Least mean squares (LMS):

The updated weight vector according to LMS is

where

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 106 DHT, HCMUT
 6. Smart Antennas (18)
Adaptive beamforming scheme:

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 107 DHT, HCMUT
 6. Smart Antennas (19)
The convergence of the LMS algorithm is directly proportional to
the step-size parameter μ. If the step-size is too small, the
convergence is slow. If the convergence is slower than the
changing angles of arrival, it is possible that the adaptive array
cannot acquire the signal of interest fast enough to track the
changing signal.
If the step-size is too large, the LMS algorithm will overshoot the
optimum weights of interest. If attempted convergence is too fast,
the weights will oscillate about the optimum weights but will not
accurately track the solution desired. It is therefore imperative to
choose a step-size in a range that insures convergence. It can be
shown that stability is insured provided if

where λmax is the largest eigenvalue of the input correlation matrix

Rxx.
Dept. of Telecomm. Eng. Comm II 2013
Faculty of EEE 108 DHT, HCMUT
 6. Smart Antennas (20)
See example 8.6. [3].

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 109 DHT, HCMUT
 6. Smart Antennas (21)
Magnitude weights (example 8.6):

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 110 DHT, HCMUT
 6. Smart Antennas (22)
Mean square error (example 8.6):

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 111 DHT, HCMUT
 6. Smart Antennas (23)
In the example 8.6, the LMS algorithm did not converge until after
70 iterations. 70 iterations corresponded to more than half of the
period of the waveform of interest.

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 112 DHT, HCMUT
 6. Smart Antennas (24)
Array factor (example 8.6):

Dept. of Telecomm. Eng. Comm II 2013

Faculty of EEE 113 DHT, HCMUT