Вы находитесь на странице: 1из 21

Introduction to Communication Systems

by Upamanyu Madhow Chapter 1: Introduction

Chapters 1-2, Lectures 1-5


(about 6 hours, possibly a little more)

What is communications? Growth areas in communications


Wireless
Communication: Information transmission between two points
Separated in space (telephony, web browsing,) Growth in cellular and WiFi continues, technologies keep
Separated in time (recording media--CDs, DVDs, hard drives,) evolving to higher speeds
Taken for granted in the modern world--like electricity Smart phone induced capacity crisis
When you send an email Emergence of multiGigabit wireless over 60 GHz band
Packets travel on wires (Ethernet), optical fiber (Internet core), possibly Sensor nets, RFID (ultra low-power)
wireless (e.g., if the recipient reads it on a smart phone)
When you receive a cell phone call Optical
A voice connection is established from the base station to your phone, the Push to make optical transceivers more like wireless--
backhaul may involve voice over IP, sophisticated signaling in the coherent comm, smart signal processing, sophisticated error-
background is used to hand off the call as you move control coding
Lots of communication happens in the background, without human end users Flash memory, other storage media
experiencing it
Microprocessor buses (e.g., between processors and memories) Constant push for increased capacity and reliability
Signaling between routers on the Internet Continuing evolution of Internet
Social networking and peer-to-peer apps, security, cloud
computing

1
The transition from analog to digital
Analog Communication

Inexorable transition from analog to digital


Analog cellular to digital cellular (CDMA, GSM, OFDM)
Analog TV/radio to Digital TV/radio
LPs to CDs, VHS to DVD
Content is often analog (speech, image, video)
The obvious thing to do Signals sent over physical channels are analog
--Message waveforms are analog Currents, voltages, EM waves are continuous-valued,
--Waveforms sent over the channel must be analog continuous-time functions
But not the right thing to do Why digital comm?
--Analog comm is rendered obsolete by digital comm Lets first see what digital comm is

Digital Communication Digital comm system: block diagram

Core ideas established by Claude Shannon around


1948 (The Mathematical Theory of Communication)
Two ingredients: source coding and channel coding
Any source can be represented to any desired level
of fidelity by bits: required rate depends on allowed
distortion
We can communicate bits reliably over any channel
up to its capacity
Allows separation of source coding and channel coding

2
Components in a digicom system
Why digital?
Source encoder: speech/audio/image/video to bits
Optimality and universality of source-channel separation
Large compression gains compared to nave sampled
Highly tuned source encoders optimized for source
representation
characteristics
Channel encoder: adds redundancy to help recover from
Channel coding and modulation optimized for channel
distortions/errors introduced by channel
Huge performance gains over analog comm
Modulator: maps coded bits to transmitted waveform (must fit
within spectral constraints of channel, e.g., WiFi or cellular) Can use this approach for any source and any channel
Channel: multipath propagation for wireless, nonlinear distortion Scalability
for optical, thermal noise Bits can be regenerated (no noise accumulation) !
Demodulator: lots of signal processing for sync, channel cascade links to build up networks
equalization, with ultimate output hard or soft decisions Bit pipe abstraction enables build-out of very large
Channel decoder (hard/soft decisions to information bits) communication networks (the classical public switched
telephone network, the Internet today)
Source decoder (e.g., compressed bits to image pixels)

Technology drivers
The Internet
Is there still a role for analog?
Thin waist promotes innovation above (apps) and below
(physical layer innovations)
Very much so: the physical world is analog Continued demands (video streaming, social networking)
Signal transmission: Need to convert digital data to Wireless (greatest growth is at the edge of the Internet)
analog signals that can be sent over the physical Cellular and WiFi
channel Smart phone induced capacity crisis ! big ideas needed
Signal reception: Need to convert analog received
Moores law (doubling of transistor counts on ICs every 2 yrs !
signals back into digital data
low-cost digital computation)
DSP-centric transceiver architectures
MultiGigabit comms requires innovations in mixed signal
design to overcome ADC bottleneck

3
The Internet Cellular

Big Idea: Spatial Reuse

Key challenges
--Interference management
--Mobility

Applications
Room for innovation
Transport layer
Network layer Thin waist (packet switching using IP)
MAC/Link layer
Room for innovation
(This book) Physical layer

Ingredients of a comm system? Key concepts (slide 1 of 2)


Structure of communication signals (Chapter 2)
Signals and systems review
Business: Identify target customer/market, acquire funding
Baseband and passband signals
Network design: network architecture, router design, protocols
How to send information with radio signals (Chapter 3)
and their implementation Amplitude modulation, Angle modulation
Spectrum occupancy
Important analog techniques (Chapter 3)
Physical link design: channel and interference modeling,
transceiver design, signal processing algorithms This book Superheterodyne reception, phase locked loop
Legacy analog communication systems used to provide
examples: AM radio, FM radio
Transceiver implementation: Integrated circuit implementation Digital modulation (Chapter 4)
of signal processing, RF; antenna design Nyquist criterion for signaling over a bandlimited channel
Modulation techniques: QAM, PSK, orthogonal modulation

4
Key concepts (slide 2 of 2)
Statistical models for signals and noise (Chapter 5)
Review of probability and random variables
Random processes
Signal-to-noise ratio computations
Demodulation (Chapter 6)
Gaussian noise model
Chapter 2: Signals and Systems
Detection theory basics
Signal space techniques: geometric approach to optimal detection
Optimal receiver design
Performance analysis
Channel coding (Chapter 7)
Information-theoretic performance benchmarks
Practical channel codes for approaching these benchmarks
Advanced signal processing (Chapter 8)
Handling channel dispersion
Exploiting multiple antennas, or MIMO

Signals and Systems Review: Outline


Signals and Systems: Goals
Complex numbers
Review of basics: convolution, Fourier transform Eulers identity
Targeted towards comm applications Inner product
Baseband and passband signals and systems Norm, energy
Spectrum is valuable, so we keep tabs on the spectral Fourier transform
occupancy of communication signals Formula
WiFi signals are passband (e.g., a 20 MHz band at 2.4 GHz; Time-frequency duality
away from DC) Properties: Convolution/multiplication, Parseval, linearity,
Wired communication (e.g., USB) often use baseband signals time/freq shift
Unified treatment of baseband and passband signals Pairs: Delta function/constant, sinc function/boxcar
Enabled by complex baseband representation for passband Bandwidth
signals Baseband and passband signals
Complex baseband representation
Role of complex baseband in DSP-centric transceiver design

5
Complex numbers
Complex numbers provide a compact way of describing amplitude and phase Some important signals
(and the operations that affect them, such as filtering)

Complex number (x and y real-valued, )


Basic signals that are useful building blocks for more
Often useful to interpret a complex number as a point in 2-D plane complex signals
Cartesian (rectangular) coordinates: Impulse function, or Delta function
Polar coordinates Indicator function
Sinc function
But a complex number is more than just a 2D real vector:
mainly because of complex multiplication (one complex multiply
requires four real multiplies, and corresponds to adding phases)

Eulers identity (we use this very often)

Delta (Impulse) function Indicator and boxcar functions


Indicator function of a set A Boxcar (indicator function of an interval)
Limit of tall, thin pulse with unit area (very useful for compact notation)

(t)

t
0
(a 0)
Using the indicator function for compact descriptions:

Multiplying a signal by a tall thin pulse at t0 selects

its value at t0. Now integrate to get sifting property.

Sifting Property of the Impulse

6
Sinc function
Signals are just like vectors
The Big Picture: Euclidean geometry is important for communication system designers
(sinc(0) = 1, defined as the limit)
1) Continuous-time signals are just like vectors
--standard ideas from Euclidean geometry apply
2) We typically transform continuous-time signals into discrete-time vectors (filtering
sinc(x), decays as 1/|x| with sinusoidal fluctuations and sampling) before signal processing
3) Vector manipulation therefore important for both theory and algorithms

Inner product is the key concept in defining signal geometry


sinc(x)

discrete time continuous time

(Complex-valued signals needed for our unified framework)


(note that constants
in second argument
x Linearity
get conjugated when
The boxcar and sinc are Fourier transform pairs pulled out)

Energy and Power, I Energy and Power, II


Energy and power are important concepts in communication
--How much power is needed to transmit a signal? Power = time average of energy, computed over a large interval
--What is the signal-to-noise ratio?
--How much interference do signals create for each other?

Energy = Inner product of a signal with itself


Example: Power of a sinusoid

Examples: Energy computations for complex-valued signals

Average Integrates to zero over each period


value (averages to zero over large interval)

7
Linear Time-Invariant (LTI) Systems

Completely characterized by their impulse response


--Any signal can be expressed as a linear combination of shifted impulses
--Use LTI property to get the convolution formula

LTI Systems and Convolution

LTI property

s
Sifting property of impulse Convolution

Convolution
Communications channels often modeled as LTI systems, so convolution is a key
modeling tool: modulated signal convolved with an LTI system, then noise added
Group Exercise: LTI concept
Recall the basic definition:
Linearity and shift invariance

(notational abuse in not distinguishing dummy variable is often convenient)

Lets review convolution using an example.

8
Flip and slide Convolution of two boxes = trapezoid

I[5,11] (t) I[1,3] (t) =

In general, useful to keep in mind the following:

Use linearity and time invariance to build convolutions of shifted boxes,


linear combinations of shifted boxes, etc.

Group exercise: convolution computation Digital modulation as convolution

Symbols at rate 1/T


3
1
3 6

-1
2 * -4 -3

Find the convolution using the convolution of two boxes as building block.
Use linearity to decompose into multiple boxes Sample at rate 1/Ts = M/T. Symbols are spaced by M samples.
Use time invariance: need not worry about where the signals start
(can make them start from origin, and add in delays later)
See Chapter 2 for discrete time convolution
Approximating continuous time convolution
DSP-based implementation of transceiver and receiver functionality Can view as convolution between p[n] and upsampled version of b, inserting M-1 zeros
between successive entries. See code fragment 2.3.1 in Chapter 2.

9
Modeling a multipath channel
Convolving with the delta function

Convolving with the shifted delta function inserts a time shift:

(use the sifting property to see this)

So we can use the delta function to model wireless multipath channels

Output of multipath channel Group exercise


(before adding receiver noise)

Multipath channel impulse response

Complex exponential through LTI system


Complex exponentials are eigenfunctions of LTI systems: pass through
unchanged, except for a scale factor
This is why we like working in the frequency domain.

Fourier Transforms

Fourier transform = System transfer function


of impulse response

10
Fourier Transform
Boxcar and sinc are a Fourier transform pair
Fourier Transform Inverse Fourier Transform

Timelimited signals with sharp edges decay slowly with frequency: rectangular
Notation for Fourier Transform Pair time domain pulse has a sinc spectrum, which exhibits 1/|f| decay
By duality, ideal bandlimited signal corresponds to time domain sinc (slow decay
with time has bad implications for digital comm--intersymbol interference)
Time-Frequency Duality:

(So we need to keep track of only half of the Fourier transform pairs)

Delta function Constant function (trivial but important pair)

Fourier Transform Properties,I Fourier Transform Properties, II


Convolution Multiplication
LINEARITY

TRANSLATION
Group exercise: Find the Fourier transform of sinc(10t) sinc(4t)
Time delay corresponds to linear phase in frequency domain (convolution of two boxes = trapezoid)

Frequency shift corresponds to modulation by complex exponential in time domain

TIME/FREQ SCALING
Time compression leads
to freq expansion, and
vice versa

11
Fourier Transform Properties, IV
Fourier Transform Properties, III Parsevals Identity
COMPLEX CONJUGATION
Inner product can be computed in either time or frequency domain
Complex conjugation in one domain corresponds to
flip and conjugation in the other

Important implication: the spectrum of real-valued signals is conjugate symmetric


Important implication: energy can be computed in time or frequency domain

Conjugate symmetry in rectangular coordinates:

Real part of spectrum is symmetric



Imaginary part of spectrum anti-symmetric Group exercise: Use Parsevals identity to evaluate
(sin 4 x / x 4 ) dx

0
Conjugate symmetry in polar coordinates:
Approach: think of integrating over the entire real line, and then think of what is
Magnitude of spectrum is symmetric | S( f ) |=| S( f ) | the Fourier transform pair for the square of a sinc.
Phase of spectrum is anti-symmetric arg(S( f )) = arg(S( f ))

Bandwidth
Group Exercise Bandwidth of a signal quantifies its frequency occupancy
One-sided bandwidth: We only consider positive frequencies
when computing bandwidth for physical signals
Problem 1(b) and 1(c) For example, a WiFi signal may occupy a 20 MHz bandwidth,
between 2.4-2.42 GHz
Physical signals are real-valued (in the time domain)
Hence they are conjugate symmetric in the frequency
domain, so we can specify them completely by their spectrum
over positive frequencies
We shall also consider complex-valued (in the time domain)
signals later
Complex envelope of a real-valued passband signal
It will turn out that the two-sided bandwidth of the complex
envelope equals the physical (one-sided) bandwidth of the
passband signal

12
Aside: reminder on why we need negative frequencies Example: bandwidth of a boxcar
s(t)

We like working with complex exponentials because they are 1


eigenfunctions of LTI systems
t
Need complex exponentials at both positive and negative T
frequencies to span the space of square integrable signals Energy spectral density = magnitude squared of Fourier transform
Real-valued sines and cosines with positive frequencies
alone would also work, but these are not eigenfunctions of
LTI systems, hence are less convenient
Physical signals are real-valued (time domain) Not strictly bandlimited, but can define fractional energy containment bandwidth
Hence must satisfy consistency condition of conjugate One-sided fractional energy containment bandwidth B (fraction a) satisfies:
symmetry (all the information resides in either positive or
negative frequencies, hence only need spectrum for one of
these)
Example: a=0.99 corresponds to 99% energy containment bandwidth
Hence physical bandwidth = one-sided bandwidth

Boxcar example (contd.)


Multipath channel modeling continued, I
Useful (and insightful) to normalize: can set T to convenient value (say T=1).
Equivalent to defining one unit of time as T. Impulse response
By scaling relation between time and frequency, if bandwidth for T=1 is B1, then bandwidth
for general T is B1/T.
Parsevals identity can be used to evaluate energy in whichever domain is more
convenient.
Transfer function is a sum of sinusoids in the frequency domain
symmetry

Evaluate energy in time domain for boxcar (= 1 for T=1)


Pure delay, What matters are the delay differences
Numerical results: B1=0.85 for a=0.9; B1=10.2 for a=0.99. can be dropped
When we want stricter energy containment, choosing a rectangular time domain pulse
is a very bad idea (when we are willing to be sloppy, its OK.) Channel delay spread: determines the fastest variations in the frequency domain
Channel coherence bandwidth: Inverse of delay spread
Channel can be approximated as constant over band much smaller than coherence bw

13
Multipath channel modeling contd, II

(ignoring pure delay term)

Time in microseconds; delay spread 2.5 microsec, coherence bandwidth 400 KHz Baseband and Passband
The Complex Baseband Representation
Frequency selective fading
can be as large as 18 dB!
Need diversity: frequency, spatial,
time (if channel varies across time)

Baseband Signals
Baseband and Passband Signals/Channels
Baseband signals have energy/power concentrated in a band around DC.
Channels often approximated as LTI systems
Signal passes through channel, and then noise is added
Channels allocated/described typically in terms of frequency
bands
Signals have to be designed for the corresponding frequency
band
Baseband channels/signals
Energy concentrated in a frequency band around DC
Passband channels/signals
Above: Real baseband signal of bandwidth W (U(f) obeys conjugate symmetry).
Energy concentrated in a frequency band away from DC Note that |U(f)| is symmetric for a real baseband signal.)
Unified treatment of baseband and passband systems For communication over a physical baseband channel, we consider physical (real-valued)
Complex baseband representation of passband systems baseband signals.
For communication over a physical passband channel (discussion coming up), we consider
complex-valued baseband signals which provide a convenient mathematical representation
for the corresponding passband signals.

14
Passband Signals Examples of baseband signals
Passband signals have energy/power concentrated in a band away from DC. Speech, audio are baseband signals

Carrier Bandwidth
frequency

Passband signal of bandwidth W Two-level digital signal is baseband

But we often want to send such signals over a passband channel


We only consider physical (real-valued) passband signals, hence their spectra
(e.g., a 20 MHz WiFi channel at 2.4 GHz). Need to understand how passband
always obey conjugate symmetry
signals are structured in order to accomplish this.

Modulation: baseband to passband I and Q components


Can modulate separately using cosine and sine of carrier
Consider a real-valued baseband message signal m(t)
Sinusoids are rapidly varying but predictable (contain no info)
Modulation: Translate to passband by multiplying by a sinusoid

Passband
or In-phase (I) Quadrature (Q)
signal
component component

Real baseband signals


(contain all the information)

We now know that we can start from two real baseband signals and
get a passband signal by I/Q modulation.
Should be How do we get back the I and Q components from the passband signal?
positive Can any passband signal be decomposed into I and Q components?

Carrier frequency should be bigger than the message bandwidth to keep away from DC

15
Band limitation need not be strict Upconversion: baseband to passband
Example: (Approximately) passband signal Block diagram follows directly from
equation defining the modulated signal

I and Q components are timelimited, hence cannot be strictly bandlimited

But frequency content is concentrated around DC, well away from fc = 150.

Happens at the transmitter


In practice, may do upconversion in multiple stages
Must worry about adhering to spectral masks, controlling nonlinearities

Downconversion: passband to baseband


Change of reference freq/phase: example
Works as long as receiver
is coherent (phase and
frequency of copy of carrier
at receiver same as that of
incoming signal)
Find the output when the following are passed through an LPF:

Recovering the I component (similar derivation for Q component):

Solution: Recognize that these are downconversion operations with different


2f c terms rejected by LPF freq/phase references. Find the corresponding complex envelopes and read
off the right components.

16
I and Q channels are orthogonal
(can send info in parallel on these channels) Passband modulation is 2-dimensional
Passband signal can be mapped to a pair of real baseband signals
That is, passband modulation is two-dimensional. Can also plot it on complex plane.

Any real baseband waveforms


with bandwidth less than the Complex envelope
carrier frequency Q component

Envelope
Phase
Then the passband waveforms ap and bp are orthogonal
Proof:
I component

is passband at 2fc(why?) (zero DC value)


Three equivalent representations of the passband signal
Why? Rectangular coordinates: I and Q
Polar coordinates: Envelope and phase
is baseband with BW at most equal to sum of bws of uc and us < 2fc Complex number: Complex envelope

Going back and forth between representations


Time domain expressions for a passband signal

In terms of
I and Q components

In terms of
u p (t) = e(t)cos(2f c t + (t)) envelope and phase
Each representation
of a complex number In terms of
corresponds to a complex envelope
time domain expression
for the passband signal

Starting from one representation,


can derive the rest based on
the relations depicted in
the figure
uc (t) = e(t)cos (t), us (t) = e(t)sin (t)

17
Frequency domain construction of complex envelope Why the frequency domain construction works, I

PASSBAND SIGNAL

Move to left by fc Move to left by fc


Scale by 2 Scale by 2

Throw away negative


frequencies part Throw away negative
frequencies part
COMPLEX ENVELOPE

Lets check that this is consistent with the time domain relationship.

Why the frequency domain construction works, II I and Q components in frequency domain
I and Q in time domain

I and Q in frequency domain

We see that

We already know

This gives

18
Modern transceiver architectures are based on complex baseband

Group Exercise Modern transceivers work with the complex envelope rather than
with the passband signal
Complex baseband signals can be represented accurately by
samples at a reasonable sampling rate
Inexpensive to perform complicated digital signal processing
(DSP) on the samples: Moores law scaling
This architecture has been responsible for economies of
scale in cellular and WiFi
All the action is in complex baseband for a typical wireless transceiver

Channel Encoding RF signal Filtering


Modulation Upconversion Dnconversion Synchronization
Spectral Shaping Demodulation
Decoding
Complex Complex
DSP-based envelope envelope DSP-based
processing processing
TRANSMITTER RECEIVER

Most transceiver operations can be performed in Passband filtering is equivalent to complex baseband filtering
complex baseband
sp(t) hp(t) yp(t) EQUIVALENT TO s(t) h(t) y(t)

Filtering (except for a scale factor of 2)


Proof by pictures:
Carrier frequency/phase correction
Coherent and noncoherent reception

19
Filtering in complex baseband
Modeling phase offset in complex baseband
Complex-valued convolution to implement equivalent of passband filtering operation

Time varying phase can model both phase and frequency offset
Requires four real-valued convolutions:

Downconverter can use sloppy analog passband filter


Sophisticated filtering can be implemented in baseband (scalar factors not shown)

Complex baseband representation: summary


Group exercise
Any real-valued passband signal can be represented by a
baseband signal which is in general complex-valued. This is
Find the convolution of the two passband waveforms called its complex envelope, or complex baseband
I[0,1] (t)cos(300t) and I[1,3] (t)sin(300t + 4 ) representation.
The complex envelope carries all the information in the passband
signal
Passband filtering operations can be equivalently performed in
complex baseband
Two-dimensional representation of complex envelope
Cartesian coordinates: A pair of real-valued baseband
waveforms called the in-phase (I) and quadrature (Q)
components
Polar coordinates: Envelope and phase waveforms
Most of the sophisticated signal processing action in modern
transceivers happens in complex baseband

20
Complex envelope for multiple paths
Propagation basics
Transmitted passband signal

Transmitted complex baseband signal Depend on angle of incidence, scattering surface characteristics,
Ai , i carrier wavelength, etc.
We will not model these in detail

Received signal for a single ray/path of length r and delay


Two key approaches to channel modeling
Ray tracing: when detailed models of propagation environment are available
Statistical: developed based on measurement campaigns, then draw channels at
random from statistical model for performance evaluation of candidate designs

Ray tracing for LoS + one bounce

Forms the basis for software lab on modeling links in a lamppost based network

21

Вам также может понравиться