From AM radio to Digital I/Q

modulation

Introduction to wireless networks design

Lecture 10
PhD. Alexander B. Hilario Tacuri

[Parts adapted from J . Kyle, B. Karp and N. Vaidya]

Roadmap
1. Analog I/Q modulation

2. Discrete-time processing of continuous signals

– The Digital Abstraction
– Quantization: Discretizing values
– Sequences: Discretizing time

3. Digital I/Q Modulation

2
Review: AM Modulation & Demodulation
AM Modulation and Demodulation
Transmitter Output Receiver Output

Lowpass
Filter
x(t) y(t) y(t) w(t) r(t)
H(j2πf)

2cos(2πfot) 2cos(2πfot)

• Multiplication (i.e., mixing) operation shifts in

• Multiplication
frequency (i.e. mixing) operation shifts in frequency
– Also creates undesired high frequency components at
• receiver
Low-pass filtering passes only the desired baseband
signal at thefiltering
• Lowpass receiverpasses only the desired baseband
signal at receiver

What can go wrong here? 3

Impact of a 90º phase shift
Impact of 90 Degree Phase Shift
Transmitter Output Receiver Output

Lowpass
Filter
x(t) y(t) y(t) w(t) r(t)
H(j2πf)

2cos(2πfot) 2sin(2πfot)

∿ ∿
Local
oscillator
• If receiver cosine wave turns into a sine wave, we
• Receiver usessuddenly receive
sine instead no baseband
of cosine: signal!
no output
– We apparently need to synchronize the phase of the
– sin(2πf0t)cos(2πf 0t) = ½ sin(4πf
transmitter 0t) local oscillators
and receiver
• This is called coherent demodulation
• Need to synchronize
• Some keyphase of transmitter and receiver local oscillators
questions:
(coherent demodulation)
– How do we analyze this issue?
– What would be the impact of a small frequency offset? 4
Coherent Demodulation: Frequency-domain analysis
Frequency Domain Analysis
W(j2πf)
H(j2πf)
X(j2πf) Y(j2πf)

f f f
0 -fo 0 fo -2fo -fo 0 fo 2fo
x(t) y(t) y(t) w(t) r(t)
Transmitter Output H(j2πf) Receiver Output
Lowpass
∿ 2cos(2πfot)
∿ 2cos(2πfot) Filter
1 1 1 1

f f
-fo 0 fo -fo 0 fo

• Transmitter and receiver

• When oscillators
transmitter andphase-synchronized
receiver local oscillators are
matched
• Demodulated copies in constructively
add phase: at baseband
– Demodulated signal constructively adds at baseband

5
90º Phase Shift: Frequency-domain analysis
Impact of 90 Degree Phase Shift
W(j2πf)
H(j2πf)
X(j2πf) Y(j2πf)

2fo
f f f
0 -fo 0 fo -2fo -fo 0 fo
x(t) y(t) y(t) w(t) r(t)
Transmitter Output H(j2πf) Receiver Output
=0
Lowpass
2cos(2πfot) 2sin(2πfot) Filter
1 1 j
f1
f f
-fo 0 fo -f1 0
-j

• Transmitter, receiver oscillators are offset in phase by 90º

• Demodulated • copies add destructively
When transmitter at baseband
and receiver local oscillators are
– Zero output90from
degree offset in phase:
receiver
– Demodulated signal destructively adds at baseband
• Two separate channels!
What would happen with a small frequency offset? 6
In-Phase/Quadrature Modulation
• I/Q
Modulate each with a cosine Demodulation
& sine, sum the result
1 1 1 1
Transmitter Output Receiver Output
It(j2πf) f f
-fo fo Ir(j2πf)
-fo 0 fo Yi(j2πf) 1 0
1 1 2
it(t) i (t)
f f
In-PhaseH(j2πf) (I)
r

f
0
2cos(2πfot) y(t)
-fo 0 fo
y(t) component
2cos(2πfot)
Lowpass 0
Qt(j2πf) Qr(j2πf)
2sin(2πfot) 2sin(2πfot)
1 j Yq(j2πf) 2
qt(t) qr(t)

f
fo
f
Quadrature (Q)
H(j2πf)
f
0 j -fo 0 j component
Lowpass
0
fo -j fo
f f
-fo 0 -fo 0
-j -j

• I, Q signals• occupy
Demodulate both a cosine
withfrequency
the same bandand sine wave
– One is real–(for
Both I andone
cos), Q channels are recovered!
is imaginary (for sin)
• I/Q modulation allows twice the amount of
7
In-Phase/Quadrature Demodulation
I/Q Demodulation
1 1 1 1
Transmitter Output Receiver Output
It(j2πf) f f
-fo fo Ir(j2πf)
-fo 0 fo Yi(j2πf) 1 0
1 1 2
it(t) ir(t)
f H(j2πf)
f f
0 -fo 0 fo Lowpass
2cos(2πfot) y(t) y(t) 2cos(2πfot) 0
Qt(j2πf) Qr(j2πf)
2sin(2πfot) 2sin(2πfot)
1 j Yq(j2πf) 2
qt(t) qr(t)
fo H(j2πf)
f f f
0 -fo 0 Lowpass
j j 0
fo -j fo
f f
-fo 0 -fo 0
-j -j

• Demodulate• with
Demodulate
both a sine both
withand a cosine and sine wave
a cosine
– Both I and –QBoth I and Q channels are recovered!
channels are recovered!
• I/Q modulation allows twice the amount of
8
I/Q Demodulation: 90º Phase Shift
Impact of 90 Degree Phase Shift
1 1 j
Transmitter Output fo Receiver Output
It(j2πf) f f Ir(j2πf)
-fo 0 fo Yi(j2πf) 1 -fo 0
1 1 -j 2
it(t) ir(t)
f H(j2πf)
f f
0 -fo 0 fo Lowpass
2cos(2πfot) y(t) y(t) 2sin(2πfot) 0
Qt(j2πf) Qr(j2πf)
2sin(2πfot) -2cos(2πfot)
1 j Yq(j2πf)
qt(t) qr(t)
fo H(j2πf)
f f f
0 -fo 0 Lowpass
j 0
fo -j -fo fo -2
f f
-fo 0 0
-j -1 -1

• I and Q channels getQswapped

• I and channels at receiver
get swapped at receiver
– Key observation: No information
– Key observation: is lost! is lost!
no information
• Questions 9
Summary ofSummary
AnalogofI/Q modulation
Analog I/Q Modulation
• Frequency domain view
Baseband Input Receiver Output
It(j2πf) Ir(j2πf)
1 ir(t) 2
it(t)
H(j2πf)
f f
0 Lowpass 0
2cos(2πfot) 2cos(2πfot)
Qt(j2πf) 2sin(2πfot) 2sin(2πfot) Qr(j2πf)
1 qr(t) 2
qt(t)
H(j2πf)
f f
0 Lowpass
0
• Time domain view
Baseband Input Receiver Output

t t
it(t) ir(t)
H(j2πf)
Lowpass
2cos(2πfot) 2cos(2πfot)
2sin(2πfot) 2sin(2πfot)
t t
qt(t) qr(t)
H(j2πf)
Lowpass 10
6.082 Spring 2007 I/Q Modulation and RC Filtering, Slide 9
I/Q modulation: Wrap-up
• I/Q modulation allows twice the amount of information to be sent compared
to basic AM

• Impact of phase offset is to swap I/Q

• Impact of frequency offset is I/Q swapping (small offset)

– Or catastrophic corruption (large offset) of received signal

11
Roadmap
1. Analog I/Q modulation

2. Discrete-time processing of continuous signals

– The Digital Abstraction
– Quantization: Discretizing values
– Sequences: Discretizing time

3. Digital I/Q Modulation

12
roduce some processing blocks:

n Processing in the Analog Domain

Representing information with voltage
• Encoding of pixel at location (x, y) in a Black &
White picture:
roduce some processing blocks:
– 0 Volts = Black

v Copy
– 1 Volt = White
v
– 0.37 V = 37% Gray
– etc...

• Encoding of a picture:
– Scan points in prescribed order

v Copy v
– Generate a continuous voltage waveform
(baseband signal)

v INV 1-v 13
 • Frequency domain view • Frequency domain view
n the Analog Domain
Information Processing in the Analog Domain
Baseband Input Baseband Input
Receiver Output Receiver Output
It(j2πf) It(j2πf) Ir(j2πf)

Let’sblocks:
build a P2P television network
Ir(j2πf)
1 21 2

ocessing
ir(t) it(t) ir(t)

First let’s introduce some processing blocks:

it(t)
H(j2πf) H(j2πf)
f ff Lowpass
f
0 Lowpass 00 2cos(2πfot) 2cos(2πfot)
0

Let’s build a system!

2cos(2πfot) 2cos(2πfot)
Qt(j2πf) 2sin(2πfot) 2sin(2πfot) QQ t(j2πf)
r(j2πf) 2sin(2πfot) 2sin(2πfot) Qr(j2πf)
1
qt(t) qr(t) 21 qt(t) qr(t) 2
H(j2πf)
One hop:
H(j2πf)
f
Lowpass
ff Lowpass
f
0 00 0
• Time domain view • Time domain view
Encode x(t)Input
Baseband Baseband Input
Receiver Output
Decode
Receiver Output

v t × v ×
Copy v t t t

Copy ∿INV
it(t) ir(t) it(t) ir(t)
H(j2πf) H(j2πf)

n the Analog Domain t

∿ 2cos(2πfot)
2sin(2πfot)
2cos(2πfot)
2sin(2πfot)
Lowpass

t
2cos(2πfot)
2sin(2πfot)
t
2cos(2πfot)
2sin(2πfot)
Lowpass

t
qt(t) qr(t) qt(t) qr(t)
H(j2πf) H(j2πf)

ocessing blocks: input

6.082 Spring 2007 Copy INV
Lowpass
6.082 Spring 2007 Digital Modulation (Part I), Slide 4
Lowpass
Digital Modulation (Part I), S

Many hops: (In Theory)

(Reality)
1-v Copy
v INV INV 1-v

v Encode
Copy INV Decode

?
The Digital Abstraction,
6.082 Spring 2007Slide 3
output
The Digital Abstraction, Slide 3

14
Why did our network fail?
1. Transmitter doesn’t work right
2. Receiver doesn’t work right
3. Theory is imperfect
4. System architecture isn’t right

• Answer: All of the above!

• Noise, transmitter/receiver imperfections inevitable, so must design system to

tolerate some amount of error

15
Analog communications issues
• Problem: It’s hard to distinguish legitimate analog waveforms from corrupted
ones
– Every waveform is potentially legitimate!

• Small errors accumulate at each hop

• Endpoints can’t help, so need to eliminate errors at each hop

16
Plan: Mixed Plan:
SignalMixed Signal Architecture
Architecture
Volts Digital Volts
Freq Signal Freq
Phase bits bits Phase
Processing

Analog store Digital

to xmit/rcv to
modify
Digital Analog

ANALOG DIGITAL ANALOG

Continuous time, Discrete time,

Continuous values Discrete values
6.082 Spring 2007 The Digital Abstraction, Slide 16
17
Discrete time, Discrete values
• Continuous-time, continuous-valued waveform is being converted into a
discrete-valued sequence

– Only certain values are ”allowed” to be used

• Questions yet to be answered:

1. How many discrete values should we use?

2. How rapidly in time should we sample the waveform?

18
 Digital Signaling II
Digital signaling Digital Signaling II
Encoding Attempt #2:
Encoding Attempt #2:

0 0 1 1
Forbidden Zone
• First encoding attempt: Forbidden Zone volts
0 VL
volts
V VH DD
0 VL VH VDD
This avoids “close calls”, but now we have to consider noise (i.e.
• But, what happens unavoidable perturbations to our signaling voltage)
This avoidsin“close
the presence
calls”, but of
nownoise?
we have to consider noise (i.e.
NOISE
unavoidable perturbations to our signaling voltage)
DEVICE DEVICE
NOISE #1 #2 So an output voltage just
below VL might become an
DEVICE DEVICE illegal input voltage in the
#1 V
#2 V±N forbidden zone!
So an output voltage just
below VL might become an
illegal input voltage in the
V V±N
6.082 Spring 2007 The Digital Abstraction, Slide 10
forbidden zone!

6.082 Spring 2007 The Digital Abstraction, Slide 10

19
Big Idea: Noise Margins
Big Idea: Noise Margins
Let’s leave room for bad things to happen! So we’ll design
devices restore marginally valid input signals. They must
accept marginal inputs and provide unquestionable outputs
(i.e., to leave room for noise).

0OUT 1OUT

OUTPUTS: Forbidden Zone

volts
0 VOL VOH VDD

0IN 1IN

INPUTS:
volts
0 VOL VIL VIH VOH VDD

Noise Margins
6.082 Spring 2007 The Digital Abstraction, Slide 11
20
 discrete-time samples, we can capture 2N possible values.

Analog value converted to digital ind

So we’ll divide up the range of possible sample values into
How many discrete values?
2 intervals and choose the index of the enclosing interval
N

as the encoding for the sample value.

VMAX
7 15
14
sample voltage 3
6 13
1 12
5 11
2 10
4 9
8
3 7
1 6
2 5
4
0 3
1
2
0
0 1
VMIN 0

quantized value 1 11 110 1101

1-bit 2-bit 3-bit 4-bit
6.082 Spring 2007 The Digital Abstraction, Slide 18

21
 Note that when we quantize the scaled sample values we
Quantization Error
may be off by up to ±½ step from the true sampled values.

The red shaded region shows

57 the error we’ve introduced

56

55

54

53

54 55 56 55 55

• Introduce at most ½ interval length of quantization error

6.082 Spring 2007 The Digital Abstraction, Slide 19

22
Roadmap
1. Analog I/Q modulation

2. Discrete-time processing of continuous signals

– The Digital Abstraction
– Quantization: Discretizing values
– Sequences: Discretizing time

3. Digital I/Q Modulation

23
 The Need
Need for Continuous for Sampling
to Discrete Conversion
Real World Matlab
(USRP Board)
Real world Computer
xc(t) A-to-D x[n]
Converter
xc(t) x[n]

t n
1

• The
• The boundary boundary
between between
analog analog and digital
and digital
– Real world–filled
Real with is filled with continuous-time
worldcontinuous-time signals signals
Computers (i.e. Matlab) operate on sequences
– Computers– operate on discrete-time sequences
• Crossing the analog-to-digital boundary requires
sampling of the continuous-time signals
• Crossing the analog to digital boundary requires conversion between the two
• Key questions
– How do we analyze the sampling process?
– What can go wrong? 24
Modelling Continuous-to-Discrete Conversion
An Analytical Model for Sampling
p(t)
1

t
p(t) T

xc(t) xp(t) Impulse Train x[n]

to Sequence
xc(t) xp(t) x[n]

t t n
T 1

• Two step process

• Two-step process:– Sample continuous-time signal every T seconds
1. Sample continuous-time signal
• Model every T seconds
as multiplication of signal with impulse train
– Model as multiplication of signalfrom
– Create sequence withamplitude
impulse of train
scaled impulses
2. Create sequence from amplitude
• Model of scaled
as rescaling of time impulses
axis (T → 1)
• Notation:
– Model as rescaling replace
of time axis (T impulses
à 1) with stem symbols
Can we model this in the frequency domain? 25
Fourier Transform of Impulse Train
Fourier Transform of Impulse Train
p(t) P(j2πf)
1
1 T
t t
T -2 -1 0 1 2
T T T T

• Impulse train in time corresponds to impulse train

• Impulse train in time corresponds to impulse train in frequency
in frequency
– Spacing in time of T seconds corresponds to 1/T Hz spacing in
frequency– Spacing in time of T seconds corresponds to spacing in
frequency of 1/T Hz
– Scale factor of 1/T for impulses in frequency domain
– Note: this is painful to derive, so we won’t …
• The above transform pair allows us to see the 26
Frequency Domain view of Sampling
Frequency Domain View of Sampling
xc(t) p(t) xp(t)
1

t t t
T T

Xc(j2πf) P(j2πf) Xp(j2πf)

1 A
A T T

f t f
-2 -1 0 1 2 -2 -1 0 1 2
T T T T T T T T

• Recall that multiplication in time corresponds to

convolution
• Sampling in time frequencyFourier Transform, with period
leads toina periodic
1/T

• We see that sampling in time leads to a periodic 27

Frequency Domain View of a Sequence
Frequency Domain View of Output Sequence
xp(t) x[n]

t n
T 1

Xp(j2πf) X(ej2πλ)
A A
T T

f λ
-2 -1 0 1 2 -2 -1 0 1 2
T T T T

• to
• Conversion Scaling in time
sequence leadstotoT scaling
amounts =1 in frequency
– Compression/expansion in time leads to expansion/
• compression
Resulting Fourier Transforminnow
frequency
periodic with period 1
• Conversion to sequence amounts to T→ 1 28
Summary of Continuous-to-Discrete Conversion
Summary of Sampling Process
p(t)
1

t
p(t) T

xc(t) xp(t) Impulse Train x[n]

to Sequence
xc(t) xp(t) x[n]

t t n
T 1

Xc(j2πf) Xp(j2πf) X(ej2πλ)

A A
A T T

f f λ
0 -2 -1 0 1 2 -2 -1 0 1 2
T T T T

• Sampling
• Sampling leads leads to
to periodicity periodicitydomain
in frequency in frequency domain
We need to avoid overlap of replicated 29
The Sampling Theorem
The Sampling Theorem
xc(t) p(t) xp(t)
1

t t t
T T

Xc(j2πf) P(j2πf) Xp(j2πf)

1 A
A T T

f t f
-fbw fbw -2 -1 0 1 2 -2 -1 0 1 2
T T T T T T T T
-fbw fbw
- 1 + fbw 1 - fbw
T T

• No overlap• inOverlap
frequencyin frequency domainwhen:
domain (aliasing) (i.e., aliasing) is
– !⁄" − 𝑓&' ≥avoided
𝑓&' or !if:
⁄" ≥ 2𝑓&'
• We refer to the minimum 1/T that avoids aliasing as the Nyquist sampling
frequency of a signal
• We refer to the minimum 1/T that avoids aliasing 30
Sine Wave Example: Sampling Above Nyquist Rate
Example: Sample a Sine Wave
t
p(t) T
xc(t) xp(t) Impulse Train x[n]
to Sequence

t t n
T 1

Xc(j2πf) Xp(j2πf) X(j2πλ)

f f λ
-1/T 0 1/T -1/T 0 1/T -1 0 1

Sample
• Time domain: Resulting rate is well
sequence above same
maintains Nyquist rate as input
period
signal
• Time domain: resulting sequence maintains the
same period as the input continuous-time signal 31
Sine Wave Example: Sampling at Nyquist Rate
Increase Input Frequency Further …
t
p(t) T
xc(t) xp(t) Impulse Train x[n]
to Sequence

t t n
T 1

Xc(j2πf) Xp(j2πf) X(j2πλ)

f f λ
-1/T 0 1/T -1/T 0 1/T -1 0 1

• Time Domain: Resulting sequence

Sample rate maintains same
is at Nyquist period as input
rate
signal
• Time domain: resulting sequence still maintains the
same period as the input continuous-time signal 32
Sine Wave Example: Sampling at Half the Nyquist Rate
Increase Input Frequency Further …
t
p(t) T
xc(t) xp(t) Impulse Train x[n]
to Sequence

t t n
T 1

Xc(j2πf) Xp(j2πf) X(j2πλ)

f f λ
-1/T 0 1/T -1/T 0 1/T -1 0 1

• Sequence now appears as constant

Sample rate is at half
(zero frequency) signal
the Nyquist rate
• Frequency domain: Aliasing to 0
• Time domain: resulting sequence now appears as a
DC signal! 33
Sine Wave Example: Sampling Below the Nyquist Rate
Increase Input Frequency Further …
t
p(t) T
xc(t) xp(t) Impulse Train x[n]
to Sequence

t t n
T 1

Xc(j2πf) Xp(j2πf) X(j2πλ)

f f λ
-1/T 0 1/T -1/T 0 1/T -1 0 1

• Resulting sequence is now a sine wave with a different period than the input
Sample rate is well below the Nyquist rate
• Frequency domain: Aliasing to a lower frequency
• Time domain: resulting sequence is now a sine
wave with a different period than the input 34
The IssueThe
of High
Issue Frequency Noise Noise
of High Frequency
p(t)
1

t
p(t) T

Undesired xc(t) xp(t) Impulse Train x[n]

Signal or to Sequence
Noise
Xc(j2πf) Xp(j2πf) X(ej2πλ)
A A
A T T

f f λ
-1 0 1 -2 -1 0 1 2 -2 -1 0 1 2
T T T T T T
• Typically set sample rate to exceed desired signal’s bandwidth
• We typically set the sample rate to be large
• Real systems can introduce noise/other interference at high frequencies
enough to accommodate full bandwidth of signal
– Sampling causes noise to alias into desired frequency band
• Real systems often introduce noise or other
interfering signals at higher frequencies
35
Anti-Alias Filtering
Anti-Alias Filtering
p(t)
1

t
Anti-alias p(t) T
Lowpass
xc(t) xc(t) xp(t) Impulse Train x[n]
H(j2πf)
to Sequence
H(j2πf)
Xc(j2πf) Xp(j2πf) X(ej2πλ)

f f λ
-1 0 1 -2 -1 0 1 2 -2 -1 0 1 2
T T T T T T
• Practical A/D converters include a continuous-time filter before the sampling
operation • Practical A-to-D converters include a continuous-
– Designed time
to filter out noise
filter andthe
before interference
sampling above 1/2T in frequency, prevent
operation
aliasing
– Designed to filter out all noise and interfering signals
above 1/(2T) in frequency 36
Summary: Advantages of Going Digital
• Allows error correction to be achieved
– Less sensitivity to radio channel imperfections

• Enables compression of information

– More efficient use of the channel

• Supports a wide variety of information content

– Voice, text, video can all be represented as digital bit streams

37
Roadmap
1. Analog I/Q modulation

2. Discrete-time processing of continuous signals

3. Digital I/Q Modulation

38
Digital I/Q modulation
Digital I/Q Modulation
Baseband Input Receiver Output
3 3
1 it(t) ir(t) 1
-1 H(j2πf) -1
I
-3 -3
t Lowpass t
2cos(2πfot) 2cos(2πfot)
2sin(2πfot) 2sin(2πfot)
3 3
1 qt(t) qr(t) 1
-1 H(j2πf) -1 Q
-3 -3
t Lowpass t
Sample
Times
• Leverage•analog
Leverage analog communication
communication channel to send
channel to send discrete-valued symbols
– Example:discrete-valued
Send symbol fromsymbols
set { -3, -1, 1, 3 } on both I and Q channels
– Example: send symbol from set {-3,-1,1,3} on both
• At receiver, sample
I andI/Q
Q waveforms every
channels each symbol
symbol period
period
– Associate
• Ateach sampled sample
receiver, I/Q valueI/Q
with symbols from every
waveforms same set on both I and Q
symbol
channels
period
– Associate each sampled I/Q value with symbols from 39
Constellation Diagrams
Constellation Diagrams
3
Baseband Input Constellation Diagrams
3
Receiver Output

1 it(t) Input
Baseband ir(t) 1 Receiver Output I
-1 3 H(j2πf) -1 3
-3 1 it(t) i-3
r(t) 1
-1 t Lowpass
H(j2πf) -1 tI
-3
2cos(2πfot) 2cos(2πfot) -3
2sin(2πfot)t 2sin(2πfot) Lowpass t
3 2cos(2πfot) 2cos(2πfot) 3
1 2sin(2πfot) 2sin(2πfot) qr(t) 1 3
3 qt(t)
-1 H(j2πf) q-1
Q
1 qt(t) r(t) 1
-3 -1 H(j2πf) -3 -1 Q
-3 t Lowpass -3 t
t Lowpass t

• Plot I/Q samples on x-yon axis

Sample
• Plot I/Q samples x-y axis
Sample
Times
Times
• Plot I/Q samples on x-y axis
– Example: sampledsampled
– Example: I/Q value of of
I/Q value
– As samples are plotted,
{1,-3} forms
constellation
{1,-3}aforms
dot at
diagram
x=1,
a dot y=-3y=-3
at x=1,
Q Q

eventually displays all possible symbol values

– As more samples are plotted,
3
– As more samples are plotted,
3
constellation diagram eventually
constellation diagram eventually
displays all possible symbol values1
1
I
displays all possible symbol values -1 I
• Provides a sense of• how
Constellation
easy it is todiagram provides-1
distinguish
• Constellation
between different a sensediagram
symbols of how easyprovides
it is to
-3

a sense distinguish
of how easy it isdifferent
to
-3
between -3 -1 1 3 40
Sending Digital Sending
Bits Digital Bits
3
Baseband Input
Sending Digital Bits
3
Receiver Output

1 it(t)
Baseband Input ir(t) 1 Receiver Output I
-1 3 H(j2πf) -1 3
-3 1 it(t) -3ir(t) 1
-1 t Lowpass H(j2πf) -1 tI
-3
2cos(2πfot) 2cos(2πfot) -3
2sin(2πfot) t 2sin(2πfot) 2cos(2πf t)Lowpass t
3 2cos(2πfot) o 3
2sin(2πfot) 2sin(2πfot) q (t) 1
1 3 qt(t) r 3
-1 1 qt(t) H(j2πf) -1qr(t) 1 Q
-1 H(j2πf) -3 -1 Q
-3
-3 t Lowpass -3 t
t Lowpass t
Sample
Sample
• Assign bits to eacheach
• Assign •I/QAssign
symbol
I/Qeach
symbol to a to a
I/Q symbol
Times
Times

– Example: I/Q = {1, 3}set

translates to bits
of digital bits 1110
set of digital bits
Q Q

– Example: I/Q = {1,3} translates 10

– Example: I/Q to = of
bits {1,3}
1110translates 10
– Gray coding minimizes bit errors when symbol
to bits of– 1110
11
errors are made Gray coding minimizes bit errors11 I

– Gray codingwhen minimizes bit errors

symbol errors are made 01
I
when symbol • errors
Example:are
I/Qmade
= {1,1} translates
• Example: I/Q = {3, 3} translates to bits 1010
01 00
to bits of 1010
(one bit flip• from
Example:
{1, 3})I/Q– Only
= {1,1} translates
one bit change from
00 00 01 11 10
41
The Impact of Noise
The Impact of Noise
The Impact of Noise
Baseband Input Receiver Output
3 3
1 it(t) Receiver ir(t) 1
-1 Baseband Input Noise H(j2πf) -1 Receiver Output I
-3 3 3
-3
1 t it(t) Receiver Lowpass ir(t) 1 tI
-1 2cos(2πfot) Noise t)
2cos(2πf H(j2πf) -1
o
-3 2sin(2πfot) 2sin(2πfot) -3
t Lowpass t
3 2cos(2πfot) 2cos(2πfot) 3
1 qt(t) 2sin(2πfot) 2sin(2πfot) qr(t) 1
-1 3 H(j2πf) -1q (t) 31 Q
1 qt(t) r
-3 -3 Q
-1 t Lowpass H(j2πf) -1 t
-3 -3
t Lowpass t
Sample
• Noise perturbs sampled I/Q I/Q
Sample
Times
• At receiver, noise perturbs
• Noisesampled
perturbs I/Q values
sampled Times

values
– Constellation values
points no longer consist of single Q Q

points for –each symbol– Constellation

Constellation points no points
longerno longer 10 10
consist of single dots for each
consist of single
symbol
dots for each 11
• Issue: What’s the best way of matching received I/Q
symbol
11 I
• Issue: what
samples with their corresponding is the best
transmitted symbols?
01 I
• Issue: what is the best
way to match received I/Q
01
00

way to match received

samples I/Q
with their 00 00 01 11 10
corresponding symbols? 42
Symbol Selection based on Slicing
Symbol Selection Based on Slicing
Baseband Input Receiver Output
3 3
1
-1
it(t) Symbol Selection Based on Slicing
Receiver
Noise H(j2πf)
ir(t) 1
-1 I
-3 Baseband Input -3 Receiver Output
t 3
Lowpass 3
t
2cos(2πf
1 o t) it(t) 2cos(2πf t)
Receiver
o ir(t) 1
2sin(2πf
-1 ot) 2sin(2πfoNoise
t) H(j2πf) -1 I
3 -3
t
3 -3
Lowpass t
1 qt(t) 2cos(2πfot) qro(t)
2cos(2πf t) 1
-1 3
2sin(2πfot) H(j2πf)
2sin(2πfot) -1 3
Q
-3 1 qt(t) -3 qr(t) 1
Qt
t -1 Lowpass H(j2πf) -1
-3 -3
t Lowpass
Sample t

• Receiver•matches
MatchI/QI/Q samples to samples to
Sample
Times
samples to corresponding
• Match I/Q Times

symbols based on decision

their regions
corresponding comprised of
their corresponding 00
Q
01
Q
11 10
decision boundaries symbols based on 00 01 11 10
symbols based on
00

– Choose decision boundariesdecision regionssymbol

to minimize 00

errors decision regions – Choose

Decision 01
decision regions Boundaries I
to minimize symbol Decision 01
11

– Choose decision regions

errors Boundaries 10
I
to minimize
• Decision boundaries symbol
also called– slicing
Decision boundaries
levels are 11
also called slicing levels
errors Decision
Boundaries 10 43
Transitioning Between Symbols
Transitioning Between Symbols
Baseband Input Transitioning Between Symbols
Receiver Output
3 3
1 Baseband Input
i (t) Receiver ir(t) 1 Receiver Output
t
-1 3 Noise H(j2πf) -1 3 I
1 it(t) Receiver ir(t)-3 1
-3 Noise It
-1 t Lowpass
H(j2πf) -1
-3 2cos(2πfot) 2cos(2πfot) -3
t
2sin(2πf Lowpass t
ot)
2cos(2πfot)
2sin(2πfot) ot)
2cos(2πf
3 3
2sin(2πfot) 2sin(2πfot) qr(t) 1 3
1 3 qt(t)
-1 1 qt(t) H(j2πf) qr(t)-1 1 Q
-3 -1 H(j2πf) -3-1 Q
-3 t Lowpass
-3
t
t Lowpass t
Sample
Sample
Times
Times
• Transition•behavior
Transition
• Transition behavior
influenced between
by transmit
behavior I/Q input
between
QQ

waveforms and receiveisfilter

symbols
symbolsinfluenced by by
both 0000 0101 11 10
is influenced both 11 10
00
transmit I/QI/Q input waveforms
00
transmit input waveforms
– Today: Focus on what the transmitter does
and receive
and receive filter
filter 01
01
II
– We– will
Wefocus
will focus on impact
on impact of of 11
11
– Ignore impact of noise for behavior
transition this analysis
transition behavior at at 10
transmitter
transmitter today
today
10
44
Transitions and the Transmitted Spectrum

• Want transmitted spectrum with minimal

bandwidth

• Issue: Sharply changing I/Q leads to very

wide bandwidth spectrum

45
Impact of Transmit Filter

• Transmit filter shapes the pulses,

enables reduced bandwidth for
transmitted spectrum

• Next Issue: Can lead to bit errors

46
Impact of Low Bandwidth Transmit Filter

• Lowering the transmit bandwidth

leads to increased bit errors

47
Eye Diagrams

• Eye Diagram: Wrap signal back onto itself in periodic time intervals, retaining all
“traces”

48
Looking at Many Symbols

• Increasing the number of symbols eventually reveals all possible symbol transition
trajectories
– Intuitively displays the impact of filtering on transmitted signal
49
Assessing the Quality of an Eye Diagram

Sample times:
• Eye diagram allows visual inspection of the impact of sample time and decision
boundary choices
– Large “eye opening” implies less vulnerability to symbol errors
50
Relating Eye Diagrams to Constellation

• Open eye diagrams lead to tight symbol

groupings in constellation

– Assumes proper sample time

placement

51
Impact of Low Transmit Bandwidth

• Eye diagrams show increased inter-

symbol interference (ISI)

– Also show symbol decision

sensitivity to sample timing
placement
52
Digital Modulation: Summary
• Digital modulation: Sends discrete-valued symbols through an analog
communication channel
– Receiver must sample I/Q signals at the appropriate time

– Receiver matches I/Q sample values to corresponding symbols based on decision

regions

– Constellation diagrams are a convenient tool to see likelihood of bit errors being
made

• Choice of transmit filter: Tradeoff between achieving a minimal bandwidth

transmitted spectrum and minimal ISI
– Eye diagrams: A convenient tool to see effects of ISI and sensitivity of bit errors to
sample time choice
53