EE 580

Wireless Communications Systems

Fall 2004
Digital Modulation

Richard S. Wolff, Ph. D.

[email protected]
406 994 7172
509 Cobleigh Hall
Modulation and wireless systems
• Analog:
– used in early radio systems,
– first generation cellular
• Digital:
– more efficient use of spectrum
– Better noise immunity (regeneration)
– inter-works better with other system elements and data applications,
– typical in second generation cellular
– Supports complex signal conditioning and processing techniques
• Source coding
• Encryption
• Equalization
– DSPs used to implement modulators and demodulators in software
– Software radio: alterations and improvements easy to implement!!
Representation of modulating
signal
• Message is a time sequence of symbols
• Each symbol can have m states
• Each symbol represents n bits of information

n = log 2 m bits/symbol
Representation of modulating
signal: example
Number of log2m n bits/symbol
states, m
Transmit m symbols/sec
1 0 0 Results in n bits/sec

2 1 1

4 2 2

8 3 3

16 4 4
 Choosing a modulation technique:
desirable features

• Low bit error rate at low received signal to

noise ratio
• Performs well under multipath and fading
conditions
• Occupies minimum bandwidth
• Easy and inexpensive to implement
Numerous trade offs implied by these attributes!!
Key issues in selecting a
modulation technique

• Power efficiency
• Bandwidth efficiency
Power efficiency

• Power efficiency ηp : ability of modulation technique to

preserve the fidelity of the digital message at low power
levels
• Increase in signal power can raise noise immunity (for
thermal noise)
• Trade off between fidelity (low errors) and signal power
• Metric: signal energy per bit/noise power spectral density:
Eb/N0 needed to achieve a a certain probability of error
(BER)
BER versus EB/N0 for different PSK
and FSK modulation techniques
Example: BER versus Eb/N0
Effects of Rayleigh fading
Bandwidth efficiency
• bandwidth efficiency : ability of modulation technique to
accommodate data within a limited bandwidth
• Increase in signal rate will decrease pulse width of digital
symbol, which increases the signal bandwidth (think of
Fourier transform: ∆T~1/∆f)
• Result is a trade off between symbol rate R and bandwidth
occupancy B
• Metric: ηB=R/B bps/Hz
• System capacity of digital mobile system directly related to
bandwidth efficiency of the modulation technique
Bandwidth efficiency: examples

• Occupied bandwidth B ~ 1/R, R=symbol rate

• BPSK: 2 states (m=2), n=1bit/symbol,
η=R/B=1/1=1 bps/Hz
• DPSK: 4 states (m=4), n=2 bits/symbol
η=R/B=2/1=2 bps/Hz
• 8PSK: 8 states (m=8), n=3 bits/symbol
η=R/B=3/1=3 bps/Hz
Channel capacity: Shannon limit

• Upper bound on achievable bandwidth efficiency

– Assumes non-fading channel
– Assumes additive white Gaussian noise (AWGN)

C S
η B max = = log 2 (1 + )
B N
C = channel capacity, bps
B = RF bandwidth
S / N = signal to noise ratio
 Shannon limit, examples
• AMPS channel, B = 30kHz. Assume SNR=20dB.
What is maximum channel capacity C?
S
C = B log 2 (1 + ) = 30,000 log 2 (1 + 100)
N
C = 199.7 kbps

Note that data rate actually used in an AMPS channel is well

below the Shannon limit!
Voice message signal = 64kbps
Voice band data signal = 48kbps
Shannon limit, examples
• Deep space link, B = 30kHz. Assume SNR=0.1dB.
What is maximum channel capacity C?
S
S S dB / 10
= 0 . 1 dB , ratio = 10 N
N N
S
ratio = 10 0 . 1 / 10 = 10 0 . 01 = 1 . 023
N
S
C = B log 2 ( 1 + ) = 30 , 000 log 2 ( 1 + 1 . 023 )
N
C = 30 . 49 kbps

Earth-space links are typically power limited: low SNR do to:

1. Limited spacecraft transmit power
2. Large path loss
Other factors affecting choice of
modulation technique

• In addition to power and bandwidth

efficiency, must consider:
– Cost and complexity of subscriber equipment
– A modulation that is simple to detect
– Performance under Ricean and Rayleigh fading
– Performance under multipath conditions
– Interference environment
– Effects of timing jitter
 Bandwidth and power spectral
density

• Signal spread in frequency inversely proportional to

symbol rate R
• Narrow pulses (high symbol rate) lead to large spectral
BW
• Define absolute bandwidth of signal as frequency
range over which the signal has a non-zero power
spectral density PSD
For symbols with rectangular pulse shape,
PSD ~ (sin f ) 2 / f 2
extends over ∞ frequency range
 PSD for rectangular pulses
Amplitude

Tp Time
 Definitions of signal bandwidth

• Null-to-null bandwidth: width of main spectral

lobe
• Half-power bandwidth: frequency where PSD has
dropped to half (3dB) of its peak value
• Occupied bandwidth: defined by FCC as band
which leaves 0.5% of power below lower and
above upper band limit. 99% of power in occupied
bandwidth
Line coding

• Line coding is the scheme used to represent

successive zeros and ones
• Return to zero (RZ): pulse returns to zero
within each symbol period

1 1 0 1
 Line coding II
• Non-return to zero (NRZ): pulse does NOT return
to zero within each symbol period
• Manchester coding: special case of NRZ
– “1” represented by a positive pulse followed by
a negative pulse (within a symbol period)
– “0” represented by a negative pulse followed by
a positive pulse (within a signal period
• Line codes can be
– unipolar (0 or +V)
– Bipolar(-V or +V)
 Line coding examples

Unipolar NRZ

Bipolar NRZ

Unipolar RZ

Bipolar RZ.

Manchester code
Line coding trade offs

• RZ: return to zero every symbol adds transitions

– Leads to spectral broadening
– Improves timing synchronization
• NRZ codes
– More spectral efficiency
– Poorer synchronization
• Manchester code
– No DC component
– Simple synchronization
– More spectral spreading than NRZ
Power spectra of Unipolar NRZ

frequency is normalized with respect to the bit rate 1/Tb, and the
average power is normalized to unity
Power spectra of bipolar NRZ

frequency is normalized with respect to the bit rate 1/Tb, and the
average power is normalized to unity.
 Power spectra of unipolar RZ

frequency is normalized with respect to the bit rate 1/Tb, and the
average power is normalized to unity.
Power spectra of Bipolar RZ

The frequency is normalized with respect to the bit rate

1/Tb, and the average power is normalized to unity.
Power spectra of Manchester code

frequency is normalized with respect to the bit rate 1/Tb, and the
average power is normalized to unity.
Comparison of line coding spectra

Unipolar NRZ

Bipolar RZ

Manchester NRZ
Effects of bandwidth limiting a
digital signal

• Typically band pass limit the transmitted

signal
• Cutting off some of the energy in the
frequency domain leads to spreading in the
time domain
• Net effect is inter-symbol interference
(ISI)!!
Graphical “proof” of ISI due to
band limiting
Transmit rectangular pulses in time domain

time

Fourier transform of a pulse

frequency
Graphical “proof” of ISI due to
band limiting II
Pass transmitted pulse through a band pass filter

frequency

Transmitted pulse is truncated in the frequency

domain

frequency
Graphical “proof” of ISI due to
band limiting III

Fewer frequency components result in wider pulse in

time domain

time

Intersymbol interference results where received

pulses overlap in time domain
Pulse shaping

• Technique to limit the bandwidth of the

received signal while at the same time
minimizing ISI
• Typical approaches: round off the “sharp
edges”
– Raised cosine roll-off filter
– Gaussian filter
Raised Cosine transfer function

1⎡ ⎡ π ( f 2Ts − 1 + α ) ⎤ ⎤
H RC ( f ) = ⎢1 + COS ⎢ ⎥⎥
2 ⎢⎣ ⎣ 2α ⎦ ⎥⎦
(1 − α ) (1 + α )
≤ f ≤
2Ts 2Ts

1 B
Rs = =
Ts 1+α

RF bandwidth increases with α, but ISI goes down

Spectrum of Raised Cosine pulse
Virtue of pulse shaping
 Representing digital signals in
vector space
• Consider a set S of waveforms representing
discrete symbols
• For binary modulation, bit=symbol, S has two
symbols
• For higher order modulation, S>2

n = log 2 M , n = number of bits

that can be transmitt ed in a symbol
Example: QPSK has 4 symbols (0, π/2,π,3π/2), log2(4)=2 bits/symbol
 BPSK constellation diagram

•Two symbols (0, π)

•Distance between the symbols represents how different the signals are
•Large symbol separation makes it easier for receiver to detect symbol
correctly
Noise effects on constellation

As the noise increases, the symbol smears out, and can overlap
with another symbol, causing a symbol error