Basic M-ary Digital Communication

Digital Modulation Techniques

- Overview - System

 Digital Modulation Basics

 Vocabulary/Notation Modulate;
(analog) Collect k bits;
 Basic Phase Modulation: BPSK, QPSK, MPSK, DPSK
A/D symbols →
information bits → symbols
 Basic Frequency Modulation: FSK
waveforms
channel
 Performance Measures & Fundamental Limits

 QPSK Variations (π/4 QPSK, OQPSK, MSK, DQPSK)

 QAM & OFDM Demodulate;
(analog)
 Pulse Shaping information
D/A bits ← symbols waveform →
 Conclusions symbols

10/2/07 van Alphen & Katz 1 10/2/07 van Alphen & Katz 2

M-ary Communication System:

Symbol –Level Considerations Simplified QPSK Example

 Transmits one of M possible waveforms  Say we use Quaternary Phase Shift Keying (QPSK) as our
 Each symbol modulation.
 corresponds to a message mi, i = 1, 2, …, M  We need M = 4 waveforms, with 4 different phase angles:
 can represents k bits of information, where M = 2k sin(2πf0t) Bits-to-Waveforms
 is associated with a waveform si(t), of duration T seconds 01
 T = Ts is called the symbol time or symbol duration (Gray-coded)
11 00
 To send message mi: cos(2πf0t)
 transmit waveform si(t), 0 < t ≤ T
 receiver guesses which of M possible messages was sent
10
10/2/07 van Alphen & Katz 3 10/2/07 van Alphen & Katz 4
 QPSK Receiver (Generalized) QPSK Example for Digital Image
2
sin(!0 t ) Consider a vector, v, taken from a digital color image with 8
T quantization levels (3 bits) for R, G and B:
01 Dotted lines ⇔ receiver decision
s1 boundaries (for equally likely messages in [5 7 4 … 4 0 3 5 … 6 2 1 0 … 5]
additive white noise)

red green blue

11 s3 s4 00 2
cos(!0 t )
T
E Vector Value 5 7 4 …
Note that the mapping between bits and Binary-Coded 101 111 100
waveforms is done using a Gray Code:
Tx Bit Sequence for M = 4:
s2 Noise that pushes the received point past a
single decision boundary causes only a (re-grouping) 10 11 11 10 …
10 (Grouping k = log2(M) = log2(4) = 2 bits per symbol)
single bit error.
10/2/07 van Alphen & Katz 5 10/2/07 van Alphen & Katz 6

Quaternary Example, Continued Relating Bit and Symbol Parameters

Bit Sequence 10 11 11 10  1 symbol ⇔ k = log2(M) bits
 T = Ts = kTb
-sin(2πf0t) -cos(2πf0t)  e.g., if M = 4, then k = 2 and T = Ts = 2Tb
-cos(2πf0t) -sin(2πf0t)
 Baud Rate R symbols/sec = 1/T
t  Bit Rate Rb bits/sec = 1/Tb
0 T 2T 3T 4T
 E = S T (E: energy in the waveform of duration T; S:
Say f0 = 1000 Hz, T = 1 ms average signal power)
10 11 11 10 Note: abrupt  Eb = S Tb (Eb: energy allocated per bit; Tb is time allocated per bit)
1
phase
0 changes  Bandwidth proportional to 1/T
increase the
 Decreasing T ⇔ Faster signaling ⇔ Wider Bandwidth
0 0.001 0.002 0.003 0.004
-1 bandwidth
10/2/07 van Alphen & Katz 7 10/2/07 van Alphen & Katz 8
 BPSK: Optimum Receiver for AWGN
Binary Phase Shift Keying (BPSK) (Assume equally likely, equal energy signals)

 M = 2 ⇒ 2 waveforms, 180° out of phase

2E
s1 ( t ) = cos(!0 t ) T I
T r(t) x !
s1 ^s
0 <t ≤T Correlation I>0 i
0
2E
s2 ( t ) = " cos(!0 t ) Receiver
T 2
"1 ( t ) = cos(!0 t )
T
 Signal Space Diagram (1-dimensional)
! E E
2
X X cos(!0 t )
s2 0 T
s1

 Throughput: Rb = 1/T = 1/Tb bps

10/2/07 van Alphen & Katz 9 10/2/07 van Alphen & Katz 10

Quaternary Phase Shift Keying (QPSK) QPSK Modulation

x
 M = 4 ⇒ 4 waveforms, 90° out of phase
deven
2E 2E "1 ( t ) = 2 cos(!0 t )
s4 ( t ) = cos(!0 t ) s3 ( t ) = " cos(!0 t ) T
T T I-Q Serial
0 <t ≤T Transmitter di -90°
Σ s(t
2E 2E Parallel )
s1 ( t ) = sin(!0 t ) s2 ( t ) = " sin(!0 t )
T T dodd
2 x
sin(!0 t )
 Signal Space Diagram (2-dim.) T
x
d(t) 0 1T 2T 3T
t

 2 bits/symbol
Data stream
x x 2 dodd(t) t
cos(!0 t ) Example 0 1T 2T 3T
E T
 Throughput: Rb = 2/T bps deven(t) t
0 1T 2T 3T
x
10/2/07 van Alphen & Katz 11 10/2/07 van Alphen & Katz 12
 MPSK: Optimum Receiver for AWGN
8-ary PSK (Assume equally likely, equal energy signals)

T I
 M = 8 ⇒ 8 waveforms, 45° out of phase x !
0
2
Signal Space Diagram (2-dim.) sin(!0 t ) Choose si
 T I-Q r(t) "1 ( t ) = 2 cos(!0 t ) θ = tan-1(Q/I) s^i
x T w/nearest θi
x Receiver
 3 bits/symbol x T Q
x !
0
x x 2
cos(!0 t )
E T " 2 ( t ) = 2 sin(!0 t )
T
x x ψ2(t)
x r
I: Projection of r(t) onto ψ1 axis Q
Q: Projection of r(t) onto ψ2 axis θ
!1 ( t )
 Throughput: Rb = 3/T bps 0 I

10/2/07 van Alphen & Katz 13 10/2/07 van Alphen & Katz 14

Bit Error Probabilities for MPSK QPSK vs. BPSK

Signals in AWGN (The “Almost Free” Lunch)
Bit Error Probabililty, MPSK
1.E+00
Signaling  QPSK has twice the throughput as BPSK
1.E-01

1.E-02  QPSK and BPSK have

1.E-03
M=2, 4

1.E-04 M=8
M=16
 the same transmission bandwidth
1.E-05 M=32

the same bit error probability, PB

M=64
1.E-06

-2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Eb/N0, dB

 even though QPSK has a higher symbol error

probability

10/2/07 van Alphen & Katz 15 10/2/07 van Alphen & Katz 16
 Differential Phase Shift Keying (DPSK)
- Motivation - (Binary) DPSK

 Consider a BPSK system in which the transmitter local oscillator  Concept: Differentially encode the information bits, prior to
and the receiver local oscillator are out of phase by angle θ. transmission; differentially decode the received bits
Tx Rcvr  m2 = 0 → no change in phase of the sinusoid, relative to the previous
burst
2 sin(!0 t ) 2 sin(!0 t )
T T
 m1 = 1 → 180° change in phase of the sinusoid, relative to the
previous burst
! E E X θ
X X 2 cos(!0 t )  Tx phase angle θi = θi-1 + θi where θi denotes the i-th
T
X information bit: Δθi = θi – θi-1 = θi
 Note: Differential schemes always require the transmission of
 Note that some phase offset between the tx and the rcvr is
one additional reference bit prior to transmitting the data.
always present, and usually necessitates a PLL at the rcvr
10/2/07 van Alphen & Katz 17 10/2/07 van Alphen & Katz 18

Example: Transmitted Waveforms

for BPSK and DPSK DPSK

 Consider the bit stream  Phase information is only relative to that of the previous
pulse ⇒ no need to generate phase reference at the
0 1 1 0 1 receiver (No PLL required).
BPSK

1.5  Performance is degraded from that of BPSK in AWGN, due

s(t)

0
-1.5 to induced dependence of errors from bit-to-bit.
0 1 2 3 4 5
t, s
 However, DPSK eliminates degradation due to phase offsets
DPSK between the transmitter and the receiver.
1.5

0
 Note that we have only discussed Binary DPSK; Quarternary
-1.5
DPSK (QDPSK) is also commonly used, with similar
0 1 2 3 4 5 6 encoding and decoding at the tx and the rcvr.
10/2/07 van Alphen & Katz 19 10/2/07 van Alphen & Katz 20
 M-ary Frequency Shift Keying: MFSK Binary Frequency Shift Keying (BFSK)

 M = 2 ⇒ 2 waveforms, at 2 different frequencies

 As M increases
2E
s1 ( t ) = cos(2!f 0 t )
 Bandwidth increases (bandwidth efficiency decreases) T
0 <t ≤T
2E
 Pb decreases (if the frequencies are chosen to yield s2 ( t ) =
T
cos(2"(f 0 + !f ) t )
orthogonality)
⇒ power efficiency increases  Signal Space Diagram (for the special case of
orthogonal signaling)
Constant envelope signaling E X s2
Δf = 1/T, NC Rcvr
⇒ No performance degradation from the use of 1/(2T), Coh. Rcvr
s1 2
X cos(2!f 0 t )
non-linear amplifiers 0 E T

 Throughput: Rb = 1/T = 1/Tb bps

10/2/07 van Alphen & Katz 21 10/2/07 van Alphen & Katz 22

BFSK Optimum (Coherent) Receiver PSD: NRZ Baseband Signaling, BPSK

1
0.9
T
x ! 0.8
0 0.7
Choose
s^i
2 cos(2!f 0t ) 0.6
r(t)
G(f)/E

T largest 0.5
T 0.4
x ! 0.3
0
0.2
2
cos(2" ( f 0 + !f )t ) 0.1
T
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

fTs
10/2/07 van Alphen & Katz 23 10/2/07 van Alphen & Katz 24
 PSD: NRZ Signaling, BPSK
(dB Scale, with Bandwidth Definitions) Communications Link: the Channel

 Channels are characterized by their capacity

0
-2
 Capacity: An inherent limit on the rate at which
-4 W3dB
-6 information can be sent “error free”
G(f)/E, dB

-8
-10
 Capacity increases with bandwidth (W) and signal-to-
-12
-14 noise ratio (SNR or S/N)
-16
WB,16 dB
-18
-20
-22 WN-N Info C = W log(1 + S/N)
0 2 4 6 8 10 12 14 16
fTs

Bandwidth,
SNR
10/2/07 van Alphen & Katz 25 10/2/07 van Alphen & Katz 26

Reviewing Shannon’s Theorem Efficiency Definitions and the

Bandwidth/Communications Efficiency Plane
& S#
 Shannon-Hartley Theorem: C = W ' log 2 $1 + !  Defn: Bandwidth Efficiency (or Normalized Throughput) =
% N"
Rb/W, bps/Hz
 Communication (with arbitrarily small error probability)

is . . .  Communications Efficiency: Eb/N0 required to Attain a

 Possible at rates Rb < C Particular Bit Error Probability
 Impossible at rates R > C
b  Bandwidth/Communications Efficiency Plane: Plots Rb/W vs.
 Establishes absolute theoretical limit on tx rate Eb/N0 for particular modulation techniques, assuming a
 Therefore, particular pulse shaping. Goal: to be as near as possible to
S the theoretical limit
R b / W < C / W = log 2 &$1 + #!
% N"  Note: Separate plots are usually done for each required bit
error probability
10/2/07 van Alphen & Katz 27 10/2/07 van Alphen & Katz 28
 The Bandwidth/Communications Efficiency Bandwidth Efficiency Plane, continued
Plane PB = 10-5
log2(Rb/W)
Shannon 4
14
Boundary: Rb = C Rb = C
12 Rb > C Region: No Bandwidth
Modulation/Coding M=4
10 0 M=2
-limited
Techniques w/P(E) 0 Region
8
Rb/W,
bps/Hz
6 Rb < C Region:
4 Modulation/Coding -4
Techniques Exist w/P(E) 0 Power-
2 limited
0 -8 Region
-5 0 5 10 15 20 25 30 35 -2 2 6 10 14 18 22 26 30
Eb/N0, dB MPSK
Eb/N0, dB
10/2/07 van Alphen & Katz 29 10/2/07 van Alphen & Katz 30

Bandwidth Efficient Digital Modulation QPSK Variations – Restricting Alowable Phase

Techniques Transitions

 Phase Modulation QPSK: OQPSK:

o o
 Discrete Phase  max phase transition 180  max phase transition 90

 loses constant envelope after  preserves constant

 MPSK

 QPSK, OQPSK, π/4-QPSK

More later on how filtering envelope after filtering

 DQPSK, ODQPSK, Oπ/4-QPSK

to do these, if the
basic modulation
 Continuous Phase techniques aren’t Compromise - π/4-QPSK:
max phase transition 135o, min 45o
 MSK, GMSK “efficient enough”
• preserves constant envelope:
 QAM • better than bandlimited QPSK
 OFDM • not as well as bandlimited OQPSK
• same BER in AWGN as QPSK, OQPSK

10/2/07 van Alphen & Katz 31 10/2/07 van Alphen & Katz 32
 QPSK Variations, cont. QPSK Variations, cont.

QPSK π/4-QPSK OQPSK MSK - reduction of phase

Discrete Phase Transition Techniques: transitions to obtain less
- reduction of phase
sidelobe regrowth and
QPSK π/4-QPSK OQPSK transitions to obtain less therefore less ISI
max 180o max 135o max 90o sidelobe regrowth and
therefore less ISI Special Case of MSK: GMSK (Gaussian Pulse Shape, time-bandwidth product BT)
• further narrows the spectrum, at the cost of re-introducing ISI (increasing BER)
Why not completely eliminate abrupt phase transitions? • smaller BT product more compact spectrum, but larger ISI
“Continuous Phase” Transition Techniques (CPFSK):
Eb/N0 Penalty in AWGN (vs. Ideal QPSK)
MSK: applies sinusoidal weight to OQPSK (constant envelope) MPSK GMSK, BT=.25 GMSK, BT = .3 GMSK, BT = .5
- same BER as QPSK & OQPSK (MF detector in AWGN) Ideal 0 dB .7 dB .3 dB .2 dB
- same bandwidth efficiency (bps/Hz) as QPSK & OQPSK Practical 1 .7 dB 1.7 dB
- spectrum has wider mainlobe than QPSK & OQPSK, but faster drop-off
of sidelobes (proportional to ω-4 vs. ω-2) 1 Simulation results, assuming non-optimal receiver

10/2/07 van Alphen & Katz 33 10/2/07 van Alphen & Katz 34

Quadrature Amplitude Modulation:

Power Spectral Densities Rectangular 16-ary QAM
10

0
MSK, dB QAM: send one coordinate on in-phase carrier, one on quadrature
QPSK, dB carrier ψ2(t)
-10

-20
3A
G(f) -30

-40 Throughput: A

-50
Rb = 4/T -3A -A A 3A ψ1(t)
-60
-3 -2 -1 0 1 2 3
-A

f/Rb, Hz/bps
-3A

10/2/07 van Alphen & Katz 35 10/2/07 van Alphen & Katz 36
 Orthogonal Frequency Division Modulation
(OFDM) Motivation for Pulse Shaping
Channel Transfer Function
 Choose N orthogonal tones to be used as sub-carriers: Sin(f) H(f) Sout(f)

2E
si ( t ) = cos(2"(f 0 + !f ) t ) i = 1, 2, …, N; f
T Δf = i/(T) f f
sin(t) Narrow relative to Signal sin(t)
 Demux the user data (serial-to-parallel conversion)
 Put each sub-stream of data onto a different sub-carrier,
t t
where the sub-carriers are the orthogonal tones described 0 T 0 T
above f1 f2 … fN  Band-limiting the signal in the frequency domain leads to time
Example where
x x x domain spreading of the signal
each sub-carrier
x x x x … x x
is modulated  Result: Intersymbol Interference, due to signal in one time-
x x x
with QPSK slot overlapping into the next time slot(s).
10/2/07 van Alphen & Katz 37 10/2/07 van Alphen & Katz 38

Motivation for Pulse Shaping: ISI Commonly Used Pulse Shapes

…
t  Nyquist Pulse: Cancel ISI (ideally)
0 T
 Example: Raised Cosine, w/roll-off parameter α = 0):

 Tails of signal overlap with body of succeeding signal 1.2

1 H(f)
α=0
 Destructive interference results when the pulses are of 0.8
α = .5
0.6
opposite polarity 0.4
0.2 α=1
 Problem cannot be solved by increasing signal power or 0
fTs ,
energy – as opposed to additive noise problems, which can! -1.5 -1 -0.5 0 0.5 1 1.5

ISI 2 2
Pb AWGN Pb  Gaussian Pulse: H G (f ) = e "! f
where α = .5887/B, and B is the 3-dB bandwidth of the filter
Eb/N0, dB Eb/N0, dB
10/2/07 van Alphen & Katz 39 10/2/07 van Alphen & Katz 40
 Power Efficiency Comparison for Digital Relative Complexity
Modulation Techniques in AWGN of Modulation Techniques [Ref: Oetting]

Required Eb/N0 for Various Bit Error Probabilities

(OPT.
Modulation P B = 10 -1 P B = 10 -2 PB = 10 -3 PB = 10 -4 PB = 10 -5 PB = 10 -6 DET.)
MPSK, M = 2, 4 -0.9 dB 4.3 dB 6.7 dB 8.4 dB 9.6 dB 10.5 dB LOW HIGH
M=8 1.0 dB 7.3 dB 10.0 dB 11.8 dB 12.9 dB 14.0 dB
NCFSK DPSK BPSK OQPSK CPFSK MPSK
M = 16 4.0 dB 11.4 dB 14. 4 dB 16.2 dB 17.5 dB 18.5 dB
M = 32 7.4 dB 16.0 dB 19.1 dB 21.0 dB 22.3 dB 23.4 dB OOK CPFSK DQPSK QPSK MSK QPR QAM
M = 64 11.2 dB 20.9 dB 24.2 dB 26.1 dB 27.4 dB 28.5 dB
DQPSK, MSK 2.0 dB 6.8 dB 9.2 dB 10.8 dB 12.0 dB 12.9 dB
GMSK, BT = .25 0.8 dB 6.0 dB 8.5 dB 10.1 dB 11.3 dB 12.2 dB (ED) (DISC.
BT=infinity -.1 dB 5.0 dB 7.5 dB 9.1 dB 10.3 dB 11.2 dB DET.)
16-QAM 1.9 dB 7.9 dB 10.5 dB 12.3 dB 13.6 dB 14.2 dB
64-QAM 5.0 dB 11.5 dB 15.0 dB 16.0 dB 17.8 dB 18.5 dB

10/2/07 van Alphen & Katz 41 10/2/07 van Alphen & Katz 42