Telecommunications B (EE2T21)

Lecture 5 overview:

More modulation techniques for digital signals

- Multi-level modulation schemes , e.g.:
* M-PSK
* Quadrature Phase Shift Keying (QPSK)
* Quadrature Amplitude Modulation (QAM)
- Advanced modulation schemes: bandwidth efficiency and robustness
* Orthogonal Frequency Division Multiplexing (OFDM)
* Direct Sequence-Spread Spectrum (DS-SS) modulation

EE2T21 Telecommunications B
Dr.ir. Gerard J.M. Janssen
May 2, 2022

1
Lectures & Work-instructions Telecommunications B
Telecommunication Techniques

Lectures:
Monday 9-5 1st+2nd hour
Tuesday 3-5 1st+2nd hour
Wednesday 11-5 3rd+4th hour

Q&A/Working lectures:
Wednesday 4-5 3rd+4th hour
Thursday 12-5 5th+6th hour

2
 Multi-level modulation techniques

In M-level modulation: for every sequence of l source bits, one out

of M  2l symbols or signal states is generated by the transmitter.

Rb [bit/s] R
The symbol rate D   b [ symb / s ] or [baud]
l [bit/symb] l
Per symbol, l bits are transmitted and the symbol time is Ts  lTb.

3
 M-PSK

The complex envelope for M-PSK is:

j ( t )

g (t )  Ac e  x(t )  jy (t )  Ac 

For each symbol, one out of M phase 4
phase states i is transmitted, with  
2 (i  1)
i  i  1, 2,.., M
M  
and xi  Ac cos  i , yi  Ac sin  i 
Signal state diagram for 8-PSK
Usually, all symbols are equally likely.

4
 Quadrature Phase Shift Keying (QPSK)

QPSK can be seen as 4-PSK or as two independent BPSK signals

on quadrature carriers:
s (t )  Ac [d1 (t ) cos c t  d 2 (t )sin c t ] d1 , d 2  {1,1}
5
 
4
-DQPSK
The 4 -DQPSK states are those of
 two QPSK constellations shifted
  over 4 rad.
 The bits to be transmitted
4 determine the phase shift.
 
The next state is one of the other
QPSK states, black or red, i.e.:
 
 11   = 45o
10   = -45o
In 4 -DQPSK (Differential QPSK),
01   = -135o
a form of differential coding is
applied. 00   = 135o

6
 Offset QPSK (OQPSK)

In OQPSK, the in-phase and

  quadrature phase symbols are
time shifted over 1/(2 D ) :

4 x '(t )  x(t )
y '(t )  y (t  21D )

  Half a symbol time delayed.

The signal spectrum is not affected

by this time-delay between
In OQPSK: - no 180o phase jumps,
- no zero-crossings
I- and Q- symbols.
- less amplitude
distortion Why do we do this?

7
 Quadrature Amplitude Modulation (QAM)

In QAM we allow discrete phase values as well as discrete

amplitude values. The general QAM signal structure is:
s (t )  x(t ) cos c t  y (t )sin c t
with: g (t )  x(t )  jy (t )  R (t )e j ( t )

x(t ), y (t ) can take on a limited number of discrete values.

or
R(t ),  (t ) can take on a limited number of discrete amplitude-
and phase values.

8
Quadrature Amplitude Modulation (QAM)

9
 Quadrature Amplitude Modulation (QAM)
As shown by the I,Q-description, general QAM can be seen as
two M-ASK signals modulating the quadrature carriers:

 n 
x(t )   xn h  t     xn h  t  nTs 
n   D  n

 n 
y (t )   yn h  t     yn h  t  nTs 
n   D  n

h(t ) is the pulse shape: - rectangular

- raised cosine
- half sinewave
Rb 1
D  is the symbol rate
l Ts
10
 Quadrature Amplitude Modulation (QAM)
QAM consists of two orthogonal M-ASK signals modulating
quadrature carriers:

x(t )   x h  t  nT 
n 
n s


y (t )   y h  t  nT 
n 
n s

For 16-QAM (black dots):

xn , yn  {3, 1, 1, 3}
(shown as red dots)
Rb
Symbol rate Rs 
4
11
 Power Spectral Density (PSD)

M-PSK and M-QAM with rectangular symbol shapes have the

same spectral shape as BPSK. However, the spectrum is a
factor l  2 log M (bits/symbol) narrower since Ts  lTb .

sin 2  fTs
Pg ( f )  2 PTs 2
 2 PlTb sinc 2
flTb
( fTs )
Proof in Couch p. 397.
In which P is the transmitted signal power.

2 Rb 2
The null-null bandwidth is: BT _ 00    2 Rs ( 2 D )
l Ts

12
Power Spectral Density (PSD)
Note: this is the single sided
spectrum in baseband!

13
 Spectral efficiency: profit (1)

The spectral efficiency for rectangular pulses is defined as:

Rb [bit/s] Rb 1/ Tb l  bit / s 
      where l  2
log M
BT [Hz] 2 D 2 / lTb 2  Hz 

For raised-cosine pulses we find with BT  (1  r ) D  (1  r ) Rb / l :

l  bit / s 

1  r  Hz 

When we can increase  by a factor of two, twice the data rate

can be transmitted in the same bandwidth: twice as many users/
turnover  doubling of the profit!

14
 Spectral efficiency: cost

According to Shannon:

2  S Rb C 2  S
Rb  C  B log 1        log 1  
 N B B  N
So, for a larger  a larger S/N is required.

This is not so strange when we study Fig. 5.32 once again.

For larger l ( larger  ), the signal states will be closer together
when the signal power remains constant  larger probability of
detection errors due to noise.

Reliable detection is only possible if also S/N is increased!

15
 Minimum Shift Keying (1)

What is the minimum frequency deviation F which allows for

orthogonal signaling when using FSK?

Let the FSK states be given by:

s1 (t )  Ac cos(1t  1 )
kTb  t  ( k  1)Tb
s2 (t )  Ac cos(2t   2 )

For orthogonal signaling the following should hold:

( k 1)Tb ( k 1)Tb

 s1 (t ) s2 (t )dt   Ac2 cos(1t  1 ) cos(2t   2 )dt  0

kTb kTb

16
 Minimum Shift Keying (2)
With application of:
2cos  cos   cos(   )  cos(   )
we get:
sin((1  2 )Tb  (1   2 ))  sin(1   2 )
1  2
sin((1  2 )Tb  (1   2 ))  sin(1   2 )

1  2
sin(2 h  (1   2 ))  sin(1   2 )
 0
2 h / Tb

where we used (1  2 )Tb  2  2F  Tb  2 h , h  2 F  Tb .

17
 Minimum Shift Keying (3)
Two situations:
1.) Continuous Phase FSK (CPFSK): 1   2

Orthogonal signaling when

sin(2 h  (1   2 ))  sin(1   2 )  0

Thus the minimum h which holds: 2 h    h  0.5,

which corresponds to the minimum shift:
2 h  1 Rb
1  2  2  2F    F  
Tb Tb 4Tb 4
FSK with h = 0.5 is also indicated as FFSK (Fast FSK) or
Minimum Shift Keying (MSK).

18
 Minimum Shift Keying (4)

2.) Discontinuous phase FSK: 1   2

Orthogonal signaling when

sin(2 h  (1   2 ))  sin(1   2 )  0

Now we find: 2 h  2  h 1
1 Rb
and F   ,
2Tb 2
which is twice as large as for CPFSK.

19
 Minimum Shift Keying (5)
MSK is FSK with h = 0.5.
The complex envelope for Minimum Shift Keying is given by:

g (t )  Ac e j ( t )
t
t
with  (t )  2F m( )d  
 m(t ) 0  t  Tb
0
2Tb

A  2 phase shift per Tb .

where
Rb 1   Rb
F   , D f  2F   , and m(t )  {1, 1},
4 4Tb 2Tb 2
20
 Minimum Shift Keying (6)
 t 
Now g (t )  Ac exp  jm(t )   x(t )  jy (t ) 0  t  Tb
 2Tb 

 t 
With x(t )  Ac cos    kTb  t  (k  2)Tb
 2 Tb 

 t 
y (t )  Ac sin    (k  1)Tb  t  (k  3)Tb
 2T b 

Since s (t )  x(t ) cos ct  y (t )sin ct we can conclude that MSK
is equivalent to OQPSK with half-sine symbol shapes with
duration 2Tb . The odd bits determine y(t) and the even bits x(t).

21
Minimum Shift Keying (7)

22
 Power Spectral Density of MSK
The spectral main lobe of MSK is
1.5x wider than for QPSK due to the
 cos 2 (2 f / R ) 
[ Py ( f )]dB  10log  2 2
 [1  (4 f / R) ]  half-sine pulse shape, however, the
side lobes decrease much faster,
 sin(2 f / R )  2 
because the pulse shape contains
[ Py ( f )]dB  10 log   
 ( 2 f / R )   less high frequency components (less
abrupt changes): better adjacent
channel properties.

 cos 2 (2 f / R) 
[ Py ( f )]dB  10 log 
In GMSK (Gaussian Minimum Shift
2 2
 [1  (4 f / R ) ] 
Keying, used in GSM), Gaussian
shaped pulses are used with even
more gradual slopes and therefore
less high frequency components and
even larger side lobe attenuation and
thus even better adjacent channel
properties.
23
Generation of MSK

24
Spectral efficiency of digital signals

25
 Orthogonal Frequency Division Multiplexing (1)

Orthogonal Frequency Division Multiplexing (OFDM) belongs to the family

of multi-carrier (MC) modulation. An OFDM signal consists of a number of
N carriers indicated as subcarriers {1, 2, ... , N }.

The OFDM signal has the following characteristics:

1. Subcarrier signals are orthogonal
2. Arbitrary modulation of the subcarriers: BPSK, QPSK, QAM, etc.
3. High spectral efficiency
4. Relatively simple generation and detection

Applications: WiFi, ADSL, digital (cable) radio and TV (DAB, DVB).

26
 Orthogonal Frequency Division Multiplexing (2)

In Orthogonal Frequency Division Multiplexing, a high bit rate single

carrier signal with symbol rate Rs is split up in N parallel data sub-
streams of a lower symbol rate Rss  Rs / N .

Each of the N low rate sub-streams modulates one of N parallel

sub-carriers.

 the symbol time is N times increased: Tss  NTs

 very bandwidth efficient

27
 Orthogonal Frequency Division Multiplexing (3)
Rs
w1 N

Rs cos 2 f1t
w2 N

l  N bits cos 2 f 2t

M  2l Rs
wN N
Rb  l  Rs
cos 2 f N t

OFDM transmission
Rs
Subcarrier symbol rate Rss 
N ...
N
Subcarrier symbol time Tss  NTs 
Rs
Frequency
28
 Generation of OFDM (1)

For orthogonality between two modulated subcarrier signals: s1 (t ), s2 (t )

Due to modulation
Tss Tss
1   2
 s (t )  s (t )dt  0   w
1 2 1 cos(2 f1t  1 )  w2 cos(2 f 2t   2 )dt
0 0

ww  sin(2 ( f1  f 2 )Tss  (1   2 ))  sin(1   2 ) 

 1 2  0
2  2 ( f1  f 2 ) 

With w1e j1 , w2 e j2 the complex amplitudes of the modulated signal states.

n
Orthogonality is obtained for: f1  f 2   n  Rss with n  , n  0
Tss
For minimum separation between subcarrier center frequencies: f  Rss .

29
 Generation of OFDM (3)
Orthogonal sub-carriers (in baseband) in frequency and time domain.
1 1

0.8
0.8
0.6

0.6 0.4

0.2
0.4
0

0.2 -0.2

-0.4
0
-0.6
-0.2
-0.8

-0.4 -1
-10 -8 -6 -4 -2 0 2 4 6 8 10 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time [Ts]
Frequency [1/T
Frequency [1/Ts] Time [TSS]
SS]

The sub-carrier frequencies are integer multiples of the

sub-carrier symbol rate:
1
f k  k  f with f   Rss
Tss

30
 Generation of OFDM (4)
w1e j1
j 2 x(t )
w2 e
cos 2 f c t
l  N bits
y (t )
l wN e j N
M 2 sin 2 f c t
g (t )  x(t )  jy (t )
A simple way to generate the OFDM baseband signal is by using the
Inverse Fast Fourier Transform (IFFT):
1 N N complex samples
j { k  2 ( k 1)( i 1)/ N } for i  1, .. , N samples
gi  
N k 1
wk e
for k  1, .. , N frequencies
 per OFDM symbol of
duration Tss

{gi } is the set of time samples of a symbol of the complex envelope g (t )

of an OFDM signal. By I&Q-mixing, the signal is brought to the desired
31
center frequency.
Generation of OFDM (5)

32
 Generation of OFDM (2)
Some important parameters of OFDM:

1. The symbol rate is equal for each subcarrier: Rss

1
2. The corresponding symbol period Tss  is therefore
also fixed per subcarrier. Rss

3. An OFDM symbol consists of the sum of all subcarrier

symbols transmitted during a symbol time Tss .
4. An OFDM symbol time is therefore equal to the subcarrier
symbol time Tss .
5. For equal M-level modulation of the subcarriers, the subcarrier
symbol rate is given by:
Rb Rs
Rss   l  2 log M
lN N
33
 Power Spectral Density

The PSD of the complex equivalent baseband signal is:

2
N 1
2 2 sin[ ( f  nf  )Tss ]
Pg ( f )  A | wn | Tss 
c
n 0  ( f  nf  )Tss
where | wn |2 is the mean power in wn and it is assumed that wn  0 .
34
 Spectral efficiency
OFDM signal Single carrier signal

2 Rs

fc Frequency

A single carrier signal with a symbol rate Rs requires a transmission

bandwidth BT ,00  2 Rs
An OFDM signal with N subcarriers, each with a symbol rate Rss  Rs / N
Rs
 BT  ( N  1) Rss  ( N  1)  Rs  Ds for N  1
N
Due to overlapping subcarrier spectra:  doubling of the spectral efficiency.
35
 OFDM receiver (1)
Rss 
Rs Tss 
w
N 1
0
L-level
R cos 2 f1t  Demodulator
Rss  s Tss w2
N
 Data
0 &
cos 2 f 2t
. De-
l  N bits
Rs
. 
Multiplexer
Rss 
N
Tss wN
0

cos 2 f N t  w n
wk k k

The OFDM receiver performs basically the reverse operations of the

transmitter. A direct analog implementation becomes very complex.

This has delayed practical application of OFDM till the end of the
1990s when fast digital signal processing components (especially
hardware implemented FFT/IFFT) became available.
36
 OFDM receiver (2)


w1
x(t )

w2
cos 2 f c t

l  N bits
y (t )

wN
sin 2 f ct
 w n
wk k k

The FFT operates on N complex time samples taken uniformly over Tss .
The output of the FFT are N complex amplitudes (symbol states)
belonging to each of the subcarriers.
Accurate synchronization in time and frequency is crucial. Errors result
in ISI and inter-carrier interference (ICI), respectively.

37
 Direct Sequence Spread Spectrum
Direct Sequence Spread Spectrum (DS-SS) modulation was originally
developed for military communication systems between 1950 - 1980
to ensure robust and covered communication.
In DS-SS modulation, the information signal is spread over a much wider
bandwidth than required based on the symbol rate.
A narrowband signal is made extremely wideband !

How?
Signal spreading over a much wider bandwidth is achieved by multiplying
the signal at the transmitter by a wideband code signal: spreading code.
Why?
- to create robustness against interference
- to allow multiple access: multiple users share the same frequency band;
- coveredness: the low PSD makes it difficult to detect and eavesdrop.

38
 DS-SS Transmitter (1)

In DS-SS, spreading of the signal over a much wider bandwidth is

achieved by multiplying the signal at the transmitter by a code
sequence c (t )  {1, 1} consisting of code chips.

m(t ) m(t )c(t )

Ac m(t )c(t ) cos c t

c(t )

Transmitted signal: s (t )  Ac m(t )c (t ) cos c t where m(t) represents

the information signal, e.g. for BPSK m(t) = ±1.
39
 DS-SS Transmitter (2)
Signals in a DS-SS transmitter for a BPSK modulated data signal.

Tb
Data modulation can be
m(t )
arbitrary, e.g. BPSK, DPSK,
Tc QPSK, MSK, QAM, etc.

c (t ) The code causes ‘random’

phase jumps of {0, } rad.

m(t )c(t )
Short code: repeats every
symbol time
Long code: extends over
m(t )c(t ) cos c t many symbol times

40
 DS-SS Transmitter (3)

DS-SS signal frequency spectra:

can easily drop below noise level.

m(t ) m(t )c (t )

2 Rs c(t ) 2 Rc

The PSD of the spread signal

2 Rc is 1/ G p less than the PSD of
the un-spread signal.
1 1
The chip rate Rc  >> data symbol rate Rs  .
Tc Ts
BWDS-SS 2R T
The ratio G p   c  s is called the Processing Gain or
BWData signal 2 Rs Tc spreading factor.
41
 DS-SS Reception (1)
Ac m(t )c(t )cos ct At the DS-SS receiver the reverse
operations are performed.
m(t )c(t ) m(t )
We can recover the signal at the
receiver only iff:
c '(t )
c(t )c '(t )m(t )  m(t ) 
- the code is known,
- perfect synchronization.

m(t )c(t ) m(t )

2 Rc c '(t ) 2 Rs

2 Rc
42
 DS-SS Reception (2)
DS-SS is very robust against interference. When de-spreading the
desired signal, the interference is spread.

Only a fraction 1/ G p of the spread interference overlaps with the

de-spread data signal spectrum:
BWData 2 Rs PI
Effective interference power: I eff  PI  PI 
BWSpread  Interference 2 Rc G p

Processing
gain

c(t )m(t )  i (t ) m(t )  c '(t )i (t )

c '(t )
PI
N I  f  fc   [W/Hz]
2 Rc
43
 DS-SS Reception (3)
DS-SS is also robust against interference of other DS-SS signals: when
the SS-code or code phase is different, the signal remains spread.

Also in this situation, only a fraction 1/ G p of the interfering signal

spectrum overlaps with the desired de-spread signal spectrum:
c1 (t ), c2 (t )
BWData 2 Rs PI synchronized
Effective interference power: I eff  PI  PI 
BWDS SS 2 Rc G p

Processing
gain

c1 (t )m1 (t )  c2 (t )m2 (t ) m1 (t )  c1' (t )c2 (t )m2 (t )

c1' (t )

44
 Pseudo-Random Noise code generators (1)
A DS-SS code is a "pseudo random" sequence of {-1,1}. There are
several ways to generate such codes depending on the required
properties. One way is to use Maximum Length sequences which are
also called Pseudo-Random Noise (PRN) codes because of their noise
like spectral properties.

Maximum Length (ML) sequences can be digitally generated in a simple

way using a shift register with feedback.

s4 s3 s2 s1 s0

A ML-code generator using a 5-stage shift register with code length 31.

45
 Pseudo-Noise code generators (2)
P
A shift register length of P stages results in an ML-code of 2  1 chips.
Tcode

The autocorrelation of an ML-code is given by: Rcc ( )   c(t )c(t   )dt

0
The ML-code has very good auto-correlation features. Good auto-
correlation properties are important for reliable synchronization and
detection.
Auto-correlation of a 7 chip
2P  1 (P = 3) ML-code:
- width autocorrelation peak
is 2Tc
- outside correlation peak very
low constant correlation value
equal to -1
P
- maximum of 2  1

1
46
Pseudo-Noise code generators (3)

47
 Spectral efficiency of DS-SS

For DS-SS, with M-level modulation, the spectral efficiency is

given by:
Rb lRs lRs l  bit / s  where 2
 = = =  Hz  l  log M
BT BT 2 Rc 2G p

The price we pay for increased robustness against interference

and for coveredness (large G p) is a very low spectral efficiency.

This can be improved by letting multiple users share the same

bandwidth.

48
 Code Division Multiple Access (CDMA)
In Code Division Multiple Access (CDMA) the spectral resources
(bandwidth) are shared among multiple users using different DS-SS
codes.

Each user is identified by its own code. The cross-correlation between

the codes has to be low (ideally zero  orthogonal codes).

5 5

 c j (t ) m j (t ) m1 (t )   c1' (t )c j (t )m j (t )
j 1 c1' (t ) j 2

49
 Pseudo-Random Noise code generators (3)
The cross-correlation of two codes, c1 (t ) and c2 (t ) , of the same
length, is given by: Tcode

Rc1c2 ( )   c1 (t )c2 (t   )dt

0
The cross-correlation properties of a set of codes is very important
for the separation of users using different codes in a
Code Division Multiple Access (CDMA) system, e.g. UMTS.

Cross-correlation
1

1 0 Tcode

Codes with poor cross-correlation properties result in bad separation or

a high level of interference for certain delays between the codes.
50
 Direct Sequence Spread Spectrum – Review
In DS-SS modulation the bandwidth of the signal is increased by
multiplying the data with a code known to the receiver.

Data Data Data

Data modulator
demodulator

DS-SS code Carrier Carrier DS-SS code

generator generator generator generator

Acquisition
Carrier
&
recovery
tracking

Tc Data symbol

Tbs Tsb Code chip

Code sequence
51