Special Issue on 5G wireless networks

5G Wireless networks: Performance Analysis of Filter Bank

Multicarrier System with Nonlinear High Power Amplifiers
G.Uma Maheswari, G.Ananthi, S.J.Thiruvengadam
TIFAC CORE in Wireless Technologies
Department of Electronics and Communication Engineering
Thiagarajar College of Engineering, Madurai 625015, Tamil Nadu, India
E-mail:. umaty91@gmail com,gananthi@tce.edu
Abstract: In this paper, the impact of nonlinear distortion induced by High Power
Amplifiers (HPA) for medium and high power signals is analyzed for Filter Bank
Multi-Carrier (FBMC) systems using Offset Quadrature Amplitude Modulation
(OQAM). Memory-less HPA nonlinear distortion models such as Travelling Wave
Tube Amplifier (TWTA), Solid State Power Amplifier (SSPA) and Soft Envelope
Limiter (SEL) are normally used. A closed form expression for Bit Error Rate (BER)
expression is derived and analyzed for SSPA and SEL models for FBMC/OQAM
system with nonlinear HPA in frequency selective Rayleigh channel by varying
constellation size in OQAM modulation and Input Back-Off. The performance is
compared for the models with 64 sub channels and input back-off for 6 and 8
decibels. In lieu of validating the obtained simulation results, theoretical results have
also been compared.
Index Terms Filter bank multicarrier, nonlinearity, high power amplifier, bit error
rate, offset QAM.

1. Introduction
Filter Bank Multi-Carrier (FBMC) is a multicarrier modulation where the
subcarriers are passed through filters that suppress signals side lobes, provides better
spectral localization. FBMC modulation is considered in Fifth Generation (5G)
cellular network standards for its robustness due to network asynchronicity and
frequency misalignment between users. FBMC is usually either coupled with QAM or
with offset-QAM (OQAM) modulation techniques.
High Power Amplifiers (HPA) used in FBMC system for radio frequency upconversion, may enter the nonlinear region for medium and high power signals. It
provides high power efficiency and increased battery life for power amplifiers and it
produces peak clipping and intermodulation distortion causing Adjacent Channel
Interference (ACI) and power losses leading to increased error rates.
Intuitively, Chang and Saltzberg introduced FBMC modulation for parallel
data transmission [1][2]. A nonlinearity for Gaussian data is analyzed [3]. Cyclic
Prefix (CP)-based Orthogonal Frequency Division Multiplexing (OFDM) modulation
offers better robustness to multipath channel effects is used widely in current
technologies. However, the use of CP and the high side lobes of the rectangular pulse
shape induce a loss of the spectral efficiency.
OFDM signals may exhibit large peak-to-average-power ratio values and
frequency synchronization among subcarriers [4][5]. A theoretical analysis have been
carried out for the analysis of an OFDM system using a memory-less HPA [6]. Effect
of nonlinearity on multiple input multiple output system is observed [7]. The FBMC
system with TWTA Saleh model of HPA is analyzed [9] in the presence of nonlinear
phase distortion is performed.

The objective of this paper is to analyse the effect of nonlinear distortion

induced by memory-less HPA models such as TWTA, SSPA and SEL on bit error rate
performance of FBMC systems with OQAM modulation. This paper is organized as
follows. Section II presents the system model of FBMC-OQAM system and HPA
models. Section III evaluates analytical expression for FBMC system with nonlinear
HPA. Simulation results are discussed in section IV and section V concludes the
paper.

2. System Model
Consider an FBMC system with OQAM modulation using TWTA [8], SSPA and
SEL models with 6dB Input Back-Off (IBO) and knee voltage 1 in frequency
selective Rayleigh fading channel as shown in figure 1. The polar coordinates of the
input signal is

s re j

s f A r e j f P r , where

.Symbol at the output of the non-linear HPA is

r , amplitude

and phase of input signal. f A r

AM/AM Conversion of the HPA is a measure of how closely the input and output
transfer characteristic of an amplifier matches with a straight line. f P r AM/PM
Conversion of the HPA is a measure of the time delay or shift in phase angle
2.1 TWTA SALEH MODEL
The TWTA amplifies signal uses an electron beam and a slow wave structure.
TWTAs has the advantage of large currents and provides high output powers.AM/AM
and AM/PM conversion of TWTA characterized by Saleh model [9] is
f A r Ais2
r2
f P r 0
r 2 Ais2

(1)

r Ais2

(2)

where Ais represents input saturation voltage, r is the amplitude of signal at the
input of HPA.
2.2 SSPA MODEL
SSPA is a microwave frequency amplifier uses field-effect transistors for
amplification. It is a high-power amplifier using solid-state electronics that possess
reduced size. The AM/AM and AM/PM conversions of the SSPA model can be
expressed as
r

f Ar

1 r

Aos

1
2 2

(3)

f P (r ) 0

(4)

where Aos represents output saturation voltage and is the knee factor sharpness
of transition from linear region to saturation region.
2.3 SEL MODEL
SEL model is used for modeling a HPA with a perfect predistortion system.
The global transfer function of the predistortion followed by the HPA is a limiter.
When knee value, the SSPA model converges towards the SEL. The SEL used to
model the HPA with ideal predistortion can be described by the following AM/AM
and AM/PM functions:

r ,r Ais
f A r
Ais ,r Ais
f P (r ) 0

(5)
(6)

2.4 FBMC

Transmitted signal at
i t

instant is defined by

t th

M 1

a m, n m, n t

(7)

m 0 n

where, M is the number of subcarriers, a m,n is the real valued 4-OQAM symbols on
m th

subcarrier and

symbol frame, m, n is the time, frequency shifted prototype

n th

filter impulse response. The prototype filter for FBMC system is designed as [10],
2
nT
D

m, n t g t
m t exp m,n
exp j
2
2

where, g(t) is

the

prototype

i 0,1, , KM 1

filter

(8)

impulse

response

which

is

null

for

T is the symbol period D KM 1 is the delay term

which depends on length of prototype filter, K is the overlapping factor

m, n

m n mn is the initial phase term. In a distortion-less, noise-free

2

channel, the demodulated signal at subcarrier m 0 and time instant n 0 is given

by,
y m0 , n0

i t , m 0 , n 0 (t )

i (t ) m

, n0

(9)

(t ) dt

Substituting Equation (7), the demodulated signal becomes

y m0 , n0 am0 , n0

m m0 n n 0

a m, n

*
m, n t m , n t dt
0

y m0 , n 0 a m0 , n0 ju m 0 , n 0

(10)

(11)

,where u m0 ,n 0 is the pure imaginary intrinsic interference. Perfect Reconstruction is

possible only when

m, n t m

, n0

t dt

m , m 0 n, n 0

According to Bussgangs

theorem [11], the nonlinearly amplified signal conditioned to the transmission of a

fixed symbol can be modeled as
s (t ) K 0i t d t

(12)

K0 is a constant complex gain given by

1 S S
E

K0

(13)

d(t) is an additive zero mean noise uncorrelated with i(t), having variance

d2 E d t 2 E | S |2 | K 0 |2 E 2

(14)

Received signal after nonlinear amplification is

ym0 , n0 K 0 H m0 , n0 am0 , n0 jum0 , n0 H m0 , n0 d m0 , n0 wm0 , n0

Where, H m

,n 0

(15)

is the channel coefficient corresponding to the subcarrier m 0 and

instant n 0 . The channel is considered to be constant over neighboring symbols

contributing the ISI. Zero-forcing equalized symbol when H m

,n 0

and K 0 are

perfectly known

d
w
y
a

K0 H
K0
K0H

(16)

3. FBMC system with HPAs

An instantaneous SNR for the FBMC/OQAM system using Equation (15),

| K 0 | 2 | H | 2 x2
| H | 2 d2 w2

(17)

PDF of the instantaneous SNR is given by

2
w c
e

2
p ( )
c d2
0

w2
2
c d

, if 0 c

(18)

d2

, otherwise

BER of M-QAM in AWGN channel is given by

c
d2

BER a

erfc b

2
w
c

2
c d2

w2
c d2

(19)

where a and b are modulation specific constants given by

a

2 M 1

M log 2 M

3 log 2 M
M 1

(20)

Channel noise variance is defined by

w2

N0
1
Pout
Eb log 2 M

(21)

The signal mean power at the output of the HPA is defined by

Pout Eb

2 log 2 M
T

(22)

A closed form expression for theoretical bit error rate is defined as

c
d2

BER a 1 erf

2
w
c

c d2

w2
c d2

(23)

BER a 1 I

where I

c
d2

erf

(24)

w2 c

2
c d2

w2
c d2

(25)

The error function is defined as

erf u

u 2

du

n 0

1 n u 2n 1
n!(2n 1)

(26)

Substituting for error function and using Taylor series expansion,

w2
c d2

1 k

w2

k!
k 0

(27)

k
c d2

By using table of integrals,

u

x 1

1 x v

dx

2 F1 v, ;1

(28)

, which is valid for [| arg 1 | , ( ) 0] where 2 F1 .,.;.;. denotes the

Hypergeometric function.
Hypergeometric function is defined for |z|<1 by the power series,
2 F1 q, b; c; z

(q ) n (b) n z n
(c ) n
n!
n 0

(29)

,n 0
1
where Pochhammer Symbol (q)
n
q(q 1) (q n 1) , n 0
w2
1
2

n 0k 0 k!n! 2n 1

k n

k 1

2
d

nk

3
2

b c n 2
nk

3
2

3
5
k 2, n k ; n k ;1
2
2

2 F1

(30)
Closed form BER expression obtained by substituting Equation (31) into Equation
(24),

k 1 nk

3
2

1 kn 2 1

k!n! 2n 1 2

2 d

BER a 1

n0k0 b n

3
5

2
c

3 2F1 k 2, n k 2 ; n k 2 ;1

(31)
Substituting Equation (21), into BER expression, the closed form expression is

k 1

N
1
3

k n 0 Pout nk

1 Eb log2 M 1 2

k!n! 2n 1 2

2
BER a 1

n0k 0 1

b c n 2

3 5

F
k

2
,
n

; n k ;1
2 1

2
2

nk

(32)
3.1 TWTA MODEL
Nonlinear characteristics of TWTA amplifier is as given in Equation (1) and Equation
(2),

Soft envelope S ( ) Fa ( )e jF ( )
P

S ( ) Ais2

j 0 2 A 2
is
e

2 A2
is

(33)
1 S ( ) S ( )
E

Complex constant gain as given by eq. (13), K 0

S ( ) Ais2 Ais4 4 j0 2 2 Ais4

2

exp j0
3
2 A2

is

Ais2 2

1 Ais2 Ais4 4 j0 2 2 Ais4

2
K0 E
exp j0
3
2 A2
2

is

Ais2 2

(34)

2 A2
is

Ais2

2
2
Ais

j 0

(35)

Variance of distortion noise is given by eq. (14), d2 E | S ( ) | 2 | K 0 | 2 E 2

| S ( ) |2

Ais4 2

2 Ais2 2

(36)

Substituting Equation (37) into Equation (14),

d2 E

Ais4 2

| K0 |
2 A2 2
is

2 2 x2 e

2
2 x2

(37)

3.2 SEL MODEL

Nonlinear characteristics of SEL amplifier is as given in Equation (5) and Equation
(6), Soft envelope S ( ) Fa ( )e jF ( ) Complex constant gain given by eq. (13),
P

K0

1 S ( ) S ( )
E

10

S() 1 for Ais

0 for Ais

(38)

1 for Ais
S ( )
Ais
for Ais

(39)

Substituting Equation (38) and Equation (39) into Equation (13),

for Ais

(1)

K 0 Ais 1
2 for Ais

(40)

Rayleigh distributed PDF of is f ( )

1
1
1
f ( )d
e

x2

2
2 x2

x2

2
2 x2

, 0

(41)

From table of integrals,

j
erf
2

2 x

K0 1 e

Ais2

x2

(42)

2 x2

A2
1
is erfc
2

x2

Ais2
x2

(43)

Variance of HPA noise as in eq. (14), d2 E d (t ) 2 E | S ( ) | 2 | K 0 | 2 E 2

11

 2 , A
is
| S |2
Ais2 , Ais
( 2 )

x2

(44)
2
2 x2

(45)

From table of integrals,

e ax
dx
n

m ax n

x e

xn
1

a
a 2

for

m 1
2
n

(46)

Using Equation (48),

( 2 ) 2 2 x2 e

2
2 x2

(47)

The variance of the nonlinear distortion noise is

d2

Ais2

x2

1 e

2
x

K 02

(48)

3.3 SSPA MODEL

Nonlinear characteristics of SEL amplifier is as given in Equation (5) and Equation
(6), Soft envelope S ( ) Fa ( )e jF ( ) Complex constant gain given by eq. (13),
P

K0

1 S ( ) S ( )
E

Let, Aos 1, 1 S ( )

2

1 2

S ( )

S ( )

(49)

1 2

1 2 2
3

(50)

12

1 2 2

E
3
3
2
1 2
2 2
2 2
1

1
Substituting, K 0 2 E

(51)

Let x 2 , n 2 ,

Chi-square distributed pdf,

n
x
1
2
x 2 e 2
p x ( x)
n
n 2 n
2

(52)

K0

x
2 x2

1
x2 e
dx

3
2
2
2

x
1 x 2

(53)

From table of integrals,

k
e ax m
k p ( x)

a k 0
ak

ax
Pm ( x)e dx

(54)

Pm (x) is a polynomial of x in degree m

p k ( x)

K0

is the k th derivative of Pm (x) with respect to. x

x
2 x2

4 x2 212

p k ( x)

( 1) k

k 0

1
2 2
x

(55)

Using binomial expansion,

1 x n 1 nx n n 1 x 2 n n 1 n 2 x 3
2!

(56)

3!

This equation is valid for any n positive or negative, integer or fraction provided that
-1<x<1

3
1 x 2 1 3 x 2 2 x 2 2 2 2 x 3

2!

3!

for 1 x 1

(57)
Neglecting higher powers of x,

13

K0

x
2 x2

(1) 1 3 x 3 x 2

2 2

x 2 1 3 x
2
1

2 2
x

3 3
2 2

1 2

2 x2

(58)
Simplifying,

K0

x
2 x2

3 2

x 2 x 2 2 x2 2 3 x 12 x4

(59)

Variance of the NLD noise as in eq. (14), d2 E | S ( ) | 2 | K 0 | 2 E 2

2
| K 0 |2 2
1 2

d2

(60)

| K 0 |2 e
1 2

Using Equation (49), d2

2
2 x2

2 e
2

d 2
1 2
1 2 2 2

1 x

x 1 x

1 e

x
2 x2

2 x2

2
2 x2

2 2 x2

(61)

(62)

(63)

dx

Approximating binomial expansion,

x
1

1 x
2 x2

x1 x e

x
2 x2

(64)

dx

Using table of integrals for polynomial coefficient as given in Equation (56),

x

1 x

x
2 x2

2 x2 21 2

p k ( x)

k 0

ak

1 k

(65)

14

1 x x
x
1 1
2 x2

x 1 x
e

2

1 x
1
1

2 2
2 x
x

x
2 x2

x1 x 2 x2 1 2 x 8 x4
e
1 x

Substituting,

x
2 x2

e
| K 0 | 2
2

(66)

(67)

3 2

x 2 x 2 2 x2 2 3 x 12 x4

2

(68)
Variance of the nonlinear distortion noise is given by,
d2 e

x
2 x2

x1 x 2

2
4
x 1 2 x 8 x

| K

|2 e

2
2 x2

2 2 x2

(69)

4. Results and Discussion

In this section, The Bit Error Rate (BER) performance of FBMC/OQAM
system with nonlinear HPA in frequency selective Rayleigh channel is investigated by
varying constellation size M = 4,16,64 in OQAM modulation, number of sub-channels
N = 64,128,256 and

Input Back-Off IBO = 6,8 dB for frequency selective Rayleigh

channel through Monte Carlo simulations. This analysis is also compared with the
analytical results. The parameters considered are Knee factor for SSPA is kept as
1 , output saturation voltage Aos 1 .

64 subcarriers is considered for OFDM and FBMC simulations are shown in

Figure 2. This figure shows that the side lobe amplitude for FBMC system is
considerably lower than that of OFDM system. Figures 3, 4 and 5 show that BER is
plotted as a function of various values of SNR for TWTA, SSPA and SEL models. It is
shown that for all the models, by increasing the number of symbols in OQAM
modulation, the bit error rate is increased. It is also shown that the bit error rate is

15

minimum for SEL model when compared with the other models is shown in figure 6.
It is due to the fact that the SEL Model provides better linear amplitude
characteristics. Figure 7 shows that if the number of subcarriers increases, bit error
rate is also increases due to reduced power per bandwidth and thereby causing
estimation errors.

Conclusion
Analytical BER expression is derived and analysed for SSPA and SEL models
in FBMC systems using OQAM. Theoretical and simulated BER performance of
FBMC/OQAM system with nonlinear HPA in frequency selective Rayleigh channel is
simulated by varying constellation size in OQAM modulation, number of subchannels and Input Back-Off. BER performance is observed to be superior for SEL
model, as it has more linear amplitude characteristics than other models. Increasing
knee factor value in SSPA reduces the nonlinear distortion and thereby decreases
BER. Side lobe amplitude for FBMC system is considerably lower than that of
OFDM system. With increasing number of symbols in OQAM modulation, BER
increases due to reduced distance between constellation points. As the number of
subcarriers increases, BER increases due to reduced power per bandwidth, causing
estimation errors.

References
[1] Robert W. Chang.: Synthesis of Band-Limited Orthogonal Signals for
Multichannel Data Transmission, The Bell System Technical Journal, December
1966, pp. 1775-1796.

16

[2] Burton R. Saltzberg.:Performance of an Efficient Parallel Data Transmission

System, IEEE Transactions on Communication Technoloqy, December 1967, pp.
805-811.
[3] E. Rowe, Memoryless nonlinearities with Gaussian inputs: Elementary results,
Bell Syst. Tech. J., vol. 61, no. 7, pp. 15191525, Sep. 1982.
[4] Farhang-Boroujeny B.,OFDM versus Filter Bank Multicarrier, IEEE Signal
Processing magazine, vol.28, pp.92-112, May 2011.
[5] Banelli, Buzzi, Modulation Formats and Waveforms for 5G Networks: Who Will
Be the Heir of OFDM?, IEEE Signal Processing Magazine, pp. 80-93, Nov 2014.
[6] D. Dardari and V. T. A. Vaccari, A theoretical characterization of nonlinear
distortion effects in OFDM systems, IEEE Transaction on Communication, vol.
48, no. 10, pp. 17551764, Oct. 2000.
[7] J. Qi and S. Assa, On the effect of power amplifier nonlinearity on MIMO
transmit diversity systems, in Proc. IEEE ICC, Dresden, Germany, Jun. 2009, pp.
15.
[8] Bouhadda, Shaiek, Roviras, Zayani, Medjahdi and Bouallegue, Sensitivity
analysis of FBMC signals to Non Linear phase distortion, ICC'14 - W8:
Workshop on 5G Technologies, 2014.
[9] A. M. Saleh, Frequency-independent and frequency-dependent nonlinear models
of TWT amplifiers, IEEE Transactions on Communications, vol. 29, pp. 17151720, 1981.
[10] M. Bellanger, Specification and design of a prototype filter for filter bank based
multicarrier transmission, in Proc. IEEE Int. Conf. Acoustics, Speech, and
Signal Processing, Salt Lake City, USA, pp. 2417-2420, May 2001.
[11] JJ Bussgang, Crosscorrelation Functions of Amplitude-distorted Gaussian
Signals (1952). Research Laboratory of Electronics, Massachusetts Institute of
Technology, Cambridge. Technical Report 216 (1952)
[12] Leon-Garcia,Probability and Random Processes for Electrical Engineering,
Reading, Mass: Addison-Wesley, 1994.

17

Figure 1 FBMC system with OQAM Modulation

OFDM and FBMC symbol

-20

Amplitude in dB

-40

-60

-80

-100

-120

-140
-1

-0.8

-0.6

-0.4

-0.2

0
freq*pi radians

0.2

0.4

0.6

0.8

Figure 2. Subcarriers for OFDM and FBMC for N=64

18

BER of FBMC system with TWTA

10

M=4 IBO=6 dB phi=pi/3 Th

M=16 IBO=8 dB phi=pi/6 Th
M=64 IBO=8 dB phi=pi/6 Th
M=4 IBO=6 dB phi=pi/3 Pr
M=16 IBO=8 dB phi=pi/6 Pr
M=64 IBO=8 dB phi=pi/6 Pr

-1

10

-2

10

-3

BER

10

-4

10

-5

10

-6

10

-7

10

-8

10

Eb/N0 in [dB]

10

Figure 3. BER for TWTA model in FBMC with N=64

BER of FBMC system with SSPA having beta=1

10

M=4 Theoretical
M=16 Theoretical
M=64 Theoretical
M=4 Practical
M=16 Practical
M=64 Practical

-1

10

-2

10

-3

BER

10

-4

10

-5

10

-6

10

-7

10

-8

10

Eb/N0 in [dB]

10

19

Figure 4. BER for SSPA model in FBMC with N=64, beta=1

BER of FBMC system with SEL

10

-1

10

M=4,IBO=6dB Theoretical
M=16,IBO=8dB Theoretical
M=64,IBO=8dB Theoretical
M=4,IBO=6dB Practical
M=16,IBO=8dB Practical
M=64,IBO=8dB Practical

-2

10

-3

10

-4

BER

10

-5

10

-6

10

-7

10

-8

10

-9

10

Eb/N0 in [dB]

10

Figure 5. BER for SEL model in FBMC with N=64

BER of FBMC system

10

TWTA Th.
SSPA Th.
SEL Th.
TWTA Pr.
SSPA Pr.
SEL Pr.

-1

10

-2

10

-3

10

-4

BER

10

-5

10

-6

10

-7

10

-8

10

-9

10

Eb/N0 in [dB]

10

Figure 6. BER for TWTA, SSPA, SEL models in FBMC with N=64

20

BER of FBMC system with TWTA M=4

-3

10

N=64
N=128
N=256

-4

10

-5

10

-6

BER

10

-7

10

-8

10

-9

10

-10

10

Eb/N0 in [dB]

10

Figure 7. BER for TWTA Model for IBO=6 dB for N=64,128,256

21