WSA 2015 • March 3-5, 2015, Ilmenau, Germany

Comparison and evaluation between

FBMC and OFDM systems
Qinwei He, Anke Schmeink
UMIC Research Centre
RWTH Aachen University
52056 Aachen, Germany
Email: {he, schmeink}@umic.rwth-aachen.de

Abstract—This paper studies the analytical bit error probabil- OFDM: R

A e
ity (BEP) of filter bank based multicarrier (FBMC) transmission
d m
and conventional cyclic prefix (CP) based orthogonal frequency P Channel S o
division multiplexing (OFDM) systems. We first compare and d
Input S / / v P Output
evaluate these two schemes under the additive white Gaussian / IFFT S P e FFT /
noise (AWGN) channel, and then extend the results to the P C S
Rayleigh channel. Closed form expressions of the bit error P C
probabilities for both two methods are derived and validated in P
this research. The results reveal that the performance of these two
methods are the same when the perfect reconstruction conditions Tramsmitter Receiver
of them are satisfied. However, the FBMC technique has less out- Synthesis Filter Bank Analysis Filter Bank
of-band power leakage as a result of lower side lobes. Meanwhile, FBMC:
the omission of CP improves the bandwidth efficiency of the
system with a increase in the equalization complexity.
Input S P Channel S P Output
I. I NTRODUCTION /
P
IFFT PPN /
S
/
P
PPN FFT /
S
Orthogonal frequency division multiplexing (OFDM) has
been probably the most successful multi-carrier modulation
(MCM) scheme for the wired or wireless communication in Fig. 1. OFDM and FBMC systems [3]
the past two decades and it is employed in the standard of
today’s 4G network. Meanwhile, OFDM still plays an impor- tions. However, the analytical bit error probability (BEP) ex-
tant role in many current applications such as asymmetric pressions of them have not been derived and evaluated yet. In
digital subscriber line (ADSL), digital video broadcasting this contribution, we first derive the analytical BEP expressions
(DVB) and wireless local area networks (WLANs). However, for both OFDM and FBMC under the additive white Gaussian
with the increasing number of users and devices, the need noise (AWGN) channel. Then, by combining these analytical
for energy efficiency and sophisticated spectrum utilization expressions with the characteristic of the Rayleigh channel,
at low costs drawbacks of OFDM become apparent such as the close-form BEP expressions under Rayleigh Channel are
the cyclic prefix (CP) leading to a decrease in the system also able to be obtained.
efficiency. Consequently, other MCM methods have been The rest of this paper is organized as follows. In Section
proposed and evaluated recently and latest research indicates II, we introduce the basic knowledge of OFDM and FBMC
a great potential for these methods which may have a big briefly. The system models under the AWGN channel are
influence on next generation communication systems and may introduced and analysed in Section III. Also, the BEP expres-
replace the conventional OFDM [1], [2]. sion of these two techniques are obtained in this section too.
One important, new emerging MCM method is the filter In Section IV, we incorporate the Rayleigh channel into the
bank based multicarrier (FBMC) transmission, which intro- models and the analytical BEP expressions of this scenario
duces filter banks to the OFDM system and discards the CP. In are derived. The simulation results and the evaluation based
this sense, the FBMC is an evolution of OFDM. The employed on these are provided in Section V. Finally, we conclude our
filter bank will import extract flexibility to the system to results and give some potential future work in Section VI.
cope with some drawbacks of OFDM. Intuitively, the filter
bank can be designed with different properties to satisfy the II. I NTRODUCTION TO OFDM AND FBMC
communication requirements. For example, for the OFDM In this section, the basic concepts for these two multicarrier
system, CP is used to eliminate inter symbol interference (ISI). systems are introduced. The basic structure of them are shown
In order to remove this interference completely, the length of in Figure 1.
CP must be no shorter than the length of the impulse response
of the corresponding channel. However, this CP-redundancy A. OFDM
decreases both spectral and power efficiency. For the FBMC, In the transmitter of the OFDM system, a bit stream
instead of CP, filter banks can be designed and applied to is usually mapped to a quadrature amplitude modulation
reduce the out-of-band power leakage and increase the spectral (QAM) symbol, called X, as shown in Figure 2. And then
efficiency with a cost of computational complexity. this constellation symbol stream passes through a serial-to-
In the past research activities, most comparisons of bit error parallel converter, form a set of M parallel QAM symbols
rate (BER) between these two systems are based on simula- X[0], X[1], ..., X[M −1] representing the symbols transmitted

ISBN 978-3-8007-3662-1 1 © VDE VERLAG GMBH · Berlin · Offenbach, Germany

 WSA 2015 • March 3-5, 2015, Ilmenau, Germany

on each subcarriers. By modulating these symbols with the Based on these provided coefficients, the frequency response
inverse fast Fourier transform (IFFT) at the transmitting side, of the filter with M subcarriers is obtained through the
a discrete baseband OFDM symbol can be generated, which equation
can be written as    
k
1 
M −1 K−1
 sin π f − M K
x[n] = F −1 {X[i]} = √
2π
X[i]ej M ni , 0 ≤ n ≤ M −1, MK
P (f ) = Pk    . (4)
M i=0 k
(1) k=−(K−1) M K sin π f −
MK
where, M is the number of subcarriers, X[i] is a complex
QAM symbol transmitted on the ith subcarrier. As shown in Figure 4(a), a highly frequency selective filter
Obviously, the QAM symbol can be obtained at the receiver is acquired with almost no out of band leakage. Therefore,
side utilizing Fast Fourier transform (FFT). In the perfect
reconstruction situation, the recovered signal is
M −1
1  2π
X̂[i] = F{x[n]} = √ x[n]e−j M ni , 0 ≤ i ≤ M − 1.
M n=0
(2)
QAM OQAM
T T
2 2

n−2 n−2

n−1 n−1
(a) Prototype filter frequency response (b) Prototype filter impulse response
Fig. 4. Prototype filter [4]
n n
the impulse response p[m] of the prototype filter is given by
n+1 n+1 applying the IFFT to the frequency response, which is
K−1
  
n+2 n+2 k 2πk
p[m] = 1 + 2 (−1) Pk cos m
MK (5)
Real part of QAM symbol Imaginary part of QAM symbol k=1

Fig. 2. OFDM and FBMC symbol mapping on carriers

p[0] = 0,
where m = 1, . . . , Lp − 1, here Lp = M K is the length of
B. FBMC/OQAM the filter p. The p[0] applied here is to make the number of
There are two kinds of implementations of FBMC, the coefficients an odd number, thus, the delay of the filter can be
frequency spreading filter bank multicarrier (FS-FBMC) and adjusted to an integer multiple of the sample periods [5].
the poly-phase network filter bank multicarrier (PPN-FBMC).
The latter is selected as the model here as it reduces the
high complexity which is introduced by the extra filtering
operations at the transmitter and receiver. The simple PPN
for the transmitter is shown in Figure 3.
d0(mM ) H0

d1(mM ) I H1 Z −1

F
x(n)
F
T
Fig. 5. Filter bank generated from prototype filter [6]
dM −1(mM ) HM −1 Z −(M −1) With this prototype filter, a filter bank of the system is ob-
tained with frequency shifts, as shown in Figure 5. Obviously,
Fig. 3. PPN-IFFT FBMC transmitter [4] the subcarriers with odd (or even) index are not overlapped.
In this work, the PPN filter applied is from PHYDYAS [4] Only the neighbour subcarriers have influences in a certain
with the overlapping factor K = 4. The frequency domain subcarrier. On account of this feature, one can exploit only odd
coefficients of it are or even subcarriers for communication with QAM symbols.
√
2 However, by utilizing only alternate subcarriers, the bit rate
P0 = 1; P±1 = 0.97196; P±2 = 2 ; P±3 = 0.235147.
comes down by half as half of the capacity is left unused.
And they satisfy the equation To gain the full capacity, orthogonality is necessary for the
K−1 neighbouring sub-channels. This orthogonality is satisfied with
1 
|Pk |2 = 1. (3) offset quadrature amplitude modulation (OQAM). The OQAM
K
k=−K+1

ISBN 978-3-8007-3662-1 2 © VDE VERLAG GMBH · Berlin · Offenbach, Germany

 WSA 2015 • March 3-5, 2015, Ilmenau, Germany

p[0] p[1] p[2] p[3] p[4] p[5] p[6] p[7] p[8] p[9] p[10] p[11] p[12] p[13] p[14] p[15]
y1 [n] = p[n]x1 [n%M ]
x1 [0] x1 [1] x1 [2] x1 [3] x1 [0] x1 [1] x1 [2] x1 [3] x1 [0] x1 [1] x1 [2] x1 [3] x1 [0] x1 [1] x1 [2] x1 [3]

p[0] p[1] p[2] p[3] p[4] p[5] p[6] p[7] p[8] p[9] p[10] p[11] p[12] p[13] p[14] p[15]
y2 [n] = p[n]x2 [n%M ]
x2 [0] x2 [1] x2 [2] x2 [3] x2 [0] x2 [1] x2 [2] x2 [3] x2 [0] x2 [1] x2 [2] x2 [3] x2 [0] x2 [1] x2 [2] x2 [3]

p[0] p[1] p[2] p[3] p[4] p[5] p[6] p[7] p[8] p[9] p[10] p[11] p[12] p[13] p[14] p[15]

x3 [0] x3 [1] x3 [2] x3 [3] x3 [0] x3 [1] x3 [2] x3 [3] x3 [0] x3 [1] x3 [2] x3 [3] x3 [0] x3 [1] x3 [2] x3 [3]

p[0] p[1] p[2] p[3] p[4] p[5] p[6] p[7] p[8] p[9] p[10] p[11] p[12] p[13] p[14] p[15]

x4 [0] x4 [1] x4 [2] x4 [3] x4 [0] x4 [1] x4 [2] x4 [3] x4 [0] x4 [1] x4 [2] x4 [3] x4 [0] x4 [1] x4 [2] x4 [3]

y[0] y[1] y[2] y[3] y[4] y[5] y[6] y[7] y[8] y[9] y[10] y[11] y[12] y[13] y[14] y[15] y[16] y[17] y[18] y[19] y[20] y[21]

Fig. 6. Filter bank generated from prototype filter

staggered the in-phase and quadrature components by half a 1) Divide the signal into K frames and apply the prototype
symbol period. Figure 2 implies the difference between QAM filter to each frame. It is easy to see that for ρth frame we
and OQAM symbols intuitively. In each sub-channel either the have rρ [n] = r[n + ρ−1
2 M ]. Therefore, by multiplying the ρth
real or the imaginary part is transmitted avoiding interference frame with the impulse response of the prototype filter, we
between neighbouring sub-channels. Finally in order to keep have
up the bit rate, the FFT is run at twice the rate as in OFDM. sρ [n] = p[n]rρ [n]. (9)
Now, we consider the signal at the transmitter side of
the FBMC. As show in Figure 1, after the serial-to-parallel 2) Resize the signal from the last step and sum up. Recall
conversion, the OQAM symbols are passed into the IFFT that the symbol after IFFT is duplicated K times at the
block. Be denoting Xρ [i] as the symbol on the ith carrier of transmitter side. Therefore, these K symbols must be reformed
ρth frame, we obtain the IFFT block output xρ [n], and summed up back to one symbol again. We use ŝρ [n] to
M −1 denote this symbol, then
1 
xρ [n] = F −1 {Xρ [i]} = √ Xρ [i]ej2πni/M . (6) K−1
M i=0 
ŝρ [n] = sρ [n + kM ] n = 0, 1, ..., M − 1. (10)
These symbols will be treated with the prototype filter before k=0
transmitting. Figure 6 shows how this symbol is constructed,
in which both the overlapping factor and the carrier size are At this point, we can pass the symbol ŝρ [n] to the FFT
equal to 4. Obviously, the procedure to form such a symbol block frame by frame to recreate the OQAM symbol, which
can be accomplished by two steps: is
1) Treat each frame with the prototype filter. As we can M −1
see from Figure 6, the output of IFFT xρ [n] is duplicated K 1 
X̂ρ [i] = F{ŝρ [n]} = √ ŝρ [n]e−j2πni/M . (11)
times and then multiplied with the impulse response of the M n=0
prototype filter. In order to simplify this procedure,
  we first
define the modulo operation as a%b = a − ab × b, where III. S YSTEMS UNDER AWGN CHANNEL
a, b are integers and x means the largest integer not greater
than x. Now, each filtered frame symbols yρ,n is In this section, we compare the performance of these two
systems under the AWGN channel and acquire the analytical
yρ [n] = p[n]xρ [n%M ] . (7) BEP expressions for them simultaneously. The Gaussian noise
is defined as w ∼ N (0, N0 ).
2) Overlap the frames with half symbol duration shift. From
the previous step, we get the filtered frames. Now they must
be shifted with half symbol period M/2 one by one. Thus, A. OFDM based model
the final transmitting signal is obtained by In the AWGN channel, the CP component of the OFDM
K system can be omitted as there is no ISI occurred. Hence,
 ρ−1 by denoting the received signal as r[n] = x[n] + w[n] and
y[n] = yρ [n − M] . (8)
ρ=1
2 combing equation (1) with equation (2) , we have

Because of the shift, the total length of yn will be extended, X̂[i] = F{r[n]} = F{x[n] + w[n]}
which is equal to Lp + (K − 1)M/2 = (3K − 1)M/2. = F{x[n]} + F{w[n]}
At the receiver side of the FBMC, a two steps process is (12)
= F{F −1 {X[i]}} + F{w[n]}
employed to recover the signal which will be passed into the
FFT block, as shown in Figure 7. Let r[n] denotes the received = X[i] + W [i].
signal, then:

ISBN 978-3-8007-3662-1 3 © VDE VERLAG GMBH · Berlin · Offenbach, Germany

 WSA 2015 • March 3-5, 2015, Ilmenau, Germany

y2 [n]
y1 [n]
y[0] y[1] y[2] y[3] y[4] y[5] y[6] y[7] y[8] y[9] y[10] y[11] y[12] y[13] y[14] y[15] y[16] y[17] y[18] y[19] y[20] y[21]

p[0] p[1] p[2] p[3] p[4] p[5] p[6] p[7] p[8] p[9] p[10] p[11] p[12] p[13] p[14] p[15]

s1 [0] s1 [1] s1 [2] s1 [3] s1 [4] s1 [5] s1 [6] s1 [7] s1 [8] s1 [9] s1 [10] s1 [11] s1 [12] s1 [13] s1 [14] s1 [15]
s1 [n] s1 [n + M ] s1 [n + 2M ] s1 [n + 3M ]

s1 [0] s1 [1] s1 [2] s1 [3]

s1 [4] s1 [5] s1 [6] s1 [7]

s1 [8] s1 [9] s1 [10] s1 [11]

s1 [12] s1 [13] s1 [14] s1 [15]

K−1
ŝ1 [0] ŝ1 [1] ŝ1 [2] ŝ1 [3] ŝ1 [n] = k=0
s1 [n + kM ] n = 0, 1, ..., M − 1.

Fig. 7. FBMC receiver

ρ−1
From equation (12) , we can see that W [i] is the noise term in which r[n + kM + 2 M] can be simplified as
which is corrupting the QAM signal. Thus, the power of it ρ−1
should be obtained for the BER expression, which is r[n + kM + M]
2
K
⎧ ρ−1 ρ−1
2 ⎫ = p[n + kM + M− M ]·
 ⎨
1
N
 −1 ⎬ 2 2
2π ρ=1
E W 2 [i] = E √ w[n]e−j N ni  
⎩ N n=0 ⎭ ρ−1 ρ−1
(13) xρ [ n + kM + M− M %M ] (16)
N −1 2 2
1  
=
4π
E w2 [n] e−j N ni = N0 . ρ−1
N + w[n + kM + M]
n=0 2
K
= p[n + kM ]xρ,n%M + wn+kM + ρ−1 M .
2
B. FBMC based model ρ=1

Thus, by substituting Equation (16) into Equation (15), we

have,
Similar to the OFDM model, in the AWGN channel, the
K−1

received signal of the FBMC system can be expressed as
ŝρ [n] = p[n + kM ]
k=0
K 
K

ρ−1  ρ−1
r[n] = y[n] + w[n] = yρ [n − M ] + w[n] p[n + kM ]xρ [n%M ] + w[n + kM + M]
2 ρ=1
2
ρ=1
K
 K−1
 K

ρ−1 ρ−1
= p[n − M ]xρ [(n − M )%M ] + w[n]. = p2 [n + kM ] xρ [n%M ]
ρ=1
2 2 k=0 ρ=1
(14)   
p̂2 [n]
K−1
 ρ−1
After acquiring the sequence, the receiver of FBMC will + p[n + kM ]w[n + kM + M]
2
conduct the two steps discussed in Section II. By resizing this k=0
  
signal and adding them together in the way of Equation (9) ŵρ [n]
and Equation (10), the signal enter the FFT block is K

2
= p̂ [n] xρ [n%M ] + ŵρ [n]
ρ=1
K−1
 K−1

ŝρ [n] = sρ [n + kM ] = p[n + kM ]rρ [n + kM ] = p̂2 [n]xρ [n] + ŵρ [n].
k=0 k=0
(17)
K−1
 Therefore, the demodulator received signal should be,
ρ−1
= p[n + kM ]r[n + kM + M] ,
2 X̂ρ [i] = F{ŝρ [n]} = F{p̂2 [n]xρ [n] + ŵρ [n]}
k=0 (18)
(15) = F{p̂2 [n]xρ [n]} + F{ŵρ [n]}.

ISBN 978-3-8007-3662-1 4 © VDE VERLAG GMBH · Berlin · Offenbach, Germany

 WSA 2015 • March 3-5, 2015, Ilmenau, Germany

Where, F{p̂2 [n]xρ [n]} can be expressed as C. BER derivation for both systems
From the two previous subsectoins, the power of the noise
F{p̂2 [n]xρ [n]} that influences QAM and OQAM demodulators are derived,
M −1
1  2 both are N0 . And it is clear that OQAM can be treated as a
=√ p̂ [n]F −1 {Xρ [i]}e−j2πni/M special case QAM because it only separates a QAM symbol
M n=0 into two symbols. This means that for I-ary rectangular QAM
M −1 M −1 or OQAM symbol, the BER expression is the same under the
1  2 1 
=√ p̂ [n] √ Xρ [i]ej2πni/M e−j2πni/M AWGN channel for both OFDM and FBMC, which is
M n=0 M i=0 √  
M −1 I −1 3 log2 I Eb
1  2 (19) Pb ≈ √ √ erfc ·
2 (I − 1) N0
, see [7]. (24)
= Xρ [i] p̂ [n] I log2 I
M n=0
M −1 K−1
IV. S YSTEMS UNDER R AYLEIGH C HANNEL
1   2 In this section, the two systems under Rayleigh channel are
= Xρ [i] p [n + kM ]
M n=0 studied. We use h[n] to denote the impulse response of the
k=0
Lp−1
Channel.
1 
= Xρ [i] p2 [n]. A. OFDM
M n=0 In this case, the received signal is r[n] = h[n] ∗ x[n] + w[n],
1
Lp−1 2 where ∗ meas convolution. At the receiver side for OFDM, a
As in equation (19), the term M n=0 p [n] = 1. Thus, we frequency domain equalizer will be applied to mitigate the
can recover the signal with the added noise, which is, distortion introduced by the channel, denoted as C[i]. The
equalizer will satisfy the constraint which is
X̂ρ [i] = Xρ [i] + F{ŵρ [n]} = Xρ [i] + Ŵρ [i].
(20) 1
C[i] = , (25)
H[i]
The equation above shows that the noise term at the OQAM where H[i] is the frequency response of the channel. There-
demodulator is Ŵρ [i]. As a result, the power of it should be fore, the signal entering the QAM demodulator is
obtained to get the analytical BER expression.
  X̂[i] = C[i] · F{r[n]} = C[i] · F{h[n] ∗ x[n] + w[n]}

E Ŵρ2 [i] = E F 2 {ŵρ [n]} = C[i]F{h[n] ∗ x[n]} + C[i]F{w[n]}
⎧ 2 ⎫
(26)
⎨ M −1 ⎬ = C[i]H[i]F{F −1 {X[i]}} + C[i]F{w[n]}
1 
=E √ ŵρ [n]e−j2πni/M = X[i] + C[i]W [i] = X[i] + Ŵ [i].
⎩ M ⎭
n=0
Thus, the channel attenuated noise power is
M −1 (21)    
1   2 −j4πni/M 2
= E ŵρ [n] e E Ŵ 2 [i] = E (C[i]W [i])
M n=0  
M −1 = E C 2 [i] · E W 2 [i] (27)
1   2 
= E ŵρ [n] . = N0 E C 2 [i] .
M n=0
As a result, the average SNR per bit is
It can be derived as Eb Eb  2 Eb
 γ= = E H [i] = Ω. (28)
E ŵρ2 [n] N0 E {C 2 [i]} N0 N0
⎧ 2 ⎫
⎨ K−1 ⎬ B. FBMC
 ρ−1
=E p[n + kM ]w[n + kM + M] For the PPN based FBMC system, a time domain equalizer
⎩ 2 ⎭ will be applied for the compensation of channel fading. By
k=0
K−1   denoting the impulse response of the equalizer as c[n] with
 ρ−1 property of
= p2 [n + kM ]E w2 [n + kM + M]
2 c[n] ∗ h[n] = δ[n], (29)
k=0
K−1
 where δ[n] is the Dirac delta function, we can achieve the
= N0 p2 [n + kM ]. received signal as rn = h[n] ∗ y[n] + w[n]. Thus, by passing
k=0 it through the equalizer, the mitigated sequence r̂[n] is
(22)
r̂[n] = c[n] ∗ r[n] = c[n] ∗ (h[n] ∗ y[n] + w[n])
By substituting Equation 22 into Equation 21, we get, = y[n] + c[n] ∗ w[n] . (30)
  
  M −1 K−1 v[n]
1  
E Ŵρ [i] = N0 p2 [n + kM ] = N0 . (23) As we define c[n] ∗ w[n] = v[n], then r̂[n] has the same
M n=0
k=0 structure as equation (14) in Section III, which means the

ISBN 978-3-8007-3662-1 5 © VDE VERLAG GMBH · Berlin · Offenbach, Germany

 WSA 2015 • March 3-5, 2015, Ilmenau, Germany

same derivations are obtained here. Using equation (20), the V. S IMULATION AND EVALUATION
received signal of the OQAM demodulator are,
In this section, the theoretical and simulated BER will
be compared under both AWGN and Rayleigh channels for
X̂ρ [i] = Xρ [i] + F{v̂ρ [n]} = Xρ [i] + V̂ρ [i]. (31) FBMC and OFDM. The carrier size M of them are 64. And
the 64-QAM (or OQAM) is applied in the simulation. For the
Now, we consider V̂ρ,i , which is OFDM system, the length of CP is 16. And in the FBMC
system, the overlapping factor K = 4 is used. The Rayleigh
V̂ρ [i] = F{v̂ρ [n]} multipath channel applied here is a 5-tap channel. We assume
K−1  that the perfect channel state information is available at the
 ρ−1 receiver.
=F p[n + kM ]v[n + kM + M]
2
k=0
M −1 K−1 (32) A. Under AWGN Channel
1  
=√ p[n + kM ]· For the AWGN channel, the CP part of OFDM system can
M n=0 k=0 be neglected as there is no ISI. Therefore, the CP are not taken
ρ−1 into the simulation.
v[n + kM + M ]e−j2πni/M .
2 0
10
OFDM
Simulation
Similar as equation (22), we get Theoretical

1
10
   
2 2 ρ−1
E V̂ρ [i] = E v [n + kM + M]
2
2

   10
2
= E v 2 [λ] = E (c[λ] ∗ w[λ])
Bit Error Rate
⎧ 2 ⎫
3
10
⎨  ⎬
=E w[η]c[λ − η]
⎩ η ⎭
4
10
  
= E (w[η]c[λ − η])
2 (33)

5
η 10
  
= E w2 [η] E c2 [λ − η]

6
η 10
  0 2 4 6 8 10 12 14 16 18
= N0 E c2 [λ − η] EbN0 (dB)

η (a) OFDM under the AWGN

 2
0
= N0 E c [λ] . 10
FBMC
Simulation
 2
 2
Theoretical
It is easy to prove that E c [λ] = E C [i] when C[i] =
1
10
F {c[λ]}. Thus, the average SNR per bit for FBMC is the
same as OFDM like equation (28) shows.

2
10
Bit Error Rate

C. BER derivation for both systems

3
10
The BER expression for both systems can be achieved base
on the previous conclusion. As FBMC and OFDM have the
4
same average SNR per bit for the demodulator, the BER 10
expression for them are the same. For I-ary rectangular QAM
or OQAM modulation over Rayleigh channel, the BER is [8],
5
10
√   
2 I −1 1.5γs
Pb (e) = √ 1−
6
10
log2 I I I − 1 + 1.5γs 0 2 4 6 8 10 12 14 16 18
√ 2 
EbN0 (dB)

I −1 1.5γs (b) FBMC under the AWGN

− √ 1− , (34)
I M − 1 + 1.5γs Fig. 8. BER under the AWGN channel
 
4 I − 1 + 1.5γ Figure 8. shows the simulation results of OFDM and FBMC
s
tan− 1 in AWGN channel and compared with the derived BER ex-
π 1.5γs pression. Both two simulations match the analytical expression
perfectly. This proves that their performances in the AWGN
where γs = γ log2 I denotes the average SNR per symbol. channel is exactly the same and validates our expression too.

ISBN 978-3-8007-3662-1 6 © VDE VERLAG GMBH · Berlin · Offenbach, Germany

 WSA 2015 • March 3-5, 2015, Ilmenau, Germany

B. Under Rayleigh Channel

0
10
Firstly, the FBMC system with different time domain equal- OFDM
Simulation
izers are simulated. The taps’ length of these equalizers varies Theoretical
from 6 to 1200, as shown in Figure 9.
0
10

Bit Error Rate

1
10
Bit Error Rate

1
10

6
taps
60
taps
2
10
200
taps 0 2 4 6 8 10 12 14 16 18 20
600
taps EbN0 (dB)
1200
taps Fig. 11. OFDM without CP in the Rayleigh channel
Theoretical

2
10
0 2 4 6 8 10 12 14 16 18 20
EbN0 (dB) VI. CONCLUSION
Fig. 9. FBMC with 200-tap equalizer in the Rayleigh Channel In this paper, the analytical BER expressions for OFDM and
PPN-FBMC systems under AWGN and Rayleigh channels are
From the figure above, we can tell that in order to reach derived and validated. We prove that the BER performance of
the theoretical BER performance for FBMC system, a very the two systems are the same when the perfect recover condi-
long time domain equalizer is necessary. However, this will tions are satisfied. In order to achieve this, OFDM requires the
introduce extra time delay to the system with growth in the cyclic prefix to eliminate the inter symbol interference, while
computational complexity. FBMC needs a long time domain equalizer.
Secondly, the OFDM system is simulated and evaluated. Hence, the further research can concentrate on the new
Figure 10 and Figure 11 show the BER performance of this equalization technique for PPN-FBMC which reduces the
system in the situation with and without CP correspondingly. complexity for mitigating the signal.
0
10
OFDM
Simulation
Theoretical
R EFERENCES
[1] A. Sahin, I. Guvenc, and H. Arslan, “A survey on multicarrier communi-
cations: Prototype filters, lattice structures, and implementation aspects,”
Communications Surveys Tutorials, IEEE, vol. 16, no. 3, pp. 1312–1338,
Third 2014.
[2] Q. He, C. Schmitz, and A. Schmeink, “BEP of fourier transform and
Bit Error Rate

discrete wavelet transform based OFDM,” in Wireless Communication

Systems (ISWCS 2013), Proceedings of the Tenth International Sympo-

1
10 sium on, Aug 2013, pp. 1–5.
[3] F. Schaich, “Filterbank based multi carrier transmission (fbmc); evolving
ofdm: Fbmc in the context of wimax,” in Wireless Conference (EW), 2010
European, April 2010, pp. 1051–1058.
[4] M. Bellanger, FBMC physical layer: a primer, PHYDYAS, Jan 2010.
[5] M. Tanda and T. Fusco, Data-aided synchronization and initialization
(single antenna), PHYDYAS, July 2008.
[6] M. Bellanger, “Physical layer for future broadband radio systems,” in
Radio and Wireless Symposium (RWS), 2010 IEEE, Jan 2010, pp. 436–
439.

2 [7] K. Cho and D. Yoon, “On the general BER expression of one- and
10
0 2 4 6 8 10 12 14 16 18 20 two-dimensional amplitude modulations,” Communications, IEEE Trans-
EbN0 (dB) actions on, vol. 50, no. 7, pp. 1074–1080, 2002.
[8] M. K. Simon and M.-S. Alouini, Digital Communication over Fading
Fig. 10. FBMC with 200-tap equalizer in the Rayleigh Channel Channels. John Wiley & Sons, Inc., 2005, pp. 223–309. [Online].
Available: http://dx.doi.org/10.1002/0471715220.ch8
Unlike in the AWGN channel where the CP does not really
matter at all, in fading channels CP plays a vital role to remove
the ISI. As we can see from it, when the length of CP is longer
than the channel impulse response, namely it can eliminate
the ISI perfectly, the analytical BER and the simulated one
are matching ideally. Thus, the OFDM system without the
cyclic prefix has a very poor performance. However, the
implementation of CP will decrease the bandwidth efficiency.

ISBN 978-3-8007-3662-1 7 © VDE VERLAG GMBH · Berlin · Offenbach, Germany