You are on page 1of 6

Evaluation of WCP-COQAM, GFDM-OQAM and FBMC-OQAM

for Industrial Wireless Communications with Cognitive Radio

Aitor Lizeaga Pedro Manuel Rodrı́guez


and Mikel Mendicute and Iñaki Val
Signal Theory and Communications area Department of Communications
Mondragon Unibertsitatea IK4-Ikelan
Loramendi 4, Mondragón, Spain P J.M. Arizmendiarrieta 2, Mondragón, Spain
Email: alizeaga@mondragon.edu

Abstract—The main challenges for current industrial wireless large and metallic obstacles within the propagation channel.
solutions are the stringent real-time requirements of industrial Most of the current industrial wireless communications are
automation applications and the harsh propagation channel based on existing standards like IEEE 802.11 (WiFi), IEEE
conditions in this kind of scenarios. Most of the current 802.15.1 (Bluetooth) or IEEE 802.15.4 (ZigBee) and ad-
industrial wireless communications are based on existing stan- ditionally proprietary protocol extensions are applied over
dards like IEEE 802.11, IEEE 802.15.1 or IEEE 802.15.4 and them. Some of the most extended industrial wireless so-
additionally proprietary protocol extensions are applied over lutions are WirelessHART, ISA 100.11a, WIA-PA, WISA,
them. These communication systems cover a wide range of WSAN-FA and some solutions based on the IEEE 802.11
industrial applications but they lack robustness when it comes standards. Although these communication systems cover a
to tight real-time requirements for factory automation. Hence, wide range of industrial applications, they lack robustness
new schemes for the physical layer, along with new media to cover all [3], specially those involving tight real-time
access techniques, for industrial wireless communications must requirements for factory automation. Considering this, new
be proposed. Among the upcoming technologies, modulation schemes for the physical (PHY) layer, along with new media
candidates for 5G and cognitive radio might be promising access techniques, for industrial wireless communications
solutions. must be proposed.
In this article we analyse FBMC-OQAM, GFDM-OQAM
Among the upcoming technologies, modulation candi-
and WCP-COQAM modulation candidates for 5G in terms of
dates for the fifth generation (5G) and Cognitive Radio
(CR) might be promising solutions [4] [5]. Considering the
bit error rate, power spectral density and spectral efficiency
aforementioned timeliness and spectrum occupation issues,
over highly dispersive channels. From this analysis, on the one
effective medium access control (MAC) protocols are crucial
hand, we assess the suitability of these modulation systems for
to ensure collision free and timely data delivery. In this
industrial wireless communications based on cognitive radio.
aspect, the application of the CR paradigm might provide
On the other hand, we provide additional details about how
significant benefits by managing the use of available bands.
windowing affects the protection against highly dispersive mul- As for the 5G modulation candidates, their filtered multi-
tipath channels and the spectral efficiency in WCP-COQAM. carrier structure provides an efficient use of the spectrum
and makes them suitable modulation systems for CR and
1. Introduction other frequency domain MAC protocols. Moreover, con-
sidering the robustness against multipath effect of a multi-
For the last years, the use of wireless communications in carrier modulation (MCM) like Orthogonal Frequency Divi-
industrial automation has been constantly increasing and has sion Multiplexing (OFDM), 5G MCM candidates might be
gained great scientific interest [1]. Despite the benefits that suitable schemes for industrial multipath channel scenarios.
wireless communications provide to the industrial automa- Although most of the research carried out on PHY
tion sector, they also present significant drawbacks regarding proposals for 5G is oriented to mobile communications,
communications robustness. The main challenges for current 5G is actually meant to cover a wide range of application
industrial wireless solutions are the high requirements of fields [4]. Among all of them, our research focuses on
industrial automation applications concerning latency, syn- wireless industrial communications. Specifically, we inves-
chronism and reliability [2]. Besides, the industrial scenarios tigate robust PHY layers designed for industrial wireless
are characterized by the presence of high impulsive noise, communications based on CR [6] [7]. According to the
high occupation of the spectrum in the Industrial, Scientific ultra-reliability and low latency requirements (more than
and Medical (ISM) bands and fading channels due to severe high data rates and massive number of users) of industrial
multipath effect and shadowing caused by the numerous, communications [8], in this work we simulate some of the

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE SANTA CATARINA. Downloaded on July 19,2021 at 19:35:16 UTC from IEEE Xplore. Restrictions apply.
MCM candidates for 5G in lowband transmissions (carrier where p(n) is the prototype filter of length Lp and n =
frequencies below 6 GHz and a bandwidth narrower than 0, ..., Lp − 1. The exponential factor corresponds to the mth
100 MHz) through large indoor spaces and severe multipath sub-carrier, where M is the total number of sub-channels.
channels that emulate industrial halls. In the synthesis filter bank the xm,k symbols are in-
In this article, among the aforementioned industrial envi- terpolated, so that before filtering they are zero-padded by
ronment characteristics, we focus on severe multipath effect. M/2.
We analyse Filter Bank Multi-Carrier (FBMC) and Gener- X
alised Frequency Division Multiplexing (GFDM) based on am (l) = xm,k δ(l − kM/2), l ∈ Z. (2)
Offset Quadrature Amplitude Modulation (OQAM) and also k∈Z
Windowed Cyclic Prefix-Circular OQAM (WCP-COQAM)
modulation systems. Although WCP-COQAM [9] is not as Then filtering is performed and the FBMC-OQAM signal
extended as the other two schemes in the bibliography, it can be expressed as
shares several similarities with GFDM, which make it an- M
X −1 X
other candidate for 5G [10] and an interesting MCM scheme y(n) = am (l)gm (n − l), n ∈ Z. (3)
to be analysed. We evaluate their performance in terms of bit m=0 l∈Z
error rate (BER), power spectral density (PSD) and spectral
efficiency (SE) under highly dispersive channels. More details about FBMC-OQAM can be found in [11].
From the analysis and the assessment of the aforemen-
tioned MCM systems we provide the following contribu- 2.2. GFDM-OQAM
tions:
1) This article complements the work in [9], by bring- GFDM-OQAM modulation is similar to FBMC-OQAM,
ing additional details about how windowing affects the difference is that in GFDM-OQAM filtering is per-
the protection against multipath channels and the formed by a circular convolution instead of the linear con-
SE in WCP-COQAM. We state the conditions to volution used in FBMC-OQAM. This way, the modulation
provide full orthogonality in multipath channels and system now adopts a block based signal structure, so that
we show how windowing reduces SE in WCP- a cyclic prefix (CP) can be added to provide orthogonality
COQAM compared to GFDM-OQAM. against multipath channels without affecting signal’s time-
2) While most research about 5G is focused on mobile frequency localization (TFL).
communications, here we simulate and assess some We consider a GFDM-OQAM system with M sub-
of the main 5G modulation candidates under differ- channels and K complex valued data symbols in each
ent conditions. We model lowband transmissions block (i.e. 2K OQAM symbols in each block). Thus, if
through large indoor spaces and severe multipath we consider y(n) as one GFDM-OQAM block, it can be
channels and we show the results we obtained from expressed as
Matlab simulations and we state our conclusions.
M −1 MX
K−1
The article is organised as follows: in Section 2 we X
explain the theoretical background of the MCM systems we y(n) = am (l)g̃m (n − l + M K − 1), (4)
m=0 l=0
are going to analyse; in Section 3 we give more detailed
explanations about WCP-COQAM and make some clarifi- where n = 0, ..., M K − 1 and g̃m filters are obtained by the
cations with respect to the bibliography; in Section 4 we periodic repetition of gm filters defined in Equation 1. So
present the results obtained from our simulations; finally, in that g̃m (k) = gm [mod(k, M K)].
Section 5 we state the conclusions we obtained from our More details about GFDM-OQAM can be found in [9],
research. where the idea of GFDM-OQAM is introduced (the authors
refer to this scheme as FBMC-COQAM or just as COQAM).
2. Multi-carrier modulations
First of all, in this section we explain the theoretical 2.3. WCP-COQAM
background of the MCM systems we are analysing in this
article. WCP-COQAM is introduced in [9]. This technique can
be considered as a GFDM-OQAM system with a windowing
2.1. FBMC-OQAM process, which improves the PSD with respect to GFDM-
OQAM.
FBMC-OQAM consists in filtering each sub-channel of
In WCP-COQAM both the CP and the windowing are
the time-frequency lattice of a set of OQAM data symbols
related. The CP is divided in two parts: the guard interval
xm,k in order to get well localized sub-carriers. The key
(GI) and the window interval (WI) (referred to as RI in [9]).
components of the FBMC-OQAM scheme are the synthesis
So that the total length of the CP will be equal to the sum
and analysis filter banks, which are defined as
  of the lengths of the GI and WI parts:
2πmn
gm (n) = p(n)exp j , (1) LCP = LGI + LW I , (5)
M

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE SANTA CATARINA. Downloaded on July 19,2021 at 19:35:16 UTC from IEEE Xplore. Restrictions apply.
Figure 1. Overlapping of adjacent WCP-COQAM blocks.

where GI is the part of the CP aimed at preventing against positions within the original GFDM-OQAM block, a cyclic
multipath channels and WI is the part which will be used shift is carried out, which can be expressed as
for windowing.
Let us consider a GFDM-OQAM signal sG 1 as a queue s0G (k) = s[mod(k + LW I , M K)], (9)
of several GFDM-OQAM blocks, each one with its CP where k = 0, ..., M K − 1 and s0G (k) is the estimation of
extension of length LCP . Equally, we will consider M the original sG (k) signal.
to be the total number of sub-carriers and K the number
of symbols in each block. Then the lth block of a WCP-
COQAM signal is defined as 3. Orthogonality condition and spectral effi-
ciency analysis of WCP-COQAM
l+1
X
sW (k) = sG (k + rLW I )w(k − rQ), (6) WCP-COQAM is introduced in [9] and it is explained as
r=l−1 we did in Section 2.3. In this section, however, we contribute
with additional details regarding orthogonality against mul-
where k = 0, . . . , M K + LCP − 1; Q = M K + LGI and tipath channels and SE in WCP-COQAM. Specifically, we
the window function w(k) is defined by state the condition that the GI region of the CP extension

window coeff., k = 0, . . . , LW I − 1 must fulfil in order to provide full orthogonality against mul-
tipath channels. Besides, we also compare WCP-COQAM



 1, k = LW I , . . . , M K + LGI − 1 and GFDM-OQAM in terms of SE and we show that, in
w(k) =

 window coeff., k = Q, . . . , M K + LCP − 1 this aspect, the former is outperformed by the latter if equal
multipath protection is assumed.

0, otherwise.

(7)
In our work we use a Hamming window for the window 3.1. Orthogonality condition
coefficients.
Note that the samples of the CP corresponding to the In WCP-COQAM multipath channel protection is di-
WI part and also the last LW I samples of the block are rectly related to the windowing process. In order to explain
multiplied by the window coefficients and they are over- this fact, we show in detail the steps to form a WCP-
lapped with the WI regions of the previous and next blocks, COQAM block starting from a basic GFDM-OQAM block.
respectively, in order to not reduce SE by adding extra LW I
samples, as shown in Figure 1 and Equations 6 and 7.
WCP-COQAM demodulation process consists in remov-
ing the CP and the symbols affected by the window in order
to get the original GFDM-OQAM blocks. In the first place,
the first LGI and the last LW I samples of a WCP-COQAM
block are removed. So that

s = s0W [LCP − LW I , ..., M K + LCP − LW I ], (8)

where s0W is the received WCP-COQAM block and s is the


resultant block after removing the CP and the windowing
effect.
This way, the CP and the windowing effect are removed. Figure 2. Windowing operation of a WCP-COQAM block. a) A WCP-
Finally, in order to get all the samples back to their original COQAM block before windowing is equal to a GFDM-OQAM block with
CP extension. b) The first and the last LW I samples of a WCP-COQAM
block are distorted by the window.
1. For the sake of readability, from this point on parameters correspond-
ing to GFDM-OQAM will be expressed with the sub-index G, parameters
corresponding to WCP-COQAM will be expressed with the sub-index W
Once a GFDM-OQAM block is formed the CP extension
and parameters corresponding to OFDM will be expressed with the sub- is added to it. Figure 2.a shows the details of this first stage
index O. in the construction of WCP-COQAM signal. CP’, WI’ and

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE SANTA CATARINA. Downloaded on July 19,2021 at 19:35:16 UTC from IEEE Xplore. Restrictions apply.
Figure 3. Overlapping between adjacent WCP-COQAM blocks and CP extension’s structure.

GI’ are those last samples of the block which are copied in In order to make a fair analysis of SE in WCP-COQAM
order to form the CP extension, so that, at this point, CP = we compare it to GFDM-OQAM, because both modula-
CP’, WI = WI’ and GI = GI’. tion schemes only differ in windowing and overlapping
Figure 2.b represents the WCP-COQAM block right processes. In order to maintain a homogeneous notation
after windowing is applied. As shown in Figure 1 and between GFDM-OQAM and WCP-COQAM, we assume the
Equations 6 and 7, the window function is applied on the following considerations:
first and the last LW I samples, so that these samples are
CPG = GIG = GI1G
distorted. Because of that, CP, WI and GI are no longer
equal to CP’, WI’ and GI’ respectively. CPW = [W IW , GIW ] (11)
At this point, since the last LW I samples are affected GIW = [GI1W , GI2W ]
by the window, we define a new region for them which we
The criterion we applied for this comparison is to pro-
call GI2’. Thus, now the original GI and GI’ regions are
vide equal protection against multipath channels for both
divided in GI1, GI2 and GI1’, GI2’ respectively. At this
modulations, so that LGI1G = LGI1W . Thus
stage we have that WI 6= WI’, GI1 = GI1’, GI2 6= GI2’;
and as for the length of these regions LW I = LW I 0 = MK MK
LGI2 = LGI20 , LGI1 = LGI10 . Note that GI1 and GI1’ are SEG = =
LGIG + M K LGI1G + M K
now the only really redundant parts. Because of that, only (12)
GI1 and not the whole GI (as it is explained in [9]) acts
MK MK
as a real guard interval against multipath channels when SEW = =
FDE is performed. Hence, LGI1 ≥ LCH − 1 must be LGIW + M K LGI1W + LGI2W + M K
fulfilled in order to maintain full orthogonality through a From this analysis it is clear that if full orthogonality is
multipath channel. Thus, considering that LGI2 = LW I and to be maintained, the windowing process brings an extra GI2
so LGI1 = LGI − LGI2 = LGI − LW I , we can define the region to the CP which has effects on the SE of a WCP-
full orthogonality condition for WCP-COQAM as COQAM signal. In particular, SE is reduced by a factor
of (LGI1 + LGI2 + M K)/(LGI1 + M K) compared to a
LGI ≥ LCH − 1 + LW I , (10)
windowless system with equal multipath channel protection.
where LCH is the transmission channel length.
4. Simulation results and discussion
3.2. Spectral efficiency analysis
We assess the performance of the simulated systems in
Figure 3 shows how overlapping between adjacent terms of BER, PSD and SE. In our analysis, we use the BER
blocks is performed in WCP-COQAM. The WI region of of transmissions through Rayleigh fading multipath channels
each block is overlapped with the GI2’ region of the previ- in order to assess the robustness of the MCM systems; we
ous block. This way, the samples within WI regions do not take into consideration the PSD of the transmitted signals,
suppose an overload and they do not affect the SE. Based for it is an important factor when it comes to CR systems;
on this fact, in [9] the authors claim that the use of the and we use SE to represent the cost of the CP extensions
window function has no effect on SE. While it is true that and the windowing process, so that we can obtain a fair
the samples in WI have no effect on SE, claiming that the comparison between different MCM systems.
windowing does not affect the SE is not precise, at least Since we are evaluating MCM systems, we consider
if the orthogonality condition defined in Equation 10 is to appropriate taking OFDM as the reference system in order
be fulfilled. Note that the use of the window function and to assess the modulation schemes we are analysing in this
overlapping of WI and GI2’ regions distort those samples work. Table 1 shows the configuration parameters we used in
in GI2’ region. This and the fact that FDE is performed our simulations. We choose parameters suitable for lowband
between GI2 and GI1’, inclusive, implies that the GI2 region transmissions, so we choose OFDM values similar to some
is necessary in order to recover those samples originally of the IEEE 802.11 standards.
placed in the GI2’ region. At this point, we conclude that All the simulations corresponding to the results shown in
both GI1 and GI2 are essential parts within the GI region this section have been carried out in Matlab. The waveforms
of WCP-COQAM. introduced above and the calculations for the results shown

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE SANTA CATARINA. Downloaded on July 19,2021 at 19:35:16 UTC from IEEE Xplore. Restrictions apply.
TABLE 1. BASIC SIMULATION PARAMETERS .

FBMC- GFDM- WCP-


OFDM
OQAM OQAM COQAM
Signal
20 MHz
bandwidth
IQ modulation QPSK, 16-QAM, 64-QAM
M (sub-
64
carriers)
K (sym-
1 - 4 4
bols/block)
LCP
16 - 16 48
(samples)
LW I
- - - 16
(samples)
LGI
16 - 16 32
(samples)
LGI1
- - - 16
(samples) Figure 4. BER curves of the analysed MCM systems with different mod-
LGI2 ulation orders in a multipath channel.
- - - 16
(samples)
Prototype fil-
- PHYDYAS [11]
ter
Lp (samples) - 256
Here we show that, in terms of error rate, FBMC-
OQAM performs worse than the rest of MCM systems under
the simulated conditions. Thus, we can conclude that for
in this section have been programmed by the authors. Spe- MCM schemes in lowband transmission and under severe
cific functions from the Signal Processing toolbox have been multipath conditions, FDE equalization with CP is more
used only for bits and IQ complex data symbol mapping. robust than MMSE linear equalization.
On the other hand, although OFDM, GFDM-OQAM
and WCP-COQAM provide similar performance in terms of
4.1. Robustness against multipath channels
BER providing equal GI length, OFDM slightly outperforms
the other two MCM schemes due to its perfect orthogonality.
In this subsection we analyse the robustness of each
However, it is worth considering the difference in the SE
MCM system against multipath channels. In order to make
of each MCM system, since for equal GI protection against
a fair comparison we simulate every system with equal
multipath channels GFDM-OQAM and WCP-COQAM pro-
protection against multipath channels, matching effective GI
vide higher SE than OFDM and vice versa. Equally, GFDM-
lengths (LGIO , LGIG and LGI1W ). These parameters are
OQAM also provides higher SE than WCP-COQAM, due
selected as shown in Table 1.
to the windowing of the latter.
While for GFDM-OQAM and WCP-COQAM we em-
ploy CP extensions and FDE, just as for OFDM, FBMC-
OQAM could be considered a special case in this aspect. 4.2. Power Spectral Density analysis
Since it is not a block-wise modulation, no CP extension
is used, in order to keep good TFL. Therefore, with no PSD is a relevant characteristic in order to design a
CP extension, other equalization techniques must be used CR communication system suitable for scenarios with high
in FBMC-OQAM systems. In this work, we perform a spectrum occupation. In this subsection we compare the
minimum mean square error (MMSE) linear equalization spectrums of the signals of the analysed MCM systems,
for FBMC-OQAM with an ICI cancellation stage based on in order to assess their potential suitability for CR based
[12]. systems.
We simulated transmissions through a Rayleigh fading Figure 5 shows the spectrums of OFDM, GFDM-
channel model with exponential power delay profile (PDP), OQAM, WCP-COQAM and FBMC-OQAM sub-carriers
whose rms delay spread and channel length are trms = 150 one on top of the other. For this test we activate only the
ns and LCH = 16 taps. Besides, we simulate different central sub-carrier, in order to make the power leakage into
modulation orders to evaluate how larger constellations may adjacent sub-carriers visible.
affect the BER performance of the filtered MCM systems in As shown in Figure 5, every filter bank based MCM
comparison with the reference OFDM system. In Figure 4 system carries significant reduction of out of band (OOB)
we compare the BER vs. Eb /N0 of the MCM systems over radiation compared to OFDM. However, even between these
these channel models, assuming perfect synchronization and MCM schemes there is still a remarkable difference. In
full channel state information (CSI). We got these BER this sense, FBMC-OQAM clearly outperforms both GFDM-
curves by simulating transmissions through 104 Monte Carlo OQAM and WCP-COQAM. On the other hand, thanks to
iterations for each MCM system and each channel model. the windowing, WCP-COQAM shows significantly lower
We transmit 12800 random data bits and we generate new OOB radiation than GFDM-OQAM, at the cost of lower
random channels at each Monte Carlo iteration. SE.

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE SANTA CATARINA. Downloaded on July 19,2021 at 19:35:16 UTC from IEEE Xplore. Restrictions apply.
hand, PSD is improved in WCP-COQAM by means of
windowing at the expense of SE with respect to GFDM-
OQAM. Therefore, we conclude that GFDM-OQAM will
be preferable in situations where SE is crucial, while WCP-
COQAM will be preferable when better PSD must be en-
sured, like in the considered CR scenarios.

Acknowledgements
This work was partly supported by TIPOTRANS
(ETORTEK) and CIIRCOS (PC2013-68) projects of
the Basque Government (Spain), and by COWITRACC
(TEC2014-59490-C2-1-P/2-P) project of the Spanish Min-
istry of Economy and Competitiveness.

Figure 5. Spectrum of a single sub-carrier of the analysed MCM waveforms. References


[1] S. Vitturi, P. Pedreiras, J. Proenza, and T. Sauter, “Guest Editorial
5. Conclusion Special Section on Communication in Automation,” IEEE Transac-
tions on Industrial Informatics, vol. 12, no. 5, pp. 1817–1821, 2016.
[2] C. Lu, A. Saifullah, B. Li, M. Sha, H. Gonzalez, D. Gunatilaka,
In this article we make a comparison between three C. Wu, L. Nie, and Y. Chen, “Real-time wireless sensor-actuator
candidates for PHY layer modulation in 5G and a ref- networks for industrial cyber-physical systems,” Proceedings of the
erence OFDM system, in order to assess their suitability IEEE, vol. 104, no. 5, pp. 1013–1024, 2016.
for industrial wireless communications based on CR. We [3] A. Frotzscher, U. Wetzker, M. Bauer, M. Rentschler, M. Beyer,
simulate lowband communications and highly dispersive S. Elspass, and H. Klessig, “Requirements and current solutions
indoor channel scenarios. Under these conditions, we show of wireless communication in industrial automation,” in 2014 IEEE
International Conference on Communications Workshops (ICC), June
the performance and features of these MCM systems from 2014, pp. 67–72.
different points of view.
[4] M. Agiwal, A. Roy, and N. Saxena, “Next generation 5g wireless
Regarding FBMC-OQAM, although it offers the most networks: A comprehensive survey,” IEEE Communications Surveys
restrained spectrum, we consider that it does not fulfil our Tutorials, vol. 18, no. 3, pp. 1617–1655, thirdquarter 2016.
robustness requirements when it is used in environments as [5] M. Zheng, W. Liang, H. Yu, and M. Song, “Smcss: A quick and
previously described, as we showed in Section 4.1. Its BER reliable cooperative spectrum sensing scheme for cognitive industrial
through a multipath channel is too high compared to the rest wireless networks,” IEEE Access, vol. 4, pp. 9308–9319, 2016.
of modulation systems. [6] P. Rodrı́guez, I. Val, A. Lizeaga, and M. Mendicute, “Evaluation of
On the other hand, OFDM, GFDM-OQAM and WCP- cognitive radio for mission-critical and time-critical WSAN in in-
COQAM provide similar BER for all the simulated mod- dustrial environments under interference,” in Factory Communication
Systems (WFCS), 2015 IEEE World Conference on, May 2015, pp.
ulation orders. More precisely, results show that OFDM 1–4.
performs slightly better due to its perfect orthogonality.
[7] P. Rodrı́guez, R. Torrego, F. Casado, Z. Fernández, M. Mendicute,
However, it must be borne in mind that the other two A. Arriola, and I. Val, “Dynamic spectrum access integrated in a
systems provide higher SE for the same GI length, thanks to wideband cognitive RF-ethernet bridge for industrial control appli-
their longer block structures. Inversely, if equal SE is con- cations,” Journal of Signal Processing Systems, vol. 83, no. 1, pp.
sidered, GFDM-OQAM and WCP-COQAM can use longer 19–28, 2016.
GI and provide higher multipath protection than OFDM. [8] P. Popovski, “Ultra-reliable communication in 5g wireless systems,”
Besides, regarding PSD and OOB radiation, these filtered in 1st International Conference on 5G for Ubiquitous Connectivity,
Nov 2014, pp. 146–151.
multi-carrier schemes provide more restrained spectrum than
OFDM. So we conclude that GFDM-OQAM and WCP- [9] H. Lin and P. Siohan, “Multi-carrier modulation analysis and WCP-
COQAM proposal,” EURASIP Journal on Advances in Signal Pro-
COQAM clearly outperform OFDM in terms of PSD, while cessing, vol. 2014, no. 1, 2014.
they also can adapt the GI length to the channel requirements
[10] ——, Signal Processing for 5G: Algorithms and Implementations.
in order to provide equal or higher SE and robustness against Wiley, 2016, ch. 8, pp. 169–187.
multipath channels than OFDM.
[11] A. Viholainen, M. Bellanger, and M. Huchard, “Prototype filter and
As for the comparison between GFDM-OQAM and structure optimization,” PHYDYAS - PHYsical layer for DYnamic
WCP-COQAM, the differences between these two systems AccesS and cognitive radio, Tech. Rep., 2009.
are caused by the windowing of WCP-COQAM. Results [12] A. Ikhlef and J. Louveaux, “An enhanced MMSE per subchannel
show similar overall performance for both systems in terms equalizer for highly frequency selective channels for FBMC/OQAM
of robustness. But because of the windowing, for equal systems,” in Signal Processing Advances in Wireless Communica-
effective protection against multipath channels, GFDM- tions, 2009. SPAWC ’09. IEEE 10th Workshop on, June 2009, pp.
186–190.
OQAM provides higher SE and vice versa. On the other

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DE SANTA CATARINA. Downloaded on July 19,2021 at 19:35:16 UTC from IEEE Xplore. Restrictions apply.

You might also like