See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/321114782

A Comparison of OFDM and GFDM-Based MFSK Modulation Schemes for Robust

IoT Applications

Conference Paper · June 2017

DOI: 10.1109/VTCSpring.2017.8108191

CITATIONS READS
9 800

4 authors, including:

Ghaith Al-Juboori Evgeny Tsimbalo

University of Bristol Toshiba Telecommunications Research Laboratory
7 PUBLICATIONS 37 CITATIONS 18 PUBLICATIONS 244 CITATIONS

SEE PROFILE SEE PROFILE

Angela Doufexi
University of Bristol
255 PUBLICATIONS 3,503 CITATIONS

SEE PROFILE

All content following this page was uploaded by Ghaith Al-Juboori on 20 March 2018.

The user has requested enhancement of the downloaded file.

 A Comparison of OFDM and GFDM-based MFSK
Modulation Schemes for Robust IoT Applications

Ghaith Al-Juboori, Evgeny Tsimbalo, Angela Doufexi and Andrew R. Nix

Communication Systems and Networks Group-Department of Electrical and Electronic Engineering
University of Bristol, Bristol, United Kingdom.
Email: Ghaith.al-juboori, e.tsimbalo, a.doufexi, Andy.nix@bristol.ac.uk
Abstract—Orthogonal Frequency Division Multiplexing based In this paper, we investigate using the Generalized Fre-
M-ary Frequency Shift Keying (OFDM-MFSK) is a non-coherent quency Division Multiplexing (GFDM) waveform, which is
modulation scheme that was proposed for robust transmission one of the important candidates for the 5G waveform, with
over fast fading environments. It combines the MFSK modulation the MFSK modulation. GFDM has considerable features such
scheme with OFDM. In this paper, the Generalized Frequency as low OOB radiation because of the filtering process at each
Division Multiplexing (GFDM) waveform is considered to use
sub-carrier, better spectral efficiency due to a single Cyclic
with MFSK to take the advantages of the low out of band
radiation and relaxed synchronisation requirements. Different Prefix (CP) for the GFDM block, which consists of multiple
methods to combine MFSK and GFDM are proposed and their sub-symbols, and its simple synchronization requirements [6].
performances are evaluated and compared with OFDM-MFSK. Different methods to combine GFDM and MFSK have been
The mixed sub-carrier and sub-symbol method, which is one proposed and the performances of both GFDM based MFSK
of the proposed methods, is the most promising technique. (GFDM-MFSK) and OFDM-MFSK have been evaluated and
Additionally, gain margin is also seen for MFSK modulation compared in different channel types.
scheme compared to ordinary OFDM or GFDM. This gain can
be used to radically improve the performance for some 5G The remainder of this paper is arranged as follows: Section
applications, such as IoT, at low Signal to Noise Ratio (SNR) II gives a brief description of the GFDM air interface and
values. OFDM-MFSK and its transceiver model. Section III describes
the proposed methods that have been used to combine GFDM
Keywords—GFDM, OFDM-MFSK, GFDM-MFSK, Non-
with the MFSK modulation scheme, whereas in Section IV, the
coherent detection.
simulation parameters and results are presented and discussed.
Finally, conclusions are presented in Section V.
I. I NTRODUCTION
II. SYSTEM MODEL OVERVIEW
Channel estimation is an important issue in certain wireless
communication systems operating in fast fading environments A. GFDM Overiew
such as high speed trains applications [1]. Non-coherent mod- GFDM is one of the multicarrier modulation schemes
ulation schemes can be used to provide robust transmission which are proposed to address the requirements of 5G. In
in these applications. OFDM based MFSK (OFDM-MFSK), contrast with OFDM, GFDM can transmit up to P symbols,
which is the combination of non-coherent detection for M-ary at different time slots, per each sub-carrier. An oversampling
Frequency Shift Keying (MFSK) and Orthogonal Frequency process by at least K, which represents the total number of the
Division Multiplexing (OFDM) transmission, was proposed sub-carriers, is done per each sub-carrier before applying the
and analysed in [2]. This combination leads to very modest pulse shape filtering process to decrease the OOB radiation.
receiver structure without any need for equalisation and chan- Non-orthogonal and orthogonal prototype filters can be used
nel estimation. in this process which increases the flexibility of the GFDM
Some IoT applications in 5G, such as smart meters, need [7]. Finally, the up-conversion process is executed before the
to operate in low signal strength conditions since they are final GFDM signal is formed by adding the sub-carrier signals
installed in basements of buildings. Moreover, the information simultaneously. The basic structure of the GFDM transmitter
which they need to exchange with the network is very low [3]. is shown in Fig.1 and the time domain of the GFDM signal
Therefore, OFDM-MFSK is one of the promising solutions for can be written as:
these applications. K−1
X PX−1
x[n] = gk,p [n]dk,p , (1)
OFDM has great features that enable it to be used in k=0 p=0
many applications and standards such as its robustness to Inter
Symbol Interference (ISI) and its simple implementation due where dk,p is the complex data symbol which is transmitted on
to the use of the Fast Fourier Transform (FFT) algorithms [4]. the sub-carrier k and the sub-symbol p. The gk,p is the shifted
However, OFDM suffers from many drawbacks, for example, version of the prototype filter in time and frequency and can
its high Out-Of-Band (OOB) radiation, its sensitivity to the be express as:
time and carrier frequency synchronization and its high Peak k
to Average Power Ratio (PAPR) [5]. gk,p [n] = g[(n − pK)modN ]e−j2π K n , (2)
where n is the sampling index ( from 0 to N-1) and N equals
to K by P. The GFDM signal, (1), can be re-expressed as:

Fig. 2. Basic structure for the OFDM-based MFSK.

Fig. 1. The basic structure of the GFDM transmitter [8].

Fig. 3. Principle of OFDM-MFSK modulation (M=4).

→
− →
−
x =Ad, (3)
→
−
where d is the data block vector that contains N symbol MFSK is designed to improve receivers sensitivity at the
and A symbolizes the transmitted matrix of the GFDM with cost of bandwidth efficiency. In this modulation scheme, the
dimension of KP × KP and its structure is mentioned by [6] higher the value of M, the better of the sensitivity, but at the
as: expense of lower information transmitted per unit of time. This
is in contrast with bandwidth efficient modulation schemes,
such as M-QAM, in which the higher the value of M, the
A = [−
g−
→ −−−−→ −−→ −−−−−−→
0,0 ......, gK−1,0 , g0,1 ............, gK−1,P −1 ] (4) more information is transmitted. The bandwidth utilisation for
MFSK is equal to (log2 (M )/M ), while it is equal to (log2 (M ))
At the receiver side, GFDM demodulation can be express for M-PSK/M-QAM. Table-I- illustrates the values for the
as: bandwidth utilisation in (bit/sec./Hz) for MFSK for different
values of M.
→
−
de = B →
−
y, (5) TABLE I. BANDWIDTH U TILISATION FOR THE MFSK M ODULATION
S CHEME .
where →−
y is the received signal after removing the CP , B is the Modulation M=2 M=4 M=8 M=16 M=64 M=256
→
− MFSK 1/2 1/2 3/8 1/4 3/32 1/32
GFDM demodulation matrix and de is the received data after
the GFDM demodulation. Different methods to implement B
can be used such as Match Filter (MF), Zero Forcing (ZF) Nevertheless, this leads to poor spectral efficiency, espe-
and Minimum Mean square Error (MMSE). ZF, in which B = cially with the increase of M, which is main disadvantage
A−1 , is used in this paper to implement the receiver due to of this modulation scheme. Some methods to improve this
its simplicity. Moreover, the ZF performance loss due to the issue were suggested such as the hybrid transmission method
noise enhancement is zero when orthogonal pulse shape filter which implies exploiting the phases of the occupied sub-
is used [9]. carriers to send additional information by combining OFDM-
MFSK and Differential Phase Shift Keying (DPSK) [1]. This
B. OFDM-MFSK Overview is because non-coherent detection scheme (OFDM-MFSK)
allows a random phase selection for all the sub-carriers in the
The OFDM-MFSK applies a grouping of M sub-carriers, transmitter side.
in OFDM, and applies the MFSK modulation scheme to each
of these subsets (groups). This type of modulation scheme In multipath propagations environment (frequency selective
permits non-coherent detection which is especially interesting channels), some sub-carriers can totally fade out, which leads
for many scenarios where no channel estimation is needed such to an error floor. Channel coding in conjunction with an
as fast fading environment case and long range communication interleaver is used to mitigate that. Soft decision detection is
applications [2]. used to obtain the best performance by providing a degree of
reliability for each bit to the decoder. An appropriate metric
The basic structure of the OFDM-based MFSK transceiver (Lj ) for the j th bit of a coded symbol in a transmission can
is shown in Fig. 2. The idea of this modulation type, using M=4 be calculated based on the components of the received signal
(OFDM-4FSK), is illustrated in Fig. 3 where the sub-carriers vector rn as follows [1]:
are grouped into subsets of four, and only one sub-carrier per
each group is chosen for transmission (the solid arrow) while Lj = maxn∈Sj1 | rn |2 −maxn∈Sj0 | rn |2 , (6)
no energy is transmitted on the other sub-carriers in the group
(the red dots). As shown in Fig. 3, log2 (M ) bits, 2 bits in this where Sj0 is the subset of all components indices where the
example, are assigned for each subset using Grey code. codes symbols have ’0’ at the j th digit of the bit mapping.
Thus, there is a ’1’ at the j th digit of the bit mapping in case
of Sj1 .

III. GFDM BASED MFSK (GFDM-MFSK)

Three methods where GFDM can be combined with MFSK
will be discussed in the following sub-sections. The same
transceiver structure, as in Fig. 2, is used in this study except
that the OFDM modulator and demodulator blocks are replaced
by GFDM modulator and demodulator blocks, respectively.
The details of each of these methods are summarised below:
1. The Sub-Carrier Based Combination Method: In this
method, the MFSK is applied on the group of sub-symbols
corresponding to each GFDM sub-carrier. If P is the number of
symbols per each sub-carrier, K is the number of sub-carriers
and N (P × K) is the number of samples in the GFDM symbol,
then in this case M = P (M in the MFSK =P in the GFDM). Fig. 5. GFDM-based MFSK with M=P=K=4 (Method-2) .
The basic structure for this case is depicted in Fig. 4 with M
= 4, P = 4 and K = 4.

Fig. 6. GFDM based MFSK with M=K=4,P=2 (Method-3).

IV. SIMULATION PARAMETERS AND RESULT

Fig. 4. GFDM-based MFSK with M=P=K=4 (Method-1) . A. Simulation Parameters.

The performance of the OFDM-MFSK and GFDM-MFSK
2. The Sub-Symbol Based Combination Method: In this systems was simulated using the Monte Carlo method. The
case, the MFSK is applied on the group of symbols for each simulation parameters are listed in Table II. A rate-1/2
GFDM sub-symbol which are distributed over all active sub- (408,204) LDPC code was employed as a channel code with a
carriers (e.g. the sub-symbol ’i’ in each sub-carrier). In other LogMAP Sum-Product decoder [10]. To get an integer number
words, in this method, M of MFSK is equal to K in GFDM. for the OFDM and the GFDM symbols for each data packet,
Fig. 5 shows the basic structure for this case. some zeros (NZP ) were padded after encoder stage.

3. The Mixed Sub-Carrier and Sub-Symbol Combination TABLE II. S IMULATION PARAMETERS FOR THE GFDM AND
Method: As it is clear from the results of the previous two OFDM-MFSK.
methods (shown in the next section) that the performance of Parameter value
GFDM-MFSK becomes better as the number of sub-carriers K Input Data Block Size(N-data block) 204 bit
OFDM/GFDM Symbol Size (N) 256
increases (in the Rayleigh channel). Therefore, in this method Cyclic Prefix (CP) Length 32
there is no direct relation between the MFSK modulation Channel Coding LDPC
parameters and the GFDM parameters. MFSK is applied for Coding Rate (Rc) 1/2
-Additive White Guassian Noise
the sub-symbols that belong to the successive sub-carriers, Fig. (AWGN).
6 illustrates the basic structure of this method for M=K=4 Channel Types
and P=2. The GFDM parameters in this method, the number -Wide-band Rayleigh channel
with six equal power taps.
of sub-carriers and the number of sub-symbols per each sub- MFSK alphabet size (M) 2, 4, 8, 16, 64, 256
carrier, are fixed and independent on the M-size of the MFSK Prototype Filter (GFDM) Dirichlet
(they have been selected to be 128 and 2 respectively in this
work). The number of data symbols after the MFSK modulator
(and after zero padding) is equal to 0
10
  M
NDataSymbol = Ndatablock
Rc + N ZP . (7)
log2 M
−1
10
Following that, the number of OFDM (or GFDM) symbols
required for a single data block can be obtained by dividing
(7) by N. Table III summarises the results for different values

BER
−2
of M. 10
OFDM−M=2
OFDM−M=4
OFDM−M=8
TABLE III. N UMBER OF THE OFDM/GFDM S YMBOLS FOR OFDM−M=16

D IFFERENT MFSK A LPHABET S IZE . −3

OFDM−M=64
OFDM−M=256
10 GFDM−M=2
Alphabet Size Number of Padded Number of OFDM GFDM−M=4
GFDM−M=8
(M) Zeros (Nzp) (GFDM) Symbols Required GFDM−M=16
2 104 4 GFDM−M=64
GFDM−M=256
4 104 4 −4
8 72 5 10
−20 −10 0 10 20
16 40 7 SNR(dB)
64 0 17
256 0 51
Fig. 8. BER comparison between OFDM and GFDM-MFSK (method-1) in
Rayleigh channel.

B. Simulation Results and Discussion

The results of three methods that proposed to combine the 2. The Sub-Symbol Based Combination Method: The per-
GFDM with the MFSK are shown and discussed below: formance of both waveforms in the AWGN channel model is
identical and it is similar to the results of the sub-carrier based
1. The Sub-Carrier Based Combination Method: The per- combination method (Method-1), see Fig. 7.
formance (BER versus SNR) comparison between OFDM-
MFSK and GFDM-MFSK in the AWGN channel is shown 0
in Fig. 7. Obviously, the performance of the two waveforms 10
is fairly similar.
−1
0 10
10
BER

−2
−1 10
10 OFDM−M=2
OFDM−M=4
OFDM−M=8
OFDM−M=16
OFDM−M=64
BER

−2 −3 OFDM−M=256
10 OFDM−M=2
10 GFDM−M=2
OFDM−M=4 GFDM−M=4
OFDM−M=8 GFDM−M=8
OFDM−M=16 GFDM−M=16
OFDM−M=64 GFDM−M=32
OFDM−M=256
−3 −4
GFDM−M=256
10 GFDM−M=2
GFDM−M=4 10
GFDM−M=8 −20 −10 0 10 20
GFDM−M=16
GFDM−M=64
SNR(dB)
GFDM−M=256
−4 OFDM−BPSK

10 Fig. 9. BER comparison between OFDM and GFDM-MFSK (method-2) in

−20 −15 −10 −5 0 5
Rayleigh channel.
SNR(dB)

Fig. 7. BER comparison between OFDM and GFDM-MFSK (method-1) in Furthermore, Fig.9 depicts the BER versus SNR perfor-
AWGN channel. mance comparison in Rayleigh channel. For high values of M
( M≥ 64), the performance of the two waveforms is the same,
On the other hand, Fig. 8 shows the comparison in the while OFDM surpasses GFDM for M ≤ 16. Moreover, it is
wide-band Rayleigh channel type. It is clear that OFDM interesting to note that as M decreases, the difference between
outperforms GFDM and the difference between them increases the two waveforms increases (with the SNR). The difference
with M . Less than 1 dB can be easily seen for low values of is less than 1 dB when M = 16. Additionally, the difference is
M (M ≤ 16) and for high values of M (64 and 256), there increasing rapidly when M = 8 before the GFDM curves level
is an error floor in GFDM. The interpretation for this result at 6×10−2 . For M = 4 and 2, there is an error floor for GFDM-
is that as M increases, the sub-carriers number (K) decreases MFSK. The reason for this behaviour for GFDM-MFSK is as a
(assuming N is constant and N = K ×P ). This means that the result of decreasing M, which also represents the number of the
sub-carriers bandwidth increases and it is affected more due sub-carriers for GFDM, the sub-carrier bandwidth increases.
to the frequency selectivity property of the channel. This will The sub-carriers are affected more, due to relatively large
affect more symbols as M increases. Moreover, the situation bandwidth, by the channel whose effects doesn’t be equalised
becomes worse due to the absence of the equalisation stage as by the receiver (according to the specifications of the MFSK
was previously mentioned. modulation type).
 3. The Mixed Sub-Carrier and Sub-Symbol Combina- in the same conditions. The results show that the GFDM-
tion Method: The performance of OFDM-based MFSK and MFSK performance is similar to that of OFDM-MFSK in
GFDM-based MFSK using this method in AWGN channel is the AWGN channel, whereas it depends on the combination
similar and the same as the previous two methods performance method between GFDM and MFSK in the wide-band Rayleigh
in AWGN, see Fig. 7. channel. The sub-carrier based combination method gives
similar results as OFDM-MFSK at low values of M for MFSK
Fig.10 shows the performance of the two waveforms in
and the performance become worse as M increases due to
the wide-band Rayleigh channel. In contrast to the previous
the decrease of the number of sub-carriers in this method.
two methods, the results are similar for all values of M. In
On the other hand, the sub-symbol based combination method
this case, the GFDM parameters are fixed to certain values
gives similar results to OFDM-MFSK for high values of M
(P = 2, K = 128) that do not depend on the MFSK parameter
and the performance degrades as M decreases. Eventually, the
M. This means that the sub-carriers bandwidth will remain
mixed sub-carrier and sub-symbol combinations method for
constant and also smaller compared to the low number of sub-
GFDM-MFSK shows similar performance to that of OFDM-
carriers cases (less than 128).
MFSK regardless of M. These results, combined with the other
It is interesting to note that the MFSK modulation scheme major advantages of the GFDM waveform such as the low
achieves a remarkable SNR gain margin that would bring OOB emission and the relaxed synchronisation requirements,
significant improvements in low-rate IoT applications, such as will lead to a superior performance of GFDM-MFSK when
smart meters, compared to standard OFDM or GFDM scheme compared with OFDM-MFSK. Moreover, important SNR gain
at low SNR. Based on Fig.7 & Fig. 10, Table-IV shows the margins can be obtained in MFSK modulation scheme which
gain margin values for MFSK modulation scheme for different can provide significant improvements in low date rate 5G
M compared to OFDM-BPSK at BER level of 1 × 10−4 in applications such as IoT.
AWGN and wide-band Rayleigh channels. Please note that
channel equalisation is used in OFDM-BPSK in the wide- ACKNOWLEDGMENT
band Rayleigh channel case. However, this is not required for
OFDM-MFSK or GFDM-MFSK It is clear that gain margins Ghaith Al-Juboori would like to thank the Higher Com-
of 14 & 11 dB can be achieved in the MFSK-256 modulation mittee for Education Development (HCED) in Iraq, Ministry
scheme in AWGN and Rayleigh channel respectively. of Oil and the University of Baghdad for sponsoring his Ph.D.
studies.
TABLE IV. GAIN MARGIN BETWEEN OFDM-BPSK & MFSK WITH
DIFFERENT M VALUES IN dB. R EFERENCES
Channel Type M=8 M=16 M=64 M=256
[1] M. Wetz, I. Peria, W. G. Teich, and J. Lindner, “Robust transmission
AWGN 1.7 4.2 9 14
over fast fading channels on the basis of OFDM-MFSK,” Wireless
Wide-band
Rayleigh
0 1.5 6 11 Personal Communications, vol. 47, no. 1, pp. 113–123, 2008.
[2] M. Wetz, I. Perisa, W. G. Teich, and J. Lindner, “OFDM-MFSK with
differentially encoded phases for robust transmission over fast fading
0
channels,” in Proc. 11th International OFDM Workshop, Conference
10 Proceedings, pp. 313–317.
[3] Nokia, “LTE Evolution for IoT Connectivity,” December-2016 2016.
[Online]. Available: http://resources.alcatel-lucent.com/asset/200178
−1
10 [4] L. L. Hanzo, Y. Akhtman, L. Wang, and M. Jiang, MIMO-OFDM for
LTE, WiFi and WiMAX: Coherent versus Non-coherent and Cooperative
Turbo Transceivers. John Wiley Sons, 2011, vol. 26.
OFDM−M=2
BER

−2 OFDM−M=4 [5] N. Michailow, I. Gaspar, S. Krone, M. Lentmaier, and G. Fettweis,

10 OFDM−M=8
“Generalized frequency division multiplexing: Analysis of an alternative
OFDM−M=16
OFDM−M=64 multi-carrier technique for next generation cellular systems,” in 2012
OFDM−M=256
GFDM−M=2
International Symposium on Wireless Communication Systems (ISWCS),
−3 GFDM−M=4 Aug 2012, pp. 171–175.
10 GFDM−M=8
GFDM−M=16 [6] N. Michailow, M. Matthe, I. S. Gaspar, A. N. Caldevilla, L. L.
GFDM−M=64 Mendes, A. Festag, and G. Fettweis, “Generalized frequency division
GFDM−M=256
OFDM−BPSK multiplexing for 5th generation cellular networks,” IEEE Transactions
−4
10 on Communications, vol. 62, no. 9, pp. 3045–3061, 2014.
−20 −10 0 10 20 [7] N. Michailow, M. Lentmaier, P. Rost, and G. Fettweis, “Integration of
SNR(dB) a GFDM secondary system in an OFDM primary system,” in Future
Network Mobile Summit (FutureNetw), 2011, Conference Proceedings,
Fig. 10. BER comparison between OFDM and GFDM-MFSK (method-3) pp. 1–8.
in Rayleigh channel.
[8] G. R. Al-Juboori, A. Doufexi, and A. R. Nix, “System level 5G
evaluation of GFDM waveforms in an LTE-A platform,” in 2016
International Symposium on Wireless Communication Systems (ISWCS),
V. C ONCLUSION Conference Proceedings, pp. 335–340.
In this paper, the GFDM-MFSK modulation scheme which [9] I. Gaspar, L. Mendes, M. Matth, N. Michailow, A. Festag, and G. Fet-
tweis, “LTE-compatible 5G PHY based on generalized frequency divi-
is based on the combination of GFDM and MFSK was sion multiplexing,” in 2014 11th International Symposium on Wireless
investigated. Three different methods to combine MFSK with Communications Systems (ISWCS), Aug 2014, pp. 209–213.
GFDM were proposed and described. Their performance in [10] D. J. Mackay, “Encycloedia of sparse random graph codes.” [Online].
two channel types, AWGN and wide-band Rayleigh, was Available: http://www.inference.eng.cam.ac.uk/mackay/codes/data.html
evaluated, analysed and compared with that of OFDM-MFSK

View publication stats