Sie sind auf Seite 1von 4

The 21st International Symposium on Wireless Personal Multimedia Communications (WPMC- 2018)

Performance Comparison of OFDM-QAM and


FBMC-QAM Filtered in Multi-User Massive
MIMO Channel Characterization
Chakree Teekapakvisit†, and Sarun Duangsuwan‡
Department of Engineering, Information Engineering
King Mongkut’s Institute of Technology Ladkrabang, Prince of Chomphon Campus
17/1 Choomco, Pathio, Chomphon, Thailand 86000

t.chakree@gmail.com, ‡ sarun.du@kmitl.ac.th

Abstract—The current 4G standard or long-term evolution ISI if this delay spread cannot be neglected in comparison with
(LTE) system is based on the use of orthogonal frequency division the symbol period. One problem of OFDM is a large side-
multiplexing-quadrature amplitude modulation (OFDM-QAM) lobe of the rectangular-pulsed signal that can potentially lead
for the downlink and uplink communications. OFDM is an
orthogonal block transmission scheme which, in the presence to high inter-carrier interference (ICI) and severe adjacent-
of multi-path fading channels, plain orthogonal multicarrier channel interference (ACI). To flexibly support multi-user
modulation formats are not able to maintain orthogonality due scenario, orthogonal frequency division multiplexing access
to inter-symbol interference (ISI) among consecutive multicarrier (OFDMA) is applied for OFDM based system [7]. In uplink,
symbols. A new modulation format is a filter bank multicarrier however, a synchronization of multipath users that transmit
(FBMC). In this paper, we compare a performance of OFDM-
QAM and FBMC-QAM to filter the ISI problem at the receiver. different subsets of subcarrier is difficult to achieve, resulting
We evaluate the performance under the assumption of multi- in unacceptable ICI.
user massive MIMO channel as a simplicity. The experimental There are several works to resolve the drawback of OFDM,
results are depicted with eye-diagrams and the performance of bit the enhanced schemes such as time-domain windowing [7]-
error rates (BERs). As the result, the FBMC-QAM is a potential [8], and active interference cancelation (AIC) [9]-[10], have
successor of the dominant OFDM-QAM where BER is acceptable
with no ISI and it is definitely a promising modulation to use in been proposed. On the other hand, the research community
5G signal processing. has begun to explore other advanced multicarrier modulation
Index Terms—ISI cancellation, OFDM-QAM filtered, FBMC- technique for using in 5G system, with FBMC the most
QAM filtered, MU-massive MIMO channel characterization, 5G promising one.
signal processing. The application of FBMC has been studied in [11], so-called
self-equalization property leading to a channel flattening effect
I. I NTRODUCTION was reported through simulation. In [12], FBMC multi-tap
Massive MIMO has been studied as a feature of the equalization was presented to improve the equalizing accuracy
foundational 5G system [1]-[3] and is a key role of the compared to the 1-tap equalization per subcarrier at a higher
smart connectivity in 2020. The massive MIMO is originally computational complexity. In addition, reported in [13] discuss
introduced by Marzetta [4], in other word, it is large scale that a pilot contamination problem in massive MIMO can
antenna system or very large MIMO. Normally the massive be resolved by using blind equalization. This studies claim
MIMO is synchronous with a regime where the number of that FBMC is an appropriate match for massive MIMO in
antennas up to infinity. In fact, massive MIMO can be used as 5G system. Recently, a linearly proposed FBMC (LP-FBMC)
an smart antenna at higher frequency bands, as large numbers system based on MIMO channel has been presented in [14].
of operational antennas with a small footprint [5]. These The proposed is a singular value decomposition (SVD) as
large arrays can allow higher power consumption and can a linear transformation where the overlapped FBMC data
compensate the effect of propagation loss. without the interference free. It well-know that FBMC reduces
One of the most important challenges in massive MIMO [6], out-of-band emission (OOB). Therefore, in [14] BER show
is no acquisition data to channel state information (CSI) at the that the conventional FBMC-QAM and conventional-OFDM
base station (BS) between the massive BS antenna and multi- are inferior to their proposed between LP-FBMC and FBMC-
user (MU). The conventional MIMO system was used in LTE OQAM (offset quadrature amplitude modulation). They dis-
4G standard, which used OFDM is based transmission system. cuss that the conventional OFDM is worst BER performance
In multipath channels, the delay spread of radio signals causes in the asynchronous multi-user uplink MIMO channel, but not
massive MIMO channel.
In this paper, we present a performance comparison of

978-1-5386-5757-7/18/$31.00 ©2018 IEEE 616


The 21st International Symposium on Wireless Personal Multimedia Communications (WPMC- 2018)

OFDM-QAM and FBMC-QAM multicarrier filtered in multi-


user massive MIMO channel characterization. By the ana-
lyzed potential of ISI cancelation at the uplink, we simulate
a channel characterization as a frequency selective fading
channel under massive one hundred and twenty-eight antennas
at the BS receiver side and multi-user as ten and twenty
users of accessible. The performance results are discussed by
using BER regime to compare the efficiency of the successive
cancelation of ISI between OFDM-QAM and FBMC-QAM
filtered.
The rest of this paper is organized as follows. The mas-
sive MIMO channel is mentioned in Section II. Section III Fig. 1. A digital implementation of FBMC based on PPN.
describes the comparison of OFDM-QAM and FBMC-QAM
filtered signal model. The performance evaluation presents in
Section IV and finally concluded in Section V respectively.
III. C OMPARISON OF OFDM-QAM AND FBMC-QAM
II. M ULTI -U SER M ASSIVE MIMO C HANNEL F ILTERED S IGNAL M ODEL
Consider a multi-user MIMO in the paper by Marzetta [4]. The OFDM scheme can be obtained as a rectangular pulse
Assume that K users and single cell at BS. Each user is with N̂ sub-channels and spacing as 1/T . The filter of
provided with a transmit-receive antenna, communicating to transmitter and receiver are assumed identical, with a real-
cell BS with time division duplex (TDD). BS is provided valued, symmetric and time-limit to [− (1+β)T
2 , (1+β)T
2 ], where
with M transmit antennas and N receive antennas, which β is obtained to the roll-off factor. Each (1 + β)T seconds,
are link to K users in cell simultaneously. We assume that the transmitter takes M̂ -QAM symbols {ap (n)}N̂−1 p=0 , and
multicarrier modulation is used for transmission data. The generates a waveform as
channel observed at the K users obtain receive antenna during
the downlink can be represented as X
N̂ −1 X


x(t) = ap (n)g(t − n(1 + β)T )ej T pt
(5)
X
K−1
p=0 n=−∞
yk = hTk,j xj + nk (1)
k=1 where g(t) denotes the transmitter filter.
where xj denotes the signal transmit vector of BS j, and nk is The channel is assumed identical.In sub-channel p at the
noise sample at receiver K users. The signals transmitted are receiver K users, the received signal is demodulated by using
constrained by E[kxj k2 ] ≤ Pj , where Pj is a transmit power the subcarrier frequency Tp offset fe (normalized with sub-
of BS j to send Sj = |ψj | as a downlink data stream per one channel spacing T1 ), then, filtered at the receiver v(t) (identical
active user. In addition, hTk,j denotes the channel gain between g(t)) and sampled at instant n(1 + β)T to get the received
the k user and transmit-receive antennas at the BS. symbol
With linear precoding, BS j forms its transmitted signal as 2π
(p−fe t)∗v(t)
X ãp (n) = {(x(t)e−j T }|t=n(1+β)T (6)
xj = vk,j uk (2) X
N −1

k∈ψj = ej2πn(1+β)fe (ap (n)Γ0 + am (n)Γm−p )


m=0,m6=p
where uk denotes mutually uncorrelated zero-mean symbols
with the average energy per-symbol for all k ∈ ψj . where ⋆ stands for the convolution, and
Regarding uplink transmission, one use of the channel Z (1+β)T
observed at the j BS array can be rewritten as 2 2π
Γp = ej2πpβn v 2 (t)e−j T (p+fe )t
dt (7)
XX − (1+β)T
yjul = hk,j uul ul
k + zj (3) 2

j∈ℑ k∈ψl For OFDM, Cyclic prefix (CP) is used to eliminate the ISI
where uul denotes the bearing signal stream transmitted from from the previous symbol, where the input signal with the
k
user terminal k, power constrained to E[kuul 2 insertion of CP can be expressed as
k k ] ≤ Pul for some
power level Pul , ψl is the set of indices of active users in cell
x = [xN −Lcp +1 (t), . . . , xN −1 (t), xN (t)] (8)
l, and zul
j is the additive noise. For decoding the stream of
user k, BS provides vk as a linear filter, rewritten as where Lcp represents the CP length of OFDM block transmis-
ykul = vH yjul (4) sion.
On the other hand, FBMC uses polyphase network (PNN)
where (·)H denotes the hermitian matrix. to filter ISI where the structure of digital implementation as

617
The 21st International Symposium on Wireless Personal Multimedia Communications (WPMC- 2018)

Fig. 2. A comparison of between OFDM-QAM and FBMC-QAM channel model.

shown in Fig. 1. The z-transform of the n transmit filter can TABLE I


be expressed as L INK - LEVEL SIMULATION PARAMETERS

X
L−1 System bandwidth 10 MHz
gn (z) = gn (l)z −l (8) Number of N̂ subcarriers 1200
Subcarrier separation 10 kHz
l=0 Subframe length 1.0 ms
where gn (l) denotes the discrete-time transmit filters can be Data rate 10 Mbps
M̂ -QAM 16-QAM
obtained by sampling gn (t) with a sampling rate of Ts , as Symbol duration 66.67 µs
follows Cyclic Prefix (CP) 4.69 µs
Antenna configuration 128 elements at BS
gn (l) = gn (lTs ) = pT (l)e2πjn∆f l+jφn (8) K users 10 and 20
Number of j cells 1 single
The n polyphase decomposition is actually used by an Channel equalizer MMSE [13]
finite impulse response (FIR) filter. Based on discrete fourier
transform (DFT) modulator and PPN, the channel model of
FBMC-QAM is illustrated in Fig. 2. as well as of OFDM-
QAM.
IV. P ERFORMANCE E VALUATION
A. Link-level Simulation
In this following, we describe link-level simulation of our
investigation on the performance comparison of OFDM-QAM
and FBMC-QAM. The number of M transmit antennas and N
receive antennas are 128 elements, where K users as 10, and
20 respectively in MU-massive MIMO channel. We assume a
channel model as instantaneous Rayliegh fading model with
seven-paths exponentially decayed power delay profile (PDP)
in term of root mean square (r.m.s) delay spread value as 0.29
µs, where the path power decayed is relative to 2 dB and each
path and Rayleigh-faded with maximum Doppler frequency of Fig. 3. Performance of eye diagram under OFDM-QAM experimentally
filtered, where (a) K users = 10 and (b) K users = 20.
10 Hz. We provide that the impact of channel estimation is
also taken into account by using minimum mean square error
(MMSE) based equalizer. The table 1 shows the link-level
simulation parameters. from noise level when users more increased in the massive
MIMO channel. Likewise, the crossing point duration in time
B. Eye Pattern is started from 5 ms to 15 ms so that to analyze one-zero level.
We perform an eye diagram for understanding signal impair- We can see the eye diagram in Fig. 3 (a) both I-channel and Q-
ment at the BS receiver. By taking high-bandwidth instanta- channel, the one-level amplitude is 1.4 mV and -1.2 mV for the
neous of high-speed data rate, eye diagram demonstrates sum zero-level respectively. The eye height is calculated by the dis-
of logic one and logic zero transition measurement. To analyze tance between one-level and zero-level as 2.6 mV. In addition,
eye diagram parameters, the characteristic of one-zero level, the eye crossing percentage is calculated by using the follow-
eye amplitude and height, eye crossing, and signal-to-noise ing equation 100 ∗ [(crossing level − zero-level)/(eye height)],
ratio (SNR) is discussed. which is 46.15 percentage. Therefore, we can calculate the
In Fig. 3 shows the eye diagram of in-phase (I) channel and SNR level from eye pattern analyzer uses the following equa-
quadrature (Q) channel of OFDM-QAM modulation, where (a) tion 20∗log10 [(one-level−zero-level)/(one-level+zero-level)],
as K = 10 users and (b) K = 20 users. It can be seen that the which is 22.27 dB. In Fig. 3 (b), meanwhile, the one-level
logic one and zero level of OFDM-QAM pulses are affected amplitude is 0.8 mV and -1.2 mV for zero-level. Both I-

618
The 21st International Symposium on Wireless Personal Multimedia Communications (WPMC- 2018)

V. C ONCLUSION
In this paper, the comparison of OFDM-QAM and FBMC-
QAM filtering technique has been presented. Under the as-
sumption of the multi-user massive MIMO channel character-
ization, we simulate the channel model of the massive MIMO
with imperfect channel state information (iCSI). As the result,
the investigation both eye diagram and BER point out that
FBMC filtering is more suitable efficiency than the OFDM
filtering that because of the OFDM signal is vulnerable to
ISI, especially, in multi-user massive MIMO channel char-
acterization. To robust ISI, FBMC are achieved as a self-
equalization in the massive MIMO channel. Furthermore, in
case of downlink will be considered for the next.
R EFERENCES
Fig. 4. Performance of eye diagram under FBMC-QAM experimentally
filtered, where (a) K users = 10 and (b) K users = 20. [1] J. Hoydis, S. T. Brink, and M. Debbah, Massive MIMO in the UL/DL of
cellular network: How many antennas de we need?, IEEE J. Sel. Areas.
Commun., vol. 31, no. 2, pp. 160171, Feb. 2013.
[2] T. Nakamura, A. Benjebbour, Y. Kishiyama, S. Suyama, and T. Imai,
5G radio access: Requirements, concept and experimental trials, IEICE
Trans. Commun., vol. E98-B, no. 8, pp. 13971406, Aug. 2015.
[3] H. Papadopoulos, C. Wang, O. Bursalioglu, X. Hou, and Y. Kishiyama,
Massive MIMO technologies and challenges towards 5G, IEICE Trans.
Commun., vol. E99-B, no. 3, pp. 602621, Mar. 2016.
[4] T. L. Marzetta, Noncooperative cellular wireless with unlimited numbers
of base station antennas, IEEE Trans. Wireless Commun., vol. 9, no. 11,
pp. 35903600, Nov. 2010.
[5] E. Bjornson, E. G. Larsson, and T. L. Marzetta, Massive MIMO: Ten
myths and one critical question, IEEE Commun. Mag., vol. 54, no. 2,
pp. 114123, Feb. 2016.
[6] J. W. Wallace, and M. A. Jensen Modeling the indoor MIMO wireless
channel, IEEE Trans. on Antenna and Propaga., vol. 50, no. 5, pp.
591599, May 2002.
[7] K. Mizutani, T. Matsumura, and H. Harada, A comprehensive study of
universal time-domain windowed OFDM-based LTE downlink system,
in 20th International Symposuim on Wireless Personal Multimedia
Communication (WPMC), Bali, Indonesia, 2017, pp. 2834.
[8] K. Mizutani, A. Yoshito, H. Kuriki, T. Matsumura, and H. Harada,
Fig. 5. BER performance of OFDM-QAM and FBMC-QAM filtered in MU- Experimental evaluation of universal time-domian windowed OFDM-
massive MIMO channel characterization. based LTE downlink system by real-time wave-shaping, in IEEE 28th
Annual Intenational Symposium on Personal, Indoor, and Mobile Radio
Communication (PIMRC), Montreal, QC, Canada, 2017, pp. 15.
[9] R. Giuliano, F. Mazzenga, and P. Loreti, Reduction of OFDM out-of-
band emissions with a multi-step active interference cancelation tech-
channel and Q-channel, it can show that the crossing percent- nique, in IEEE Wireless Communications and Networking Conference
age is about 39 of percentage, where SNR is equal to 13.97 (WCNC), Istanbul, Turkey, 2014, pp. 735740.
[10] J. F. Schmidt, D. Romero, and R. L. Valcarce, Active interference
dB at least. In Fig. 4 shows the eye diagram of I-channel cancelation for OFDM spectrum sculpting: Linear processing is optimal,
and Q-channel of FBMC-QAM modulation. Fig. 4 (a), the IEEE Commun. Letters, vol. 18, no. 9, pp. 15431546, Sep. 2014.
calculation of eye crossing percentage is to 70 of percentage, [11] A. Aminjavaheri, A. Farhang, L. E. Doyle, and B. F. Boroujeny,
Prototype filter design for FBMC in massive MIMO channels, in IEEE
where one-level 1.4 mV and zero-level is 1.0 mV. However, International Conference on Communications (ICC), Paris, France, 2017,
the percentage of eye crossing is at least to 68 percentage in pp. 16.
Fig. 4 (b) when increasing of K = 20 users accessibly. We [12] A. Aminjavaheri, A. Farhang, N. Marchetti, L. E. Doyle, and B. F.
Boroujeny, Freuquency spreading equalization in multicarrier massive
calculate the SNR level as 44 dB in Fig. 4 (a) and as well as MIMO, in IEEE International Conference on Communication Workshop
the SNR level in Fig. 4 (b) is about 38 dB , respectively. (ICCW), London, UK, 2015, pp. 12921297.
[13] S. Duangsuwan, C. Teekapakvisit, and S. Promwong, On the uplink blind
The BER performance is depicted in Fig. 5, with an assum- equalization in a massive MU-MIMO systems to avoid pilot contami-
ing under instantaneous CSI. With K = 10 users, we observe nation, in International Symposium on Multimedia and Communication
that the performance of FBMC filtered by using PNN is able Technology (ISMAC), Ayutthaya, Thailand, 2017, pp. 117120.
[14] J. Kim, Y. Park, S. Weon, J. Jeong, S. Choi, and D. Hong A new filter-
to cancel the effect of ISI as efficiency. On the other hand, the bank multicarrier system: The linearly processed FBMC system, IEEE
performance of OFDM filtered by using CP is initialized by Trans. Wireless Commun., vol. 17, no. 7, pp. 48884898, July 2018.
ISI causing in the massive MIMO channel, likewise, compare
to FBMC at 22.27 dB of the SNR in OFDM. The performance
result of FBMC is able to recover the BER approximately as
50 percentages.

619

Das könnte Ihnen auch gefallen