Received May 9, 2019, accepted June 8, 2019, date of publication June 19, 2019, date of current version July

3, 2019.
Digital Object Identifier 10.1109/ACCESS.2019.2923774

A Comprehensive Performance Evaluation of

Universal Filtered Multi-Carrier Technique
SHANLIN WEI 1,2 , HUI LI1 , WENJIE ZHANG 1, AND WEI CHENG1
1 School of Electronics and Information, Northwestern Polytechnical University, Xi’an 710072, China
2 School of Physics and Telecommunication Engineering, Zhoukou Normal University, Zhoukou 466000, China

Corresponding author: Shanlin Wei ([email protected])

This work was supported in part by the National Natural Science Foundation of China under Grant 61571364 and Grant 61401360, in part
by the Fundamental Research Funds for the Central Universities under Grant 3102017zy026, and in part by the Innovation Foundation for
Doctoral Dissertation of Northwestern Polytechnical University under Grant CX201833.

ABSTRACT The Internet of Things (IoT) and machine-to-machine (M2M) communication will play
an important role in future communications, but there is currently no ultra-reliable low-latency wireless
communication theory that guides its design. For highly mobile vehicle-to-vehicle (V2V) wireless commu-
nication scenarios, asynchronous transmission should also be accepted. Therefore, it is particularly urgent to
study ultra-reliable low-latency wireless transmission technology that satisfies asynchronous transmission.
Universal Filter Multi-Carrier (UFMC) is a new type of filtering wireless transmission mechanism that meets
this character. Although some of its properties have been explored in recent years, there are few articles that
systematically evaluate its performance. In this paper, firstly, the performance of the UFMC system is fully
evaluated in terms of spectral efficiency (SE), bit error rate (BER), peak-to-average power ratio (PAPR),
carrier frequency offset (CFO), as well as various multipath fading channels, effects of time delay (TD),
etc. Meanwhile, a mathematical analysis model is established for the BER of UFMC system, and some
exact closed-loop expressions of bit error probabilities for UFMC are derived. Moreover, the equivalent
form of the transmitter in the frequency domain is derived. Finally, the Monte Carlo simulation results
related to UFMC are presented. The results reveal that UFMC suffers from the same problems as other
multi-carrier system, including higher PAPR, affected by CFO, but it has its own inherent advantages such
as insensitivity to time delay or energy-efficient, and the main negative factor of UFMC is inter-carrier inter-
ference (ICI) not inter-symbol interference (ISI), which may play an important role in future M2M and V2V
communication.

INDEX TERMS UFMC, spectral efficiency, BER, PAPR, carrier frequency offset, time delay.

I. INTRODUCTION no longer support all types of applications in future multi-

Over the last few years, the continuous application of var- scenarios. To address these new challenges, fifth generation
ious new wireless technologies such as multiple antennas, (5G) technology [1]–[3] emerged as the times require. Dif-
multi-carrier transmission, etc., basically meets the people’s ferent from single scenario in 4G, the IoT and M2M commu-
requirements for high bandwidth and reliability. Voice, video, nication will become the main driving forces in 5G mobile
network and other multiple transmission greatly enrich the communication. In other words, 5G will mainly address the
people’s growing spiritual and cultural life. However, with the challenges from multiple scenarios and provide a unifying
continuous improvement of image resolution and the increase technology that utilizes all the existing techniques that satisfy
of online time, mobile data traffic has grown exponentially. the diverse set of communication requirements.
Massive Internet devices and new application scenarios also In fact, the 5G application model not only focuses on
have emerged constantly, such as holographic projection, mobile broadband communication, but it also introduces
autonomous vehicle, robots, artificial intelligence and mobile large amount of novel scenarios. The wireless technology
edge computing. In these cases, the existing network can architecture white paper released by the IMT 2020 promo-
tion group summarizes the three main application scenarios
The associate editor coordinating the review of this manuscript and
for 5G [4]. In general, they are enhanced mobile broad-
approving it for publication was Tariq Umer. band scenarios (eMBB) [5], ultra-reliable and low-latency

This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/
VOLUME 7, 2019 81429
 S. Wei et al.: A Comprehensive Performance Evaluation of Universal Filtered Multi-Carrier Technique

communication scenarios (uRLLC) and large-scale machine- urgent to propel the small packet and low-latency wireless
like communication scenarios (mMTC) [6], respectively. All transmission technology that mainly provides services for
of them must be considered in future communication systems. IoT or M2M. For this purpose, UFMC [10], [11] is pro-
This enforces new technology to comprehensively consider posed as a new filtering transmission mechanism to meet
multi-technology indicators such as data rate, spectrum effi- this demand. As a promising candidate waveform for short
ciency, BER, CFO, and time delay. Meanwhile, effectively packet and low latency transmissions, UFMC has drawn more
supporting multiple types of services, provided with high data and more attentions because of its energy-efficient trans-
rate, large bandwidth, multiple access, low BER and large mission, single-tap frequency domain equalizer, suppressing
throughput also need to consider. Under these conditions, the OOB emission, and shorter filter length [12]–[14].
traditional wireless air interface technology can not meet a The UFMC technique reduces the out-of-band power and
diverse set of services in different scenarios. Therefore it is the length of the filter by changing the edge drop charac-
urgent to design a new wireless air interface to offer hetero- teristic of symbol and filtering the carriers of each group in
geneous service requirements in an efficient and flexible way. sub-bands. Furthermore, what makes us more gratified is that
Orthogonal frequency division multiplexing (OFDM) [7] UFMC employs QAM modulation rather than OQAM, which
is widely used in various wired and wireless applications such guarantees it the opportunity to integrate with MIMO. These
as LTE downlink, WiMAX, WIFI, multiuser access, optical characteristics prompt UFMC more suitable for short uplink
communication, etc., due to its anti-multipath fading, one- burst communication or low delay communication.
step equalization and seamless integration of multi-antenna At present, researchers all over the world are rushing to the
technology. However, higher frequency offset, out of band development of 5G standards and actively exploring various
(OOB) emission, and strict time synchronization limit the new services and technical solutions. Although some scholars
discontinuous spectrum of carrier aggregation. In particular, have studied the performance of multi-carrier waveforms
more localized spectrum characteristics and relaxed synchro- appearing in 5G asynchronous communication, to the best
nization will be one of the most important requirements of our knowledge, there are few articles that perform a com-
of the physical layer for the future wireless network. This prehensive evaluation for UFMC. In this paper, the authors
makes IMT-2020 seek more diverse transmission technology. comprehensively evaluate the performance of UFMC scheme
In [8], the authors propose a filter bank based multi-carrier in terms of SE, BER, PAPR, CFO, TD, and so on. The
(FBMC) technique that offers higher spectral efficiency and obtained results have certain reference value for performance
superior spectrum confinement. By filtering each subcarrier evaluation of UFMC and new waveform design. Specifically,
separately, FBMC can be allowed to have good localization it provides some exact closed-loop expressions of BER in
in both time and frequency. In addition, one-tap equalizer different channel environments. In addition, the impact of
guarantees optimal performance. However, it is difficult to CFO and TD on the system performance is also analyzed.
implement for a long filter impulse response. Meanwhile, Based on this strategy, the contributions of this paper can be
it is obvious that FBMC is not suitable for low-latency summarized as follows:
communication of the uplink and efficient energy-saving • To provide an exact closed form expression of BER for
communication system. The adoption of offset quadrature UFMC scheme;
amplitude modulation (OQAM) is also more complicated • To extend the analysis model of UFMC and derive the
when combined with multiple-input multiple-output (MIMO) equivalent analysis model in frequency domain;
technology. Meanwhile, a generalized frequency division • To evaluate the SE, BER, PAPR, and the performance
multiplexing technique (GFDM) [9] has attracted the atten- under various channel environments;
tion both in scholars and practitioners. Because a cyclic • To investigate the impact of CFO and TD on system
prefix (CP) is used for only one set of symbols, rather than performance;
employing a CP for each symbol, GFDM has higher band- • To show some simulation results related to UFMC.
width efficiency than OFDM. The sparse nature of the signal
and variable filter design make the GFDM robust to synchro- The rest of this paper is organized as follows. In Section II,
nization errors and more suitable for interactive scenarios of the principle of UFMC is introduced. Section III analyzes the
spectrum fragmentation. However, the large size FFT and performance of UFMC. Some simulation results and discus-
continuous successive interference cancellation (SIC) algo- sions are presented in Section IV. Section V concludes this
rithms result in high complexity and delay at the receiver side. paper.
On the other hand, non-orthogonal waveform design also
makes GFDM face the same dilemma as FBMC: complex II. THE PRINCIPLE OF UFMC
pilot design and incompatible with multi-antenna technology. The original idea of UFMC is to group all the allocated
After the development of previous generations of com- subcarriers into several sub-bands, independently filtered on
munication technologies, the demand for eMBB has been every sub-band differing from OFDM whose filter is per-
basically solved, but the wireless technology that satisfies formed on the whole band while FBMC filters each subcarrier
the communication scenarios of uRLLC and mMTC has not individually. Generally speaking, UFMC is a new waveform
been sufficiently developed. In other words, it is particularly designed for 5G, which combines the advantages of OFDM

81430 VOLUME 7, 2019

S. Wei et al.: A Comprehensive Performance Evaluation of Universal Filtered Multi-Carrier Technique

where Oi represents the subcarrier index set in the i-th sub-

band. Then, a filtering operation is performed on each sub-
band. Unlike FBMC, the filtered signal xi can be seen as the
result of linear convolution of FIR filter impulse response fi
and time domain signal si . We assume that the i-th sub-band
filter is f i = [fi (0), fi (1), . . . , fi (L − 1)]T , and without loss of
generality, each filter
P power is2 usually chosen to be normal-
ized to unity, i.e., L−1 l=0 |fi (l)| = 1. Due to the protection of
filter edge, UFMC relaxes the requirement of time synchro-
nization and enhances the robustness of frequency and time
misalignment. The transformed signal si can be written as a
matrix form as shown below:
0
N S i,
si = WH (2)
FIGURE 1. The typical structure diagram of the general UFMC system. where WH N is N -point IFFT matrix, and the elements of
S0i is composed of Si which is expanded by zero insertion
according to the position of sub-band. Importantly, S0i is also
and FBMC. In a sense, OFDM can be regarded as a spe- in accordance with the following formula
cial case of UFMC. Additionally, the biggest advantage of
B
selecting UFMC as the main candidate waveforms for 5G
S0 i .
X
is that a large number of research results in OFDM can be S= (3)
modified slightly and directly applied to it. At the same time, i=1

UFMC does not employ cyclic prefix to eliminate the inter- The discrete time signal comes from the sum of the signals
symbol interference, which indirectly improves the spectral emanating from the filtered sub-bands. Therefore, the super-
efficiency of the system and effectively alleviates the precious position of all sub-band symbols in the time domain, that is,
spectral resources. the signal x sent by the transmitter, which can be expressed
The typical structure diagram of the general universal filter as
multi-carrier system is shown in Fig. 1. For convenience of B
X
analysis, a simple UFMC system can be, for instance, consid- x(l) = si (l) ∗ fi (l), l = 0, 1, . . . , N + L − 2, (4)
ered with Dolph-Chebyshev filter. Suppose that single-user i=1
scenario is adopted, for multi-user, the basic idea can be easily where ∗ represents a linear convolution operation, si (l) is a
analogized in turn. As illustrated in Fig. 1, a multicarrier sys- time domain signal, and L denotes the length of filter.
tem with the total number of N subcarriers is divided into B Assuming that complete information about the channel can
sub-bands, and each sub-band consists of several consecutive be acquired at the receiver. Due to filtering, the received
subcarriers, i.e., each sub-band includes K = N /B subcarri- signal y passed through the channel can be regarded as the
ers. There are many ways to divide subcarriers, for example, convolution of signal x and channel h. This process can be
average or random distribution. Theoretically, the number expressed as
of carriers in the i-th sub-band is Ki , for i = 1, 2, . . . , B.
B
Actually, we usually default that each sub-band is divided X
y=h∗x+n=h∗ si ∗ f i + n, (5)
equally for simplicity. In this case, Ki = K . Otherwise,
P i=1
Ki 6 = K , but Ki = N , for i = 1, 2, . . . , B. At first, a certain
bit of information is mapped to symbol by Gray code M-ary where h of length r is the impulse response of channel in the
quadrature amplitude modulation (M-QAM). These symbols time domain, which can be written as [h(0), h(1),. . . , h(r-1)]T ,
are then assigned to each sub-band by means of average fi (l) is the filter impulse response on the l-th subcarrier in the
allocation of carriers. For the i-th sub-band, the assigned i-th sub-band, and [.]T is defined as the transpose of a matrix.
symbols in the frequency domain is Si , whose length is Ki . For these received signals, up-conversion and RF- oper-
These symbols are then converted to time-domain signal si ation are adopted. Due to the interference of noise, these
by N -point fast inverse Fourier transform (IFFT). It should band-pass signals need to be converted into base-band sig-
be noted that the IFFT here is slightly different from the nals. For these signals to be demodulated, the traditional
conventional transform. After the IFFT module, for the method is to perform a zero padded operation for the received
i-th sub-band, the time domain signal si can be expressed signal, that is, to fill a certain number of zeros behind the
as received signal, so that the length of the received signal
reaches twice the transmitted signal, i.e., 2N . Then 2N -point
1 X FFT is carried out. These frequency signals can be expressed
si (l) = √ Si (k)ej2πkl/N , l = 0, 1, · · · , N − 1,
N k∈O as
i
(1) Y = W2N PHFVS + W2N Pn ∈ C2N ×1 , (6)

VOLUME 7, 2019 81431

 S. Wei et al.: A Comprehensive Performance Evaluation of Universal Filtered Multi-Carrier Technique

where W2N is a 2N-point FFT matrix, P is an extended unit be used. In a practical implementation, only even carrier
matrix that can be written as a 2N × (N + L + r−2) matrix, components can be restructured independently, which can be
whose elements can be written as pi,j = {1, 0}, (if i = j, represented as
pi,j = 1, else pi,j = 0). Meanwhile, H is a Toeplitz
Ye = Pe W2N PHFVS + Pe W2N Pn ∈ CN ×1 , (9)
matrix whose elements are related to the impulse response
of the channel h, and the first row is [h(0), 01x( N +L−2)]. where Pe is a 2N ×N dimensional matrix whose elements
More precisely, F = [F1 , F2 , . . . , FB ], where Fi is a (N + satisfies pe = {1, 0}, (if i = 2j, pe = 1, else pe = 0).
L − 1)×N Toeplitz matrix, and each column is obtained by Meanwhile, P, H,F, V and S are defined as mentioned above.
cyclically shifting the filter impulse response coefficient at We can recover the transmitted signal Ye , for e = 0,
the corresponding sub-band position. Fi can also be adjusted 1. . . 2N − 1, from the received signal by linear equalization.
according to the propagation conditions of the scheme and the For the sake of understanding, we only discuss the receiver
time-frequency offset. In addition, V = 3[V1 , V2 , . . . , VB ], of ZF equalization. For MMSE equalization, since it also
where 3 denotes the diagonal matrix, and Vi is an N × Ki belongs to the category of linear equalization, the derivation
matrix for i = 1, 2, . . . , B, based on the corresponding process is similar to ZF, we will not discuss the case of MMSE
sub-band position over the entire available frequency range. here.
Vi is constructed by extracting Ki -column vectors from the In the following section, we first analyze the spectral effi-
inverse Fourier transform matrix, and a complete inverse ciency of several waveforms. In order to ensure reliable oper-
Fourier transform matrix consists of all the transform matri- ation of the device, we evaluate the PAPR of several typical
ces of sub-bands. waveforms, and then explore the demodulation process of
To pave the way for a better understanding of the derivation UFMC under Gaussian and Rayleigh multipath channel. The
behind, the property of Fourier transform is first introduced as impact of CFO and TD on UFMC system is presented at the
follows. According to the convolution property of a linear and end.
time-invariant system and linearity of the Fourier transform,
we can deduce the following property. A. SPECTRAL EFFICIENCY
Property 1: The Fourier transform of time discrete signal Offering ultra-reliable data transmission with ultra-low
that is composed of summing the convolution of multiple latency and small packet transmissions are of crucial impor-
discrete signals is equal to the sum of the frequency domain tance for New Radio (3GPP named in Release 15). Spectral
signals after multiplication, which can be expressed as follow: efficiency, is a measurement parameter of importance, which
X X can be defined as symbol transmission rate in a unit band.
F( xi [n] ∗ xj [n]) = Xi (ejω )Xj (ejω ), (7) For OFDM, the spectral efficiency is a function of the FFT,
i,j i,j CP and modulation order, and it has the same spectral effi-
where Xi (ejω ) and Xj (ejω ) are the Fourier transforms of the ciency as single carrier. The spectral efficiency of OFDM can
input signal xi [n] and xj [n], respectively. be expressed as
According to (5) and (7), the equivalent frequency signals ROFDM N log2 M
ηOFDM = = , (10)
can be rewritten as WOFDM N + NCP
B
X where ROFDM represents the transmission rate of OFDM,
Y=H Fi Si + W2N n, (8) WOFDM is the transmission bandwidth of system, and M
i=1 defines the modulation order. In OFDM, the insertion of CP
where H is the channel impulse response of h in the frequency improves the performance of OFDM against ISI and ICI.
domain, Fi is the filter impulse response of fi in the frequency However, there is also a spectral loss due to the insertion
domain, and Si is 2N-point FFT ofsi . In addition, W2N is of cyclic prefix. As the available spectrum becomes more
2N-point FFT matrix, and n is Gaussian white noise in the and more difficult to obtain, improving spectral efficiency is
time domain. considered as an effective means for 5G to meet high-speed
At the receiver, Frequency domain equalization such data transmission.
as linear equalization and nonlinear equalization or time For UFMC technology, assuming that the length of the
domain equalization, for instance, minimum mean squared filter is set to L = NCP + 1 and the same number of carriers,
error (MMSE) equalization and zero-forcing (ZF) equaliza- this implies that UFMC has the same spectral efficiency
tion can be used to eliminate the adverse effects of the as OFDM. Then, the spectral efficiency of UFMC can be
channel. In general, the different performance between the defined as
frequency domain equalization and time domain equalization RUFMC N log2 M
ηUFMC = = , (11)
is not too obvious. For frequency domain signal Y of length WUFMC N +L−1
2N , only half of the carriers bring the information to transmit. where RUFMC represents the transmission rate of UFMC,
As mentioned in other articles, all the odd carrier components which is related to the modulation mode and the size of FFT
contain part of available information and inter-carrier inter- in baseband transmission. L is the length of FIR filter, and M
ference from other carriers in the sub-band, which cannot denotes the order of modulation.

81432 VOLUME 7, 2019

S. Wei et al.: A Comprehensive Performance Evaluation of Universal Filtered Multi-Carrier Technique

For OFDM and UFMC, the burst duration is independent where n is Gaussian white noise in the time domain, and B is
of spectral efficiency. However, FBMC closely depends on total number of sub-bands. The equation above can be rewrite
the frame duration. The spectral efficiency of FBMC has been as:
given in [15], [16], which can be written as
y0 = FVS + n, (16)
RFBMC Slog2 M
ηFBMC = = , (12)
WFBMC S + O − 1/2 where the elements of F, V and S are same as the value in (6)
where S is the number of symbols, O is overlapping factor, above. The conventional operation for these received signals
and M is bit per symbol. is zero padding, in other words, that is, the length of signal is
Through (10), (11) and (12), it is obvious that with the extended to 2N by inserting zero. The FFT operation with a
reduction of cyclic prefix or the length of filter or the increas- length of 2N -point is performed to the zero padded signal,
ing of modulation order, the spectral efficiency of multi- then the time domain signals are converted to frequency
carrier modulation can be improved. domain. This process can be expressed as

Y0 = W2N PFVS + W2N Pn ∈ C2N ×1 , (17)

B. PEAK-TO-AVERAGE POWER RATIO
The high PAPR of the transmitted signal is one of the signif- where P remains the same value as previous matrix. Because
icant drawbacks of multi-carrier modulation (MCM) which only the even carrier can be used to recover the original data,
degrades the digital-to-analogue converter and the efficiency we can take out the required data by a matrix Pe . This can be
of the power amplifier [17], [18]. A low PAPR is of great expressed as
importance for multi-carrier modulation, as high PAPR also
brings about lower efficiency and higher energy consump- Y0 e = Pe W2N PFVS + Pe W2N Pn ∈ C2N ×1 , (18)
tion. In fact, a number of solutions have been proposed to where Pe is defined as (9). Let 8 = Pe W2N PFV. The ZF
deal with the high PAPR in OFDM, such as deep learning estimator detects signals by performing the inverse operation
(DL) [19], tone injection (TI) [20], partial transmit sequence at the receiver which can be given as
(PTS) [21], selected mapping (SLM) [22] and so on. The
peak-to-average power ratio of the UFMC can be defined WZF = (8H 8)−1 8H . (19)
as
After equalization, the adverse effects of the channel are
max |x (l)|2

PAPR =  , (13) removed, leaving the original signal superimposed by noise.
E |x (l)|2

Hence, the signal equalized can be expressed as
where E[|x(l)|2 ] represents the average power of the signal, Ỹ0 = S + ((Pe W2N PFV)H (Pe W2N PFV))−1
max[|x(l)|2 ] represents the maximum power of the signal,
× (Pe W2N PFV)H Pe W2N Pn. (20)
and x(l) is as shown in (4). It is our desired goal to reduce
the maximum value as much as possible while keeping the The even bits need to be demodulated, and the odd bits are
average value unchanged. Because UFMC is a multi-carrier interference. In general, it is necessary to discard the signal
in essence, it is inevitable that the same-phase amplitudes are in the odd bit and demodulate only the signal in the even bit.
superimposed on each other in peak power. The PAPR is usu-
ally described by the complementary cumulative distribution D. BIT ERROR RATE OVER RAYLEIGH FADING CHANNEL
function (CCDF), that is, the probability that the PAPR is From (9) and (19), the expression of the received signal of
greater than the given threshold PAPR0 : UFMC under the Rayleigh channel can be obtained. It can be
CCDFPAPR = Prob(PAPR > PAPR0 ), (14) written as

where PAPR0 is the power threshold. In theory, UFMC also Ỹ = (8H 8)−1 8H 8S + (8H 8)−1 8H Pe W2N Pn, (21)
faces higher PAPR, but many PAPR reduction methods in
where 8 = Pe W2N PHFV. The above equation can be further
OFDM can be adopted with minor modifications.
expressed as
C. BIT ERROR RATE OVER ADDITIVE GAUSSIAN WHITE Ỹ = S + ((Pe W2N PHFV)H (Pe W2N PHFV))−1
NOISE CHANNEL
× (Pe W2N PHFV)H Pe W2N Pn ∈ CN ×1 . (22)
In this subsection, a UFMC system over additive Gaus-
sian white noise channel have been considered. As shown It can be seen that the original transmitted data has been
in Fig. 1, the transmitted signal passes through the Gaussian completely restored, except for a part of the noise. Moreover,
channel. After arriving at the receiver, then the signal can be it is also shown that the bit error is the main result of the
expressed as equivalent additive noise which includes the filter impulse
B response and multipath channel. If we can reduce the average
y0 = x + n =
X
si ∗ f i + n, (15) power in the second half, then we can improve the reliability
i=1 of the system.

VOLUME 7, 2019 81433

 S. Wei et al.: A Comprehensive Performance Evaluation of Universal Filtered Multi-Carrier Technique

E. CARRIER FREQUENCY OFFSET Although UFMC does not insert cyclic prefix to eliminate
For multi-carrier systems, they are more sensitive to carrier ISI between signals, the sub-band filtering mechanism can
frequency offset than single-carrier systems [23]. The CFO, generally remove most ISI. Therefore, for UFMC, ISI has less
which gives rise to ICI [24], may be produced by the phase negative effect, and the main negative factor of system is ICI
noise of transceiver local oscillator, the Doppler spread of caused by CFO. From (29), it is shown that the ICI which
the time-varying fading channel, frequency synchronization cannot be accurately estimated at the receiver side causes an
errors, or the relative position movement of transceiver and increase of BER in UFMC system.
receiver, thereby reducing the performance of multi-carrier
systems [25]. UFMC, as a member of multi-carrier systems, F. TIME DELAY
is of course also affected by CFO [26]. For receiving signal, it contains not only direct wave but also
The received signal of the UFMC system suffering from reflected waves from obstacles, such as mountains, buildings,
carrier frequency [27] can be expressed as etc. As the extra path increases, the time it takes for the
XB reflected signal to reach the receiver is delayed compared
y00 = h(l) ∗ ci (l)si (l) ∗ fi (l) + n(l), (23) to the direct signal. This effect is commonly referred to as
i=1 multipath dispersion. Time delay, a key parameter is involved
where h(l) is impulse response of the time domain channel. here. In a point-to-point communication system, the time
ci (l) = ej2πεl/N , which denotes the time domain frequency delay usually leads to inter-symbol interference because the
offset of sub-band i. In addition, ε represents the relative delayed multipath signal overlaps with the following sig-
CFO of sub-band i, which is normalized with respect to nals. Inter-symbol interference would bring about significant
the subcarrier spacing. For simplicity, assuming that all sub- errors in high bit rate transmission system, especially in time
bands are subjected to the same CFO, and the channel is division multiplexing scheme. It also increases rapidly as
perfectly known at the receiver. the transmitted bit rate increases. Before discussing the time
Taking these factors into account, the above formula can delay, let’s introduce the second property of Fourier transform
be expressed in matrix as first, i.e., the time domain shifting of FFT:
Property 2: The N -point sequence x(n) is cyclically shifted
Y00 = HFCVS + n. (24)
left or right by m sampling periods, then performing an
After zero padding and 2N -FFT transforming, we take out the N -point Fourier transform. It can be equivalently expressed as
even carrier data and rewrite it as the Fourier transform of sequence multiplying a shift factor,
Y00 = Pe W2N PHFCVS + Pe W2N Pn = 8S + Pe W2N Pn. which can be expressed as
(25) F[x(n + m)] = WN−km X (k),
Assuming that the receiver can accurately estimate the CFO, F[x(n − m)] = WNkm X (k), (30)
then let 8 = Pe W2N PHFCV. We can get where WN = e−j2π/N , and X (k) is the Fourier transform
Ỹ = S + (8 8)
00 H −1
8 Pe W2N Pn.
H
(26) of x(n). The impulse response for multipath time-varying
channels [28] can be written as
The above equation can be further expressed as
L−1
Ỹ00 = S + ((Pe W2N PHFCV)H (Pe W2N PHFCV))−1 X
h(t, τ ) = hl (t)δ(t − τl ), (31)
× (Pe W2N PHFCV)H Pe W2N Pn. (27) l=0

However, in a practical implementation, the CFO emerges where hl (t) is the time-variant path gain of the l-th multipath
unconsciously, which may be not found by the receiver. Thus, component, which is a complex Gaussian random process.
ICI inevitably came into being. In order to differ from conven- L is the number of propagation paths, and τl is the corre-
tional research, therefore, we evaluate performance assuming sponding path delay. If we convert it to the frequency domain,
that the receiver side does not know the CFO introduced at the we can get frequency domain expression according to the
transmitter. Hence, making 8 = Pe W2N PHFV, the signals above theorem, which can be expressed as
containing ICI can be expressed as L−1
Hl (t)e−j2πf τl ,
X
Yˆ00 = (8H 8)−1 8H Pe W2N PHFCVS H (t, f ) = (32)
l=0
+ (8 8)
H −1
8 Pe W2N Pn.
H
(28)
where Hl (t) is hl (t) in frequency domain. Here we find that
The above equation can be further rewritten as the multipath dispersion of channel is converted into the
Yˆ00 frequency domain and the time delay in frequency domain
causes tailing of symbols. In other words, the dispersion
= ((Pe W2N PHFV)H (Pe W2N PHFV))−1 (Pe W2N PHFV)H
characteristic of time is the main cause of inter-symbol inter-
× Pe W2N PHFCVS + ((Pe W2N PHFV)H ference. The difference between ISI and ICI is that the ISI
× (Pe W2N PHFV))−1 (Pe W2N PHFV)H Pe W2N Pn (29) is caused by trailing of symbols which can be eliminated

81434 VOLUME 7, 2019

S. Wei et al.: A Comprehensive Performance Evaluation of Universal Filtered Multi-Carrier Technique

by accurate channel estimation, and ICI is caused by carrier where K is the total subcarriers of each sub-band, and B is the
frequency offset which causes interference between adjacent number of sub-bands. Although the UFMC is not an orthog-
carriers. They all lead to a decline in system performance. onal waveform as a whole, the sub-bands and the subcarriers
Furthermore, for UFMC system, if the delay parameter of in the band are still orthogonal. Therefore, the overall BER
the channel is accurately estimated, the negative effect of of the system is equal to the average bit error rate of the
time delay can be eliminated. Thus, there is no ISI but respective sub-bands under the condition that each sub-band
only frequency-selective fading. Due to the filter operation is independent of each other.
of UFMC, the symbol edge of UFMC is not as steep as
that of OFDM. By reasonably setting the parameters of the B. BIT ERROR RATE OVER RAYLEIGH FADING CHANNEL
filter, adequate buffer space can be provided for time delay. The BER of Gauss white noise system is the upper limit of
Therefore, we can conclude that the UFMC system is not the performance of UFMC system. In fact, a communication
sensitive to the effect of time delay introduced by multipath system cannot simply suffer from additive noise, but it is often
channel, and does not require strict time synchronization as accompanied by multiplicative noise, this is exactly what we
orthogonal waveforms, e.g., OFDM. In other words, UFMC need to consider.
is more suitable for asynchronous transmission systems. For UFMC, the main factor affecting the BER of system
III. ANALYZING THE PERFORMANCE OF UFMC
is the signal to noise ratio. After equalization compensation,
For a communication system, its bit error rate performance the effect of channel on noise is the same for OFDM and
is the focus of the designer. For OFDM, there has been a lot UFMC. The only difference is that the filtering of the UFMC
of research in literature due to its booming in the 4G era. For changes the power of the noise. Thus, the noise variance on
UFMC, due to its novelty, there are few articles on its bit error the k-th subcarrier in the i-th sub-band for UFMC system can
rate. However, for the new IoT and V2V scenarios of 5G, be written as
σk2 = σn |λk |2 ,
2
.
whether the BER of UFMC can meet the needs of reality (37)
is a problem worth exploring. In this section, we introduce
the closed-form BER analysis of UFMC in additive Gaussian where λk is corresponded to the equivalent filter impulse
white noise channel and Rayleigh fading channel. response and the Rayleigh fading channel response in the
k-th subcarrier, and σn is the variance of Gaussian white
A. BIT ERROR RATE OVER ADDITIVE GAUSSIAN WHITE
noise. Bringing (37) to (36), the BER of UFMC can be
NOISE CHANNEL
rewritten as
For AWGN channel, the BER of OFDM using M-QAM √
B K
in [29] is given as 0 1 XX 2 M − 1
Pe1 = √ √ E
√ s √  KB M log2 M
2( M − 1)  6Eb log2 M  i=1 k=1
Pe = √ √ Q . , (33) √ 
 s
M log2 M N0 M − 1 6E b log M 
× Q  |λk |2 . 2 . (38)
N0 M −1
where M represents the modulation order, Eb /N0 is signal to
noise ratio and Q(x) is Q-function which is defined as From the above formula, we can see that BER of UFMC
Z +∞
1 t2 is mainly determined by modulation order, signal to noise
Q (x) = √ e− 2 dt. (34) ratio, equivalent filter response, and fading channel response,
2π x
which is different from OFDM. Simultaneously, the number
The difference between UFMC and OFDM is that each
of subcarriers K and the number of sub-bands B also impact
subcarrier of UFMC is filtered. At the receiver side, each
on BER for UFMC. Under certain conditions, |λk |2 = 1.
sub-band needs to perform the inverse process of the filter.
Then, OFDM and UFMC should have the same performance.
In this process, the noise variance on the k-th subcarrier of
But, in a practical implementation, we can’t guarantee that
UFMC system in the i-th sub-band is
this situation will be met. Under these circumstances, the per-
σk2 = σn |rk |2 ,
2
.
(35) formance of UFMC will not coincide with OFDM. This is
due to the amplification of the noise caused by the filtering
where rk indicates the equivalent filter response correspond- operation and channel equalization. If we can try to reduce the
ing to the k-th subcarrier, and σn represents Gaussian white noise amplification, we can improve the performance of the
noise variance. Therefore, the BER of M-QAM in the UFMC UFMC and even improve its BER. Following this direction,
system under the AWGN channel can be expressed as we will find the right direction to design filter in the UFMC.
B K √
0 1 X X 2( M − 1)
Pe = √ √ Q IV. SIMULATION RESULTS AND DISCUSSIONS
KB M log2 M
i=1 k=1 In this section, some simulations and discussions are illus-
s √  trated. We firstly evaluate the spectral efficiency of three
6Eb log M 
× |rk |2 . 2 , (36) waveforms, and then, the PAPR of OFDM, UFMC, SC-
N0 M −1 FDE [30], FBMC, and LFBMC [31] is simulated. Some

VOLUME 7, 2019 81435

 S. Wei et al.: A Comprehensive Performance Evaluation of Universal Filtered Multi-Carrier Technique

TABLE 1. Key parameters.

FIGURE 2. The spectral efficiency of UFMC compared with OFDM and

FBMC.

BER vs. Eb /N0 performance of UFMC is also comprehen-

sively evaluated whether it is under a Gaussian channel or a
Rayleigh multipath channel. Furthermore, in the case of CFO,
the BER performance is also performed. Finally, the impact
of time delay in UFMC and the performance in multiple
channels for different scenes are eventually explored. In order
to clearly show the simulated environment, in Table 1, some
key simulation parameters are summarized.
The spectral efficiency of UFMC, OFDM, and FBMC is
shown in Fig. 2, where 512 carriers and 16QAM modulation
are adopted. From the figure we can see that the spectral
efficiency of UFMC and OFDM is independent of burst
duration, but the spectral efficiency of FBMC increases loga-
rithmically with the burst duration. When the burst duration is
greater than 15ms, FBMC has the highest spectral efficiency FIGURE 3. The PAPR of UFMC compared with several typical waveforms.
compared with UFMC and OFDM under the same length
of protection interval. For instance, in the duration of burst burst communication scenarios, and this is consistent with the
T = 30ms, the spectral efficiency of FBMC can be about previous analysis.
0.3 dB higher than OFDM, and UFMC can achieve a benefit Fig. 3 shows the PAPR of UFMC compared with OFDM,
of 0.5 dB when OFDM is selected as a reference. In theory, SC-FDE, FBMC, and LFBMC. From the figure we can see
OFDM and UFMC have equal spectral efficiency when the that the single-carrier system has the better PAPR perfor-
filter length is equal to the length of the cyclic prefix. The mance compared to the multi-carrier system. This can be
reason why the spectral efficiency of UFMC is better than attributed to the factor that the single carrier system is not
OFDM here is that the UFMC filter length is L = 43, which is similar to multi-carrier system that often has multiple sym-
much smaller than the cyclic prefix CP = 128 in OFDM sys- bols with the same phase in one signal, which in turn increases
tem. At the same time, we can also see that in the lower burst the peak power of the signal. Meanwhile, LFBMC has sub-
duration, UFMC has higher spectral efficiency, which means optimal performance because it adopts some techniques that
that UFMC is more suitable for small packets short-term reduce the PAPR of FBMC, so the PAPR of LFBMC is

81436 VOLUME 7, 2019

S. Wei et al.: A Comprehensive Performance Evaluation of Universal Filtered Multi-Carrier Technique

FIGURE 4. The BER of UFMC under AWGN channel with different FIGURE 5. The BER of UFMC under rayleigh fading channel with different
modulations. modulations.

significantly reduced relative to FBMC. From Fig. 3, we can these four modulations are significant except for BPSK and
also see that OFDM has a higher PAPR that has been widely QPSK. Meanwhile, by comparing Fig. 4 and Fig. 5, we can
discussed in other articles. For example, when PAPR0 = 11, conclude that the BER of UFMC in Rayleigh channel is far
Prob(PAPR>PAPR0 ) ≈10−2 . Although UFMC has many worse than that in Gaussian channel due to the interference
advantages, compared to OFDM, its PAPR is somewhat too of multiplicative noise. For example, at BER = 10−3 , for
high, what needs to be taken seriously. Conservatively speak- BPSK, to achieve the same performance, the Rayleigh chan-
ing, there is a performance loss of 0.5 dB in UFMC compared nel requires up to 23dB of additional signal-to-noise ratio.
to OFDM. From Fig. 5, it also can be seen that the bit error rates of
From Fig. 4, we can see that UFMC has better performance BPSK modulation and QPSK modulation are equal, which is
under Gaussian channel, which can provide high-reliability the same as that under Gaussian channel, and it demonstrates
and low-delay packet transmission services within the error that the simulation results are reasonable. At the same time,
tolerance. It is also found in Fig. 4 that the UFMC scheme has BPSK has the best system performance. For instance, when
the same BER for BPSK modulation and QPSK modulation, Eb /N0 = 30, the BER can achieve to 10−4 . This threshold
which is consistent with our traditional theoretical knowl- value can meet the requirements of most scenarios. Compared
edge. This phenomenon can be interpreted as that QPSK with 64QAM, BPSK modulation outperforms by about 5dB
modulation can be regarded as two independent BPSK. signal-to-noise ratio at BER = 10−3 , which ensures reliable
Although the signal error rate of QPSK has a 3dB difference communication with each other in certain energy-constrained
compared with BPSK, the bit error rate of QPSK is still situations.
equal to BPSK. Moreover, it is interesting to note that as the The BER of UFMC under ITU-VehA channel subjected
modulation order increases, the BER performance of UFMC to different CFO is shown in Fig. 6. We can see from the
will deteriorate. The reason is that as the modulation order figure that as the CFO increases, the BER performance of
increases, the Euclidean distance between the constellations the UFMC decreases. To clarify, for ε = 0.02 and Eb /N0 =
becomes smaller. Eventually, it is easy to misjudge after 10 dB, the BER is close to 10−2 . However, for ε = 0.08
superimposing noise. Therefore, the 16QAM modulation for and Eb /N0 = 10 dB, the BER is close to 10−1 . This phe-
UFMC has better BER performance than the 64QAM mod- nomenon implies that the orthogonality is lost within the
ulation. For example, at BER = 10−4 , the system can save same sub-band due to CFO, which is the main reason for
about 4dB of signal-to-noise ratio, which greatly reduces the performance degradation of UFMC system. Compared
the transmit power of the signals. At the same time, it also with OFDM, although CFO has less impact on UFMC,
provides us with some compromises between performance we can see that CFO still causes a large performance loss on
and power. UFMC when CFO compensation technology is not used at
Figure 5 shows the BER of UFMC under Rayleigh fading the receiver. Furthermore, it can be seen that when the CFO
channel with different modulations, and ZF equalization is is small, the system performance changes relatively slowly.
adopted. As seen in Fig. 5, the performance gap between As the CFO reaches a certain level, the performance drops

VOLUME 7, 2019 81437

 S. Wei et al.: A Comprehensive Performance Evaluation of Universal Filtered Multi-Carrier Technique

FIGURE 6. The BER of UFMC under ITU-VehA channel subjected to FIGURE 8. The BER of UFMC under different channels with 16QAM.
different CFO.

TABLE 2. The performance comparison of several waveforms.

For instance, the fluctuation of BER does not exceed 0.1 dB

in the case of 10−4 . This is consistent with the previous
theoretical analysis, which means that the time delay has
less impact on UFMC. In other words, UFMC can guarantee
ISI-free transmission when the filter design is reasonable,
and UFMC transmitter performs time domain filter which
plays the role of soft protection for multipath delay signals.
Therefore, for UFMC system, although it does not use the
cyclic prefix operation as OFDM, it can still achieve better
system performance. At the same time, in Fig. 7, we can also
see that UFMC is not very strict about time synchronization,
FIGURE 7. The BER of UFMC under ITU-VehA channel with different time in other words, it is more suitable for asynchronous transmis-
delay. sion scenarios.
In order to see the performance comparison of several
dramatically. As the CFO increases again, the performance waveforms more intuitively, in Table 2 we give the perfor-
degradation becomes smaller, which can be attributed to that mance comparison of several waveforms mentioned in the
UFMC has reached the maximum error that does not dete- paper. For details, please refer to Table 2.
riorate as the CFO increases. This is what we usually call The horizontal line represents the need for further research
the floor effect. Of course, the impact of the CFO on UFMC in table 2.
can be reasonably removed by effective compensation at the
receiver, but this is beyond the scope of this paper. V. CONCLUSION
Figure 7 shows the effect of time delay on the UFMC In this paper, a comprehensive evaluation has been done
under ITU-VehA channel. From the figure we can see that on the performance of universal filter multi-carrier system
the impact of TD on the UFMC system is relatively small. that have been explored in terms of SE, BER, PAPR, CFO,

81438 VOLUME 7, 2019

S. Wei et al.: A Comprehensive Performance Evaluation of Universal Filtered Multi-Carrier Technique

as well as different channels, effects of time delay, etc. Par- [14] L. Zhang, A. Ijaz, P. Xiao, K. Wang, D. Qiao, and M. A. Imran, ‘‘Opti-
ticularly, a mathematical analysis model was established for mal filter length and zero padding length design for universal filtered
multi-carrier (UFMC) system,’’ IEEE Access, vol. 7, pp. 21687–21701,
the BER of UFMC, and some exact closed-loop expressions 2019.
of bit error probabilities for UFMC system were also derived. [15] R. Gerzaguet, D. Kténas, N. Cassiau, and J. Doré, ‘‘Comparative study of
Moreover, some key waveform technologies have also been 5G waveform candidates for below 6 GHz air interface,’’ in Proc. ETSI
Workshop Future Radio Technol. Focusing Air Interface, 2016, pp. 1–16.
discussed that may be expected in different scenarios to fulfill [16] J.-B. Doré, R. Gerzaguet, N. Cassiau, and D. Ktenas, ‘‘Waveform con-
the potential performance desires, like OFDM, SC-FDE and tenders for 5G: Description, analysis and comparison,’’ Phys. Commun.,
FBMC. In addition, some equivalent forms of the transmitter vol. 24, pp. 46–61, Sep. 2017.
[17] S. H. Han and J. H. Lee, ‘‘An overview of peak-to-average power
in the frequency domain and properties of FFT were also ratio reduction techniques for multicarrier transmission,’’ IEEE Wireless
presented. It is concluded that UFMC is the most promising Commun., vol. 12, no. 2, pp. 56–65, Apr. 2005.
candidate for V2V or M2M scenario as it is not sensitive to [18] T. Jiang and Y. Wu, ‘‘An overview: Peak-to-average power ratio reduction
techniques for OFDM signals,’’ IEEE Trans. Broadcast., vol. 54, no. 2,
TD. In particular, this paper may provide a comprehensive pp. 257–268, Jun. 2008.
reference for better understanding of waveform design for [19] M. Kim, W. Lee, and D.-H. Cho, ‘‘A novel PAPR reduction scheme for
researchers of New Radio in 5G wireless systems especially OFDM system based on deep learning,’’ IEEE Commun. Lett., vol. 22,
no. 3, pp. 510–513, Mar. 2018.
about UFMC. [20] J.-C. Chen and C.-K. Wen, ‘‘PAPR reduction of OFDM signals using cross-
entropy-based tone injection schemes,’’ IEEE Signal Process. Lett., vol. 17,
no. 8, pp. 727–730, Aug. 2010.
ACKNOWLEDGMENTS
[21] L. J. Cimini and N. R. Sollenberger, ‘‘Peak-to-average power ratio reduc-
The authors gratefully acknowledge the helpful comments tion of an OFDM signal using partial transmit sequences,’’ IEEE Commun.
and suggestions of the reviewers, which have improved the Lett., vol. 4, no. 3, pp. 86–88, Mar. 2000.
[22] S. Y. L. Goff, S. S. Al-Samahi, B. K. Khoo, C. C. Tsimenidis, and
presentation. B. S. Sharif, ‘‘Selected mapping without side information for PAPR
reduction in OFDM,’’ IEEE Trans. Wireless Commun., vol. 8, no. 7,
REFERENCES pp. 3320–3325, Jul. 2009.
[23] L. Rugini and P. Banelli, ‘‘BER of OFDM systems impaired by carrier
[1] J. G. Andrews, S. Buzzi, W. Choi, S. V. Hanly, A. Lozano, A. C. K. Soong, frequency offset in multipath fading channels,’’ IEEE Trans. Wireless
and J. C. Zhang, ‘‘What will 5G be?’’ IEEE J. Sel. Areas Commun., vol. 32, Commun., vol. 4, no. 5, pp. 2279–2288, Sep. 2005.
no. 6, pp. 1065–1082, Jun. 2014. [24] Q. Ma, P. Yang, Y. Xiao, H. Bai, and S. Li, ‘‘Error probability analysis of
[2] F. Boccardi, R. W. Heath, Jr., A. Lozano, T. L. Marzetta, and P. Popovski, OFDM-IM with carrier frequency offset,’’ IEEE Commun. Lett., vol. 20,
‘‘Five disruptive technology directions for 5G,’’ IEEE Commun. Mag., no. 12, pp. 2434–2437, Dec. 2016.
vol. 52, no. 2, pp. 74–80, Feb. 2014. [25] C. Chen, Y. Chen, Y. Han, H.-Q. Lai, and K. J. R. Liu, ‘‘High resolu-
[3] X. Zhang, L. Chen, J. Qiu, and J. Abdoli, ‘‘On the waveform for 5G,’’ IEEE tion carrier frequency offset estimation in time-reversal wideband com-
Commun. Mag., vol. 54, no. 11, pp. 74–80, Nov. 2016. munications,’’ IEEE Trans. Commun., vol. 66, no. 5, pp. 2191–2205,
[4] M. Shafi, A. F. Molisch, P. J. Smith, T. Haustein, P. Zhu, P. D. Silva, May 2018.
F. Tufvesson, A. Benjebbour, and G. Wunder, ‘‘5G: A tutorial overview [26] Y. Li, B. Tian, K. Yi, and Q. Yu, ‘‘A novel hybrid CFO estimation
of standards, trials, challenges, deployment, and practice,’’ IEEE J. Sel. scheme for UFMC-based systems,’’ IEEE Commun. Lett., vol. 21, no. 6,
Areas Commun., vol. 35, no. 6, pp. 1201–1221, Jun. 2017. pp. 1337–1340, Jun. 2017.
[5] A. Osseiran, F. Boccardi, V. Braun, and M. Fallgren, ‘‘Scenar- [27] M. Wu, J. Dang, Z. Zhang, and L. Wu, ‘‘An advanced receiver for uni-
ios for 5G mobile and wireless communications: The vision of the versal filtered multicarrier,’’ IEEE Trans. Veh. Technol., vol. 67, no. 8,
METIS project,’’ IEEE Commun. Mag., vol. 52, no. 5, pp. 26–35, pp. 7779–7783, Aug. 2018.
May 2014. [28] J. Li and M. Kavehrad, ‘‘Effects of time selective multipath fading on
[6] S.-Y. Lien, S.-L. Shieh, Y. Huang, B. Su, Y.-L. Hsu, and H.-Y. OFDM systems for broadband mobile applications,’’ IEEE Commun. Lett.,
Wei, ‘‘5G new radio: Waveform, frame structure, multiple access, vol. 3, no. 12, pp. 332–334, Dec. 1999.
and initial access,’’ IEEE Commun. Mag., vol. 55, no. 6, pp. 64–71, [29] M. Agrawal and Y. Raut, ‘‘BER analysis of MIMO OFDM system for
Jun. 2017. AWGN & Rayleigh fading channel,’’ Int. J. Comput. Appl., vol. 34, no. 9,
[7] F. Schaich and T. Wild, ‘‘Waveform contenders for 5G—OFDM vs. FBMC pp. 33–37, 2011.
vs. UFMC,’’ in Proc. 6th Int. Symp. Commun., Control Signal Process. [30] F. Pancaldi, G. M. Vitetta, R. Kalbasi, N. Al-Dhahir, M. Uysal, and
(ISCCSP), New York, NY, USA, 2014, pp. 457–460. H. Mheidat, ‘‘Single-carrier frequency domain equalization,’’ IEEE Signal
[8] R. Nissel, S. Schwarz, and M. Rupp, ‘‘Filter bank multicarrier modulation Process. Mag., vol. 25, no. 5, pp. 37–56, Sep. 2008.
schemes for future mobile communications,’’ IEEE J. Sel. Areas Commun., [31] D. Na and K. Choi, ‘‘Low PAPR FBMC,’’ IEEE Trans. Wireless Commun.,
vol. 35, no. 8, pp. 1768–1782, Aug. 2017. vol. 17, no. 1, pp. 182–193, Jan. 2018.
[9] G. Fettweis, M. Krondorf, and S. Bittner, ‘‘GFDM—Generalized fre-
quency division multiplexing,’’ in Proc. IEEE Veh. Technol. Conf.,
Apr. 2009, pp. 1–4.
[10] V. Vakilian, T. Wild, F. Schaich, S. ten Brink, and J.-F. Frigon,
‘‘Universal-filtered multi-carrier technique for wireless systems beyond
LTE,’’ in Proc. IEEE Globecom Workshops (Gc Wkshps), Dec. 2013,
pp. 223–228.
[11] J. Nadal, C. A. Nour, and A. Baghdadi, ‘‘Novel UF-OFDM trans-
mitter: Significant complexity reduction without signal approxima- SHANLIN WEI received the M.S. degree from
tion,’’ IEEE Trans. Veh. Technol., vol. 67, no. 3, pp. 2141–2154, Dalian Polytechnic University, Dalian, China,
Mar. 2018. in 2014. He is currently pursuing the Ph.D.
[12] F. Schaich, T. Wild, and Y. Chen, ‘‘Waveform contenders for 5G— degree with Northwestern Polytechnical Univer-
Suitability for short packet and low latency transmissions,’’ in Proc. IEEE sity, Xi’an, China. His research interests include
79th Veh. Technol. Conf. (VTC Spring), Seoul, South Korea, May 2014, wireless communication and signal processing.
pp. 1–5.
[13] L. Zhang, A. Ijaz, P. Xiao, and R. Tafazolli, ‘‘Multi-service system:
An enabler of flexible 5G air interface,’’ IEEE Commun. Mag., vol. 55,
no. 10, pp. 152–159, Oct. 2017.

VOLUME 7, 2019 81439

 S. Wei et al.: A Comprehensive Performance Evaluation of Universal Filtered Multi-Carrier Technique

HUI LI received the B.S. degree in electrical WEI CHENG received the B.S., M.S., and Ph.D.
engineering, the M.S. degree in circuits and sys- degrees in communication and information sys-
tems, and the Ph.D. degree in system engineer- tem, M.S. degree from Northwestern Polytechni-
ing from Northwestern Polytechnical University, cal University, Xi’an, China, in 2003, 2006 and
Xi’an, China, in 1991, 1996, and 2006, respec- 2011, respectively. From April 2011 to April 2013,
tively. He joined the School of Electronic Infor- he worked in the postdoctoral research station of
mation, Northwestern Polytechnical University, in School of Electronic Information, Northwestern
1993 and was promoted to an Associate Professor, Polytechnical University. Since 2013, he has been
in 2002. Since 2008, he has been a Professor of a Lecturer with the School of Electronic Informa-
School of Electronic Information, Northwestern tion, Northwestern Polytechnical University and
Polytechnical University. He is the author of many books, more than 80 was promoted to Associate Professor, in 2016. His research interests include
articles, and many inventions. He is a well-known teacher and had won wireless sensor networks, ad hoc networks, and radar signal processing.
several provincial and ministerial awards and hosted several National Natural
Science Funds, Aerospace Science and Technology Key Funds, and provin-
cial projects. He is a member of many academic institutions, such as Fire and
Command Control Research Association, Chinese Association for System
Simulation, and so on. His research interests include radar target tracking and
identification, communication signal processing, avionics integrated system
simulation, and multi-sensor information fusion.

WENJIE ZHANG received the B.S. degree in elec-

tronic and information engineering from Yangtze
University, Jinzhou, China, in 2011, and the M.S.
degree in communication and information sys-
tem from Guizhou University, Guiyang, China, in
2015. He is currently pursuing the Ph.D. degree
in electronic science and technology with North-
western Polytechnical University, Xi’an, China.
His research interests include wireless communi-
cations, massive MIMO, Millimeter wave commu-
nication, and signal processing for communication.

81440 VOLUME 7, 2019