Experimental Study of Simplified UTW-OFDM Receiver Technology for Application to 5G Using Software-Defined Radio Platform

The simplified universal time-domain windowed-orthogonal frequency-division multiplexing (Simplified UTW-OFDM) has been proposed to improve spectral efficiency. This study proposed a novel receiving method using an optimal inter-carrier interference (ICI) cancellation technique to improve communication quality during the application of the simplified UTW-OFDM to 5G. The simplified UTW-OFDM suppresses out-of-band emission (OOBE) through the application of very long time-domain windowing to the conventional OFDM with a cyclic prefix (CP-OFDM) symbol. However, in exchange for the OOBE suppression performance, a large ICI is generated due to the symbol distortion caused by the application of the time-domain window, resulting in the degradation of the reception quality. Therefore, in this study, we proposed a new method for calculating the log-likelihood ratio that takes into account the effect of the time-domain window. Further, an ICI canceller that included a process for suppressing the noise enhancement effect of the time-domain window was proposed. The effectiveness of the proposed method was evaluated through computer simulations and experiments using software-defined radio. The experimental evaluation showed that the proposed ICI canceller could suppress the OOBE by 23.5 dB compared to CP-OFDM under the condition that BLER of 0.1 is achieved even when 64QAM is applied.


I. INTRODUCTION
The 5th generation mobile communication system (5G) operated by telecommunications carriers [2] and local 5G (L5G) operated by private operators [3], [4], [5] are becoming increasingly popular. The number of users to be accommodated is expected to increase, and it is necessary to improve spectral utilization efficiency [5], [6], [7]. One solution for this issue is the suppression of out-of-band emission (OOBE) of the transmitted signal. In both 5G and L5G, cyclic-prefix orthogonal frequency division multiplexing (CP-OFDM) [8], [9] has been applied to the physical layer. However, the large OOBE caused by the discontinuity between OFDM symbols in CP-OFDM necessitates using a large guard band between frequency channels, thereby decreasing spectral efficiency [10].
This study proposes a 5G system that uses simplified UTW-OFDM and a receiver-side signal processing technique that mitigates the receive quality degradation due to the effects of UTW applied at the transmitter side. Although the OOBE suppression performance itself is solely determined by the signal processing at the transmitter side, there is a limit to the applicable window length under the condition of achieving a certain receive quality. The purpose of this article is to propose a receiver technology that can improve the suppressible OOBE by increasing the applicable window length under conditions that can achieve a BLER of 0.1 or less, thereby suppressing the effect of ICI generated by the window applied at the transmitter side. To obtain appropriate coding gain, the 5G-compliant LDPC requires additional processing to consider the effect of the window applied at the transmitter side (e.g., ICI) in the process of log-likelihood ratio (LLR) calculation performed during decoding [22]. Therefore, this study investigates the LLR adjustment effect in the simplified UTW-OFDM applied 5G system. In addition, as the mitigation effect of ICI using only LDPC coding is limited, particularly in highorder modulation schemes such as 64QAM, we also propose and apply an ICI canceller based on the scheme proposed in [1], with additional improvements to suppress noise enhancement effects. These proposed receiving technologies for the simplified UTW-OFDM utilized 5G are evaluated through not only computer simulations but also experimental studies using software-defined radio (SDR).
In our prior and shorter version of this study [1], we proposed the basic concept of the simplified UTW-OFDMutilized 5G system and ICI canceller. In addition, the proposed system was evaluated with QPSK and 64QAM [1] via developed link-level computer simulations. The proposed ICI canceller has adjustment coefficients with different optimal values for each UTW length. In [1], these optimal values were determined by simulation in a fully exploratory manner. In this article-the extended version of that study-, we express the signal-to-interference/noise ratio (SINR) after ICI cancellation in a mathematical expression and propose an analytical method to find the adjustment factor of the ICI canceller that can increase this SINR the most. The proposed system was evaluated with not only QPSK and 64QAM but also 16QAM. Furthermore, the proposed system was implemented based on a 5G-based SDR platform [28] to evaluate its actual performance with QPSK and 64QAM, and its effectiveness was demonstrated on actual equipment.
The main contributions of this study can be summarized as follows: r A novel receiving technique (optimal ICI canceller to mitigate SINR degradation due to time window) when Simplified UTW-OFDM was applied to 5G. r OOBE suppression performance and block error rate (BLER) characteristics of the proposed receiving technique were evaluated via computer simulations. r A 5G physical layer with simplified UTW-OFDM, including the proposed receiver technology, was implemented in an SDR-based actual system, and the effectiveness of the proposed system was experimentally demonstrated. The remainder of this article is organized as follows. Section II provides an overview of the conventional waveform shaping techniques. Section III presents the proposed simplified UTW-OFDM receiver-side processes, including optimal ICI canceller. Section IV presents an evaluation of the proposed scheme by developing 5G-based link-level computer simulations. Further, Section V demonstrates the results of SDR. Finally, Section VI offers the conclusion of this article.

II. CP-OFDM AND SIMPLIFIED UTW-OFDM
This section describes the transceiver configuration and signal processing of the conventional CP-OFDM and simplified UTW-OFDM.
A. CP-OFDM 1) TRANSMITTER Fig. 1 shows a block diagram of the CP-OFDM transmitter. First, the transmission bit sequence is encoded by LDPC code and modulated to obtain the complex transmission symbols.  Next, the obtained symbols are allocated to the frequencydomain subcarriers. Thereafter, the m-th CP-OFDM symbols, s m ∈ C (N+N CP )×1 , are generated via the application of inverse fast Fourier transform (IFFT) to the subcarriers and by inserting CP: where N is the IFFT size, N CP is the number of sample points for the CP, x m = [x (m,0) , . . . , x (m,N−1) ] T ∈ C N×1 denotes the transmission complex symbols in the frequency domain (i.e., subcarrier symbols), F −1 ∈ C N×N is the IFFT matrix, whose elements are Here, 0 i× j is an i-by-j matrix with all zero elements and I N is an N-by-N identity matrix.
2) RECEIVER Fig. 2 shows a typical block diagram of the CP-OFDM receiver. First, the CP is removed from the received signal, and then the frequency-domain complex symbols, y m ∈ C N×1 are obtained by applying fast Fourier transform (FFT): where n m ∈ C (N+N CP )×1 denotes the zero-mean complex AWGN, H ∈ C (N+N CP )×(N+N CP ) represents the channel matrix, D = [ 0 N×N CP , I N ] T ∈ Z N×(N+N CP ) is the CP removal  matrix, and F ∈ C N×N denotes the FFT matrix, whose elements are F(p, q) = exp(−j2π pq/N )/ √ N. If the multipath delay is longer than the CP length, or if carrier frequency offset (CFO) occurs, ICI will occur, and the reception quality will be degraded. However, several ICI compensation methods have been proposed for these ICIs [29], [30], [31].
Following the application of channel estimation and equalization to y m , the received bit sequence is obtained through demodulation and channel decoding.

B. SIMPLIFIED UTW-OFDM
In systems operating in multipath low-latency environments (e.g., local 5G), the maximum Doppler delay is sufficiently shorter than the CP length, resulting in a surplus of CP length. Considering this extra CP length, UTW-OFDM, wherein overlap margin (OM) processing is omitted as shown in Fig. 3(b), is referred to as simplified UTW-OFDM [1], [23]. This section describes the transmitter and receiver configurations for applying Simplified UTW-OFDM to 5G systems.
where W ∈ C (N+N CP )×(N+N CP ) is time-domain windowing matrix defined as follows: where diag(·) denotes the diagonal matrix with the components in the parentheses as diagonal components, w ∈ C (N+N CP )×1 represents the simplified UTW vector, and w TR ∈ C N TR ×1 and w I TR ∈ C (N TR −1)×1 are the UTW and inverse UTW transition vectors, respectively, N TR represents transition length of the time-domain window, where 0 ≤ N TR ≤ N/2. Further, here, 1 i× j is an i-by-j matrix with all elements set to one. When N TR = 0, the transmitted signal is identical to that of a CP-OFDM. This study uses a raised cosine window defined by the following equation as the time-domain window: where w m is the m-th component of w TR .
2) RECEIVER a) Overview of the receiver configuration: The block diagram of the conventional simplified UTW-OFDM receiver is shown in Fig. 6. The m-th received symbol sequence, y UTW m , is expressed as follows: After demodulating y UTW m with channel estimation and equalization, the received bit sequence can be obtained through LLR adjustment [18] and channel decoding. b) LLR adjustment technique: The absolute value of the LLR (i.e., the certainty of the code) used for LDPC decoding is inversely proportional to the noise power [22]. Therefore, the absolute value of LLR is considerably large in high SNR environments. However, in the case of simplified UTW-OFDM, ICI is caused by the loss of OFDM orthogonality owing to the application of the time-domain window with a transition length that exceeds the CP length at the transmitter side. Consequently, this ICI results in the absolute value of LLR to be calculated being calculated mainly for the wrong direction. The solution to this problem is LLR adjustment, which normalizes the absolute value of LLR. When the average value of LLR is normalized to m, the adjusted LLR, λ j , is expressed as follows: where λ j is the j-th original LLR. In this article, m is assumed to be 4 [22].

III. PROPOSED SIMPLIFIED UTW-OFDM RECEIVING TECHNIQUE WITH OPTIMAL ICI CANCELLER
In this section, we propose an optimal ICI canceller with noise enhancement suppression to improve the reception quality of 5G systems using simplified UTW-OFDM. Since simplified UTW-OFDM is designed to operate in a short multipath delay environment, ICI due to delayed waves exceeding the CP length does not occur. Furthermore, because frequency synchronization between the base station and user equipment is precisely implemented by using the global navigation satellite system (GNSS) in 5G, it is safe to assume that ICI due to CFO is not generated. However, even in such a situation, ICI is generated by setting a UTW length that exceeds the CP length. The purpose of the proposed method is to suppress the effect of ICI caused by UTW.

A. PROPOSED ICI CANCELLER WITH NOISE ENHANCEMENT SUPPRESSION
The proposed method suppresses the ICI caused by the UTW multiplied at the transmitter side through the multiplication of the ICI cancellation matrix in the frequency domain at the receiver side. The ICI cancellation matrix is generated via the multiplication of the inverse function of the known UTW, obtained based on the minimum mean square error (MMSE) method, through an adjustment weight to reduce the noise enhancement effect. Fig. 7 shows a block diagram of the simplified UTW-OFDM receiver applying the proposed ICI canceller. As the demodulation reference signal (DM-RS), which is the reference signal when performing channel estimation, is also affected by ICI due to the application of UTW, the proposed ICI canceller is applied prior to performing channel estimation and equalization. The m-th received symbol, y IC m ∈ C N×1 , with the proposed ICI canceller is expressed as follows: where P ∈ C N×N is the ICI cancellation matrix, h 0 is the time-domain channel gain of the first arrival wave, H D ∈ C (N+N CP )×(N+N CP ) is the channel matrix with the delayed wave components (i.e., h 0 I N+N CP + H D = H), and n IC m is the effective noise when the ICI canceller is applied to the UTW-OFDM-applied 5G system. Originally, the purpose of multiplying P is to cancel the ICI generated by HW when applying a time window with a transition length that exceeds the CP length. Here, as mentioned in Section II-B, the simplified UTW-OFDM is originally designed for multipath environments with low delay. Therefore, the receiver design in this study suppresses the ICI generated by the combination of UTW and the main path (first arrival wave), which is expected to have a dominant effect on the reception quality. In other words, the P is designed to reduce the ICI effect generated by UTW in the component of h 0 PFDWCF −1 x m in (9). Then the ICI cancellation matrix, P, is calculated by using MMSE as follows: where (·) H denotes the complex conjugate transpose of the matrix in the bracket, N 0 is the noise power spectral density, E s is the energy per one symbol transmission, α is the adjustment factor to mitigate noise enhancement at the OFDM symbol edge (NEOSE), and A −1 is the zero-forcing weight to obtain x m (when α = 1, A = FDWCF −1 ). Although it is difficult to remove the ICI component due to the combination of UTW and delayed wave components with the designed P, the ICI effect suppression can be performed without using the channel estimation results. Therefore, channel estimation using DM-RS can be performed after suppressing the ICI effect. In addition, as mentioned above, since an environment with small delayed waves is assumed, the residual ICI that cannot be suppressed is expected to be small. First, let us consider the case where α is 1 (i.e., no adjustment). Before ICI cancellation. In this case, the amplitude of the OFDM symbol edge is depressed by UTW multiplication at the transmitter side, as shown in Fig. 8(a-1). These symbol edge depressions are amplified via the multiplication of the ICI cancellation matrix at the receiver side, and a flat amplitude characteristic is attempted, as shown in Fig. 8(a-2). However, as shown in (9) and (10), the ICI cancellation matrix is applicable to both the desired signal and the noise components. Therefore, the noise in (10) is over-amplified at the symbol edge (i.e., NEOSE is generated) by W used in (12), as shown in Fig. 10(a-2). Consequently, the communication quality is degraded.
However, α is appropriately adjusted (Fig. 8(b-1)), this NEOSE generation can be suppressed. The smaller the value of α, the greater NEOSE suppression; however, the ICI cancellation performance degrades (Fig. 8(b-2)). Therefore, there exists a trade-off between the reduction of NEOSE and ICI cancellation performance, and it is necessary to adjust α considering these trade-offs.

B. DESIGN OF OPTIMAL NEOSE ADJUSTMENT FACTOR α
Owing to a trade-off relationship between NEOSE generation and ICI cancellation performance, an appropriate α must be designed. Therefore, the relationship between α and SINR after ICI cancellation is formulated. Using the formulation to be developed, the α that achieves the highest SINR is determined as the optimal α.
In the formulation, to calculate SINR, the desired signal components are identical matrices with each modulation subcarrier symbol as one. In addition, additive white Gaussian noise (AWGN) is used as the noise component. The propagation path is assumed to be ideal (i.e., H = I N+N CP ) because the purpose is to express in SINR only the effect of residual ICI and enhanced noise when signal processing is performed to cancel ICI generated by the application of a time window with a transition length that exceeds the CP length. The matrix with desired and interference signal components for each subcarrier, Y α ∈ C N×N , is expressed as follows: where the diagonal component of Y α (y α k, k , 0 ≤ k < N) denotes the desired signal component, and remaining component (y α i, k , 0 ≤ i < N, i = k) denotes the interference component. Therefore, the SINR, β(α), which is defined as a function of α, is expressed as follows: Therefore, if the number of subcarriers to be used is known, the optimal α can be calculated in advance using (14) as follows; α opt = arg max (β (α)) .
Note that these calculations are not performed during communication, but rather the α opt for the UTW length is calculated in advance.

C. COMPUTATIONAL COMPLEXITY OF PROPOSED ICI CANCELLER
The ICI canceller proposed in this article is a technique to cancel ICI (i.e., ICI artificially generated to suppress OOBE) caused by a time window with a transition length longer than the CP length applied at the transmitter side. As can be seen from (11) and (12), the ICI cancellation matrix P is uniquely determined once the type of time window, the transition length, and α are determined and can be calculated in advance. Also, as shown in (9), ICI cancellation is completed simply by multiplying the subcarrier signal (the signal sequence after FFT) obtained by ordinary CP-OFDM by the pre-computed P. The size of this calculation is about the same as that of the FFT, which is a perfectly acceptable increase in computational complexity given the capabilities of recent signal-processing hardware.

IV. EVALUATION BY COMPUTER SIMULATIONS
In this section, we evaluate the proposed receiving method applied to simplified UTW-OFDM using 5G-based link-level computer simulations. First, the optimal α is calculated employing the proposed numerical model (i.e., (14) and (15)) and  compared with the optimal α obtained by the link-level simulation to validate the proposed model and to determine the α to be used in the evaluation. Subsequently, the OOBE suppression performance and BLER characteristics are evaluated assuming L5G applications. The simulation parameters [32], [33], [34], [35], [36] are listed in Table 1. Here, the modulation scheme and coding rate are set according to the modulation and coding scheme (MCS) defined by the third-generation partnership project (3GPP), and the required BLER is set to 10 −1 [36]. The channel model applied is the extended pedestrian-A (EPA) specified by the 3GPP.

A. CALCULATING THE OPTIMAL α
The relationship between α and β(α) in QPSK, 16QAM, and 64QAM calculated based on (14) is shown in Fig. 9. The calculation was performed under the condition of E s /N 0 to achieve BLER = 10 −1 when the ICI canceller with α = 1 was applied to CP-OFDM. In the case of QPSK, for small N TR , the variation in SINR with respect to the variation of α is small. This is because the effect of ICI owing to timedomain window multiplication is small and the improvement in communication quality by LDPC codes is large. However, when N TR is large, the effect of NEOSE generated by the ICI canceller is large. Consequently, the variation in SINR with respect to α is also large. For QAM, the optimal α was close to 1. This is because the E s /N 0 for the calculation was large (i.e., the original noise power is small), and the NEOSE generated by the application of the ICI canceller was also small.  The optimal α obtained using (14) and (15) and that obtained by computer simulation (α with the minimum BLER) are presented in Table 2. The optimal values obtained were equivalent to those obtained through the computer simulations, thus validating (14) and (15). In this study, the ICI canceller applicable for the optimal α for each MCS and N TR is referred to as the optimal ICI canceller. Fig. 10 shows the transmitted signal spectrum of the simplified UTW-OFDM-applied 5G system. The vertical axis represents the averaged power spectral density (PSD) normalized by the maximum value, and the horizontal axis represents the offset frequency from the center frequency. Further, the resolution bandwidth and oversampling factor were set to 100 kHz and 4, respectively. Moreover, N TR = 0 is equivalent to CP-OFDM. Comparisons of the PSD at the channel edge (i.e., offset frequency of ±2.5 MHz) confirmed that the larger N TR yielded the higher OOBE suppression performance. When N TR was set to 256, the OOBE at the channel edge was suppressed by 54.8 dB compared to CP-OFDM. This result is independent of the MCS.   Fig. 11 shows the BLER characteristics of the simplified UTW-OFDM using QPSK, 16QAM, and 64QAM without using the LLR adjustment technique and ICI canceller (i.e., conventional CP-OFDM receiver). For QPSK, BLER = 10 −1 was achieved when 0 ≤ N TR ≤ 128. However, BLER characteristics degraded at the high E s /N 0 environment in the case of N TR = 128, and BLER again increased to a value greater than 10 −1 . The BLER degradation is attributed to the very large LLRs in the wrong direction caused by ICI generated by the UTW. The impact of this problem is particularly large at high MCSs. In the case of 16QAM, BLER = 10 −1 was achieved with N TR ≤ 64. In the case of 64QAM, only N TR = 32 achieved BLER = 10 −1 . However, even in these cases of QAM, the BLER was degraded for high E s /N 0 . Fig. 12 shows the BLER characteristics of the simplified UTW-OFDM using QPSK, 16QAM, and 64QAM with LLR adjustment technique and no ICI canceller. Through the use of LLR adjustment, the BLER degradations in high E s /N 0 environments were improved. Consequently, BLER = 10 -1 was achieved when N TR ≤ 256 for QPSK and when N TR ≤ 64 for 16QAM and 64QAM. In addition, the BLER degradation problems were improved.

2) BLER CHARACTERISTICS WITH LLR ADJUSTMENT AND WITHOUT ICI CANCELLER
These results indicate that upon the application of the simplified UTW-OFDM to 5G, the degradation of the communication quality can be reduced using the LLR adjustment technique in the LDPC decoding [22]. However, when achieving OOBE suppression of 54.8 dB for QPSK and 45.0 dB for 16QAM and 64QAM, respectively, certain degradation of E s /N 0 to achieve BLER = 10 -1 was still observed owing to ICI generated by the time-domain window multiplying. As the impact of the ICI is still not small, the ICI should be reduced at the receiver by the ICI canceller. Fig. 13 shows the BLER characteristics of the simplified UTW-OFDM using QPSK, 16QAM, and 64QAM with the LLR adjustment and proposed optimal ICI canceller. For QPSK, the E s /N 0 value required to achieve BLER = 10 -1 at N TR = 256 was improved by 2.9 dB through the application of the optimal ICI canceller, as shown in Figs. 12(a) and 13(a). Whereas, for 16QAM, the E s /N 0 value required to achieve BLER = 10 -1 at N TR = 64 was improved by 0.3 dB, as shown in Figs. 12(b) and 13(b). Moreover, the BLER characteristics were also significantly improved, as BLER = 10 -1 was achieved when N TR ≥ 128. In the case of 64QAM, the E s /N 0 value required to achieve BLER = 10 -1 at N TR = 64 was improved by 7.0 dB through the application of the optimal ICI canceller, as shown in Figs. 12(c) and 13(c). Furthermore, the BLER characteristics were significantly improved, achieving BLER = 10 -1 when N TR ≥ 128. Fig. 14 shows the relationship between the normalized PSD at the channel edge and the E s /N 0 value required to achieve BLER = 10 −1 . The solid and dashed lines indicate the results with the proposed optimal ICI canceller (i.e., the relationship between Figs. 10 and 13) and without the proposed optimal ICI canceller (i.e., the relationship between Figs. 10 and 12), respectively.

D. OOBE SUPPRESSION PERFORMANCE VS E S /N 0 TO ACHIEVE REQUIRED BLER
In the case of QPSK, the PSD at the channel edge was suppressed by 33.4 dB with nearly identical communication quality as that of CP-OFDM, regardless of the usage of the optimal ICI canceller when N TR ≤ 64. However, when N TR = 128 and 256, upon accepting 1.0 and 5.9 dB degradation of the E s /N 0 to achieve BLER = 10 −1 , respectively, the simplified UTW-OFDM with the optimal ICI canceller improved the communication quality by 0.3 and 3.0 dB, respectively. Simultaneously, the PSD at the channel edge was suppressed by 45.0 and 54.8 dB, respectively, compared to that of CP-OFDM.
In the case of 16QAM, when N TR = 32 and 64, upon accepting 0.3 and 1.4 dB degradation of the E s /N 0 to achieve BLER = 10 −1 , respectively, the simplified UTW-OFDM with the optimal ICI canceller improved the communication quality by 0.2 and 0.4 dB, respectively. Simultaneously, the PSD at the channel edge was suppressed by 19.5 dB and 33.4 dB, respectively, compared to that of CP-OFDM. Moreover, when N TR = 128 and 256, BLER = 10 −1 was achieved via the application of the optimal ICI canceller. The simplified UTW-OFDM suppressed the PSD at the channel edge by 45.0 and 54.8 dB, respectively, compared to that of CP-OFDM; in this case, the E s /N 0 degradation values were 4.9 and 14.1 dB, respectively.
In the case of 64QAM, when N TR = 32 and 64, upon accepting 1.5 and 3.8 dB degradation of the E s /N 0 to achieve BLER = 10 −1 , respectively, the simplified UTW-OFDM with the optimal ICI canceller improved the communication quality by 0.5 and 0.7 dB, respectively. Further, the PSD at the channel edge was suppressed by 19.5 and 33.4 dB, respectively, compared to that of CP-OFDM. Moreover, when N TR = 128 and 256, BLER = 10 −1 was achieved by applying the optimal ICI canceller, with the simplified UTW-OFDM suppressing the PSD at the channel edge by 45.0 dB and 54.8 dB, respectively, compared to that of CP-OFDM; the E s /N 0 degradation values were 10.9 and 21.6 dB, respectively.

V. EXPERIMENTAL EVALUATION BY USING SOFTWARE-DEFINED RADIO PLATFORM
In this section, the simplified UTW-OFDM and the proposed receiving technique, whose effectiveness in 5G was demonstrated via computer simulations in the previous section, are implemented using an SDR-based evaluation platform [28]. Consequently, their basic characteristics are experimentally evaluated. First, the configuration of the transmitter and receiver in the constructed experimental evaluation system and the configuration of the entire system are described. Thereafter. the simplified UTW-OFDM and the proposed receiving technique were experimentally evaluated using the constructed evaluation system, focusing on the OOBE suppression performance and BLER characteristics. Fig. 15 shows the SDR-based transmitter configuration. The transmitter comprised a chassis with a high-speed communication bus (NI PXIe-1082), an RF module (NI 5791), an FPGA module (NI PXIe-7975R), a timing module (NI PXIe-6674T), and an embedded controller (PXIe-8861). The synchronization module generated a high-precision clock, and all modules shared the same operating clock. First, 5G-based baseband signals modulated by the simplified UTW-OFDM were generated using a signal processing software (NI Lab-VIEW) in the embedded controller, and 16-bit I/Q digital signals were written to the FPGA module. Subsequently, the signal was converted to an analog signal using the analog-todigital converter (ADC) and up-converted to an RF signal in the RF module. Fig. 16 shows the SDR-based receiver configuration. The basic configuration was identical to that of the transmitter. First, the received RF analog signal was down-converted to  baseband signal in the transceiver module. Subsequently, the baseband signal was sampled via a digital-to-analog converter (DAC) to a 14-bit I/Q digital signal in the FPGA module. The sampled digital I/Q signal was then passed to the embedded controller. Thereafter, the demodulation process of the received baseband signal was performed in the signal processing software (NI LabVIEW). Further, receiver diversity in the demodulation process was achieved through the combination of the received signals obtained at different times to perform pseudo-maximum ratio combining (MRC). Fig. 17 shows the developed experimental evaluation platform and its system configuration. The transmitted downlink (DL) signal was input to the fading emulator (FE, NJZ-1600D, JRC). The FE output and AWGN signals generated by the signal generator (SG, MS2830A, Anritsu) were input to a combiner (PD510-OS, R&K). Consequently, the combined signal is input to the SDR-based receiver. The operating clock was shared by the transmitter and receiver by connecting a synchronization module via an RF cable. Therefore, the experiment was conducted assuming that the frequency offsets did not influence the outcome. The FE used in this experiment supported frequency bands up to 3 GHz and bandwidth up to 20 MHz. In addition, the upper limit of the number of paths that could be output in one channel was 6 waves.

D. EVALUATION
The parameters of the baseband signal in this evaluation were the same as those presented in Table 1, whereas the parameters specific to the SDR-based evaluation platform are presented in Table 3. Because the FE specification limits the operating frequency, the center frequency was set to 2.15 GHz, and the moving speed was applied such that the fading was equivalent to the frequency in the evaluation in Section IV (i.e., 4.7 GHz). This is valid because when modulating and demodulating baseband signals in a multipath fading environment, the only effect of differences in the frequency band is the relationship between moving speed and Doppler frequency. Consequently, as the FE could only be set up to 6 paths, the EPA model with the 7th path removed was used as the channel model. Fig. 18 shows the transmitted signal spectrum of the simplified UTW-OFDM-applied 5G system obtained through experimental evaluation. Through comparisons of the PSD at the channel edge it was confirmed that larger N TR yielded a higher OOBE suppression performance when N TR ≤ 128. When N TR was set to 128, the OOBE at the channel edge was suppressed by 23.5 dB compared to CP-OFDM. Further, compared to the results of the computer simulation (Fig. 10), the noise floor was increased and observed at approximately -50 dB in the case of the experiment. This is because of the quantization noise from the transmitter's DAC and the receiver's ADC. Moreover, the PSD at the channel edge was degraded at N TR = 256 compared to N TR = 128, where the waveform shape exerted a larger effect.

2) BLER CHARACTERISTICS WITH LLR ADJUSTMENT AND WITHOUT PROPOSED ICI CANCELLER
The BLER characteristics of QPSK and 64QAM with LLR adjustment and no proposed ICI canceller are shown in Fig. 19(a) and (b), respectively. In Fig. 19, the solid and dashed lines represent the BLER obtained by the SDR-based experiment and computer simulation, respectively. The experimental results show that BLER = 10 −1 was achieved for all N TR conditions where BLER = 10 −1 was achieved in the computer simulation. Thus, the essential characteristics of the 5G system with simplified UTW-OFDM were successfully verified on the actual RF device. However, compared to the computer simulation results, the SNR to achieve BLER = 10 −1 is degraded by 0.8 dB at worst, and the slope of the curve was slightly different for N TR = 64 in the case of 64QAM. This can be attributed to the difference owing to the aforementioned FE setting limits.

3) BLER CHARACTERISTICS WITH PROPOSED ICI CANCELLER
The BLER characteristics of QPSK and 64QAM with the LLR adjustment and the proposed optimal ICI canceller are shown in Fig. 20(a) and (b), respectively. The effectiveness of the proposed optimal ICI canceller can also be confirmed in the experimental evaluation. In the case of QPSK, a comparison of Figs. 19(a) and 20(a) shows that the application of the optimal ICI canceller improved the SNR to achieve BLER = 10 −1 for all N TR . For example, it was improved by 0.2 dB FIGURE 20. Experimentally evaluated BLER characteristics with LLR adjustment and proposed optimal ICI canceller.
at N TR = 128. In the case of 64QAM, Figs. 19(b) and 20(b) show that the SNR to achieve BLER = 10 −1 was improved by 4.7 dB at N TR = 64 owing to the application of the optimal ICI canceller. Moreover, the BLER characteristics were significantly improved at N TR ≥ 128, achieving BLER = 10 −1 .
These results indicate that the proposed method is effective for the actual experiment with the SDR.

VI. CONCLUSION
This study proposed a 5G system that applies simplified UTW-OFDM and a receiver-side signal processing technique that mitigates the degradation of receive quality due to the effects of UTW applied at the transmitter side. First, we investigated the LLR adjustment effect in the simplified UTW-OFDM applied 5G system. In addition, an ICI canceller based on the scheme proposed with additional improvements was developed and applied to suppress noise enhancement effects. Computer simulations and experimental studies using SDR evaluated these proposed receiving technologies for the simplified UTW-OFDM-utilized 5G. Computer simulations showed that simplified UTW-OFDM suppressed PSD at the channel edge by up to 54.8 dB compared to CP-OFDM. Furthermore, by using the LLR adjustment technique and the proposed optimal ICI canceller, the problem of BLER degradation at high SNR was solved, and BLER = 10 -1 was achieved even when applying the maximum value of N TR = 256 with MCS = 18. Further, experiments with the SDR system confirmed that the OOBE suppression was sufficiently effective, although the OOBE suppression performance was degraded owing to the quantization error. The proposed optimal ICI canceller was also shown to improve the BLER characteristics in SDR experiments and to be applicable to time-domain windows with long transition lengths.