I. INTRODUCTION
THE initial aim of a communication system is solely to transmit data in a sufficient and reliable manner. This view has been changed via the advancement of wireless communication and heavy data-driven services. The recent 5G deployment and technical roadmap for beyond 5G and 6G showed that wireless communication systems should have additional functionality and purposes instead of simple data transfer [1]. Since beyond 5G and 6G are on their way, various communication and signal processing schemes are possible candidates for becoming part of new wireless communication systems.
One promising scheme that paid significant attention is the in-band full-duplex (IBFD) system. This system can substantially increase spectral efficiency with simultaneous transmission and reception on the same frequency band. The IBFD can also significantly improve a wide range of applications and interests, including network capacity and throughput, latency in bidirectional communication, spectrum utilization, relay network performance, and efficient cognitive radio systems. With this significance, IBFD becomes a possible key enabler for emerging applications such as Integrated Sensing and Communications (ISAC) systems aiming to achieve both sensing and data transmission simultaneously [2]. In ISAC scenarios, such as autonomous driving, IBFD can facilitate simultaneous sensing of the environment and high-speed data transmission for vehicle localization and high data transmission, improving overall system efficiency [2], [3].
While the IBFD system can resolve various issues and utilize the entire capacity via simultaneous transmitting and receiving, significant self-interference power from the transmit signal causes the low signal-to-interference-plus-noise ratio (SINR) issue in the IBFD system resulting in severe degradation in receiver performance [4]. Especially considering a case of mobility as a service (MaaS) scenario where multiple small mobility devices are present in the urban or indoor environments, IBFD systems in multiple mobility devices can cause these particular low SINR issues due to doubling multiple interference signals with significant high power presence on the spectrum, as shown in Fig. 1. Several solutions were presented to mitigate the low SINR case at the IBFD system. Especially considering the low SINR of the data symbols caused by the self-interference from the transmit signal, self-interference (SI) cancellation schemes have been proposed to improve SINR in the IBFD system [5]–[7].
Possible mobility-as-a-service scenario with multiple IBFD-equipped mobility devices
The existing literature also proposes possible solutions for self-interference cancellation and the IBFD system using a wide range of approaches. For instance, the analog circuit approach using balanced feed networks can shift the phase angle at the antenna for self-interference cancellation so that the transmit and receive can be performed at one antenna [8]. Learning approaches using convolutional neural network (CNN) and simple neural networks were also proposed to minimize self-interference and noises in intertower communication [9], [10]. In the beamforming case, the multi-layer perception method offers the design of the beamforming for the IBFD system to remove self-interference and co-channel interference [11]. Even in the photonic field, self-interference cancellation is suggested as the main critical component to improve the IBFD system performance via circuit-based cancellers and optical processing [12], [13].
Conventional SI cancellation schemes are assumed to have the ideal channel estimation for significant performance gain [5]–[7]. However, when we consider applying a channel estimator to SI cancellation in real condition cases, the performance gain of SI cancellation is much smaller than the ideal case due to the poor-quality channel and parameter estimations at the IBFD system [14]. Improving the performance of the SI cancellation receiver requires accurate channel estimation since the performance of SI cancellation is highly dependent on the qualities of channel and parameter estimations. While existing methods cannot accurately estimate the channel condition with the presence of significant self-interference, the successive interference cancellation (SIC) approach iteratively estimates the signal components to cancel the self-interference effects. This approach is a possible solution to achieve significantly better channel estimation quality than other channel estimation methods, especially in the high interference scenarios typical of IBFD systems.
In this paper, we present SIC-based channel estimation for the enhancement of channel estimation quality at the IBFD system. Our proposed SIC-based method enables the SI cancellation to operate closer to its ideal performance, significantly improving the overall system throughput and spectral efficiency of the IBFD system. Since our proposed method recursively estimates the channel via reference signals for channel accuracy, the overall performance of SI cancellation can be improved using enhanced channel estimation.
The rest of the paper is organized as follows: Section II describes A system model of the SIC-based channel estimation. In Section III, we explain the SI cancellation procedure using SIC-based channel estimation. Section IV presents the simulation environments and results of our proposed SI cancellation using the SIC-based channel estimation. In Section V, we conclude this paper.
II. SYSTEM MODEL
In this section, we present a system model based on an NR system [17] with IBFD communication systems due to their recent popularity. IBFD systems can transmit and receive signals simultaneously at the same time and frequency resources. Depending on the method of antenna connection between the transmit radio frequency (RF) chain and the receive RF chain, the types of IBFD can be divided into an antenna sharing method and an antenna separation method.
This antenna-sharing method requires a circuit to separate the received and transmitted signals. In the conventional frequency division duplexing (FDD)-based communication system, it was possible to distinguish between a received signal and a transmitted signal in an antenna-sharing method through filtering technologies such as a duplexer. However, since the IBFD transmits and receives simultaneously in the same frequency band, it is impossible to distinguish between the received and the transmitted signals with the conventional duplexer device, which requires the SI cancellation method.
Figs. 2 and 3 show the system model composed of two IBFD systems with different antenna configurations [14]. Especially Fig. 2 shows how to configure the IBFD of the antenna-sharing method using a circulator with three input/output ports. In the radar system, the circulator element is used to restrict the direct flow of the transmit signal from the transmit RF chain to the receive RF chain. Therefore, when applied to the IBFD system of the antenna-sharing method using this characteristic of the circulator, this system can limit the flow of the transmission signal into the reception RF chain. However, the performance of the conventional circulator is much smaller than in the ideal case.
Fig. 3 shows the IBFD of the antenna separation method. This method physically separates a transmitting antenna from a receiving antenna, reducing the amount of received SI due to the frequency attenuation of the radio waves. However, it is impractical to use in an environment with spatial restrictions such as user equipment (UE) because SI remains regardless of its own reduction.
IBFD System with RF chain by antenna sharing.
IBFD System with RF chain by antenna separation.
Note that, in our system model, we assumed that the transmission timings of uplink (UL) and downlink (DL) are aligned. Time alignment of UL and DL offers several operational and performance advantages [15], [16]. In real deployments, multiple base stations are already deployed after aligning the timings of UL and DL by timing advance. Therefore, we investigated SIC-based channel estimation in an NR system under time-aligned environments for an ISAC system.
In an existing NR system, to avoid the collision of demodulation reference signals (DMRS), it is usual that UL and DL DMRS use different symbol time shift values. However, our method proposed to align the positions of the DMRSs in the same symbol time with the SIC-based channel estimation for better channel estimation quality. Under this setting, the DMRSs of UL and DL are received in the same resource element. Therefore, in that specific resource element, it is possible to assume that only DMRSs from UL and DL and additive white Gaussian noise (AWGN) exist.
Since neither method can remove SI completely, as illustrated previously, SI impacts a receive RF chain [18]. Each transmitter sends DMRSs for reliable channel estimation; in the conventional NR, it is usual to use different symbol time shift values for DMRSs from UL and DL caused by using a time division duplexing (TDD)-based communication system.
In SIC operation, we also consider synchronizing the UL and DL transmission timings so that all OFDM symbols are time-aligned at the receiver and the DMRSs are received at the same resource elements. Furthermore, the cell IDs are determined to align DMRS symbols in frequency and time domains.
For a simple description of the proposed channel estimation algorithm, let us consider a simple scenario with one desired signal and only one SI. In this case, the received signal [TeX:] $$y(t)$$ at time t in a receiver can be:
where [TeX:] $$h_d(t) \text { and } h_i(t)$$ are the channel of the desired signal and the SI at time t, respectively. [TeX:] $$x_d(t) \text { and } x_i(t)$$ are the desired signal and the SI at time t respectively. [TeX:] $$n(t)$$ is AWGN.
As mentioned in the previous section, our IBFD system considered the heavy mobility cases, including mobility services, that contain various scenes of mobility devices, vehicles, and pedestrians. In other words, we assume that our UE can be non-stationary under various channel conditions, including Rayleigh fading and EPA for IBFD system design.
III. SI CANCELLATION USING SIC-BASED CHANNEL ESTIMATION
A. Receiver Structure
For this section, we present our receiver structure model based on an NR system [17], as shown in Fig. 4. After a fast Fourier transform (FFT), the receiver estimates the block, noise, and channel responses. The receiver applies SI cancellation to mitigate the SI using SIC-based channel estimation. In addition, the receiver demodulates the signal in a MIMO processing & Demodulation block and decodes it in a Decoding & De-rate matching block. Since the SI cancellation performance is highly dependent on channel estimation accuracy, the receiver can achieve accurate channel estimate if the SI included in the DMRS is lowered via SIC. The performance of SI cancellation can be further improved using accurate channel estimation.
Receiver structure of IBFD with SIC-based channel estimation.
B. Channel Estimation
Fig. 5 shows the block for channel estimation. Channel estimation consists of cross-producting the received signal and the reference signal (RS), frequency filtering, and interpolation.
1) Mapping to physical resource of DMRS: The UE shall assume the DMRS being mapped to physical resources according to configuration type 1. The UE shall assume the RS sequence [TeX:] $$R(s)$$ is scaled by a factor [TeX:] $$\beta^{D M R S}$$ to conform with the transmission power specified in [21], where [TeX:] $$\beta^{D M R S}=1$$ is assumed. Mapped to resource elements [TeX:] $$X(k, l)$$ located in the kth frequency domain and lth time domain according to
where [TeX:] $$k=4 n+2 k^{\prime}+\Delta, \Delta=0,1, k^{\prime}=0,1, n=0,1, \cdots, N / 2 \text{ and } N$$ is the number of subcarriers.
Block diagram of SIC-based channel estimation.
Fig. 6 shows the mapping of the DMRS. If is 0, DMRS is mapped at two intervals starting from the 0th element of the frequency domain of the resource grid. Otherwise, it is mapped starting from the 1st element of that. The elements between which the DMRS is mapped are not used. In the time domain of the resource grid, the DMRS is mapped to the lth element, and the DMRS is mapped to the second element in general.
Mapping of the DMRS to the resource grid for channel estimation when using two antenna ports.
2) Filtering on the frequency domain: In channel estimation, the next step after the cross product [TeX:] $$\bar{R}(k)=R^{\prime}(k) \times R^*(k)$$ of the RS from the received signal [TeX:] $$R^{\prime}(k) \text { and } \operatorname{RS} R(k)$$ by a predefined standard is filtering on the frequency domain. Fig. 7 shows the RS locations along the frequency domain of the resource block. The channel response for the resource block is estimated by extracting only the RS from the received signal according to a predefined standard, as described in the previous section. The number of RSs used on both sides of the RS is to be estimated depending on the delay profile. The channel response [TeX:] $$\hat{R}(k)$$ is estimated as
where, [TeX:] $$N_\mu$$ is the number of received RSs to be used for channel estimation, and 1, 2, 4, 8, or 16 are used depending on the channel status. [TeX:] $$\alpha_f$$ is a constant value depending on the location of the received RS on the frequency domain, 0, 1, 1/2, 1/4 or 1/8 are used. [TeX:] $$\beta=1+2\left(\alpha_f \times N_\mu\right).$$
Location of the RS along the frequency domain of the resource block for channel estimation.
3) Filtering on the frequency domain: We apply interpolation to estimate the remaining part based on the channel estimate values of the RS location. Interpolation method include Lagrange interpolation method and linear interpolation method. Given two points [TeX:] $$\left(w_0, l_0\right) \text { and }\left(w_1, l_1\right),$$ the Lagrange interpolation formula to find the value of l by w is as
However, the performance of the Lagrange interpolation method is degraded due to the nonlinearity of Lagrange interpolation at low signal-to-noise ratios (SNR). On the other hand, the performance gain at high SNR is only slightly larger than that of the linear interpolation method. Since we need to investigate with the detailed comparison over various methods, we also focus on linear interpolation method in our paper. We consider linear interpolation as equivalent to the case for linear functions of Lagrange interpolation above.
C. Successive Interference Cancellation
SIC is employed in communication systems, particularly in scenarios where multiple signals interfere. The primary goal of SIC is to mitigate the impact of interference and improve the overall system performance, especially in scenarios similar to wireless communication where signals can overlap.
In a communication system, multiple signals are transmitted over the same channel or frequency spectrum. These signals may interfere with each other, leading to a degradation in the quality of received signals. The received signals, which include both the desired signal and interfering signals, are received by the receiver. These signals are typically received in a particular order or priority.
The interference cancellation process begins with decoding or demodulating the signal with the highest priority, usually depending on the SINR or decoding complexity levels. The decoded signal is then subtracted or canceled from the overall received signal. The above step is repeated. After the cancellation of the first signal, the remaining signals are processed again in the same manner.
This process continues until all the significant interference has been canceled or until a certain predefined threshold is reached. The final decoding is complete on the remaining signals, which now experience reduced interference due to the successive cancellation process.
SIC is effective when dealing with scenarios where signals have varying strengths or when the interference pattern is known. It exploits the fact that not all signals interfere equally with each other. By canceling the more dominant interference first, the subsequent signals can be more accurately decoded.
While SIC can significantly improve the performance of communication systems in the presence of interference, it is essential to note that it introduces complexity to the receiver, especially in terms of processing power and latency. The success of SIC depends on the accuracy of interference estimation and the order in which signals are processed.
The diagram of the SIC is in Fig. 8, where the SIC consists of [TeX:] $$N_{\text {total }}=e \times m$$ interference cancellation units (ICUs) that have a predetermined number indicating the maximum iteration, arranged in a multistage structure. The basic ICU is in Fig. 9.
Multistage structure of the SIC process.
D. Proposed SIC-based Channel Estimation
The overall channel estimation is composed of multiple channel estimations and SICs from Step 1 to Step 5 as follows: Step 1: A variable m, indicating the number of performed iterations is initialized to zero.
Step 2: The receiver determines the signal with a stronger channel response between a desired signal and a SI. After operating K-point FFT, the receiver composes the m-th estimated channel response [TeX:] $$\hat{H}_d^{(m)}(k)$$ of the desired signal and [TeX:] $$\hat{H}_i^{(m)}(k)$$ of the transmit self-interference signal as
where [TeX:] $$\operatorname{LPF}(\cdot)$$ is a low pass filter used for a component channel estimation, [TeX:] $$Y(k)$$ is the frequency domain representation of the received signal [TeX:] $$Y(t)$$ and [TeX:] $$R_d(k) \text{ and } R_i(k)$$ are the DMRSs of the desired signal and the transmit self-interference signal, respectively. Then, m is increased by one.
The power [TeX:] $$P_d \text { and } P_i$$ of the channel estimation response [TeX:] $$\hat{H}_d(k) \text { and } \hat{H}_i(k)$$ are obtained, respectively.
Go to step 3 if the received power of the desired signal [TeX:] $$P_d$$ is larger than that of the SI [TeX:] $$P_i.$$ Otherwise, step 4.
Step 3: The receiver reduces the desired signal by using [TeX:] $$\hat{H}_d^{(m)}(k)$$ of (5) and new signal [TeX:] $$\hat{Y}_i(k)$$ is received as
The receiver composes a first channel response estimate with m = 1 of SI using [TeX:] $$\hat{Y}_i(k)$$ of (9) as
The more accurate channel response of the SI [TeX:] $$\hat{H}_i^{(m)}(k)$$ can be obtained since the power of the desired signal [TeX:] $$H_d(k) R_d(k)$$ is reduced. Then, go to step 4.
Step 4: After increasing m by one and stop the channel estimation, if m reaches [TeX:] $$N_{\text {total }}.$$ Otherwise, go to step 5.
Step 5: The receiver reduces the SI by using [TeX:] $$\hat{H}_i^{(m-1)}(k)$$ and new received signal [TeX:] $$\hat{Y}_d(k)$$ is obtained as
The receiver composes the channel response estimation of the desired signal using [TeX:] $$\hat{Y}_d(k)$$ of (11) as
Then, m is increased by one. Next, the receiver reduces power by using [TeX:] $$\hat{H}_d^{(m-1)}(k)$$ of (12) and [TeX:] $$\hat{Y}_i(k)$$ is received as
The receiver composes the channel response estimation of the SI using [TeX:] $$\hat{Y}_i$$ of (13) as
Then, go to step 4 after increasing m by one.
The overall channel estimation consists of several component channel estimations and SIC. A trade-off between performance and complexity depends on the number of component channel estimations used for the overall channel estimation. Based on the above discussion, we summarized the proposed SIC-based channel estimation algorithm in Algorithm 1.
SIC-based Channel Estimation Algorithm for SI Cancellation in IBFD Systems
E. SI Cancellation
SI cancellation can be represented by only one ICU, as shown in Fig. 9. SI cancellation calculates the interference value by multiplying the final estimated channel response obtained by SIC in (14) with the transmitted signal in the frequency domain known to the receiver instead of the reference signal in Fig. 9, and the received signal [TeX:] $$\tilde{Y}_d(k)$$ with this value removed is obtained as
As shown in Fig. 8, by repeating this operation several times, the SIC method can remove much of the interference [7]. However, as described in the previous section, the trade-off between performance and complexity depends on the number of component ICUs used for the overall SI cancellation.
IV. SIMULATION RESULTS
Given the system model and details of SI Cancellation, we provide simulation results with the proposed SI cancellation receiver using SIC-based channel estimation. To evaluate the performance of the SI cancellation receiver in the practical environment, we implemented an NR link level simulator, complying with NR standards [17], [19]–[21].
In our simulator, we designed the interference signal that we modified cell ID in one of its blocks to transmit desired signal so that this ID can contain DMRS including the difference sequence signal. After producing the interference signal, we implemented in a way that both desired signal and interference signal transmitted in each fading channel and added AWGN noise. Since the IBFD system receives a desired signal with a large amount of SI, the simulator applies QPSK modulation and a low-rate, low-density parity check code (LDPC) code. In this simulation, we applied ideal finite impulse response (FIR) low pass filters over time and frequency domains for component channel estimation. The receiver has two receiver antennas and transmits data with transmit diversity mode. The center frequency of transmission and reception is set to 3.5 GHz. The speed of the receiver is set to 0 km/h, 30 km/h, 60 km/h, and 90 km/h, respectively. Also, we select the four channels including two path fading channels with delay [TeX:] $$1 \mu s \text { and } 5 \mu s$$ based on the Extended Typical Urban model (ETU) [22]. Table I shows the parameters used for the simulation.
SIMULATION PARAMETERS FOR EVALUATING THE PERFORMANCE OF SI CANCELLATION WITH SIC-BASED CHANNEL ESTIMATION IN IBFD SYSTEMS.
The SINR is defined for a particular receiver as
where [TeX:] $$N_0$$ is the power of AWGN.
Since the main purpose of this paper aims to evaluate the performance of self-interference cancellation via interference, we considered signal-to-interference ratio (SIR) as the main parameter with high SNR for the accurate channel estimation. For high SNR, we assumed that SNR is 10 dB as equivalent to the QPSK bit error rate (BER) with [TeX:] $$10^{-5}.$$ Thus, our [TeX:] $$\operatorname{SNR}\left(P_d / N_0\right)$$ is assumed to be 10 dB. [23]
The self-interference and noise ratio (INR) is defined as [24]
The SI cancellation ratio (SICR) is defined as [7]
Figs. 10, 11, and 12 compare the mean square error (MSE) performances of the conventional channel estimation and the proposed SIC-based channel estimation under a path Rayleigh fading channel, an two equal path Rayleigh fading channel with [TeX:] $$1 \mu s \text { and } 5 \mu s$$ delay and EPA channel, respectively.
As shown in Fig. 10, the MSE performance of the SIC-based channel estimation under one Rayleigh fading case is about 1/14 of the conventional channel estimation performance at −4 dB SINR. Also, we observe that the MSE of the SIC-based case preserves under 0.1. This shows that SIC can mitigate substantial damages caused by one-path fading channel.
In Fig. 11, the MSE performance of the SIC-based channel estimation is about one-third of the conventional channel estimation performance at −4 dB SINR. Also, the result showed that the MSE correction of SIC is worse than one path fading channel because the MSE of the conventional case does not reduce well over SINR compared to one path fading case. It indicates that SIC can improve different types of fading channel cases, but the improvement result depends on the types of fading channels and environments.
MSE performance comparison of conventional channel estimation and proposed SIC-based channel estimation in an one path Rayleigh fading channel.
MSE performance comparison of conventional channel estimation and proposed SIC-based channel estimation in an equal-gain two path Rayleigh fading channel with [TeX:] $$1 \mu d$$ tap delay.
In Fig. 12, the MSE performance of the SIC-based channel estimation is about one-third of the conventional channel estimation performance at −4 dB SINR. Since the length of the cyclic prefix excluding the 0th symbol is about [TeX:] $$4.7 \mu s$$ based on the subcarrier spacing of 15 kHz, the equal gain [TeX:] $$5 \mu s$$ delay Rayleigh fading channel affects the next symbol by about [TeX:] $$0.3 \mu s$$. Since the minimum chip time is 32.55 ns, 300 ns is only about 10 samples. Therefore, the performance of the equal gain [TeX:] $$5 \mu s$$ delay Rayleigh fading channel is similar to or slightly worse than that of [TeX:] $$1 \mu s$$ delay.
MSE performance comparison of conventional channel estimation and proposed SIC-based channel estimation in an equal-gain two path Rayleigh fading channel with [TeX:] $$5 \mu s$$ tap delay.
In Fig. 13, the MSE performance of the presented SIC-based channel estimation is one-tenth of the conventional channel estimation performance at −4 dB SINR. This result also follows a similar description we provided previously, which states that the fading channel condition is one of the significant factors for channel estimation and SIC performance. However, we can see that the one-path Rayleigh and EPA cases, which are not heavy fading cases, can archive substantial performance improvements from SIC. Even in the two-path cases, the improvement can be possible but slightly less than in the other two cases.
MSE performance comparison of conventional channel estimation and proposed SIC-based channel estimation in an EPA channel.
In Fig. 14, it can be seen that as the speed increases, the performance decreases because the Doppler frequency increases. When the speed of the UE is 60 km/h, the MSE performance of the presented SIC-based channel estimation is one-seventh of the conventional channel estimation performance at −4 dB SINR. In the SINR −1 to 0 dB range, the power of the interference is similar to the power of the desired signal, and the performance of SIC-based channel estimation is greatly reduced due to the influence of the Doppler frequency. Therefore, our observation can explain that the performance of MSE of conventional channel estimation and MSE of SICbased channel estimation are reversed.
MSE performance comparison of conventional channel estimation and proposed SIC-based channel estimation in an one path Rayleigh fading channel.
In Fig. 15, it can be seen that as the speed increases, the performance decreases because the Doppler frequency increases. When the receiver speed is 60 km/h, the MSE performance of the presented SIC-based channel estimation is one-third of the conventional channel estimation performance at −4 dB SINR.
MSE performance comparison of conventional channel estimation and proposed SIC-based channel estimation in an equal-gain two path Rayleigh fading channel with [TeX:] $$5 \mu s$$ tap delay.
To demonstrate the function of the SIC-based channel estimation, Fig. 16 presents the performance of the SI cancellation of the proposed SIC-based channel estimation under various SINRs with or without the SIC-based channel estimation. With SIC-based channel channel estimation, it can be seen that the SI can be canceled by about 5 dB at an INR value of 36 dB. However, with conventional channel estimation, the performance of SI cancellation shows a SINR value of 3 dB at an INR value of 36 dB. For a SINR value of 1 dB, compared with the SI cancellation receiver using conventional channel estimation, achieving about 3dB performance gains with an SI cancellation receiver using the SIC-based channel estimation is possible. The proposed SI cancellation with the SIC-based channel estimation shows a performance close to that of SI cancellation with an ideal channel (with perfect channel and parameter estimation) within 0.5 dB performance difference at an SINR value of 5 dB.
SINR performance comparison of SI cancellation with the existence of the proposed SIC-based channel estimation.
Data acquisition from IoTs to MEC server using a UAV.
Figs. 17, 18, 19, and 20 compare the NR downlink BER for several receivers for a one-path Rayleigh fading channel, an equal-gain two path Rayleigh fading channel with [TeX:] $$1 \mu s \text{ and } 5 \mu s$$ tap delay, and EPA channels, respectively. For comparison, we measure the BER performances of receivers, including the conventional receiver, a receiver with SIC-based channel estimation (without SI cancellation), an SI cancellation receiver with conventional channel estimation, and an SI cancellation receiver with SIC-based channel estimation. We also compared them to an ideal SI cancellation receiver with perfect channel and parameter estimation.
As shown in Fig. 17 with a path Rayleigh fading channel, we observed that the BER performance of our proposed system can achieve about 0.75 dB performance gains with an SI cancellation receiver and SIC-based channel estimation as compared to the conventional channel estimation receiver at [TeX:] $$\text { BER }=10^{-1}.$$ It also shows that our proposed system’s performance is close to that of an ideal SI cancellation receiver (with perfect channel and parameter estimation) within a 1 dB performance difference.
BER performance comparison of conventional receiver and and SI cancellation receiver over conventional channel estimation, SIC-based channel estimation and ideal channel estimation in an one path Rayleigh fading channel.
In Fig. 18, using an equal-gain two path Rayleigh channel with [TeX:] $$1 \mu s$$ tap delay, we observed that the BER performance of our proposed system can achieve about 2.5 dB performance gains with an SI cancellation receiver and SIC-based channel estimation as compared to the conventional channel estimation receiver at [TeX:] $$\text { BER }=10^{-2}.$$ This shows that our proposed system’s performance is close to that of an ideal SI cancellation receiver within a 1 dB performance difference.
BER performance comparison of conventional receiver and and SI cancellation receiver over conventional channel estimation, SIC-based channel estimation and ideal channel estimation in an equal-gain two path Rayleigh fading channel with [TeX:] $$1 \mu s$$ tap delay.
In Fig. 19, using an equal-gain two path Rayleigh channel with [TeX:] $$5 \mu s$$ tap delay, we observed that the BER performance of our proposed system can achieve about 2.5 dB performance gains with an SI cancellation receiver and SIC-based channel estimation as compared to the conventional channel estimation receiver at [TeX:] $$\text { BER }=10^{-1}.$$ This shows that our proposed system’s performance is close to that of an ideal SI cancellation receiver within a 1 dB performance difference.
BER performance comparison of conventional receiver and and SI cancellation receiver over conventional channel estimation, SIC-based channel estimation and ideal channel estimation in an equal-gain two path Rayleigh fading channel [TeX:] $$5 \mu s$$ delay.
Fig. 20 compares the BER performance in an EPA channel. For a BER value of 0.6, compared with the conventional receiver, we can achieve about 2 dB performance gains with a receiver using an SI cancellation and SIC-based channel estimation. The proposed SI cancellation receiver with SICbased channel estimation shows a performance close to that of an ideal SI cancellation receiver within a 1 dB performance difference.
BER performance comparison of conventional receiver with and without SIC-based channel estimation, and SI cancellation receiver with conventional channel estimation, SIC-based channel estimation and ideal channel estimation in an EPA fading channel.
The BER values of the EPA channel for Fig. 20 are shown in Table II.
FOR FIG. 20, BER PERFORMANCE VALUE CONVENTIONAL RECEIVER AND SI CANCELLATION RECEIVER WITH CONVENTIONAL CHANNEL ESTIMATION, SIC-BASED CHANNEL ESTIMATION AND IDEAL CHANNEL ESTIMATION IN EPA CHANNEL.
As shown in Fig. 21 by the speed of the UE with a one-path Rayleigh fading channel, it can be seen that as the speed increases, the performance decreases because the Doppler frequency increases. When the speed of the UE is 60 km/h, we observed that the BER performance of our proposed system could achieve about 4.25 dB performance gains with a SI cancellation receiver and SIC-based channel estimation compared to the conventional channel estimation receiver at BER = 0.2.
In an one path Rayleigh fading channel, BER performance comparison of SI cancellation with and without the proposed SIC-based channel estimation for UE speeds of 30 km/h, 60 km/h, and 90 km/h, respectively.
The BER values of a one-path Rayleigh fading channel by a speed of UE for Fig. 21 are shown in Table III. When comparing processing speeds of 0 km/h and 90 km/h in Table III, we observed that UE speed becomes an important factor in performance degradation for both the conventional channel estimation and proposed SIC-based channel estimation. For both the conventional method and the proposed SIC-based method, speed and performance are inversely proportional. It can be seen that as the SINR increases when the UE moves, the performance gap between the conventional method and the proposed SIC-based method decreases
FOR FIG. 21, BER PERFORMANCE VALUE OF SI CANCELLATION WITH AND WITHOUT THE PROPOSED SIC-BASED CHANNEL ESTIMATION IN AN ONE PATH RAYLEIGH FADING CHANNEL.
As shown in Fig. 22, according to a speed with an equal gain two-path Rayleigh fading channel, it can be seen that as the speed increases, the performance decreases because the Doppler frequency increases. When the speed of the UE is 60 km/h, we observed that the BER performance of the presented SIC-based channel estimation is two-third of the conventional channel estimation performance at −4 dB SINR.
In an equal-gain two path Rayleigh fading channel with [TeX:] $$1 \mu s$$ tap delay, BER performance comparison of SI cancellation with and without the proposed SIC-based channel estimation.
For comparison with one path Rayleigh fading channel case, we listed the result of an equal-gain two path Rayleigh fading channel with [TeX:] $$1 \mu s$$ tap delay case from Fig. 22 shown in Table IV. For both the conventional method and the proposed SIC-based method, speed and performance are inversely proportional. This result shows that the performance gap is almost similar regardless of speed and SINR.
FOR FIG. 22, BER PERFORMANCE VALUE OF SI CANCELLATION WITH AND WITHOUT THE PROPOSED SIC-BASED CHANNEL ESTIMATION IN AN EQUAL-GAIN TWO PATH RAYLEIGH FADING CHANNEL WITH [TeX:] $$1 \mu s$$ TAP DELAY.
V. CONCLUSION
In this paper, we presented a SIC-based channel estimation method for self-interference in an IBFD system. This method recursively estimated the channel response of the desired signal and SI component, progressively canceling out the highpowered signal at each iteration to enhance the estimation of the low-powered signal. Compared to conventional methods, our proposed SIC-based channel estimation scheme significantly improved the MSE performance by up to 5 dB at the fading channel models similar to EPA and Rayleigh channels.
For additional research, we need to explore the applicability and practicality of the SIC-based estimation technique in specific scenarios and other self-interference cancellation methods, not limited to ISAC systems, to validate further its effectiveness in real-world conditions, including cellular networks, cognitive radio systems, and autonomous driving. We also need to investigate the complexity analysis and enhancement of this approach with other advanced signal processing schemes to establish even more robust and comprehensive IBFD systems capable of meeting the diverse needs of next-generation wireless networks.