Open Access
Add to BibSonomy
Abstract
Non-orthogonal multiple access (NOMA) has recently attracted significant attention as a promising multiple access scheme for the 5th generation (5G) wireless communication due to its superior spectral efficiency, which has also been studied and shown to achieve a superior performance in visible light communication (VLC) networks. However, the error propagation (EP) problem due to successive interference cancellation (SIC) decoding has not yet been resolved, which degrades the system BER performance and causes user unfairness. In this work, symmetric superposition coding (SSC) and symmetric SIC (SSIC) decoding are proposed for a downlink NOMA-based VLC network, in which the distribution of the demodulation regions of the user allocated with more power will be symmetrical in terms of the decision threshold of the user allocated with less power. Furthermore, the proposed method is experimentally tested and the results show that more than 90% demodulation errors caused by EP are eliminated compared with traditional NOMA VLC.
© 2017 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Visible light communication (VLC) has recently attracted significant attention in both academia and industry as an effective complement to RF communication due to the advantages of license-free, easy implementation into existing infrastructure, low cost front-ends and radio frequency (RF) interference-free. As a wireless broadband technology, an efficient multiple access (MA) method to provide multi-user with network access simultaneously is essential for VLC [1]. In the 4th generation (4G) mobile communication, orthogonal frequency multiple access (OFDMA) was adopted and has achieved great system-level throughput performance. To further enhance the spectral efficiency, non-orthogonal multiple access (NOMA) has recently been proposed for 5G wireless networks as a promising MA technique [2–5]. In NOMA network, user messages are superposed in the power-domain at the transmitter and separated by successive interference cancellation (SIC) receiver so that multiple users can share the same time-frequency (TF) resources. However, it is unpractical for all of the users to share the same TF resources. In general, the users in the network are divided into multiple groups, to which orthogonal resources are allocated, and there are only a few users in each group to perform NOMA jointly. The effect of user pairing/grouping is investigated in [6]. It was shown that NOMA can achieve higher throughput by 30% to 40% compared with orthogonal multiple access (OMA) with a group factor (the number of users in each group) of two in [7]. More throughput can be achieved when the group factor is larger according to the research of [8]. Studies have shown that NOMA is also suitable for VLC network and was shown to achieve a superior performance [9-10]. In [11], a bidirectional NOMA-OFDMA VLC network was proposed and experimental demonstrated, which can offer a high throughput, flexible bandwidth allocation and a higher system capacity for a larger number of users.
However, the previous works have not yet solved error propagation (EP) problem due to SIC decoding, which degrades the system BER performance and cause user unfairness. The research of [2] and [12] have shown that EP can be limited by decreasing the power allocation ratio (PAR), P1/P2 (P1<P2, P1is the transmission power allocated for the user with good channel quality and P2 is for the user with poor channel quality). However, the system throughput will decrease as the decrease of PAR according to the research of [12]. On the other hand, the value of the PAR is usually set according to the difference in channel qualities between the users to guarantee the user QoS and fairness (the larger the channel difference, the smaller the PAR value should be). It is unpractical to keep all of the users have a large difference in channel qualities, especially for VLC network because the cell size is usually small, which means a large PAR is more suitable in many scenes.
In this paper, the symmetric superposition coding (SSC) and symmetric SIC (SSIC) decoding are proposed to overcome the EP problem for downlink NOMA-based VLC network. By adjusting the way of superposition and demodulation of the signal of different users, the distribution of the demodulation regions of the user allocated with more power are symmetrical in terms of the decision threshold of the user allocated with less power, which can effectively overcome the EP problem. Experiments with a user grouping factor of two and three are demonstrated and the results show that more than 90% demodulation errors caused by EP are eliminated compared with traditional NOMA VLC.
2. Principle of SSC/SSIC
OFDM modulation is adopted for our downlink NOMA-based VLC network and the signals of different users are superposed in the power domain before inverse fast Fourier transform (IFFT). According to the traditional superposition coding (TSC) of NOMA, the superposed signal of k user can be written as:
where P represents the total transmitted power of all the signals; Xi is the frequency-domain OFDM signal of user i; αi is the power allocation coefficient of the i-th user and satisfies:In general, users with poor channel qualities are assigned with more power. Without loss of generality, the channel qualities are sorted as h1 ≥ h2 ≥ ... ≥ hk, leading to α1 < α2 < αk. Therefore, the received frequency-domain signal at the m-th user can be written as:
where Hm and Nm are channel coefficients and noises represented in frequency-domain of user m respectively. The SIC decoding is utilized after channel estimation and can be represented as:where H is the result of channel estimation; Si is the demodulation result of user i; DeQAM(•) and QAM(•) represent QAM demodulation processing and QAM modulation processing respectively. According to Eq. (4), we can recover the signal of user m after removing the signal of user k to user m + 1. When the signal of user k to user m + 1 cannot be accurately demodulated, the demodulation of user m will be affected, which results in EP.The main source causing EP is analyzed based on a two user NOMA network and both users are 4QAM modulated, leading to 16 points on the constellation of the superposed signals [13]. Figure 1(a) is the constellation of the superposed signals. According to Eq. (1), the data of the two users mapped by each constellation point is represented beside the constellation point in Fig. 1(a), and Fig. 1(b) shows the demodulation region for the two users at the receiver.
According to Eq. (4), the decision result of each region is represented in Fig. 1(b). The four regions of the same quadrant have the same decision result for user 2 and the four regions of the same color have the same decision result for user 1. Considering that channel noise is mainly additive white Gaussian noise, the main source of the demodulation error of user 2 is that the constellation point near the axis at the transmitter moves to the adjacent region of another quadrant after passing through the channel, e.g. point A3 moves to region II-4 or IV-2 and point A2 moves to region II-1. Comparing the original data with the demodulation result of these constellation points and demodulation regions, the demodulation error of user 2 caused by the above situation will lead to the demodulation error of user 1, which is the main source causing EP.
The process of proposed SSC is represented as follows:
- 1. Determine the row and column where the constellation point of user 2 (X2) is located;
- 2. Inverse the real part of the data of user 1 (X1) if the constellation point of X2 is located in an even column and inverse the imaginary part of the X1 if the constellation point of X2 is located in an even row;
- 3. Add the processed X1 to X2 according to the power allocation coefficients. Regard the superposed data of X1 and X2 as a new data S2.
- 4. Process the constellation point of user 3 (X3) in the same way as step 1. Process S2 in the same way as step 2 according to the constellation point of X3. Add the processed S2 to X3 according to the power allocation coefficients. Regard the superposed data of S2 and X3 as a new data S3.
- 5. Repeat this process until all user data are superposed.
The “even column” and “even row” mentioned in step 2 is defined as follows: Consider a constellation of QAM modulation, the column of the constellation is numbered from right to left as “column 1”, “column 2”, “column 3” etc. The row of the constellation is numbered from top to bottom.
The superposed signal of k users can be written as:
where Qir and Qii represent the maximum value of the R-coordinate and I-coordinate of the constellation of user i respectively, e.g. Qir = a and Qii = a for 4QAM modulation, and a is the unit distance of the constellation (a = 1 in this paper).The process of proposed SSIC for user m is represented as follows:
- 1. Acquire the data of user k by demodulating the received and equalized signal Yk;
- 2. Re-QAM-modulate the demodulated data of the user k and determine the row and column where the constellation point is located;
- 3. Acquire Yk-1 by removing the re-QAM-modulated data of user k from Yk, inverse the real part of Yk-1 if the constellation point is located in an even column and inverse the imaginary part of Yk-1 if the constellation point is located in an even row;
- 4. Acquire the data of user k-1 by demodulating Yk-1, and repeat these steps until the data of user m is demodulated.
The formula of SSIC can be written as:
Similarly, the analysis for EP of proposed NOMA network is given based on Fig. 2. Figure 2(a) is the constellation of the superposed signals with SSC. According to Eq. (5), the data of the two users mapped by each constellation point is represented beside the constellation point in Fig. 2(a). The bold words in the data expression of the two users in Fig. 2(a) shows the difference from that in Fig. 1(a). Figure 2(b) shows the demodulation region for the two users. According to Eq. (6), the decision result of each region is represented in the figure. Same as Fig. 2(b), the four regions in the same quadrant have a same decision result for user 2 and the four regions with the same color have a same decision result for user 1. Compared to Fig. 1(b), due to SSC and SSIC processing, the distribution of the demodulation regions of the user 1 in Fig. 2(b) are symmetrical in terms of the decision threshold of user 2 (R-axis and I-axis in Fig. 2). As analyzed in Fig. 1, the main source of the demodulation error of user 2 is that the constellation point near the axis at the transmitter moves to the adjacent region of another quadrant after passing through the channel. Therefore, when such errors of user 2 appears, the demodulation result of user 1 will still be correct, e.g. when point A3 moves to region II-4 or IV-2, the demodulation result of X1 is -1-j, the same as the data sent at the transmitter. Therefore, most of the demodulation errors from user 2 will not be propagated to user 1.
3. Experiments and discussion
Figure 3 shows the experimental setup of our downlink NOMA VLC with two users and three users, respectively. The signal is firstly transmitted by an arbitrary waveform generator (AWG, Tektronix AWG5012) operating at 10 M/S, then amplified and superposed onto the LED (OSRAM LUW W5SM) by aid of Bias-T. The bias voltage is fixed at 5.5V and the amplitude of the superimposed signal is about 2V. The optical signal is received and converted back into the electrical signal by a commercially available avalanche photodiode (Hamamatsu C12702-12), and then captured by an oscilloscope (LeCroy SDA760Zi). In the experiments, 4QAM modulation is adopted for all of the users. The FFT size is 512, Containing 120 effective subcarriers, and the CP size is 8. Each frame consists of 200 OFDM symbols, 10 of which are training symbols for channel estimation.
In addition, we have PAR = α1/α2 for the experiments of two users, and fixed power allocation strategy (FPA) is selected for the experiments of three users, i.e. PAR = α1/α2 = α2/α3 for the experiments of three users. On the other hand, only the signal received by user 1 is analyzed since only user 1 has the EP problem in two user case and the analysis of the EP problem of user 2 in three user case is similar to that of user 1 in two user case. The “user 2” and “user 3” mentioned later in the experimental analysis refers to the signal of user 2 and user 3 that received by user 1.
Figures 4 and 5 shows the BER difference as a function of the PAR between the users applying SSC/SSIC and the users applying TSC/TSIC. Figures 4(a) and 4(b) shows the experimental results with 2 users at transmission distances of 30cm and 40cm, respectively. As shown in Fig. 4, the user 2 with SSC/SSIC and the user 2 with TSC/TSIC have a same BER performance for the reason that they have the same channel and SSC/SSIC has no effect on the decoding of the user allocated with the maximum power. The BER performance gap between the user 1 with SSC/SSIC and the user 1 with TSC/TSIC is quiet small when the PAR is set at about 0.22. However, when the PAR increases from 0.22 to 0.32, the BER performance of the user 1 with SSC/SSIC becomes much better, but the BER performance of the user 1 with TSC/TSIC only achieves a small gain at first, and then begins to deteriorate. This can be explained as follows. When the PAR is small, the power allocated to user 2 is much more than user 1, resulting in that the BER of user 2 is much smaller than user 1. So the effect of EP is slight, leading to a close BER performance between the user 1 with SSC/SSIC and the user 1 with TSC/TSIC. When the PAR increases, there are more power allocated to the user 1 and less power allocated to user 2, which lead to the improvement of the BER performance of user 1 and the decline of the BER performance of user 2. Thus, the effect of EP gets much larger.
Furthermore, the BER of the user 1 with SSC/SSIC almost equals to the BER of the user 1 with TSC/TSIC minus the BER of the user 2 with TSC/TSIC, which means the effect of error propagation has almost been overcome with the proposed method. By comparing the position of the error data of user 1 and user 2 in multiple experiments with different conditions (PAR and L), we found that more than 95% of the error data of user 2 are propagated to user 1 in traditional NOMA VLC experiments while no more than 5% of the error data of user 2 are propagated to user 1 in proposed NOMA VLC experiments. In other words, more than 90% demodulation errors caused by EP are eliminated compared with traditional NOMA VLC.
Except for better BER performance, a more flexible resource scheduling program could be achieved with SSC/SSIC. In proposed NOMA network, when user 1 and user 2 have different requirements for BER, we can adjust the value of PAR (0.22-0.32) to find a suitable size to guarantee the BER requirements of both users. However, in traditional NOMA network, as shown in Fig. 4, increasing the value of PAR from 0.22 to 0.32 does not make any sense since the BER performance of user1 is not getting better but the BER performance of user 2 is rapidly decaying. However, when a PAR that smaller than 0.22 is selected, the proposed method is not needed anymore since the effect of the EP problem is slight enough in such conditions.
Figures 5(a) and 5(b) shows the experimental results of the NOMA VLC with 3 users at transmission distances of 10cm and 20cm, respectively. The proposed technology in the experiments of three users still has a good performance. It is noteworthy that the BER of user 2 with SSC/SSIC only have slightly changes as the increase of the PAR. The explanation is represented as follows. Assuming that the interference from user 3 is successively canceled, the SINR of user 2 is represented as
where h is the DC channel gain between the transmitter and the receiver. N is the average power of channel noise and N′ = N/P|h|2. In this experiment, the reduction of (PAR + 1/PAR + 1)N′ is close to the increase in PAR, resulting in only a slight change in SINR2 and BER performance.4. Conclusions
Symmetric superposition coding and symmetric successive interference cancellation decoding are proposed for downlink NOMA-based VLC network to overcome error propagation problem in this paper. Compared with traditional superposition coding and traditional successive interference cancellation decoding, a more reasonable joint constellation is obtained with the proposed method, which can eliminate most of the EP. The experimental results show that with the proposed method, more than 90% demodulation errors caused by EP are eliminated. As a result, better BER performance and more flexible resource scheduling can be achieved for NOMA-based VLC network.
Funding
National Natural Science Foundation of China (61401032); National 973 Program (2013CB329205).
References and links
1. H. Haas, L. Yin, Y. Wang, and C. Chen, “What is LiFi?” J. Lightwave Technol. 34(6), 1533–1544 (2016).
2. A. Benjebbour, Y. Saito, Y. Kishiyama, A. Li, A. Harada, and T. Nakamura, “Concept and practical considerations of non-orthogonal multiple access (NOMA) for future radio access,” in International Symposium on Intelligent Signal Processing and Communications Systems (IEEE, 2013), pp. 770–774.
3. L. Dai, B. Wang, Y. Yuan, S. Han, I. Chih-Lin, and Z. Wang, “Non-orthogonal multiple access for 5G: solutions, challenges, opportunities, and future research trends,” IEEE Commun. Mag. 53(9), 74–81 (2015).
4. Z. Ding, Z. Yang, P. Fan, and H. V. Poor, “On the performance of non-orthogonal multiple access in 5G systems with randomly deployed users,” IEEE Signal Process. Lett. 21(12), 1501–1505 (2014).
5. Y. Wang, B. Ren, S. Sun, S. Kang, and X. Yue, “Analysis of non-orthogonal multiple access for 5G,” China Commun. 13(Supplement2), 52–66 (2016).
6. Z. Ding, P. Fan, and H. V. Poor, “Impact of user pairing on 5G nonorthogonal multiple-access downlink transmissions,” IEEE Trans. Vehicular Technol. 65(8), 6010–6023 (2016).
7. A. Benjebbour, A. Li, Y. Saito, Y. Kishiyama, A. Harada, and T. Nakamura, “System-level performance of downlink NOMA for future LTE enhancements,” in Globecom Workshops (IEEE, 2013), pp. 66–70.
8. L. Lei, D. Yuan, C. K. Ho, and S. Sun, “Power and channel allocation for non-orthogonal multiple access in 5G systems: Tractability and computation,” IEEE Trans. Wirel. Commun. 15(12), 8580–8594 (2016).
9. L. Yin, W. O. Popoola, X. Wu, and H. Haas, “Performance evaluation of non-orthogonal multiple access in visible light communication,” IEEE Trans. Commun. 64(12), 5162–5175 (2016).
10. H. Marshoud, V. M. Kapinas, G. K. Karagiannidis, and S. Muhaidat, “Non-orthogonal multiple access for visible light communications,” IEEE Photonics Technol. Lett. 28(1), 51–54 (2016).
11. B. Lin, W. Ye, X. Tang, and Z. Ghassemlooy, “Experimental demonstration of bidirectional NOMA-OFDMA visible light communications,” Opt. Express 25(4), 4348–4355 (2017). [PubMed]
12. A. Benjebbour, K. Saito, A. Li, Y. Kishiyama, and T. Nakamura, “Non-orthogonal multiple access (NOMA): Concept, performance evaluation and experimental trials,” in International Conference on Wireless Networks and Mobile Communications (IEEE, 2015), pp. 1–6.
13. S. R. Islam, N. Avazov, O. A. Dobre, and K. S. Kwak, “Power-domain non-orthogonal multiple access (NOMA) in 5G systems: potentials and challenges,” IEEE Commun. Surveys Tuts. 19(2), 721–742 (2017).
