Open Access
Add to BibSonomyAbstract
We have implemented a modified Low-Density Parity-Check (LDPC) codec algorithm in ultraviolet (UV) communication system. Simulations are conducted with measured parameters to evaluate the LDPC-based UV system performance. Moreover, LDPC (960, 480) and RS (18, 10) are implemented and experimented via a non-line-of-sight (NLOS) UV test bed. The experimental results are in agreement with the simulation and suggest that based on the given power and 10−3bit error rate (BER), in comparison with an uncoded system, average communication distance increases 32% with RS code, while 78% with LDPC code.
© 2014 Optical Society of America
1. Introduction
The ultraviolet communication system is an optical communication mode that uses ultraviolet radiation in the solar-blind region (wavelengths between 200 nm and 280 nm) as the information carrier. Due to absorption in the ozone layer, this wavelength range has negligible solar noise, and the strong scattering effect of the particles, aerogels, and dust in the atmosphere [1,2] makes it possible – compared with other wireless optical communication systems – to realize a non-line-of-sight (NLOS) propagation mode. This appealing characteristic makes UV communication useful in some special conditions, such as confidential communication.
UV light is severely affected by scattering, refraction, absorption, and even path blockage during its propagation through air. Experiments demonstrate that, when transmitting 100 m in LOS mode, a UV channel results in 66-dB energy attenuation, which implies that a UV channel could be dramatically less effective in NLOS mode [3–5]. However, the UV communication system relies upon these wavelengths’ huge attenuation and strong scattering to ensure communication safety. To advance this system, it is important to look for effective methods and techniques to facilitate better performance of the UV channel. For example, forward error correction (FEC) is widely used in wireless communication and fiber communication to guarantee the quality of communication.
In a wireless channel, the dominant channel coding techniques are the Reed-Solomon (RS) code and the newly developed Low-Density Parity-Check (LDPC) code; both are powerful FEC codes that add constant size redundancy to correct random and burst errors [6]. The RS code is easier to implement [7] in hardware, while the LDPC code must use more complicated decoding algorithms, but the LDPC code can greatly reduce the BER to meet the required signal-to-noise ratio (SNR) in wireless channels. In this paper, LDPC codes will be used to verify whether these codes are the appropriate channel coding to guarantee the quality of communication or to extend the transmission distance in UV communication systems.
The structure of this paper is as follows. The basic structure of the UV communication system is introduced in Section 2. RS and LDPC codes are constructed, and their encoding and decoding algorithms are developed in Section 3; in this section, we also simulate the performance of the LDPC code in MATLAB based on measured channel parameters. Detailed hardware experimental systems and field test work are shown in Section 4. Section 5 draws conclusions based upon the previous sections.
2. Basic structure of an ultraviolet communication system with an FEC module
A basic UV communication system with an FEC module is built, as shown in Fig. 1, and this system consists of a transmitter (Tx), a UV channel, and a receiver (Rx). A semiconductor laser or a light-emitting diode (LED) is always chosen as the optical source, while a photodiode or a photomultiplier tube (PMT) is typically used as the UV detector at the receiver. The data encoded with the FEC code modulates the light source, and then the modulated data is transmitted through the UV channel, where it experiences severe effects from strong scattering, refraction, and path blockage. The receiver collects the UV light and proportionally converts the light to an electrical signal to be processed easily. At the receiver, both the optical noise and also the electronic equipment introduce dark current and thermal noise.
The NLOS mode is the most important and useful feature of the UV communication system, whereas common wireless optical communication is operated in LOS mode. In the NLOS mode, angles in the transmitter and receiver will be changed as an important variable parameter in our system. Due to high attenuation in the NLOS mode, the FEC module is important for extending the communication distance.
Compared with other FEC codes, RS code is easy to implement and has better error correction characteristic. Thus, we have picked RS code to compare with other FEC codes. This code is formed by sequences of m-bits and was developed based on finite field theory, known as the Galois Field (GF). The code is identified as RS (n, k), where n is the total number of symbols in a frame and k is the number of data symbols. This code is capable of correcting random and burst errors, where t = (n - k)/2 [8].
The LDPC code possesses excellent performance and compared with all other similar codes comes closest to Shannon limit [9]. The LDPC decoder can achieve higher throughput due to a high degree of parallelism in its hardware design, which satisfies the high data rate transmission demand in modern communication. Although the hardware design of the LDPC codes is somewhat complex and consumes abundant hardware resources, many communication standards – such as IEEE 802.16e and DVB-S2 – have adapted LDPC as their error correction coding. The performance and structure of the RS and LDPC codes will be discussed in the following sections.
3. Performance and structure of the RS and LDPC codes in the UV communication system
3.1 RS encoding and decoding algorithm employed in the UVC system
Normally, code rate of RS code is mostly above 80%, and error correction ability increases with lower code rate. We have applied RS (18, 10) code, it has 10 information symbols and 8 check symbols, which means that the code’s efficiency is about 55%, and thus this code could be better adapted to the poor situation of UV channel and compared with LDPC (960, 480) whose code efficiency is 50%. In selecting the RS code’s symbol bit number, the system’s SNR demands are decreased with an increase in code length from m = 2 to m = 8 under the condition of a certain BER. Usually a symbol with 8 bits is widely used in computer or other fields and 28 is a coefficient commonly used in GF(2m), thus, it is convenient for our hardware implement and experiment to select a symbol with 8 bits.
According to primitive polynomials:
We thus obtain the generating polynomial:
Therefore, we can implement the encoding process by using a division circuit, as shown in Fig. 2. As soon as the information m(x) is entered into the circuit, the check symbols are in the register.
Assuming the received code is R(x), Fig. 3 shows the basic flow chart of a decoder for RS codes [10].
A complete architecture of RS (18, 10), as mentioned above, was constructed in a practical UVC test bed to verify its feasibility and performance in this paper [11].
3.2 The source of the parity-check matrix of LDPC code and its encoding and decoding architecture in the UVC system
The LDPC code is a special linear group of codes whose parity-check matrix is a sparse matrix. Many constructor methods for the parity-check matrix have been developed [12–15]. In this paper, we employ the base check matrix in the IEEE 802.16e standard as the UVC system’s parity-check matrix shown in Fig. 4 for the following analysis [16]. We select the base check matrix, whose bit rate is 1/2 and whose code length is 960 bit, considering the comparability with the RS code (bit rate 55%) and the complexity of the hardware implementation.
Conventional encoding algorithm employ the generator matrix to encode the message. Gauss elimination method has to been conducted to obtain the generator matrix resulted in increasing the algorithm complexity of the LDPC codes encoder. According to the double diagonals structure of the base check matrix, a fast encoding algorithm [17] was introduced to simplify the encoding complexity in our experiment.
The Belief Propagation (BP) algorithm [9] is the primary decoding algorithm for the LDPC codes. However, for hardware implementation, the BP algorithm’s check-node computation is complex. To reduce this complexity, Fossorier [18] proposed a simplified and easy-to-implement algorithm named Min-Sum (MS) by sacrificing BER performance using log-likelihood rate belief propagation (LLR BP) [19]. In this paper, the Offset Min-Sum (OMS) algorithm [20] is employed for better performance.
In order to get a proper value of the offset factor β in OMS algorithm, the decoding performance simulation measured by bit error rate (BER) is conducted at different signal-noise ratio (SNR) using MATLAB. The simulation result is shown in Fig. 5.
In this paper, the offset factor is set to 0.125 considering the decoding performance and the convenience of hardware implementation.
3.3 Simulation analysis of LDPC codes applied to the UVC system
In order to analyze the performance of the soft-decision algorithm and to conduct hardware design, a coding and decoding system simulation platform (based on the Monte Carlo method [21,22]) was built in MATLAB. The key values in the simulation are based on measured channel parameters, as shown in Table 1.
Table 1. Measured UVC System Parameters
After finishing design, we conduct an experiment with the elevation angle of 10°~10° between Tx and Rx in real channel to compare with the simulation. The results of experiment and simulation are presented in Fig. 6.
The experiment result in the trend is in consistent with the simulation, although there are some discrepancies. We can conclude from the comparison that the simulation result is a little better than the experiment work for the practical background noise and the system noise. But it is demonstrated that the system is working properly.
4. Experimental verification for application of the RS and LDPC codes in the UVC system
4.1 Introduction of experiment platform
In consideration of eye-safety, multiple chips from a UV LED array are chosen as our UV signal source. Because of the varying angles for Tx and Rx, the path loss in the NLOS mode may reach 100 dB. This loss makes the PMT – which can achieve 105 gain – necessary to detect the weak UV signal, shown as Fig. 7.
A computer is used to input data and to transmit data in packets into the field programmable gate array (FPGA) buffer. Data are encoded in the RS (18, 10) or LDPC (960, 480) code, which both have similar coding rates, and these data are then modulated by the OOK scheme. Finally the FPGA sends modulated data to the LED driver, and the power control module changes the LED’s current to emit pulsed laser beams that carry the information to the UV channel.
The optical signal is detected by the PMT and is then converted to an analog electrical signal. Because of its wide sensitivity range, an optical filter is used before the PMT is used to reduce the impact of background light in other wavebands. Then, the analog signal is amplified and converted to a digital signal in the ADC circuit. A digital filter is used to reduce noise and improve SNR and to reduce the probability of errors. Received RS-coded data will be “one” or “zero” after the hard decision. After demodulation and decoding, we will receive the original data to calculate the BER. Considering tradeoff between better performance and probability of future work, we applied well used offline test method to decode LDPC code in PC. Thus we can obtain reliable data and accelerate process of experiment. Received LDPC-coded data will be stored in the PC and subsequently decoded in the PC using a soft-decision algorithm to obtain the data’s BER.
4.2 Design of experiment
The transmitter light source was a hemispherical packaged LED array with 36 UV LED chips (UV TOP 260 TO 3 HS with nominal center wavelengths of 265 nm, SETI corp.). Though the UV source is an arrayed-LED, it is obviously isotropy and comes from the same signal. Its output optical power range is 0~15.5 mW. A pseudorandom sequence is generated using polynomial y = x2 + x9. The transmission rate is 500 kbps.
We chose a UV filter with a 266-nm center wavelength, and also selected a HAMAMATSU PMT (R7145 model) whose spectral response range is 160 nm~320 nm and whose largest cathode gain is 105 ~107 as receiver. Equipment used in experiment including hemispherical packaged LED array and PMT are easy to buy and total cost is under $5,000. Besides, the price of UV LED is kept on falling and the whole system cost will be equal to the FSO system with infrared laser. So the UV system can be applied to some special cases with reasonable cost, such as radio-silent communication.
The elevation angle in the transmitter is the angle between LED and horizon, and the elevation angle in the receiver is the angle between PMT and horizon, meanwhile, LED and PMT keep on same height, shown as Fig. 8. To meet the requirements of the digital filter, the sampling frequency of the A/D circuit is 8 MHz. The hard-decision gate limit is half of the peak value. The BER is displayed in the PC using the UART interface.
Due to the processing delay and the distance between Tx and Rx, we designed an embedded error bit counter to measure the BER of the UV channel online. The transmitted data is composed of two parts: the frame header and the pseudorandom sequences. The frame header is used for detecting the beginning of the effective signal and for the synchronization of the pseudorandom sequences; the pseudorandom sequences are encoded data.
According to the frame header, the receiver begins to synchronize received data; if the frame header is not detected, this frame is abandoned. After synchronization and demodulation, the data will be systematically decoded. The RS code is output using two methods: one method is to decode and correct errors, and the other method is to extract information bits from the RS code without error correction. Finally, we compare pseudorandom sequences from these two methods with the output bit error rate until the last frame is received. Similarly, for comparison with the RS code, received LDPC-coded data will be stored in the PC and decoded in the PC with the soft-decision algorithm.
4.3 Analysis of experiment
In the UV communication system, the receiver’s SNR is severely affected by communication distance and elevation angle. To test the BER both with and without the FEC code in different SNR situations, we changed both the communication distance and the elevation angle of Tx and Rx in our experiment to obtain different SNRs. Considering experimental error and randomness of the results, the following BER results are averaged values collected during repeated measurements. This paper uses communication distance gain in the same BER situation to demonstrate the impact of the LDPC and RS codes on the UV channel.
These experiments test several classic elevation angles for Tx and Rx: 0°~0°, 10°~10°, 20°~20°, and 30°~30°, both with and without the FEC code. Comparisons of the BER with respect to the distance are shown in Fig. 9. We can conclude from Fig. 9 that the RS and LDPC coding schemes have better performance than the uncoded system in the poor SNR environment common to UV communication.
Fig. 9 Experimental results of the LDPC and RS codes with different angles. (a) 0°~0° (BER = 0 with the LDPC code). (b) 10°~10°. (c) 20°~20°. (d) 30°~30°.
In the LOS situation, an elevation angle of 0°~0° between Tx and Rx and a BER of the RS code at same distance are better than an uncoded system, while there are no error bits in the system with the LDPC code when the distance increases to 50 m in Fig. 9(a).
In the NLOS mode, we apply an elevation angle of 10°~10° between Tx and Rx and a BER of the RS code at the same distance, which still always outperforms the uncoded system. After the transmission distance reaches 40 m, error bits start to emerge in the system with the LDPC code, and the BER increases to 10−3 in Fig. 9(b).
Similarly, when we apply 20°~20° elevation angles, the communication distance is 17 m in the uncoded system, 22 m with the RS code, and 27.5 m with the LDPC code in the situation where the BER is 10−3, as in Fig. 9(c).
When the elevation angle is as high as 30°~30°, the communication distance is reduced to 11 m in the system with the LDPC code, 9 m in the system with the RS code, and only 6 m in the uncoded system where the BER is 10−3, as in Fig. 9(d).
Based on 10−3 BER, in comparison with an uncoded system, communication distance increases 25% with RS code, while 42% with LDPC code in the 10°~10° elevation angles and increases 46% with RS code, while 91% with LDPC code in the 30°~30° elevation angles. We can easily find that FEC codes are more effective in the high elevation angles.
5. Conclusion
The UV communication system faces very noisy environments, and channel coding is an effective means of improving the BER of received data. This study’s simulation and experimental results show that, with the FEC code, this system outperforms uncoded systems; meanwhile, at a similar code rate, the LDPC code can extend the communication distance at the same BER with respect to the RS code. According to repeated measurements, the average communication distance of the system with the RS code and the LDPC code extend 32% and 78% at 10−3 BER, respectively. The only challenge facing this system is that the LDPC code exhausts more hardware resources, which hampers system application.
Acknowledgment
This work was supported by the NSFC project (61101110).
References and links
1. F. Xu, A. Khalighi, P. Caussé, and S. Bourennane, “Channel coding and time-diversity for optical wireless links,” Opt. Express 17(2), 872–887 (2009). [CrossRef] [PubMed]
2. G. Chen, Z. Xu, and B. M. Sadler, “Experimental demonstration of ultraviolet pulse broadening in short-range non-line-of-sight communication channels,” Opt. Express 18(10), 10500–10509 (2010). [CrossRef] [PubMed]
3. G. A. Shaw, A. M. Siegel, and J. Model, “Extending the range and performance of non-line-of-sight ultraviolet communication links,” Proc. SPIE 6231, 2310 (2006). [CrossRef]
4. G. Chen, Z. Xu, H. Ding, and B. Sadler, “Path loss modeling and performance trade-off study for short-range non-line-of-sight ultraviolet communications,” Opt. Express 17(5), 3929–3940 (2009). [CrossRef] [PubMed]
5. Y. Tang and G.-Q. Ni, “Study of channel character of solar blind UV communication,” Proc. SPIE 6622, 6620 (2008).
6. A. Joseph, “Design of the high-speed framing, FEC, and interleaving hardware used in a 5.4km free-space optical communication experiment,” Proc. SPIE 7464, 33 (2009).
7. L. Chang, L. Song, and L. Zhang, “Design and realization of FEC-based RS code in the atmospheric laser communication system,” Proc. SPIE 7160, 716038 (2008). [CrossRef]
8. W. Li, N. Qunfeng, and Z. He, “Wireless Communications Link with RS Channel Code for Setting Electronic Fuze,” in Computing, Communication, Control, and Management, 2008 CCCM '08 ISECS International Colloquium (2008), pp. 398–402.
9. R. G. Gallager, “Low-density parity-check codes,” IEEE Trans. Inf. Theory 8(1), 21–28 (1962). [CrossRef]
10. L. Chang, L. Song, and L. Zhang, “Design and realization of FEC-based RS code in the atmospheric laser communication system,” Proc. SPIE 7160, 716038 (2008). [CrossRef]
11. T. C. Barbosa, R. L. Moreno, and T. C. Pereira, “FPGA implementation of a Reed-Solomon CODEC for OTN G.709 standard with reduced decoder area,” in 6th International Conference on WiCOM, Chengdu, China (2010). [CrossRef]
12. J. Hagenauer, E. Offer, and L. Papke, “Iterative decoding of binary block and convolutional codes,” IEEE Trans. Inf. Theory 42(2), 429–445 (1996). [CrossRef]
13. D. J. C. MacKay, “Good error-correcting codes based on very sparse matrices,” IEEE Trans. Inf. Theory 45(2), 399–431 (1999). [CrossRef]
14. Y. Kou, S. Lin, and M. P. C. Fossorier, “Low density parity check codes: construction based on finite geometries,” in Global Telecommunications Conference, GLOBECOM '00 IEEE2000 (2000), Vol. 2, pp. 825–829. [CrossRef]
15. B. Vasic, E. M. Kurtas, and A. V. Kuznetsov, “Kirkman systems and their application in perpendicular magnetic recording,” IEEE Trans. Magnetics 38, 1705–1710 (2002).
16. IEEE Standard for Local and Metropolitan Area Networks–Part 16: Air Interface for Fixed and Mobile Broadband Wireless Access Systems. IEEE Std P802.16e/D12 (2005).
17. Z. Cai, J. Hao, P. H. Tan, S. Sun, and P. S. Chin, “Efficient encoding of IEEE 802.11n in LDPC codes,” Electron. Lett. 42(25), 1471–1472 (2006). [CrossRef]
18. M. P. C. Fossorier, M. Mihaljevic, and H. Imai, “Reduced complexity iterative decoding of low-density parity check codes based on belief propagation,” IEEE Trans. Commun. 47(5), 673–680 (1999). [CrossRef]
19. E. Eleftheriou, T. Mittelholzer, and A. Dholakia, “Reduced-complexity decoding algorithm for low-density parity-check codes,” Electron. Lett. 37(2), 102–104 (2001). [CrossRef]
20. J. Chen and M. P. C. Fossorier, “Density evolution for two improved BP-based decoding algorithm of LDPC codes,” IEEE Trans. Commun. 6(5), 208–210 (2002). [CrossRef]
21. D. Han, X. Fan, K. Zhang, and R. Zhu, “Research on multiple-scattering channel with Monte Carlo model in UV atmosphere communication,” Appl. Opt. 52(22), 5516–5522 (2013). [CrossRef] [PubMed]
22. P. Luo, M. Zhang, D. Han, and Q. Li, “Performance analysis of short-range NLOS UV communication system using Monte Carlo simulation based on measured channel parameters,” Opt. Express 20(21), 23489–23501 (2012). [CrossRef] [PubMed]
