Open Access
Add to BibSonomyAbstract
Combination of overlapping pulse position modulation and pulse width modulation at the transmitter and grouped bit-flipping algorithm for low-density parity-check decoding at the receiver are proposed for visible Light Emitting Diode (LED) indoor communication system in this paper. The results demonstrate that, with the same Photodetector, the bit rate can be increased and the performance of the communication system can be improved by the scheme we proposed. Compared with the standard bit-flipping algorithm, the grouped bit-flipping algorithm can achieve more than 2.0 dB coding gain at bit error rate of 10−5. By optimizing the encoding of overlapping pulse position modulation and pulse width modulation symbol, the performance can be further improved. It is reasonably expected that the bit rate can be upgraded to 400 Mbit/s with a single available LED, thus transmission rate beyond 1 Gbit/s is foreseen by RGB LEDs.
© 2012 Optical Society of America
1. Introduction
Nowdays, the visible light communication (VLC) has attracted much more attention in various fields of applications, such as lighting, automobiles, transportation, communication, imaging, agriculture and medicine [1]. The VLC system employs a signal transmitter as an illumination module such as light emitting diode (LED) lighting [2–6]. Usually, on-off keying (OOK) modulation can be used in VLC system to build a basic communication system. However, as data rate becomes high, more advanced modulation and encoding techniques of good power and bandwidth efficiency may be required due to the power and bandwidth limitations. With discrete multitone (DMT) modulation, VLC systems can provide a transmission rate of 200+ Mb/s with commercial high-power lighting LEDs [7]. Based on wavelength-division multiplexing (WDM) and DMT modulation, a VLC system with a single RGB-type white LED can operate under an aggregate rate of 803 Mbit/s [8]. However, the DMT modulation needs an arbitrary waveform generator and a high-speed IFFT/FFT operator. The overlapping pulse position modulation (OPPM) method can increase the throughput with the same complexity as the OOK modulation. However, since Euclidean distance between signal points in the constellation was reduced for OPPM, the performance degradation was observed [9]. Fortunately, error correcting codes can be employed to improve the performance. In this paper, to further improve the transmission rate, we combine OPPM and pulse-width modulation (PWM) modulation, which will be referred to as OPPM-PWM thereafter. At the transmitter, the information is encoded with low-density parity-check (LDPC) coding [10–12], and then the light from LED is OPPM-PWM modulated before being transmitted. Because the minimum pulse-width of OPPM-PWM symbol is the same as the pulse-width of OPPM symbol, the same photo-detector is used at the receiver as that for OPPM system.
LDPC coding has been commonly used in wireless and optical communications to correct errors caused by channel noise and interference. At the receiver, soft decision decoding algorithm is usually used [13]. The algorithm of soft decision decoding is based on detailed knowledge of the probability distribution of channel state and noise [14], which is complex especially if the channel varies with time. Due to the simplicity and low cost of the system, the bit-flipping (BF) decoding algorithm can be adopted in various situations. However, as the modulation level of OPPM is increased, the merit of BF algorithm contracts. To solve this problem, we propose to use the grouped bit-flipping decoding algorithm at the receiver.
The rest of this paper is organized as follows. In section 2, the modulation format combining OPPM and PWM is described and the motivation to lower bit error rate (BER) of the LED communication system is analyzed. In section 3, the grouped bit-flipping algorithm (GBF) is proposed along with the simulation results. In section 4, an experiment of LED indoor communication system is carried out which demonstrates the performance improvement. In section 5, the optimization of OPPM-PWM symbol is researched, and followed by the conclusions in section 6.
2. Combination of PPM and PWM modulation
The rise and fall time of the available RC-LED-650-02 (produced by ROITHNER LASERTECHNIK) is 3.0 ns. By OOK modulation format, transmission rate no more than 166 Mbit/s can be achieved because one bit takes 6.0 ns. OPPM allows pulses to overlap when they represent different information, so that each pulse carries increased amount of information on the average [15]. If the response time of photo-detector (PD) at the receiver end is 1.0 ns, the minimum pulse-width of RC-LED-650-02 can be divided into 6 slots. When RC-LED-650-02 is overlapping pulse position modulated by a symbol representing 2 bits data, 9.0 ns of 9 slots per symbol are required as shown in Fig. 1. In this case, the duration per bit is decreased from 6.0 ns to 4.5 ns.
As the pulse-width is not less than 6 ns, it can be expanded to 7 ns or more. If RC-LED-650-02 is modulated by a signal with possible pulse-width of 6, 7, 8, and 9 ns, 2 bits per symbol of pulse width modulation is achieved. To improve the transmission rate, we combine OPPM of 2 bits per symbol and PWM of 2 bits per symbol by extending a symbol to 10 slots, i.e., 10ns. In this way, we can get a higher bit rate because per symbol duration of 10 ns represents 4 bits. It predicts that the transmission rate with RC-LED-650-02 can be improved to 400 Mbit/s by combining OPPM and PWM modulation,thus transmission rate beyond 1 Gbit/s is foreseen by wavelength-division multiplexing RGB LEDs. For the realization convenience and simplicity, the symbol of all zeros in 10 slots is included together with the OPPM-PWM symbols. Table 1 lists 16 possible symbols and the corresponding 4-bit codes. The inclusion of all zeros in 10 slots makes the modulation somewhat like OOK. Compared to the OOK modulation format with 50% brightness on average, the brightness of LED is also improved to 68.75% by OPPM-PWM modulation, which makes sense because the VLC system is also operated as an illumination system. It is worth mentioning that the trade-off of OPPM-PWM modulation is the requirement of higher frequency clock for the encoding and decoding. However, it is not the problem for the currently available digital signal processing chips.
Table 1. OPPM-PWM symbol with minimum pulse-width of 6 slots
3. Grouped bit-flipping decoding algorithm
The configuration model of LED indoor communication system is shown in Fig. 2. The data from information source is encoded by an LDPC code. According to the Mackay’s Construction 1A, the low-density parity-check matrices H (512, 256, 0.5) are created [13]. A 256 by 512 matrix (256 rows, 512 columns) is created at random with weight of 3 per column, weight per row as uniform as possible, and overlap between any two columns no greater than 1. Then, the light from LED is OPPM-PWM modulated, and transmitted over the additive white Gaussian noise (AWGN) channel. The AWGN channel adds the noise in the transmission system. At the receiver end, the received signal is expressed as y = x + n, where y is the received signal, x is the transmitted signal from the transmitter, and n is the AWGN with zero mean. The photon detector (PD) converts the optical signal to electrical signal, the OPPM-PWM modulation symbols are obtained by the hard decision, and the 4-bit codes are derived by the rule of maximum likelihood. Then the 4-bit sequence is sent to the LDPC decoding module. In doing so, one error in a 10-slot symbol may cause multiple errors in the derived 4-bit code. To solve this problem, we propose the grouped bit-flipping (GBF) decoding algorithm for LDPC decoding, which has similar complexity as the standard BF algorithm.
The flow chart of GBF algorithm is shown in Fig. 3, and described as follows.
- 1) Calculate the parity-check sum of the 4-bit sequences by the standard BF algorithm. If all the parity-check check sums are zeros, the decoding is completed.
- 2) For any bit, record the number of parity-check equations with non-zero sum, denoted by , for i = 1, ···,n, where i is the index of the bit, and n is the length of an LDPC code.
- 3) Subtract the which is greater than zero by one, then every log2M-bit is grouped. The sum of for bits in a group is denoted as , for j = 1,···,n/log2M, where M is the modulation order of OPPM-PWM.
- 4) Define the groups for which is the greatest as the set of S. Flip the bits corresponding to greater than one in S.
- 5) Repeat 1)–4) until all the parity-check sums are zeroes, we call this process as inner iteration. Otherwise, go to step 6) if the preset maximum number of inner iterations (we set it to 10 in this paper) is reached. Step 6) is referred to as the inner iteration.
- 6) Recalculate the as the same way in step 2) and flip the bits corresponding to greater than 1, then go to 1). If a preset maximum number of outer iterations is reached (in this paper, we set it to 3), the decoding is stopped, and the present values of bits are considered as the output of the LDPC decoding module.
In contrast to the case that the bit is checked one by one in BF algorithm, every 4 bits in a symbol is considered as a whole in GBF algorithm as described above, thus more than one bit error in a symbol can be corrected at a time and lower BER can be obtained. By simulation, the left in Fig. 4 shows the BER of 4-order OPPM-PWM modulated LED communication system. At BER of 10−5, GBF algorithm has more than 1.0dB coding gain over BF algorithm, and more than 3.0dB coding gain over the uncoded transmission. If the OPPM-PWM modulation order increases to 16, the right in Fig. 4 shows that more than 2.0dB coding gain can be obtained by GBF algorithm over BF algorithm at BER of 10−6. It can be expected that the GBF algorithm can get more coding gain if the modulation order increases. What is more, the calculation time of GBF algorithm is shorter, because less iterations are needed for error correcting due to the lower BER.
4. Experiment results
The experimental setup is shown in Fig. 5. At the transmitter, the FPGA (EP2C8Q208C8) is used as the driver of LED. To verify the OPPM-PWM modulation scheme in this paper, the red LED with the wavelength of 650nm is used in the experiment. The rise and fall time of the LED is about 20ns and 10ns respectively, and the power of light from the LED is about −6dBm. At the receiver, the photon-detector is PDA10A provided by THORLABS. The specified bandwidth of PDA10A is 150MHz, which is 0.75 of 200MHZ slot rate, so the photo-detector can sufficiently respond to the slot of 5ns. Because the response time of the used LED is 30ns, the pulse-width of a 16-order OPPM-PWM symbol can be set as 30, 35, 40, 45 and 50ns respectively, and the minimum pulse-width of 30ns corresponds to 6 slots. Oscilloscope (OSC) produced by LeCroy Corp. measures the received waveform from the photo-detector. Due to the power limitation of the used LED, the transmission distance is shorter than 20 centimeters.
The 16 possible symbols in the experiment can be listed as in Table 1. If every symbol is considered as one 10-bit binary string and ordered by the natural number that a 10-bit binary string represents, 4-bit codes representing the order is listed in the right column of Table 1. For example, if the 4-bit codes are 0001-0011-1010-1101-···, the OPPM-PPM modulated output pulse can be expressed as 0000111111- 0001111111-0111111111-1111111100-···. At the transmitter, the FPGA board transfers the user’s data to the 10-bit binary string pulses, which modulates the output light of LED. At the receiver, the optical pulses carrying the OPPM-PWM symbols are converted to electrical signals. In this way, the duration of a OPPM-PWM symbol is 50ns, thus the corresponding bit rate is 80Mbit/s. In fact, the encoding of OPPM-PWM symbol can be changed according to the requirement of bit rate. For example, if the pulse-width of a 16-order OPPM-PWM symbol is set as 40, 50, 60, 70, and 80ns respectively, the duration of a symbol is 80ns and 4 bits is represented by a symbol, thus the bit rate of 50Mbit/s can be obtained.
The measured waveforms of OOK and 16-order OPPM-PWM modulation optical signals are shown in Fig. 6. Figure 6(a) and Fig. 6(b) are the waveforms of 50Mbit/s and 80Mbit/s OOK modulation optical signals. Figure 6(c) and Fig. 6(d) are waveforms of 16-order OPPM-PWM modulated optical signals when the duration of one symbol is 80 ns and 50 ns, corresponding to the transmission rate of 50 Mbit/s and 80 Mbit/s respectively. Figure 6(e) and Fig. 6(f) are the magnified waveforms of Fig. 6(c) and Fig. 6(d). It is observed that the output optical power of LED is −9.5dBm for 30Mbit/s OOK modulation, while decreased to −10.5dBm for 50Mbit/s OOK modulation. For OPPM-PPWM modulation of LED, the optical power maintains about −8.5dBm with bit rate up to 80Mbit/s. Figure 6(b) shows that the modulation depth decreases significantly when the bit rate becomes higher. However, for OPPM-PWM modulation, the modulation depth remains almost the same with bit rate up to 80Mbit/s. Figure 6(d) indicates that transmission rate of 80Mbit/s with good signal quality can be obtained by OPPM-PWM modulation.
Fig. 6 Received waveforms of OOK modulated optical signals(a, b) and 16-order OPPM-PWM modulated signals (c, d) after transmitting 3.0cm in free space. Left: 50Mbit/s. Right: 80Mbit/s. (e, f) are the magnified figures of (c, d).
For OPPM-PWM modulated optical signal, it is not hard to understand that the eye-diagrams measured by the oscilloscope is completely closed even if the signal quality is good, because that the position and width of the pulse changes with the 16 possible symbols. Nevertheless, the eye-diagram of the symbols is clearly open by picking out and shifting the pulses with the same width. In this way, the eye-opening and Q factot can be determined by the symbols with the minimum pulse-width. In Fig. 7, Q factors of the received light after propagating different distance in free space are shown. It can be seen that, 80Mbit/s opical signals from OPPM-PWM modulated LED can transmit about 15cm in free space with Q factor of about 5.0, i.e., BER of 10−6. But for OOK modulated LED, 50Mbit/s or lower bit rate optical signal can transmit 15cm with BER of 10−6. When the distance is longer, power of the optical signal from LED is very low and the channel noise is dominant in the received signal, so the difference in Q factor and BER is not obvious for different modulation signal. Considering the optical power of available visible LED can be 20dB higher than the LED we used, the prolonging of the transmission distance by the light focusing apparatus, and the coding gain of GBF algorithm we propose, we can expect that 80Mbit/s OPPM-PWM modulated optical signal can transmit 10 meters in free space, satisfying the requirement of indoor communication.
5. Encoding of OPPM-PWM modulation symbol different than natural way
In Table 1, the 16 possible OPPM-PWM modulation symbols are encoded into 4 bits in a natural way as described in section 2. However, the natural coding cannot guarantee the lowest BER. Ungerboeck introduced the trellis-coded modulation to deal with the similar problem in order to achieve maximum free Euclidean distance and improve error performance [16]. If the brute force way is used to find the maximum free Euclidean distance for 16 possible OPPM-PWM modulation symbols, 16 factorial process calculation is needed, which will overwhelm the ability of computer. The better way to achieve maximum free Euclidean distance will be explored in our near future work. Here, we introduce one modified encoding of OPPM-PWM modulation symbols different than the natural way, as listed in Table 2. The encoding in Table 2 is obtained by comparing the BER of 3000 encoding arranged in a random way. The systematic way to optimize the encoding will be researched in the future.
Table 2. Modified encoding of OPPM-PWM modulation symbol
The left in Fig. 8 shows the BER for the OPPM-PWM modulated LED communication system. For the system employing LDPC codes at the transmitter, the GBF algorithm is used at the receiver. It is found that, with modified encoding of the OPPM-PWM modulation symbol, the BER is lower than that with natural coding, no matter whether error correction scheme is employed or not. To define the ratio of BER measured from the modified encoding system over that from natural encoding system, the right in Fig. 8 shows the improvement achieved by the modified encoding. For OPPM-PWM modulated LED communication system without any error correcting, BER can be lowered to 70% of that with natural encoding. If LDPC codes and GBF are employed, BER can be further lowered to 60% of that with natural encoding. When BER is even lower due to high SNR, the improvement of modified coding becomes small.
Fig. 8 BER improvement achieved by the modified encoding of OPPM-PWM symbols. Left: Comparison between BER of modified encoding and that of natural encoding. Right: The BER ratio of modified encoding over natural encoding.
6. Conclusion
In this paper, combination of OPPM and PWM modulation format was researched to improve the transmission rate of LED indoor communication system, and the GBF algorithm at the receiver was proposed to improve the performance.The result showed that the proposed GBF algorithm has 2.0 dB coding gain over the standard BF algorithm at BER of 10−5 with the similar complexity. Proof experiment of LED communication system has demonstrated the improvement of transmission rate and signal quality by the scheme we propose in this paper. It is expected that 400 Mbit/s transmission rate can be achieved with a single available LED, thus transmission rate beyond 1 Gbit/s is foreseen by RGB LEDs. Moreover, better performance of the system can be achieved if the block length of LDPC coding increases. By optimizing the coding of OPPM-PWM modulation symbols, the BER of LED communication system is lowered. The results in this paper make sense for the bandwidth limited LED indoor communication system to upgrade the transmission rate and improve the performance.
Acknowledgments
This work was financially supported by the National Natural Science Foundation of China under Grant 60972017, 60978007 and 61177067, and State 863 Plans under Grant 2012AA121604.
References and links
1. E. F. Schubert and J. K. Kim, “Solid-state light sources getting smart,” Science 308, 1274–1278 (2005).
2. E. Cho, J. H. Choi, C. Park, M. Kang, M. Shin, Z. Ghassemlooy, and C. G. Lee, “NRZ-OOK signaling with LED dimming for visible light communication link,” in Proc. NOC (2011), pp. 32–35.
3. G. Ntogari, T. Kamalakis, and T. Sphicopoulos, “Performance analysis of space time block coding techniques for indoor optical wireless systems,” IEEE J. Sel. Areas Commun. 27, 1545–1552 (2009).
4. L. Zeng, D. O’Brien, H. L. Minh, G. E. Faulkner, K. Lee, D. Jung, Y-J. Oh, and E. T. Won, “High data rate multiple input multiple output (MIMO) optical wireless communications using white LED lighting,” IEEE J. Sel. Areas Commun. 27, 1654–1662 (2009).
5. Z. X. Wang, C. Y. Yu, W. D. Zhong, J. Chen, and W. Chen, “Performance improvement by tilting receiver plane in M-QAM OFDM visible light communications,” Opt. Express 19, 13418–13427 (2011).
6. Z. X. Wang, C. Y. Yu, W. D. Zhong, J. Chen, and W. Chen, “Performance of a novel LED lamp arrangement to reduce SNR fluctuation for multi-user visible light communication systems,” Opt. Express 20, 4564–4573 (2012).
7. J. Vucic, C. Kottke, S. Nerreter, A. Buttner, K. D. Langer, and J. W. Walewski, “White light wireless transmission at 200+ Mb/s net data rate by use of discrete-multitone modulation,” IEEE Photon. Technol. Lett. 21, 1511–1513 (2009).
8. J. Vucic, C. Kottke, K. Habel, and K.-D. Langer, “803 Mbit/s visible light WDM link based on DMT modulation of a single RGB LED luminary,” in” Proc. OFC/NFOEC (2011), paper OWB6.
9. H. M. H. Shalaby, “Capacity and cutoff rate for optical overlapping pulse-position modulation channels,” IEEE Trans. Commun. 43, 1284–1288 (1995).
10. R. G. Gallager, Low-Density Parity-Check Codes (MIT Press, 1963).
11. X. Li and M. R. Soleymani, “A proof of the Hadamard transform decoding of the belief propagation algorithm for LDPCC over GF(q),” in Proc. VTC (2004), Vol. 4, pp. 2518–2519.
12. B. Rong, T. Jiang, X. Li, and M. R. Soleymani, “Combine LDPC codes over GF(q) with q-ary modulations for bandwidth efficient transmission,” IEEE Trans. Broadcast. 54, 78–84 (2008).
13. D. J. C. MacKay, “Good error-correcting codes based on very sparse matrices,” IEEE Trans. Inf. Theory 45, 399–431 (1999).
14. J. Hamkins, “Performance of binary turbo coded 256-ppm,” TMO Progress Report (1999), pp. 42–138.
15. B. Bai, Z. Y. Xu, and Y. Y. Fan, “Joint LED dimming and high capacity visible light communication by overlapping PPM,” in Proc. WOCC (2010), pp. 1–5.
16. G. Ungerboeck, “Channel coding with multilevel/phase signals,” IEEE Trans. Inf. Theory 28, 55–67 (1982).
