Performance analysis of short-range NLOS UV communication system using Monte Carlo simulation based on measured channel parameters

Pengfei Luo; Min Zhang; Dahai Han; Qing Li

doi:10.1364/OE.20.023489

1. Introduction

With its unique channel characteristics, ultraviolet (UV) communication proves to be a natural candidate for short-range non-line-of-sight (NLOS) communication technology [1,2]. A review of the previous relevant studies leads to the summary of three approaches to UV NLOS communication channel modeling.

The first is an analytical model similar to the NLOS single-scatter propagation model in a prolate-spheroidal coordinate system [3,4] or simplified and modified single-scatter models [5–8]. Ref [8–10]. extended the existing single-scattering coplanar analytical model to noncoplanar geometry where the transmitter and the receiver cone axes can be pointed in arbitrary directions. This type of model is able to predict the path loss and describe the temporal characteristics of single scattered radiation, but can’t model solar UV irradiation. It isn’t widely used because for the small to medium elevation angles operation, the majority of the received photons are multiple scattering photons, whereas single scatter prevails during optical wave propagation when $τ < 0.1$ (short range and large elevation angles), where $τ$ is the optical depth defined as a product of the scattering coefficient and the range [6,11].

The second channel model uses Monte Carlo (MC) method to track the transportation of photons in the atmosphere medium [11–16]. And it is a multiple-scattering model used to improve the accuracy of channel simulation by modeling the migration of photons that experienced multiple scattering interactions. It is able to simulate many UV communication atmosphere channel characteristics. However, the MC algorithm is complex and time-consuming because the prediction model require selecting groups of parameter including atmospheric model parameters and parameters characterizing the measurement system, and thus some mismatch is expected (for example see Ref [17].).

The third channel model is an empirical path loss model for a communication range of up to a few hundred meters. This model is validated by field experiments [18,19]. It is based on extensive measurements of path loss by varying the transmitter (Tx) elevation angle, the receiver (Rx) elevation angle, and baseline separation. In [18,19], the author used analytical bit error rate (BER) performance of the corresponding UV system with simplified noise analysis, and gave on-off keying (OOK) and K-ary pulse-position modulation (K-PPM) performance.

For the performance simulation of communication systems, the MC method is one of the most common methods. It has several advantages over the analytical method and experimental method. First, MC method requires only a little mathematical analysis and can be applied to almost any communication system, while sometimes analytical method may not be a good solution [20]. Second, compared with experimental method, by using MC method the related parameters can be changed conveniently according to real configurations [21]. Therefore, the MC method is chosen in the study of performance analysis of short-range NLOS UV communication system.

In this paper, we are motivated by the recent channel model work to address the following questions. Is there a method of accurate system performance predicting that applies an efficient, accurate channel model with non-simplified treatment of noise? And can one predict BER performance of different modulation schemes including symbol-length-variable modulation such as differential pulse-interval modulation (DPIM) and differential PPM (DPPM) for an arbitrary geometry and communication range without conducting time-consuming experiments? It is expected that the questions will be addressed with our proposed method.

The organization of this paper is as follows. In Section 2, the UV communication system structure is presented and BER performance derivations are given between different modulation schemes. In Section 3, the experimental path loss and background noise measurements are provided and discussed. In Section 4, Monte Carlo simulation results of UV communication system performance are given. Conclusions and future research topics are presented in Section 5.

2. System structure and BER performances of four modulation schemes

The schematic diagram of our UV communication system is illustrated in Fig. 1 . The system comprises three basic components: the transmitter, the propagation channel and the receiver. In the transmitter, a UV LED is modulated to carry information. The output UV light is scattered through atmosphere channel to establish a NLOS communication link. At the receiver, a Photomultiplier Tube (PMT) aided by a solar-blind optical filter is employed for light detection.

Fig. 1 Non-line-of-sight UV communication system schematic diagram.

Download Full Size | PDF

As aforementioned, the 4 modulation schemes used in the performance analysis are OOK, PPM, DPIM and DPPM, among which the MC-based BER comparison will be conducted by using integrate-and-dump filter and threshold decision method. The integrate-and-dump filter integrates the received symbol over the bit interval T, dumps this value to the slicer and then resets to start with the next symbol again [22]. The BER derivations are provided here while the simulation results will be presented and discussed in Section 4. Denote the received signal by $y (t)$ which can be regarded as the convolution of the UV NLOS channel impulse response function $h (t)$ and the transmitted signal $x (t)$ corrupted by noise $n (t)$ [23]

y (t) = h (t) \otimes x (t) + n (t)

where x(t) is a non-negative and power-limited signal. Denote

s (t) = h (t) \otimes x (t),

and then Eq. (1) becomes

y (t) = s (t) + n (t) .

The ADC converts the analog signal

y (t)

to digital form

y [k],

so that

y [k] = s [k] + n [k] .

After the integrate-and-dump filter which is a matched filter for square pulses for maximizing the signal to noise ratio (SNR) in the presence of additive stochastic noise [22,24], the signal

z [k]

is given by

z [k] = (\sum_{k = 1}^{m} y [k]) / m = (\sum_{k = 1}^{m} s [k]) / m + (\sum_{k = 1}^{m} n [k]) / m

where k is the kth bit and m is the number of samples per bit interval.

According to the experimental results from [25], the full-width at half-maximum (FWHM) at the receiver is less than 300ns, when the elevation angles of both Tx and Rx are less than 60°. Consequently, the NLOS channel can be simplified to a path-loss-limited and inter-symbol interference (ISI) free channel [19], provided that the bit rate is less than 500kbps and the communication distance is less than 1Km. In intensity-modulation (IM) system, $x (t)$ is a constant number during one bit interval, the digital signal $s [k]$ accordingly remains the same during one bit interval, and also $(\sum_{k = l}^{m} s [k]) / m = m \times s [k] / m = s [k], k \in [1, m] .$ Let $S = s [k]$ ( $S = A$ if send “1”; $S = 0$ if send “0”) and $N = (\sum_{k = l}^{m} n [k]) / m .$ Thus:

z [k] = S + N

A more detailed expression of Eq. (3) is

z = S + N = {\begin{matrix} A + N \\ N \end{matrix} \begin{matrix}  \end{matrix} \begin{matrix} s e n d " 1 " \\ s e n d " 0 " \end{matrix}

where A and N represent the voltage of received bit “1” and the noise output of the filter, respectively.

According to the central limit theorem (CLT), the mean of a sufficiently large number of independent random variables, each with finite mean and variance, will be approximately normally distributed [12,13,26,27]. In this system, $n [k]$ is composed of solar irradiation noise and transimpedance amplifier (TIA) noise floor, and hence it is independent and identically distributed. Before we begin the BER derivations for 4 modulation schemes, we define the mean and the variance of sampled noise $n [k]$ as μ and $σ_{n}^{2} .$ Therefore, the mean of N is still μ, but the variance of N becomes to $σ_{n}^{2} / m .$

For an OOK system, if data rate is $r_{b},$ and sample rate is $r_{s},$ then OOK bite duration is derived as $T_{O O K} = 1 / r_{b},$ and the sample points per bit period are $m_{O O K} = r_{s} / r_{b} .$ Therefore, $N \sim N (μ, σ_{n}^{2} / m_{O O K}) .$

In the OOK case, the probability of error for this modulation is expressed as follows:

P (e) = P (1) P (e | 1) + P (0) P (e | 0) = P (1) \int_{- \infty}^{V_{T}} p_{1} (y) d y + P (0) \int_{V_{T}}^{+ \infty} p_{0} (y) d y

If equal number of 1’s and 0’s are sent, the optimal detection threshold can be obtained as

V_{T O O K} = A / 2 + μ

and $P_{O O K} (e)$ is

P_{O O K} (e) = Q (\sqrt{\frac{{(A / 2)}^{2}}{{(σ_{n} / \sqrt{m_{O O K}})}^{2}}})

For a K-PPM system, if data rate is $r_{b}$ and sample rate is $r_{s},$ then K-PPM slot duration is adopted as $T_{K - P P M} = (\log_{2} K) / (r_{b} \cdot K),$ and the sample points per bit period are $m_{K - P P M} = r_{s} \cdot T_{K - P P M} = [(\log_{2} K) / K] \cdot m_{O O K} .$ Therefore, $N \sim N (μ, σ_{n}^{2} / m_{K - P P M}) .$ In the K-PPM system, the number of 0’s is K-1 times more than that of 1’s, so that $P (1) = 1 / K,$ and $P (0) = (K - 1) / K .$ The optimal detection threshold can be obtained as

V_{T_K - P P M} = μ + \frac{A^{2} + 2 {(σ_{n} / \sqrt{m_{K - P P M}})}^{2} \ln (K - 1)}{2 A}

And the BER before K-PPM demodulation $P_{K - P P M} (e)$ is

P_{K - P P M} (e) = \frac{1}{K} Q (\frac{A + μ - V_{T_K - P P M}}{σ_{n} / \sqrt{m_{K - P P M}}}) + \frac{K - 1}{K} Q (\frac{V_{T_K - P P M} - μ}{σ_{n} / \sqrt{m_{K - P P M}}})

For a K-DPIM system, if data rate is $r_{b}$ and sample rate is $r_{s}$ , then K-DPIM average slot duration is expressed as $T_{K - D P I M} = 2 [(\log_{2} K) / (r_{b} (K + 3)],$ and the sample points per bit period are $m_{K - D P I M} = r_{s} \cdot T_{K - D P I M} = [2 (\log_{2} K) / (K + 3)] \cdot m_{O O K} .$ Therefore $N \sim N (μ, σ_{n}^{2} / m_{K - D P I M}) .$ In the K-DPIM system, the number of 0’s is on average $[(K + 3) / 2] - 1$ times more than that of 1’s, so that $P (1) = 2 / (K + 3),$ and $P (0) = (K + 1) / (K + 3) .$ The optimal detection threshold can be obtained as

V_{T_K - D P I M} = μ + \frac{A^{2} + 2 {(σ_{n} / \sqrt{m_{K - D P I M}})}^{2} \ln ((K + 1) / 2)}{2 A}

And the BER before K-DPIM demodulation $P_{K - D P I M} (e)$ is

P_{K - D P I M} (e) = \frac{2}{K + 3} Q (\frac{A + μ - V_{T_K - D P I M}}{σ_{n} / \sqrt{m_{K - D P I M}}}) + \frac{K + 1}{K + 3} Q (\frac{V_{T_K - D P I M} - μ}{σ_{n} / \sqrt{m_{K - D P I M}}})

For a K-DPPM system, if data rate is $r_{b}$ and sample rate is $r_{s}$ , then K-DPPM average slot duration is given as $T_{K - D P P M} = 2 [(\log_{2} K) / (r_{b} (K + 1))],$ and the sample points per bit period are $m_{K - D P P M} = r_{s} \cdot T_{K - D P P M} = [2 (\log_{2} K) / (K + 1)] \cdot m_{O O K} .$ Therefore $N \sim N (μ, σ_{n}^{2} / m_{K - N D P P M}) .$ In the K-DPPM system, the number of 0’s is on average $[(K + 1) / 2] - 1$ times more than 1’s, so that $P (1) = 2 / (K + 1),$ and $P (0) = (K - 1) / (K + 1) .$ The optimal detection threshold can be obtained as

V_{T_K - D P P M} = μ + \frac{A^{2} + 2 {(σ_{n} / \sqrt{m_{K - D P P M}})}^{2} \ln ((K - 1) / 2)}{2 A}

And the BER before K-DPPM demodulation $P_{K - D P P M} (e)$ is

P_{K - D P P M} (e) = \frac{2}{K + 1} Q (\frac{A + μ - V_{T_K - D P P M}}{σ_{n} / \sqrt{m_{K - D P P M}}}) + \frac{K - 1}{K + 1} Q (\frac{V_{T_K - D P P M} - μ}{σ_{n} / \sqrt{m_{K - D P P M}}})

For variable symbol length modulation schemes like K-DPIM and K-DPPM, the ensuing demodulation will be affected if there is one decision error in any bit. For example, sending “00011011” would be transmitted as: “10100100010000” after the 4-DPIM modulation. A one bit error burst in the underlined bits would be received as “10100101010000”, which becomes “0001000011” after 4-DPIM demodulation. Compared to the original bits, all underlined bits of the received bits are errors. Therefore, if symbol-length-variable modulation is employed, the corresponding performance degrades, for the number of bits per frame increases.

3. Experimentally measured path loss and background noise

3.1 Experimental setup

Our experimental measurements are based on the solar blind UV communication test-bed as depicted in Fig. 2 . The transmitter includes a microphone, a voice compression circuit, a modulating and coding model, an LED current driver, a 265nm LED array, and a DC power supply. The receiver consists of a solar blind filter, a solar blind PMT, a high voltage power supply socket, a TIA, an 8bit ADC circuit, a demodulation and decoding model, a voice decompression circuit, a speaker, a DC power supply and a high speed oscilloscope.

Fig. 2 A test-bed for NLOS UV communications.

Download Full Size | PDF

Specifically, our LED array is made up of 36 UV LED chips with a nominal center wavelength of 265nm, an optical output power of 10mW and a full divergence angle of 6°. At the receiver, the UV narrowband filter has a center wavelength of 266nm, a diameter of 25mm and an FWHM of 15nm. Its transmission efficiency η_f is 20% at 266nm. The UV PMT is Hamamatsu R7154, with a TIA socket, and an active detection area of 1.92 cm². Its cathode sensitivity S_C is 62mA/W at 254nm, with a typical gain G of 1 × 10⁷. The field of view (FOV) of the PMT plus the filter is about 30°, while the TIA achieves 3 dB bandwidth of 5MHz with a gain of 22KΩ.

3.2 Path loss measurements

We detail the path loss measurements as follows. Define the LED optical output power as $P_{t},$ received light power as $P_{r},$ and hence the path loss is denoted by $L = P_{t} / P_{r}$ or $10 \log_{10} (P_{t} / P_{r})$ in decibels. By recording the TIA output voltage $V_{T I A o u t},$ we can calculate received light power $P_{r}$ before solar blind filter by Eq. (14)

P_{r} = \frac{V_{T I A o u t}}{R_{T} G S_{C} η_{f}}

where η_f is the optical filter transmission, S_C is the PMT cathode radiant sensitivity, G is the PMT gain, and R_T is the TIA gain in Ω.

In order to reduce the solar irradiation noise levels during the path loss measurements, we carry out this experiment at night, and employ the PMT directly without solar-blind filter. The measured results are shown in Fig. 3 . It is notable that the Rx elevation angle is fixed to 0° in Fig. 3(a), while the Tx elevation angle is set as 0° in Fig. 3(b). The path losses are measured by varying the Tx and Rx separation distance and Tx or Rx elevation angles, respectively. It is evident that the path loss increases by about 10 dB when Tx/Rx elevation angle pair varies from (20° - 0°) to (30° - 0°), while the path loss increases by only about 2 dB when Tx/Rx elevation angle pair shifts from (0° - 0°) to (0° - 50°).

Fig. 3 Measured optical path loss of a NLOS UV links employing a UV LED and a PMT (without solar-blind filter) under different distance and different Tx\Rx elevation angles.

Download Full Size | PDF

The empirical path loss model is employed in this paper [18,19], given by Eq. (15)

L = ξ r^{α}

The curving-fitting approach is used to derive the path loss exponent α, the path loss factor ξ, their values and the corresponding RMSE are listed in Table 1 and 2 below.

Table 1. Path loss exponent α, path loss factor ξ and the RMSE for fixed Rx elevation angles

View Table | View all tables in this article

Table 2. Path loss exponent α, path loss factor ξ and the RMSE for fixed Tx elevation angles

View Table | View all tables in this article

3.3 Noise measurements

We use the test-bed above to measure the background noise parallel to the path loss measurement. In addition, we will present the receiver's waveform and the statistical characteristics in two noise scenarios, namely, low and high noise cases [28].

In low noise case, when there is no light emitted to PMT, for instance, in a completely dark state or dark night, the output voltage of the TIA consists of thermal noise plus PMT dark current noise. The waveform for the low noise case is shown in Fig. 4 . Judging from its histogram, this kind of noise clearly follows a Gaussian distribution with a mean of 0.028 and a variance of 3.01 × 10⁻⁴.

Fig. 4 Measured receiver output noise waveform (a) and its histogram (b) under low noise situation.

Download Full Size | PDF

For the high noise condition, the PMT gain is set to 4.9 × 10⁵ when it is directly exposed to the sunlight through a UV filter, which makes the TIA output suitable to observe. The total noise is a combination of different noise contribution mechanisms: the thermal noise from TIA, the PMT dark current shot noise, the solar noise, etc. Its waveform is collected from our high-speed oscilloscope and shown in Fig. 5(a) .

Fig. 5 Measured receiver output noise waveform (a) and its histogram (b) under high noise situation.

Download Full Size | PDF

Figure 5(b) demonstrated the histogram of this kind of noise. The mean value is 0.35 and the variance value is 0.3027. Evidently, there is no known probability distribution function to fit in with this noise distribution; therefore, the BER performance analysis which is based on the Gauss distribution cannot be used directly. In order to minimize the effect of noise and make system BER performance easy to analyze, the integrate-and-dump filter is employed in our system. After received noise is processed by the filter, this unknown distribution noise is transformed to the normal distribution as we have discussed in Section 2.

As noted by G.A. Shaw [29], the UVC waveband can be called the “solar-blind” region, and it is always the darkest of nights in this region of the spectrum, regardless of the time of day. However, according to our experimental results, it is apparent that the sunlight will sometimes introduce serious errors into filter installed UV communication systems if no appropriate approach is used.

4. Simulation results and discussion

By observing the experimental results of the path loss and noise distribution of the NLOS UV communication system, there is obviously no existent analytic model suitable for the unique noise distribution and the framed modulation. Thus, the MC simulation method is employed here.

As noted previously, the basic system model is illustrated in Fig. 1, and the key parameters are listed in Table 3 corresponding to our experimental setup. It is assumed that the daytime noise or high noise stems from solar radiation and circuit noise, while the nighttime noise or low noise originates only from circuit noise, whose waveform and distribution are depicted in Section 3. Next, our experimental results and MC method [20] are applied to predict the system performance in terms of BER, such as the effects of path loss caused by communication range and Tx/Rx elevation angles, the transmitted optical power, the noise condition, the data rate and modulation methods.

Table 3. Typical UV communication system parameters

View Table | View all tables in this article

4.1 BER versus SNR

According to our test-bed, the transmitted optical power is set as 10mW and the PMT gain is set as 4.9 × 10⁵. OOK, 4PPM, 4DPIM, and 4DPPM are chosen as our system modulation schemes, and the data rate is 10kbps. The sampling frequency is 1MHz, and the sample number per OOK bit interval m_OOK is 100 to make sure that after the integrate-and-dump filter the noise follows Gauss distribution. As indicated in Eq. (6), (8), (10), and (12), the decision threshold is set to be the optimal value in simulation to make sure the best system performance is achieved.

Figure 6 presents the comparison of the system performance under different modulation schemes with experimental verification from Ref [30]. The simulated and measured results are in good agreement. It is evident that the OOK yields the best performance, while other modulation schemes have similar performance. This is because we use an integrate-and-dump filter before the decision threshold rather than using Direct Detection directly. Hence, the longer the pulse width, the smaller the filtered noise variance and the smaller the BER value. Figure 7 shows the comparison of the BER versus path loss performance with and without the integrate-and-dump filter under the OOK modulation scheme. This figure illustrates that, at BER = 10⁻³, the path loss differences are approximately 10 dB and 15 dB in cases of high noise and low noise respectively, which agrees with our analysis in Section 2.

Fig. 6 Simulated BER versus SNR performance comparison of different modulation schemes with experimental verification.

Download Full Size | PDF

Fig. 7 Simulated BER versus path loss performance comparison of OOK modulation scheme with and without integrate-and-dump filter, at 10kbps, 10mW Tx power in the presents of two noise situation.

Download Full Size | PDF

4.2 BER versus distance

From [12, 13, 28], it is clear that the coefficients of the absorption, scattering and extinction in the UVC band are higher than those in the other bands, so that even a minor distance variation will lead to abrupt change of the path loss. According to Fig. 3, the path loss increases by more than 10 dB as the distance grows from 2m −10m.

We set all the parameters in the same way as we did the experimental parameters listed in Table 3, particularly, the Tx and Rx pointing angle pairs are set to (30°, 0°) and (0°, 50°). Figure 8 demonstrates BER performance against distance for the high and low noise cases. The BER performance critically depends on elevation angle geometry and noise situation. For example, when there is low noise and pointing angle pairs of (30°, 0°), all kinds of modulation schemes can provide a range of more than 70 meters at a BER of 10⁻³ and data rate of 10kbps, but this range drops to less than 20 meters as in high noise case. In pointing angle pairs of (0°, 50°) for the high noise case, our system can provide a range of more than 60 meters at the BER of 10⁻³ and the same data rate of 10kbps, while the range in low noise case is more than 200 meters at the same BER. For fear of possible inaccurate results, the curves for the modulation schemes in this situation have been omitted.

Fig. 8 Simulated BER versus distance performance comparison of different modulation schemes at 10kbps, 10mW Tx power under (a) Tx/Rx elevation angles 30°-0°, (b) Tx/Rx elevation angles 0°-50°.

Download Full Size | PDF

4.3 BER versus data rate

As illustrated in Fig. 1, the TIA output is sent to an 8bit ADC and then passes through an integrate-and-dump filter. With the increase in the data rate, the sample number per one bit interval decreases simultaneously, and consequently, the filtered noise variance increases. In order to simulate a higher data rate, the ADC sample rate is set to 10MHz which can have enough sample points, even at the data rate of 1Mbps. Afterwards, the system BER performance is simulated against data rate in two noise cases. The MC simulation model we built has been employed to simulate 10⁷ random bits one time. If BER performances are simulated in the same path loss with different noise cases, 10⁷ random bits will not be sufficient for the two noise cases at the same time. Therefore, we set path loss at 69 dB for the high noise case, and 84 dB for the low noise case.

Figure 9 plots the BER against data rate for a UV communication system in high and low noise situations with attenuation of 69 dB and 84 dB. It can be inferred from Fig. 9 that OOK has the best BER performance. When the data rate is 10kbps and the path loss is 69 dB, the OOK BER performance is 2 orders of magnitude lower than others from less than 10⁻⁵ to 10⁻³ in high noise case. Same results occur from 10⁻⁴ to 10⁻² in the low noise case with the data rate of 10kbps and the path loss of 84 dB. We also observe in Fig. 9 that 4DPIM has the worst BER performance; 4PPM and 4DPPM have almost the same BER performance, while the 4DPPM's performance is superior to that of 4PPM when the data rate less than 10kbps.

Fig. 9 Simulated BER versus data rate performance comparison of different modulation schemes: (a) Tx/Rx elevation angles 30°-0° (path loss = 69 dB), high noise case. (b) Tx/Rx elevation angles 0°-50° (path loss = 84 dB), low noise case.

Download Full Size | PDF

4.4 Discussion

We showed that the UV LED-based communication system allows data rates on the order of kilobits per second and distance on the order of magnitude of 100 meters. In comparison, the system performance of the low noise case is better than the high noise case. This phenomenon suggests a better noise reduction method or forward-error-correction (FEC) is needed to bring end-to-end error rates to an acceptable level or to improve system performance.

5. Conclusions and future work

A large number of experimental path loss and solar noise measurements for short range NLOS UV communication channels have been presented based on UV communication test-bed. The UV communication system performances have been investigated in terms of path loss, distance, bit rate and bit error rate. Theoretical analytical and simulation results show that solar noise has a great influence on communication performance. The Results also reveal that the communication distance and the data rate can be improved by using the integrate-and-dump filter, and that these two parameters can be further improved by introducing more sample points per bit period through an ADC with higher sample rate.

Future work in this area includes characterization of fading channels in UV NLOS communication systems, research on interference suppression method, BER measurements in different geometry like non-coplanar geometry and more research on the UV channel model by combining theoretical and experimental methods.

Acknowledgments

This study is supported by NSFC Project No.61101110, Fundamental Research Funds for Central Universities (2011PTB-00-06), Beijing Municipal Education Commission Sci. & Tech. Programs (KM200610009014, KM201110009001). And the authors would like to thank Ms. Haixia Bao, not only for her skillful editing, but also for her numerous insightful comments on the content.

References and links

1. Z. Y. Xu and B. M. Sadler, “Ultraviolet communications: Potential and state-of-the-art,” IEEE Commun. Mag. 46(5), 67–73 (2008). [CrossRef]

2. D. E. Sunstein, “A scatter communications link at ultraviolet frequencies,” B.S. Thesis, MIT, Cambridge, MA, (1968).

3. D. M. Reilly and C. Warde, “Temporal characteristics of single-scatter radiation,” J. Opt. Soc. Am. 69(3), 464–470 (1979). [CrossRef]

4. M. R. Luettgen, J. H. Shapiro, and D. M. Reilly, “Non-line-of-sight single-scatter propagation model,” J. Opt. Soc. Am. A 8(12), 1964–1972 (1991). [CrossRef]

5. Z. Y. Xu, H. P. Ding, B. M. Sadler, and G. Chen, “Analytical performance study of solar blind non-line-of-sight ultraviolet short-range communication links,” Opt. Lett. 33(16), 1860–1862 (2008). [CrossRef] [PubMed]

6. H. P. Ding, G. Chen, A. K. Majumdar, and Z. Y. Xu, “A parametric single scattering channel model for non-line-of-sight ultraviolet communications,” Proc. SPIE 7091, 70910M, 70910M-6 (2008). [CrossRef]

7. H. Yin, S. Chang, X. Wang, J. Yang, J. Yang, and J. Tan, “Analytical model of non-line-of-sight single-scatter propagation,” J. Opt. Soc. Am. A 27(7), 1505–1509 (2010). [CrossRef] [PubMed]

8. H. F. Xiao, Y. Zuo, J. Wu, H. X. Guo, and J. T. Lin, “Non-line-of-sight ultraviolet single-scatter propagation model,” Opt. Express 19(18), 17864–17875 (2011). [CrossRef] [PubMed]

9. L. Wang, Z. Xu, and B. M. Sadler, “Non-line-of-sight ultraviolet link loss in noncoplanar geometry,” Opt. Lett. 35(8), 1263–1265 (2010). [CrossRef] [PubMed]

10. L. Wang, Z. Xu, and B. M. Sadler, “An approximate closed-form link loss model for non-line-of-sight ultraviolet communication in noncoplanar geometry,” Opt. Lett. 36(7), 1224–1226 (2011). [CrossRef] [PubMed]

11. H. Ding, G. Chen, A. K. Majumdar, B. M. Sadler, and Z. Xu, “Modeling of non-line-of-sight ultraviolet scattering channels for communication,” IEEE J. Sel. Areas Comm. 27(9), 1535–1544 (2009). [CrossRef]

12. R. J. Drost, T. J. Moore, and B. M. Sadler, “UV communications channel modeling incorporating multiple scattering interactions,” J. Opt. Soc. Am. A 28(4), 686–695 (2011). [CrossRef] [PubMed]

13. D. M. Junge, “Non-line-of-sight electro-optic laser communications in the middle ultraviolet,” M.S. Thesis, Naval Postgraduate School, Monterey, CA, December 1977.

14. R. J. Drost, T. J. Moore, and B. M. Sadler, “Monte-Carlo-based multiple-scattering channel modeling for non-line-of-sight ultraviolet communications,” Proc. SPIE 8038, 803802, 803802-9 (2011). [CrossRef]

15. H. Zhang, H. Yin, H. Jia, J. Yang, and S. Chang, “Study of effects of obstacle on non-line-of-sight ultraviolet communication links,” Opt. Express 19(22), 21216–21226 (2011). [CrossRef] [PubMed]

16. H. Ding, G. Chen, Z. Xu, and S. B. M., “Characterization and modeling of non-line-of-sight ultraviolet scattering communication channels,” in Proceedings of IEEE Conference on Communication on Systems Networks and Digital Signal Processing (IEEE, 2010), pp. 593–597.

17. H. Ding, G. Chen, A. K. Majumdar, B. M. Sadler, and Z. Xu, “Correction to “Modeling of non-Line-of-sight ultraviolet scattering channels for communication” [Dec 09 1535–1544],” IEEE J. Sel. Areas Comm. 29(1), 250 (2011). [CrossRef]

18. G. Chen, Z. Y. Xu, H. P. Ding, and B. M. 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]

19. Q. He, Z. Xu, and B. M. Sadler, “Performance of short-range non-line-of-sight LED-based ultraviolet communication receivers,” Opt. Express 18(12), 12226–12238 (2010). [CrossRef] [PubMed]

20. W. H. Tranter, K. S. Shanmugan, T. S. Rappaport, and K. L. Kosbar, Principles of communication systems simulation with wireless applications (Prentice Hall, 2004), Chap. 7.

21. J. Li, Y. Ma, Q. Zhou, B. Zhou, and H. Wang, “Channel capacity study of underwater wireless optical communications links based on Monte Carlo simulation,” J. Opt. 14(1), 015403 (2012). [CrossRef]

22. C. Langton, “Linear Time Invariant systems and Matched filtering,” http://www.complextoreal.com/tutorial.htm.

23. Q. He, B. M. Sadler, and Z. Xu, “Modulation and coding tradeoffs for non-line-of-sight ultraviolet communications,” Proc. SPIE 7464, 74640H, 74640H-12 (2009). [CrossRef]

24. E. Sackinger and J. Wiley, Broadband circuits for optical fiber communication (John Wiley & Sons, 2005), Chap. 4.

25. 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]

26. Wikipedia, “Central limit theorem,” http://en.wikipedia.org/wiki/Central_limit_theorem.

27. G. A. Shaw, A. M. Siegel, and M. L. Nischan, “Demonstration system and applications for compact wireless ultraviolet communications,” Proc. SPIE 5071, 241–252 (2003). [CrossRef]

28. P. Luo, M. Zhang, Y. Liu, D. Han, and Q. Li, “A moving average filter based method of performance improvement for ultraviolet communication system,” presented at the 8th International Symposium on Communication Systems, Networks & Digital Signal Processing, Poland, 18–20 July 2012.

29. 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, 62310C, 62310C-12 (2006). [CrossRef]

30. G. Chen, F. Abou-Galala, Z. Y. Xu, and B. M. Sadler, “Experimental evaluation of LED-based solar blind NLOS communication links,” Opt. Express 16(19), 15059–15068 (2008). [CrossRef] [PubMed]

	Tx elevation angles
Rx elevation angles	0°	10°	20°	30°
0°	α = 0.1833	α = 0.177	α = 0.169	α = 0.1398
0°	ξ = 29.85	ξ = 31.06	ξ = 32.76	ξ = 43.35
RMSE	0.6745	0.7587	0.4857	0.2048

	Rx elevation angles
Tx elevation angles	10°	20°	30°	40°	50°
0°	α = 0.1764	α = 0.18	α = 30.82	α = 0.1721	α = 0.1695
0°	ξ = 30.37	ξ = 30.28	ξ = 0.1778	ξ = 32.06	ξ = 32.31
RMSE	0.6341	0.7737	0.5638	0.6829	0.7230

Parameter	Value
Wavelength	265nm
Tx average power P_t	1-10mW
PMT cathode radiant sensitivity (265nm)	62mA/W
PMT Gain	4.9 × 10⁵
Transimpedance Factor	22KΩ
Optical filter transmission η_f	20%
Modulation methods	OOK\4PPM\4DPIM\4DPPM
Bit per frame	10bit
Sample number per OOK bit interval	100
FOV	30°
LED beam angle	6°

	Tx elevation angles
Rx elevation angles	0°	10°	20°	30°
0°	α = 0.1833	α = 0.177	α = 0.169	α = 0.1398
0°	ξ = 29.85	ξ = 31.06	ξ = 32.76	ξ = 43.35
RMSE	0.6745	0.7587	0.4857	0.2048

	Rx elevation angles
Tx elevation angles	10°	20°	30°	40°	50°
0°	α = 0.1764	α = 0.18	α = 30.82	α = 0.1721	α = 0.1695
0°	ξ = 30.37	ξ = 30.28	ξ = 0.1778	ξ = 32.06	ξ = 32.31
RMSE	0.6341	0.7737	0.5638	0.6829	0.7230

Performance analysis of short-range NLOS UV communication system using Monte Carlo simulation based on measured channel parameters

Abstract

1. Introduction

2. System structure and BER performances of four modulation schemes

3. Experimentally measured path loss and background noise

3.1 Experimental setup

3.2 Path loss measurements

3.3 Noise measurements

4. Simulation results and discussion

4.1 BER versus SNR

4.2 BER versus distance

4.3 BER versus data rate

4.4 Discussion

5. Conclusions and future work

Acknowledgments

References and links

Cited By

Figures (9)

Tables (3)

Equations (15)

Optics Express