Open Access
Add to BibSonomyAbstract
In this paper, we comprehensively study the NLOV UV turbulence effect through simulated and experimental results. A Monte Carlo NLOS UV turbulence channel model, which incorporates the effects of multiple scattering and turbulence attenuation, is proposed based on previous work. To validate this model, a series of outdoor experiments is conducted to investigate the received-signal energy distribution and channel path loss under the turbulence circumstance for the first time. These experimental and simulated results are valuable for studying NLOS UV channel and communication system design.
© 2015 Optical Society of America
1. Introduction
Given the unique propagation features of radiation in the deep Ultraviolet-C (UV-C) band which ranges from 200–280 nanometers, research on ultraviolet communication becomes increasingly popular. Atmospheric absorption prevents shorter wavelength of solar radiation from reaching the surface of the earth, which results in a virtual noiseless UV channel enabling the use of extremely sensitive photon-counting detectors for ultraviolet communication. Therefore, high received signal-to-noise ratio (SNR) can be easily achieved by using a very large field-of-view (FOV) photon multiplier tubes(PMTs) detector with a out-band optical filter on top. Meanwhile, significant atmospheric scattering in the UV-C band makes outdoor non-line-of-sight (NLOS) communication possible and eases the strict requirements on pointing, acquisition and tracking (PAT) [1]. As such, it is considered that NLOS UV communication theoretically can be used as a novel application to supplement conventional communication systems where radio-frequency (RF) communication is impermissible or undesirable.
Extensive research work have been performed, not only experimentally, but also theoretically on NLOS UV communication. Under the assumption of single scattering, analytical and tractable single scattering channel path loss models have been developed in [2–3]. Through a short range outdoor experiment, an empirical path loss model has been proposed in [4] and it also partially validated the correctness of single scattering channel path loss model. In [5–6], authors proposed a numerical multiple scattering channel path loss model based on Monte Carlo method (MC). After the verification of [7], this MC based multiple scattering model was shown to be more accurate than single scattering channel path loss model. Under the assumptions of those two path loss models, NLOS UV channel coding [9], modulation schemes [10], NLOS UV MAC, and network issues [11–13] have been also studied and discussed. Meanwhile, practical communication systems were built [14–15] for experimentation and validation of theoretical NLOS UV communication research.
However, these works only considered short communication range scenarios, in which turbulence effects were assumed to be negligible. In fact, as communication range and the index of refraction structure parameter increase, optical turbulence in the UV-C band may deteriorate the communication system performance with the effects of irradiance fluctuation (scintillation) and extra signal attenuation. Thus, author in [16] proposed an analytical model of NLOS UV turbulence channel for the first time. In paper [17–18], authors advanced the previous model and took extra turbulence attenuation into account. Both of those two models, however, were limited because they assumed that the photons were only scattered once before detected by the receiver.
In this paper, we propose a MC channel model to capture the multiple scattering channel behavior under turbulence condition. In addition, we present a serial experimental results and study the characteristic of NLOS UV turbulence channel with farthest distances up to 1 km. Through the experiment and simulation, we discuss the turbulence effect on NLOS UV channel with focus on received-signal energy distribution and channel path loss. In a addition, a special characteristic of NLOS UV channel is proposed and studied as well, which is turbulence strength trade off between path length and common volume size. To the best of our knowledge, this is the first experiment to study NLOS UV turbulence channel characteristic. The organization of this paper is as follows. In section 2, the NLOS UV turbulence channel model based on MC method is summarized and derived. In section 3, the NLOS UV turbulence channel test bed, experimental setup and experimental conditions are exhaustively introduced. In section 4, through field test data and simulated results, the effects of NLOS UV scintillation are examined, by the means of studying the received-energy probability density function (PDF), the NLOS UV turbulence channel path loss and a special characteristic of the NLOS UV channel. Lastly in section 5, some initial concluding remarks are given.
2. NLOS UV turbulence channel model based on monte carlo method
With the increasing of communication distance, optical turbulence effects may degrade UV communication performance because the fading irradiance significantly deteriorates the received signal in two aspects: received energy fluctuation and extra average path loss. According to the Rytov solution to the wave equation, the log-amplitude variance is proportional to λ−7/6 [19]. As a result, the irradiance fluctuations due to atmosphere turbulence might be two or three times worse in the UV band than in the visible or infrared (IR) band, implying that UV links may be much more sensitive to turbulence compared to other optical links.
However, few analytical or numerical channel models have considered the NLOS turbulence effect comprehensively. In this work, we extent previous MC multiple scattering channel [5] and introduce some critical parameters for turbulence channel. Figure 1 depicts the scenarios of multiple scattering. Assuming there are n times of scattering, one n time NLOS path is comprised of 2n segments, including n paths from the transmitting point to the scattering centers and n paths from scattering centers to the Rx. Given each scattering is self-governed, and the distances and angles for different scattering events are dependent on previous quantities. Therefore, on each segment, the photon’s propagation is assumed to follow the law of single scattering until it reaches the next scattering center or arrives at the receiver. Following this theory, the distance between each scattering interaction is given by the random variable:
where ε(s) is a uniform random variable between zero and one, and ks is the scattering coefficient. In addition to scattering, the NLOS UV communication is also affected by turbulence. By takeing advantage of that UV LED array is a non-coherent source and each scattering center is spatially separated, each scattering center can be regarded as a secondary point source emitting photons independently so that we can apply the turbulence theory to each LOS path. According to the Rytov approximation [20–21], the scintillation attenuations (dB) of each LOS path can be expressed as where is the index of refraction structure parameter, which is altitude dependent. The typical value of is 10−13 m−2/3 for strong turbulence, 10−15 m−2/3 for medium turbulence, and 10−17 m−2/3 for weak turbulence, respectively. Thus, the corresponding turbulence attenuation of each path in linear scale is where Δs can be calculated from (1).
As well as additional signal attenuation, optical turbulence effects may also lead to irradiance fluctuations. According to our experience and field test results, there is an obvious turbulence effect once the baseline distance is over 200m and the atmosphere circumstance is approximately assumed to be weak-medium condition. Therefore, here we adopt log-normal (LN) distribution to describe the probability density functoin (PDF) of LOS received intensity. With the irradiance denoted by I, the log-normal model distribution is given by following:
where I0 stands for the mean of received irradiance and variance of the log amplitude fluctuation is defined as follows: for a plane wave, where k is the vector wave number (2π/λ).Without turbulence, the w, arriving probability of each photon, can be obtained by the simulation process described in [5]. However, when taking turbulence effect into account, the arriving probability of each photon in the arriving scattering center should scale a linear scintillation attenuation LαΔs and random irradiance variable T, as shown by:
where T follows the distribution (4). By merging those above mentioned critical variable into the MC based channel model in [5], the entire photon’s immigration behavior under the turbulence circumstance can be captured. Next, we apply this model to obtain all the simulated numeric results and use a serial of experiment to verify this new model in following section.3. NLOS UV turbulence channel test bed and experiment setup
In order to observe the obvious turbulence effect, we conducted a series of experiment from June 20, 2014 to June 23, 2014 during the daytime hours from 9 am to 9 pm. According to the weather records, the outdoor temperature ranged from 17.2° C to 33.3° C. The average wind speed during the testing was from 3.5 mph to 15 mph and the average relative humidity was 39.5%. Experiments were conducted in a flat open field of University of California, Riverside, which was located a significant distance away from any considerable optical noise sources (such as illumination devices). All of the measurements reported here were collected by utilizing a collimated UV LED array test-bed shown in Fig. 2.
At the transmitter side, a 6-pact 270 nm UV LED array was deployed with 1.8 mW average output optical power. On the top of each LED, we put a collimation lens to make the output full beam angle within 1°. The transmitting pulse width was set to 1 ms, ensuring enough photons were output from LED. Since high pointing angles geometries would bring a high mount of path loss and the output power of LED array was limited, we only used UV LED array to conduct the measurements on low pointing angles cases. Based on our simulation and observation, there was no saturation issue [22] occurred in our testing cases.
Synchronization between the transmitter and the receiver was achieved via two synchronized GPS CNS II clocks that each output one 100 pulse per second (100 pps) signal. Since the frequency of atmosphere turbulence is very close to the GPS rate, each transmitted signal can be regarded as independent instance. At the receiver side, with a 15% efficiency 270 nm solar-blind filter on top, a PerkinElmer MP1922 photomultiplier tube (PMT) was deployed, which output a standard transistor-transistor-logic pulse when a photon was detected. It was responsive to wavelengths from 165 nm to 320 nm with 10 dark counts per second. The peak quantum efficiency of 15% and the peak gain of 106 occurred at a wavelength of 200 nm. The quantum efficiency decreased to 10% at 260 nm and 7% at 280 nm. The noise level of this PMT is under 10 count per second. Assuming 15% efficiency for filter and 10% quantum efficiency for PMT can be reached, another attenuation 18.24 dB was taken into account in our calculation. The digital PMTs had a circular sensing window with a diameter of 1.5 cm (resulting in an active area of 1.77 cm2). Based on measurements, the effective field of view (FOV) of each detector (combining the PMT, solar blind filter, and attenuation filters) was estimated to be 30°. The resulting signal output from PMT was recorded by MSA300, a high-speed photon counter with minimum 5 ns counting time and zero dead time, triggered by the GPS clock with an appropriate propagation time delay, which can be observed from the oscilloscope.
4. NLOS UV turbulence study through simulation and experiment
In order to roughly quantify how turbid the atmosphere was, before each geometry data collecting, we first measured the LOS signal and then fitted the received-signal distribution to a log-normal (LN) distribution with as the only variable. By this, we can get the estimated values of at that moment, which are presented in below table. To make our simulation more accurate, those data estimated values will be adopted as reasonable parameters in following simulation.
Table 1. Estimated based on field test data.
4.1. Irradiance fluctuations PDF
Due to the lacking of enough sample size, paper [23] could not give out a solid conclusion about the received energy distribution. Thus, to ensure enough sample size, we transmitted 1000 pulses in each configuration, in which the pointing angles are low. Then we collected the received photon number of each pulse and use them to plot the PDF of normalized received energy.
Figures 3(a) and 3(b) present two groups of original normalized field test results. Their configuration parameters are (r, θ1, θ2)=(400m, 10°, 10°) and (r, θ1, θ2)=(350m, 10°, 15°), respectively. For comparison, we also plot the corresponding fitting curves of experimental and simulated results in Fig. 3(c) and 3(d). It can be seen that the experimental results and simulated results are very close and the general features appear to be captured by the proposed model, although there are some small mismatch between the theoretical and measured PDFs in those specific cases. These experimental and simulated results prove that the received-signal also follow LN distribution even in NLOS scenarios when the pointing angles are low and the atmosphere is in weak-medium turbulence condition. Based on this, we predict that if the turbulence is very strong, the received signal might follow Gamma–Gamma distribution or K distribution.
Limited by our LED output optical power, we cannot receive enough photons to draw a figure with enough sample size when the pointing angles are relative high. In that cases, the number of received photon would drop into Poisson domain and dominated by shot noise. But if enough photons can be detected by the receiver, LN distribution will play a more significant role demonstrated by our results.
4.2. Turbulent NLOS UV channel path loss
Since UV LED has more stable output optical power and higher bandwidth than UV laser, we can obtain more accurate path loss results with less time by using UV LED array. Therefore, in our experiment UV LED array was used to measure the turbulence NLOS UV channel path loss.
Note that each measured received-energy exhibits random variations due to the probabilistic nature of the received signal. This can be especially problematic in configurations where the detected signal is weak, resulting in large differences between each measurement. As such, we averaged over 1000 realizations per geometry in order to obtain an accurate average path loss for each configuration.
Our previous work mainly focus on the average path loss without turbulence effect. But once taking it into account, the extra path loss become significant and can not be ignored. To investigate whether turbulence effect plays impact on the overall path loss, we conducted the experiment at different time, but with same system configuration. Figure 4 shows two groups path losses measured on daytime and nighttime respectively. Two groups system configuration is (θ1, θ2)=(10°, 10°) and (θ1, θ2)=(10°, 50°) and the baseline distance ranges from 300m to 500m. Except for the time, all the other system parameters were set as the same. Clearly, the path losses measured on nighttime appear to be 1–2 dB less than the results measured on daytime. The reasonable explanation is that the turbulence effect in daytime was stronger than in nighttime. So the daytime corresponding turbulence attenuation was greater than nighttime turbulence attenuation, resulting in the gap of 1–2 dB. Those experimental result prove the hypothesis we made in the beginning—-turbulence effect has an effect on overall channel path loss.
Then, to study how turbulence strength affects the link path loss, we then simulated the path losses with different (from 10−16 m−2/3 to 10−14 m−2/3) and the path loss without turbulence. The simulated results are presented in Fig. 5, as well as the field test results as their comparison. As we can see, in both Fig. 5(a) and 5(b), the path loss presented by curve with is apparent higher than field test results, as well as other simulated curves. And in those two testing cases, the curves of the medium turbulence condition ( ) show extremely fitness to the field test results. Obviously, the simulated path loss results with weak turbulence ( ) and those assuming no turbulence occurred are lower than the other curves. In a word, all the other curves unveil that the path losses would vary 2–10 dB on different turbulence conditions.
These measurements, along with their simulated results prove again that it is necessary to take turbulence attenuation effect into account when building the model to estimate the NLOS UV channel path loss, as well as further system link budget estimation.
4.3. Study special characteristics of NLOS UV turbulence channel
The NLOS UV channel is a complicated stochastic process that it not only includes the atmospheric effect of scattering, absorption and turbulence. Different configurations of system and deployment positions will also lead to different channel responses. Through experimental and simulated data, we found a special characteristics of NLOS UV Turbulence Channel, which is the turbulence strength trade off between channel path length and common volume size. Therefore, next we further evaluate and analyze the received-signal scintillation distribution in terms of baseline distance and pointing angles.
First of all, figure 6 once again prove the conclusion we get before that the received signal energy will follow LN distribution when turbulence is from weak to medium. The field test results in Fig. 6(a) depicts the PDFs of normalized received energy for NLOS UV links with varying baseline range. The red curve (without asterisk) represents the PDF under the system configuration (r, θ1, θ2)=(500m, 10°, 10°). As its comparator, the blue curve (with asterisk) shows the PDF under the system configuration (r, θ1, θ2)=(300m, 10°, 10°). The only different system parameter is baseline distance. It appears that the signal will suffer more turbulence if it travels through longer channel. Obviously, the left flatter curve (the red one) deviates more from its mean than the other one. This phenomenon agrees with the LOS turbulence cases where the scintillation effect is proportional to the channel length. As such, stronger scintillation is supposed to be observed in a longer NLOS UV channel.
But figure 6(b) shows another interesting phenomenon and result, which doesn’t comfort to this. The normalized received energy PDFs for the corresponding different pointing angles are illustrated by Fig. 6(b). In particular, all the system parameters are fixed except for the pointing angles for two representative geometries, (350m, 0°, 0°) and (350m, 10°, 10°). From the system configuration parameters, it’s apparent that the channel length represented by blue curve (with asterisk) is longer than that represented by red curve (without asterisk). If applying the conclusion obtained from Fig. 6(a), the signal represented by blue curve (with asterisk) is suppose to appear more fluctuational than the signal represented by red curve (without asterisk). However, the normalized variance of red curve is 0.3652, which is larger than the other curve’s variance, 0.1802. This observed phenomenon is opposite to what has been concluded from Fig. 6(a). And this mitigation effect is a special characteristic of NLOS UV channel. It is stated in [24] that the incorporation of the different potential propagation paths from the transmitter to the receiver has the potential to mitigate the turbulence effects, resulting in, e.g., reducing irradiance fluctuations than previously predicted. Therefore, the bigger the size of common volume is, the more slightly the turbulence will affect the irradiance. Figure 6(b) provides experimental evidence to prove the existing of this phenomenon.
However, the above mitigation effect does not appear to be monotone as the change of pointing angles, illustrated by simulated results in Fig. 7, in which the normalized variances of the received-irradiance for various system geometries (here θ1 = θ2) are depicted. More specifically, the appearance of those curves look like to be concave and the minimum points are around at 20°.
Although the growth of pointing angles can enlarge the common volume size, which leads that there are more paths for photons to reach the receiver, it also increases the channel length meanwhile. Therefore, the final received-energy distribution is a trade off between channel length and common volume size. Based on above simulated results, we reasonably predict that when pointing angles are in a certain range, the mitigation effect from common volume will overwhelms than the effect from the increase of channel length. However, if the pointing angles continue to grow, the scintillation effect will still improve in the end and turbulence effect caused by long channel length will dominate even with the mitigation effect.
When building a end-to-end NLOS UV communication, the choice of system geometry also becomes a trade off, because of this special characteristic. The system BER is related to two aspects: 1) average received signal energy, 2) received signal’s variation. If we use low pointing angles, of course, the average received signal energy is stronger compared to high pointing angles. However, notice that higher average received signal energy might not lead to better system performance, because of stronger variation of signal. For higher pointing angles cases, it’s opposite, less energy, but might more stabler. Therefore, in term of best end-to-end BER performance, the choice of best system geometry is still an open research topic.
5. Conclusions and future work
We conducted a series of measurements to investigate NLOS UV turbulence channel for the first time. The received-signal distributions and NLOS UV turbulence channel path losses were studied in terms of system geometry and turbulence strength. The comparison between experimental and simulated results provided the validation for the NLOS UV turbulence channel model based on MC method. Through field test data and simulation results, the received-signal energy is proved to follow LN distribution in NLOS scenarios when the pointing angles are low and the atmosphere is in weak-medium turbulence condition. In addition, we also prove that turbulence effect will induce extra path loss for NLOS UV communication. The special characteristic of NLOS UV channel was also observed and analyzed through experiment and simulation. Because of this special characteristic, the choice of system geometry becomes a trade off between the size of common volume and channel length. Those above experimental and simulated analysis is valuable to fundamentally understand the NLOS UV turbulence channel, which is an essential prerequisite to the design of a NLOS UV communication system.
Further open work in this area include the outdoor experiment and model verification under severe turbulence condition. Due to the special characteristic of NLOS UV turbulence channel, the choice of best system geometry is still an open research topic.
Acknowledgments
This work was supported in part by the United States Army Research Office under grants W911NF-09-1-0293.
References and links
1. Z. Xu and B. M. Sadler, “Ultraviolet communications: potential and state-of-the-art,” IEEE Commun. Mag. 46(5), 67–73 (2008). [CrossRef]
2. D. M. Reilly and C. Warde, “Temporal characteristics of single-scatter radiation,” J. Opt. Soc. Am. 69(3), 464–470 (1979). [CrossRef]
3. Z. Xu, H. Ding, B. M. Sadler, and G. Chen, “Analytical performance study of solar blind non-line-of-sight ultra-violet short-range communication links,” Optics Letters 33(16), 1860–1862 (2008). [CrossRef] [PubMed]
4. G. Chen, Z. Xu, H. Ding, and B. M. Sadler, “Path loss modeling and performance trade-off study for short-range non-line-of-sight ultraviolet communications,” Opt. Exp. 17(5), 3929–3940 (2009). [CrossRef]
5. 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 Commun. 27(9), 1535–1544 (2009). [CrossRef]
6. 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, Atmospheric Propagation VIII, 803802 (2011). [CrossRef]
7. G. Chen, R. J. Drost, B. M. Sadler, and L. Liao, “Long-distance ultraviolet scattering channel measurements: analog vs. digital approaches,” Conf. on Lasers and Electro-Optics, JTu4A. 88, (2013).
8. G. Chen, F. Abou-Galala, Z. Xu, and B. M. Sadler, “Experimental evaluation of LED-based solar blind NLOS communication links,” Opt. Exp. 16(19), 15059–15068 (2008). [CrossRef]
9. Q. He, B. M. Sadler, and Z. Xu, “Modulation and coding tradeoffs for non-line-of-sight ultraviolet communications,” Proc. SPIE7464 (2009).
10. M. Noshad, M. Brandt-Pearce, and S. G. Wilson, “NLOS UV communications using M-ary spectral-amplitude-coding,” IEEE Trans. Commun. 61(4), 1544–1553 (2013). [CrossRef]
11. Y. Li, L. Wang, Z. Xu, and S. V. Krishnamurthy, “Neighbor discovery for ultraviolet ad hoc networks,” IEEE J. Sel. Areas Commun. 29(10), 2002–2011 (2011). [CrossRef]
12. L. Wang, Y. Li, and Z. Xu, “On connectivity of wireless ultraviolet networks,” J. Opt. Soc. Am. A 28(10), 1970–1978 (2011). [CrossRef]
13. A. Vavoulas, H. G. Sandalidis, and D. Varoutas, “Connectivity issues for ultraviolet UV-C networks,” J. Opt. Commun. Netw. 3(3), 199–205 (2011). [CrossRef]
14. 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]
15. L. Liao, B. M. Sadler, and G. Chen., “GPS synchronized UV USRP communication system,” Proc. SPIE 8874887409 (2013). [CrossRef]
16. H. Ding, G. Chen, A. K. Majumdar, B. M. Sadler, and Z. Xu, “Turbulence modeling for non-line-of-sight ultra-violet scattering channels,” Proc. SPIE 8038, 80380J (2011). [CrossRef]
17. Y. Zuo, H. Xiao, J. Wu, X. Hong, and J. Lin, “Effect of atmospheric turbulence on non-line-of-sight ultraviolet communications,” IEEE 23rd Int. Symp. on Personal Indoor and Mobile Radio Communications1682–1686 (2012).
18. H. Xiao, Y. Zuo, C. Fan, C. Wu, and J. Wu, “Non-line-of-sight ultraviolet channel parameters estimation in turbulence atmosphere,” Asia Communications and Photonics Conf, AF3G. 5, (2012). [CrossRef]
19. S. Karp, R. M. Gagliardi, S. E. Moran, and L. B. Stotts, Optical Channels: Fibers, Clouds, Water, and the Atmosphere (Plenum Press, 1988).
20. L. Dordová and O. Wilfert, “Laser beam attenuation determined by the method of available optical power in turbulent atmosphere,” J. Telecommun. Inform. Technol.(2), 53–57 (2009).
21. M. Al Naboulsi, H. Sizun, and F. de Fornel, “Propagation of optical and infrared waves in the atmosphere,” in Proc. of the XXVIIIth General Assembly of International Union Radio Science, New Delhi, India, (2005).
22. R. J. Drost, P. L. Yu, G. Chen, and B. M. Sadler, “Receiver dead time in non-line-of-sight ultraviolet communications,” in SPIE Defense, Security, and Sensing, Baltimore, MD, (2014).
23. G. Chen, L. Liao, Z. Li, R. Drost, and B. Sadler., “Experimental and simulated evaluation of long distance NLOS UV communication,” CSNDSP904–909 (2014).
24. P. Wang and Z. Xu, “Characteristics of ultraviolet scattering and turbulent channels,” Opt. Lett. 38(15), 2773–2775 (2013). [CrossRef] [PubMed]
