Introduction
Non-line-of-sight ultraviolet (NLOS UV) communication is well matched to the problem of providing short-range links between wireless nodes, such as unattended ground sensors. The concept for NLOS UV relies upon two related phenomena. The first is that atmospheric gases, particularly ozone and oxygen, are strong absorbers of light in the spectral region between 200nm to 280nm, so that the amount of solar radiation reaching the ground in this waveband is negligible. In essence, it is always the darkest of nights in this “solar-blind” region of the spectrum, regardless of the time of day. The second related phenomenon is that in the solar blind region of the spectrum, UV light generated at the earth’s surface is strongly scattered.
Plots of the scattering, absorption, and extinction coefficients, as well as the normalized phase functions for both molecular and aerosol scattering, are shown in Figure 1. The testbed described in this article operates at a wavelength of 274nm corresponding to a nominal extinction coefficient of 1.1km-1. For ground-to-ground communication link geometries, the scattering angles of interest lie in the range of 90 to 135 degrees where isotropic molecular (Rayleigh) scattering dominates and the contribution from aerosol scattering is negligible.

Figure 1: Plots of the coefficients of absorption, scattering and extinction for combined molecular and aerosol contributions (left) and the normalized scattering phase function for aerosols at three UV wavelengths and for molecular Rayleigh scattering (right)

The concept of exploiting the absence of solar background and the strong scattering to realize non-line-of-sight (NLOS) communication links was first published by Sunstein[1]. A number of demonstration systems were subsequently pursued but never fielded, due in part to limitations of source technology, as well as operational concepts that demanded minimal operational ranges of several kilometers[2].

Figure 2: Depiction of a non-line-of-sight UV communication link between unattended ground sensors. The transmitter launches a broad UV beam upwards, and isotropic scattering by atmospheric gases returns a diffuse signal to a large area on the ground. Wide field of view receivers on the ground detect the flux from within this scattering volume and demodulate the message.

More recently, it was recognized that NLOS UV communication links are well-suited to the problem of providing short-range links between wireless nodes, such as unattended ground sensors[3-9]. Figure 2 illustrates the basic concept in which UV light from a zenith-pointed transmit beam is scattered isotropically by well-mixed atmospheric gases. As a result, a weak signal, in the form of scattered UV light, is returned to the ground in a radius around the transmitter. For a wide field of view (WFOV) receiver at ranges less than the UV extinction length (nominally 1km), the strength of the scattered signal returned to the ground decays inversely with range from the transmitter. A WFOV receiver placed anywhere within the effective radius of the scattered light intercepts a signal proportional to the active area of the detector. An important feature of these NLOS comm links is that terrain elevation changes and ground-level obstructions that would preclude ground-level line-of-sight (LOS) links, have no impact on the performance of the NLOS link. This insensitivity to ground-level obstructions suggests that NLOS UV comm links may provide higher availability than RF links, not only in undeveloped areas, but also in urban environments, where buildings and other ground-level obstructions often induce an RF propagation loss that is proportional to the sixth power of range. The key attributes of UV comm links, and examples of potential application domains, are summarized in Figure 3.

Figure 3: Attributes, functions, and representative scenarios for the application of non-line-of-sight UV comm links.

UV Transceiver Technology Transmitter Technology
Until recently, small, low-cost UV sources were limited to low-pressure mercury lamps, which are less than ideal for use in outdoor wireless communication links due to their temperature sensitivity, fragility, and limited modulation bandwidths. The DARPA Semiconductor Ultraviolet Optical Source (SUVOS) program, initiated in 2002, has focused on developing semiconductor LED and laser sources in the deep UV, including solar-blind LEDs. Several generations of UV LEDs have been developed under the SUVOS program. Figure 4 summarizes the measured performance of a 274nm 24-element array procured from Sensor Electronics Technology, Inc. Multiple array packages can be employed to implement higher-power sources, and we are presently assembling a 50mW source using this approach. These sources are research-grade LEDs, but with sufficiently long lifetimes to justify fabrication of multi-element arrays.

Figure 4: 274nm LED array mounted in TO3 package and associated drive electronics.

Figure 5: Perkin Elmer channel photomultiplier model MP-1921, and associated quantum efficiency as a function of wavelength

Receiver Technology
Detection of NLOS UV signals differs from conventional LOS FSO comm in two important respects. First, from the perspective of the detector, the signal source is not located at a point, but appears as diffuse molecular scattering occurring over a relatively wide field of view. Second, since the background flux in the solar-blind spectral region is negligible, quantum-limited detection is possible even with wide field of view optics. Consequently, photon-counting detectors with low dark count rates are required to achieve the best possible performance.
The two main components in the UV receiver are the detector and the visible blocking, or “solar-blind,” filter. For NLOS UV communication, the ideal detector should provide a large active area, a solar-blind spectral response, and an extremely low dark current density. Our testbed currently utilizes a Perkin Elmer MP-1921 channel photomultiplier (CPM) with an active area of 1.8cm2, and a dark-count rate of <10 cps. The CPM and associated quantum efficiency curve are shown in Figure 5. The CPM offers the ultimate in photon-counting detector performance for the testbed. However, in terms of size, power, and cost, semiconductor detectors offer potential advantages over photomultipliers. Consequently, there is interest in replacing the CPM with a semiconductor detector even though detection performance may degrade.
Detectors fabricated from wide-bandgap semiconductors such as aluminum gallium nitride exhibit a sharp spectral roll-off above 300nm that supplements the filter suppression of both visible and near-infrared light. In addition, the quantum efficiency of semiconductor detectors is significantly better than that of photoemissive materials. However, photoemissive detectors currently achieve 5 to 6 orders of magnitude lower dark count rate per unit area than semiconductor detectors. The dark count rate per unit area is a critical factor for NLOS communication, where background flux is negligible, and the signal flux is low. The higher dark currents in wide-bandgap semiconductor photodetectors presently preclude photon-counting implementations. However, avalanche photodiodes (APDs) are an area of active research and low dark-current APDs operated in Geiger mode may someday provide an alternative to the CPM detector.

Figure 6: Matching the UV LED sources to the spectral bandpass of the detector/filter combination improves throughput, which extends range. Overlaid on the bandpass of the solar absorption filter are the spectra of two different sources, one from Sandia National Laboratories and the other from Sensor Electronics Technology.

4.1 UV source-detector matching
In order to take full advantage of the absence of solar background below 280nm, the UV receiver must successfully block all detectable visible and IR flux outside of this band. The visible blocking filter employed in the testbed is a composite structure, consisting of a specially grown hydrated nickel sulfate crystal backed up by a near-IR blocking layer and an organic filter layer. The visible band transmission is below 10-10, however transmission increases to around 10-6 at 900nm. By combining this filter with a detector with low sensitivity in the near-IR, the overall rejection ratio exceeds 1010 in the visible through IR bands.
The spectral bandpass of the solar-blind absorption filter is shown in Figure 6, overlaid with the emission spectra of two different UV LEDs. Proper wavelength matching between the LED and the solar-blind filter can significantly improve throughput. The 5nm spectral mismatch between the SNL LED and filter passband shown in Figure 6 reduces the system throughput by a factor of two.

Figure 7: NLOS field measurements (solid lines) agree well with single-scatter model predictions (dotted lines). Differences between data and model predictions are due in part to variations in atmospheric scattering and absorption. Data were collected using a variety of source arrays. Weather influences the received signal, as shown by the differences between the 5/7/04 dataset (clear sky conditions) and the 6/1/04 dataset (heavy overcast).

Field-Measurements
Over a period of a year, field measurements have been made under a variety of atmospheric conditions with increasingly more powerful solar-blind UV source arrays. The array sources used and associated measurements are summarized in Figure 7. Also shown in Figure 7 are propagation model predictions of the received photons per pulse. Note that the propagation model predicts, and the data confirms, that the volumetric scattering associated with NLOS comm links leads to a range-dependent propagation loss proportional to R-1, rather than the usual R-2 or higher order exponential. The linear dependence of propagation loss on range is an important feature of NLOS comm, since it eases the constraints on the placement (range separation) of unattended ground sensors. From the data of Figure 7, we conclude that with appropriate waveform modulation and channel coding, the 24-element UV array can support communication links to 50m or more. The 50mW array currently under development will enable a ~10x increase in bandwidth over the same range, a 10x increase in range at the same bandwidth, or any proportional combination of the two.

Summary and Conclusions
A modular solar-blind UV communication testbed employing 274nm LED sources has been developed and employed to make field measurements. The measurements have shown good agreement with a single-scatter propagation model and lead us to conclude that UV LED arrays can support NLOS UV communication links at kilobit data rates and at ranges up to 100m using omni-directional, zenith-pointed beams. A 50mW array is under construction and will be used to demonstrate NLOS links at ranges of 100m and nominal bit rates of 1kbps.

Acknowledgements
This work was sponsored by DARPA under contract FA8721-05-C-0002. Opinions, interpretations, conclusions and recommendations are those of the authors, and not necessarily endorsed by the United States Government.

References
[1] D. E. Sunstein, “A Scatter Communications Link at Ultraviolet Frequencies,” Thesis, MIT, Cambridge, MA, 1968.
[2] B. Charles, B. Hughes, A. Erickson, “ An ultraviolet laser based communication system for short range tactical applications,” Proceedings SPIE Vol. 2115, pp. 79-86, 25-26 January 1994.
[3] G. A. Shaw, M. L. Nischan, M. Iyengar, S. Kaushik, and M. K. Griffin, “NLOS UV Communication for Distributed Sensor Systems”, Proceedings SPIE Vol. 4126, 2000.
[4] G. A. Shaw, M. L. Nischan, “ Short-Range NLOS Ultraviolet Communication Test Bed and Measurements,” Proceedings SPIE Vol. 4396, 2001.
[5] G. A. Shaw, A. M. Siegel, M. L. Nischan, “Demonstration System And Applications for Compact Wireless Non-Line of Sight UV Communication,” Proceedings SPIE Vol. 5071, 2003.
[6] A. M. Siegel, G. A. Shaw, J. Model, “Short-Range Communication with ultraviolet LEDs,” Proceedings SPIE Vol. 5530, August 2004.
[7] G. A. Shaw, A. M. Siegel, J. Model, M. L. Nischan , “Field Testing and Evaluation of a Solar-Blind UV Communication Link for Unattended Ground Sensors,” Proceedings SPIE, Vol. 5417, 2004.
[8] D. Reilly, D. Moriarty, J. Maynard, “Unique properties of solar blind ultraviolet communication systems for unattended ground sensor networks,” Proceedings SPIE, Vol. 5611, 2004.
[9] G. A. Shaw, A. M. Siegel, J. Model, D. Greisokh, “Recent Progress in Short-Range Ultraviolet Communication,” Proceedings SPIE, Vol. 5796, Orlando, FL, 2005.