Mid-infrared (MIR) free space optical communication has seen renewed interest in recent years due to advances in quantum cascade lasers. We present data from a multi-wavelength test-bed operated in the New York metropolitan area under realistic weather conditions. We show that a mid-infrared source (8.1 μm) provides enhanced link stability with 2x to 3x greater transmission over near infrared wavelengths (1.3 μm & 1.5 μm) during fog formation and up to 10x after a short scavenging rain event where fog developed and visibility reduced to ~ 1 km. We attribute the improvement to less Mie scattering at longer wavelengths. We confirm that this result is generally consistent with the empirical benchmark Kruse model at visibilities above 2.5 km, but towards the 1 km eye-seeing limit we measured the equivalent MIR visibility to be > 10 km.
© 2009 Optical Society of America
1. Mid-IR free-space systems and the Kruse model
Mid-infrared (MIR) (8–10μm) quantum cascade laser (QCL) sources and detectors have recently improved in price and performance and are becoming a viable alternative to traditional near-infrared (NIR) (0.7-1.6μm) free-space optical (FSO) communication components. The interest in a MIR platform stems from an expected physical layer advantage of reduced light scattering and higher throughput in adverse weather, particularly fog. Recent publications comparing MIR to NIR sources have presented promising short-range results using quantum cascade lasers (QCL’s) in such conditions [1,2].
Attenuation of laser power in fog is principally due to absorption and strong Mie scattering of light by suspended water particles through which the beam must propagate. Absorption windows in the atmosphere pose an equivalent challenge and advantage to both NIR and MIR wavelengths. From this perspective it may be argued that no wavelength advantage exists . However, in the context of scattering, longer wavelength MIR light is known to produce lower losses as the result of a reduced size-to-wavelength ratio in hazes and fogs . The empirical Kruse model  may be used to characterize the FSO response. The model predicts the transmitted irradiance using the following expression:
Here τsi is the transmitted irradiance, x is the system propagation distance, λi the laser wavelength and V is “seeing-eye” visibility in kilometers. The exponent q was determined by Kruse to be based on the size and distribution of scattering particles and is a linear function of V when V is < 6km and otherwise a constant. The laser wavelength is represented by λi . Via this model, enhanced MIR over NIR transmission in fog can be calculated.
The visibility V is evaluated at 550 nm. From this observable the optical depth at any wavelength can be calculated based on the visible range on-site. Historically, the model has found agreement in many campaigns over a broad range of wavelengths. As it is so simplistic in treating atmospheric scattering there may be cases where it deviates from observation. The model cannot for example discriminate between atmospheric compositions, complex or simple particle type or size distributions.
Fog is known to have a bimodal droplet size distribution with the majority (>70%) of particles being (0.8–4)μm in size and the remainder at (5 – 10)μm . As MIR systems may be 6–8 times longer in wavelength than NIR systems, MIR light suffers from less resonant Mie-scattering losses in haze and fog than NIR light. Also, the growth of fog from haze, while not clearly understood, is known to have smaller particles dominate by number in early onset, which favors longer MIR system uptime at the onset of adverse events.
We present measurements from a coaxial, multi-wavelength test-bed that simultaneously compares NIR and MIR FSO systems over 550 m for six hours of fog in the New York City area. Adverse weather conditions of haze, fog and rain culminated in a final fog-only affected visibility of 1.05 km. We show that under all conditions of adverse weather encountered, the MIR source outperforms the NIR in general by a factor of 2x – 3x in power throughput. We validate our results using the FSO industrial benchmark Kruse model and demonstrate that the model and data both agree that greater throughput is possible with a MIR over NIR system. However, under low (< 2.5 km) visibility conditions, we measured the MIR system to have 10x greater transmission than is predicted by the model. We attributed this to the influence of a short rain event that impacted an established bimodal particle distribution. Though the visibility continued to decline, the theory could not account for the atmospheric impact or the greater MIR performance.
We constructed our 550 m outdoor, multi-wavelength, FSO optical link at Stevens Institute of Technology in Hoboken, New Jersey (Fig.1). The site was 2 km from mid-town Manhattan on the bank of the Hudson River. In order to ensure an equivalent atmospheric propagation channel for our measurement we co-axially aligned three laser wavelengths, (1.345, 1.558 & 8.1) μm. This was done by first combining the NIR lasers in a homemade fiber coupler and then co-aligning the MIR beam in free space using a Germanium window. The beam cross-sections and divergence were controlled to be equal for all beams and the bench-top output power was monitored to account for any drift over time. A Davis VantagePro2 Weather Station monitored the outdoor humidity, temperature, pressure and wind values in real time on site, 10 m from the emission and collection point.
The NIR lasers were independently AM-modulated (bias-T) and had matched average output powers of 20 mW. The MIR source was an 8.1 μm DFB quantum cascade laser (QCL) operated in pulsed mode, thermo-electrically cooled to -25°C with an average output power of 1 mW. The QCL was both electrically modulated and mechanically chopped, enabling a combination Lock-In-Box-Car acquisition scheme (SR 510, SR 250).
The beams were launched coaxially from a 1” plane gold mirror on the front of an f/8 Newtonian telescope towards a 5” gold coated hollow retro-reflector 275 m across the campus. The f/8 telescope collected the return signal. This was filtered with a second Germanium window to separate the NIR and MIR beams. The NIR detector was a ThorLabs PDA400 while the MIR was a liquid nitrogen cooled 2 MHz MCT (New England MPV11-0).
3. Experimental results and discussion
We present six hours of continuous multi-wavelength transmission measurements from October 19th, 2006. We recorded an overall decline in signal irradiance as the humidity rose from 70% (17:00) to 86% (22:45). By application of the Kruse model we estimate the actual visual range fell from 12 km to 1.05 km. Overall, NIR dB/km losses correlated to documented values . However, in the final hour we found the MIR data inconsistent with the model, giving a final optical depth of > 10 km.
3.1 Power transmission through haze and fog
Figure 2 shows the irradiance of each laser over the measurement period. Each data set was normalized to ensure a true comparison between wavelength performances. The attenuation sequence from blue to red (long to short wavelength) supports the prediction by Kruse and Mie theories that a longer wavelength source will propagate more effectively though micron sized suspended particles such as haze and fog. Reduced effective scattering is thought to be the principle mechanism for this effect.
At 22:00 we saw the onset of a short 2 mm/hr rain event. We believe this caused the MIR reversal by a meteorological process known as “washout” or “scavenging”. Larger MIR sensitive particles (5–10) μm are collected by larger falling rain droplets, nucleate, and then fall to the ground pushing smaller suspended NIR selective particles from their path . Thus the smaller particles remain and grow, while the larger particles are filtered out. To our knowledge this is the first report of MIR scavenging in the atmosphere for such a FSO link.
We found in general that > 2x more absolute power is transmitted for the MIR wavelength compared to the 1.558 μm and > 3x for the MIR over the 1.345μm. This MIR signal can be greater than traditional NIR wavelengths by a wide margin at the onset of fog (V = 1.05 km) by a factor of over 10 dB/km. While this leaves useful reserve power for a link budget it also translates into a two - three (peak > 10) fold increase in general for instrument range when compared to the NIR systems.
3.2 Kruse-Mie model comparisons
In Fig. 3 we present a double-log plot of the MIR against the two NIR transmission signals for the full period represented in Fig. 2. The laser irradiance was normalized for each wavelength and the natural logarithm was taken and plotted in comparative pairs as shown. The linear bisector (solid line) illustrates the case for equivalent attenuation of MIR vs. NIR wavelengths, i.e. no advantage to MIR over NIR. In this way we can compare relative attenuation coefficients in the Beer-Lambert law. We overlay the measured data with the Kruse model prediction using our wavelengths in equation 1. Each data point represents 100 m of increasing visibility, V, from 1 km on the bottom left to 12 km in the upper right. There is a slight disconnect in the top corner of the fit due to a change in the definition of q at 6 km.
One feature of this result is the fine agreement and then divergence of measured data with the model which begins at (-0.1, -0.4) 18:30, or when V = 2.3 km. This first process may be explained by a slowing of the particle formation rate as the humidity on-site decreased between 18:30 – 20:00 by 76% – 72% and thereby accounts for the deviation of the 1.558/8.1 μm observation from the prediction. We measure the average slope of data in this region to 20:00 as (2.31 ± 0.02) for the 1.558/8.1μm and (3.07 ± 0.02) in the 1.345/8.1μm.
The stronger deviation begins at 21:45 (V = 1.4 km) with the onset of the rain event. The reversal of slope in each case illustrates the loss of the larger MIR sensitive particles from the beam path but continued formation of smaller NIR sensitive ones. We believe that the increasing humidity 82 – 86% (21:45 – 22:45) provided the growth mechanism for this.
The final visibility based on the 1.345 μm NIR projection is 1.05 km. We use this value because the 1.558 μm result appears to plateau, possibly under the influence of the scavenging. The effective MIR visibility obtained in the measurement does not agree with this 1.05 km prediction, but it instead places the MIR visibility > 10 km. In this case we determine that the model in strong haze to early onset fog may be inadequate in quantifying all field measurements, especially under scavenging, but it can support a general MIR over NIR advantage.
The physical layer advantage of MIR over NIR light for FSO communication was investigated in adverse weather where visibility was reduced to 1.05 km. Our result shows that a MIR QCL consistently outperforms conventional NIR systems and reached a 2x to 3x gain in transmitted power. We attribute this to reduced scattering of MIR light in haze and fog compared with an equivalent NIR system. We validated our results using the Kruse model but found that its approximations were too rigid to account for bimodal fog under scavenging conditions. We concluded however that a MIR wavelength does performs better than NIR for FSO under adverse conditions with reduced visibility.
We wish to thank Claire Gmachl at Princeton University and George Wohlrab at Stevens Institute of Technology for their support. The U.S. Army through Picatinny Arsenal funded this project.
References and links
1. R. Martini, C. Bethea, F. Capasso, C. Gmachl, R. Paiella, E. A. Whittaker, H. Y. Hwang, D. L. Sivco, J. N. Baillargeon, and A. Y. Cho, “Free-Space Optical Transmission of Multimedia Satellite Data Streams Using Mid- Infrared Quantum Cascade Lasers,” Electron. Lett. 38, 181–183 (2002). [CrossRef]
2. C. P. Colvero, M. C. R. Cordeiro, and J. P. von der Weid, “Real time measurements of visibility and transmission in far-, mid- and near-IR free space optical links,” Electron. Lett. 4110 (2005).
3. E. Korevaar, I. Kim, and B. McArthur, “Debunking the recurring myth of a magic wavelength for free-space optics,” Proc. SPIE 4873155 (2002). [CrossRef]
4. E. J. McCartney, Optics of the Atmosphere Scattering by Molecules and Particles (John Wiley & Sons, New York1976).
5. P. W. Kruse, L. D. McGlauchlin, and R. B. McQuistan, Elements of Infrared Technology: Generation, transmission, and detection (John Wiley & Sons, New York, 1962).
6. H. Willebrand and B. Ghuman, Free Space Optics: Enabling Optical Connectivity in Today’s Networks (Sams, Indianapolis, 2001).
7. D. M. Chate and T. S. Pranesha, “Field Studies of scavenging of aerosols by rain events,” J. Aerosol Sci. 35, 695–706 (2004). [CrossRef]