A space division multiplexed free-space-optical communication system that can auto-locate and fully self align with a remote transceiver

Free-Space Optical (FSO) systems offer the ability to distribute high speed digital links into remote and rural communities where terrain, installation cost or infrastructure security pose critical hurdles to deployment. A challenge in any point-to-point FSO system is initiating and maintaining optical alignment from the sender to the receiver. In this paper we propose and demonstrate a low-complexity self-aligning FSO prototype that can completely self-align with no requirement for initial manual positioning and could therefore form the opto-mechanical basis for a mesh network of optical transceivers. The prototype utilises off-the-shelf consumer electrical components and a bespoke alignment algorithm. We demonstrate an eight fibre spatially multiplexed link with a loss of 15 dB over 210 m.


Initial location Identification
The transceiver is designed around a motorised Altazimuth mount, commercially available from Celestron (NexStar Evolution) with approximately × − 5 10 6 radian step resolution that can be computer controlled. The RPi is used for the system control, and an Arduino is used as the interface device for communicating with the GPS and low-speed radio link operating at 446 Mhz (HiLetgo SI4463; 446 Mhz chosen to match UK handheld radio operational frequency regulations). There are two optical systems within the transceiver. The first is the optical system for managing initial optical alignment comprising a RPi 3 Near IR camera with integrated lens and a 650 nm band pass filter to mitigate stray light from sunlight and environmental background, which is placed off to the side of the main optical collection aperture, see Fig. 1. Alignment is facilitated by using an independent 650 nm laser diode source, slight offset from the camera and mechanically position such that it is approximately parallel to the main optical beam over a test range of 10 m. Second, the main collection system for the optical data signal comprising a 3-inch clear aperture AR coated 200 mm focal length lens, and a fibre array of 50 μm core fibres with 127 μm spacing to provide low complexity spatially separated independent channels Fig. 2(a,b). The fibres within this array are connected to independent SFP transceivers installed within Ethernet media converters to allow for performance testing.
The low-speed low-frequency radio link shares GPS coordinates, latitude (latA and latb for transmitter, U1, and receiver, U2, respectively) and longitude (lonA and lonb for U1 and U2 respectively), to allow for the initial horizontal and vertical mount orientations to be calculated. The Altazimuth mount requires two variables for controlling alignment, Azimuth (Az), the direction of pointing, and Altitude (Alt), for any height variation between the sender and receiver systems 24,25 . Azimuth is calculated by considering the change in latitude (Δlat) and longitude (Δlon), where The altitude angular tilt variable can be calculated by using the altitude information and GPS co-ordinates for the distance, d, using the formula Figure 1. A diagram of the prototype transceiver system that is based on a commercially available motorised Altazimuth mount that has control in both Azimuth and Altitude. To auto-locate, the systems communicate their GPS locations through a low-power radio receiver. Subsequent alignment stages utilise an alignment laser and low-resolution camera to optically align the system. The main aperture collects the light carrying data from another transceiver.
where R is the radius of the earth. Note that above atan2 Δ Δ ( lon, lat) is used in place of Δ Δ arctan( lon, lat), as arctan does not distinguish between angles that differ by π. Unfortunately, the GPS system has an inherent inaccuracy that arises from factors including satellite geometry, atmospheric conditions and other receiver design limitations 26 . The GPS reported position is a centroid of a normal position distribution where the error, σ, is defined with a probability of 95% 26 . This positional error region defines the spiral search area (SA) for our mechanical alignment procedure. The horizontal error for our GPS system (MTK3339 GPS chipset) was approximately σ σ = = .

optical Alignment
Inspired by inter-satellite and ground-to-space optical communications 2,3,27 , our bespoke alignment scheme requires each of the FSO units to perform a spiral search path to locate their partner unit, Fig. 3. From the rough GPS alignment the transmitter (U 1 ) is steered by the computer to points at the receiver (U 2 ), and vice-versa. Using a compact 650 nm laser diode source with a packaged collimating lens as the beacon, and a small CCD camera (Raspberry Pi NOIR Camera v2) the U 1 moves in spiral motion predefined by the GPS accuracy limit. The spiral path has a radial function defined as beacon Figure 2. (a) The physical system used for the system testing is based on a commercial telescope mount available from Celestron and a range of low-cost off the shelf components. (b) A diagram of the system components for one of the two transceiver units. www.nature.com/scientificreports www.nature.com/scientificreports/ where, θ is the azimuthal coordinate in polar form, a is a scaling parameter and w beacon is the approximate beam waist of the beacon laser at the receiver. Equation 5 is then used to calculate effective latitude and longitude for position at the receiver, where Eqs. 2 and 4 can be used to determine the direction of pointing of the receiver. The camera on U 2 monitors for the beacon light and reports back over the radio link whether it has received any light from the beacon laser. An initial quick scan is performed where U 1 scans for a predetermined amount of time. After this time, U 2 responds with the time at which light was received, and the U 1 then moves back to that position. If no light is detected the U 1 moves along the spiral path for another predefined amount of time, where this time is determined by the hysteresis of the mechanical mount. Once initial optical alignment is determined a second, more accurate, alignment process begins, where U 1 makes a single steps of * a w beacon , and waits for a response from U 2 on whether it is collecting optical power from the beacon laser. Due to using a small camera aperture, and directional nature of the laser beacon we define a digital region of interest, and intensity threshold value through computational image processing using the OpenCV library installed on the RPi micro-computer. When the intensity reaches the target, U 2 performs a mini spiral scan to locally maximise alignment with the beacon laser and then subsequently communicates with the U 1 to perform a tighter spiral, by reducing parameter a, and setting a higher intensity threshold value, see Fig. 4. This process is repeated three times to maximise the positional alignment of the beacon laser with the camera. To prevent saturation, the shutter speed of the camera is changed with each iteration of the alignment stage.
A secondary stage of alignment is carried out to maximise the power coupled specifically into the optical fibre. In this stage of the alignment, both transceivers are required to move sequentially to fully align the beam in both position and tilt. A first power search is carried out, where U 1 moves in a tight spiral pattern and U 2 is continually monitoring power searching of the maximum power position, which repeats with small step size until a threshold power is achieved. Then inspired by the mechanical process of beam walking, the U 1 performs one step along a tight spiral path, where at each step the U 2 scans to maximises the power received at a photo-diode connected to the output of the U 2 fibres.
Finally, as our system is designed to support spatial multiplexing and spatially separated up/down data streams at matched wavelengths, our final stage of alignment assures that channels are directly aligned with of multiple fibres at the transceiver. The laser is switched for the desired channel, and the cross talk determined between any neighbouring channels. If in-balance between channels is detected, the U 2 moves in a linear path, determined by the relative orientation of the multiple fibres with each of the transceivers. To minimise the number of degrees of control we need, it is assumed that the systems are leveled at the both sender and receiver as part of the initial setup of the tripods. If required further rotational alignment could be integrated, however we did not notice any considerable issues with levelling beyond that of the blue's eye spirit level integrated into the supplied tripod from Celestron.

experimental System and testing
The system was tested in two link scenarios, first at a distance of approximately 210 m between across the campus of the University of Glasgow, see Fig. 5, and secondly at a link distance of 12.8 m equivalent to a data centre type deployment, Table 1. Over the long distance urban link, both vibrational sensitivity and atmospheric turbulence are critical concerns for the link stability. For the 210 m link the atmospheric turbulence condition during our measurements was , by computing the seeing disk in the back focal plane of the main optical lens and by monitoring the intensity variance of the source over 8 seconds respectively. The transceivers did not include a high speed centre of mass tracking systems to mitigate these effects, therefore led to increase in the variance of collected optical power over the time of measurement, Fig. 6(c). To analyse the effect of these two affects, a high speed camera (Imaging Source Skyris M618) was placed at the fibre plane in the optical system to measure the centre of mass movement of the received beam. The centre of mass was measured for 8 s at 120 frames per second, where the angular deflection was subsequently calculated based on the optical system. The variation in angle was μ . 0 09 radians and μ . 0 20769 radians for tip and tilt respectively, see Fig. 6(a). For the environmental analysis a 850 nm fibre coupled laser source (Thorlabs LPS-830-FC) with 10 dBm of optical optical power was used as 1550 nm sources are not compatible with standard cameras and power meters could be used for the system performance analysis, the main optical lens was changed to an uncoated lens of equivalent focal length to the 1550 nm system. The system is expected to support approximately 5 spatial modes at a range of 210 m and was determined to support spatially separated up and down stream data signals. The relatively small number of available spatial frequencies means higher order turbulence was not measured during our system tests. Larger optical apertures could allow a greater number of spatial mode that one could multiplex, and would lead to more noticeable higher order turbulence that may require the integrating of adaptive optics.
Shorter systems, operating across streets or in data centre environments can offer access to a larger number of spatial modes without a considerable increase in the aperture size of the system. To analyse the system for use over shorter distances, the system was tested at a range of 12.8 m, Fig. 6(b,d). Both transceivers were equipped with fibres array with eight, 50 μm fibres spaced at 127 μm, to provide low complexity set spatial channels encoded in direction space similar to plane waves, Fig. 2(b) 12 . Alternating fibres were assigned to the up and down stream for the system, allowing for bi-directional performance to be assessed using a commercial performance tester (Ideal Network Unipro Mgig1), certifying a system performance of up to 1 Gbps. We further analysed the channel cross talk between the four spatially separately multiplexed channels showing nearest neighbour crosstalk of -26 dB, indicating the performance is suitable for use with advanced modulation schemes such as Quadrature Amplitude Modulation.
The radio system used to communicate between transceivers and mechanical movement time are the two key parameters that limit alignment time for the system. For each step it takes approximately 50 ms for the mechanical mount to receive the command and execute a movement. To limit this time for performing the initial alignment steps with the beacon laser pointer, a large spot size at the U 2 of around 25 cm is used, allowing an initial quick search to approximately cover the full search area. The hysteresis within the mechanical system alignment leads to a re-positioning error of approximately × − 6 10 4 radians. A further mechanical concern is the tracking error from moving the mount at speed, i.e. the error in final position, which was measured for a movement rate of 0.06 radians s −1 as . × − 2 6 10 3 radians and × − 5 10 3 radians, for azimuthal and altitude respectively. Therefore, at a distance of 210 m this error can lead to corresponding error in position of 0.55 mm and 1.05 mm. However, moving the system at considerably slower speeds and operating in a closed loop limits the effect and enables precision adequate enough for fibre to fibre alignment.  The overall link latency of the radio connection over the 210 m was measured to be approximately 500 ms. This is because of the low bitrate of the HC-12 transceiver and the limited processing power of the Arduino. A minimum time of one second is required to send the required 16-character message between the transceivers, process the message and then subsequently perform the requested movement of the mount, assuming no packet loss. Potential routes to increase the speed of alignment would be to use a higher frequency and therefore higher bandwidth radio communication module (such as commonly available 2.4 GHz modules which support up to 1 Mbps at lower transmission distances), mounts with lower hysteresis and the ability to dynamically vary the beacon laser pointer size according to the stage of the alignment process.

conclusion
In this paper, a system that can auto-locate and fully self align to a remote transceiver over a free-space channel was outlined. This FSO system utilised spatially separated fibres at the back focal plane of the main collection lens to provide SDM channels for increased link capacity. Although a more complex system can perform this task, this paper outlines the process and performance of a system using widely available and cost-effective commercial components. The proposed five-stage method was capable of establishing a high precision, full-duplex FSO link using an image sensor, laser pointer, controller board and some mechanical parts at distances of up to 210 m within ninety minutes. With further optimisation of the radio components of the link, this time could be reduced significantly. We expect our system could be deployed at ranges over 1 Km without considerable increase in alignment time, however, the number of SDM channels would be naturally limited without an increase in the system aperture and alignment precision. Although not as advanced as space-bourne or military systems, the device presented here could be readily replicated for use in developing nations to allow network access to unconnected communities, who are often within a few hundred metres of fibre and are not provided access due to infrastructure constraints 1 . In addition, a portable, self-aligning system would be highly advantageous in providing emergency connectivity in disaster situations where existing fibre infrastructure has been damaged 28,29 .