Ultra-high-capacity wireless communication by means of steered narrow optical beams

Abstract

The optical spectrum offers great opportunities to resolve the congestion in radio-based communication, aggravated by the booming demand for wireless connectivity. High-speed infrared optical components in the 1500 nm window have reached high levels of sophistication and are extensively used already in fibre-optic networks. Moreover, infrared light beyond 1400 nm is eye-safe and is not noticeable by the users. Deploying steerable narrow infrared beams, wireless links with huge capacity can be established to users individually, at minimum power consumption levels and at very high levels of privacy. Fully passive diffractive optical modules can handle many beams individually and accurately steer narrow beams two-dimensionally by just remotely tuning the wavelength of each beam. The system design aspects are discussed, encompassing the beam-steering transmitter, wide field-of-view optical receiver and the localization of the user's wireless devices. Prototype system demonstrators are reported, capable of supporting up to 128 beams carrying up to 112 Gbit s⁻¹ per beam. Hybrid bidirectional systems which use a high-speed downstream optical link and an upstream radio link at a lower speed can provide powerful asymmetric wireless connections. All-optical bidirectional beam-steered wireless communication will be able to offer the ultimate in wireless capacity to the user while minimizing power consumption.

This article is part of the theme issue ‘Optical wireless communication’.

1. Introduction

The world of wireless communication is seeing exploding growth, and we increasingly depend on it in our professional and private life. We need to get and stay connected always and everywhere, including sometimes in urgent critical circumstances. The internet has become our lifeline, and the vast majority of communications run over it. We highly value being able to move freely, without being hampered by connections through wires, hence wireless communication has become our main modus of being connected. The majority of the traffic on the internet emerges from indoor locations (more than 85% has been reported), and today wireless local area networks such as the IEEE802.11-based ones (a.k.a. WiFi) are taking care of this. Also, public mobile networks, such as the earlier 2G (GSM) and 3G (UMTS) networks originally laid out for voice and instant messaging (sms), and the current 4G (LTE) networks supporting in addition higher data speeds, are handling much wireless traffic. The upcoming 5G networks promise another step forward, not only towards more than tenfold increase in data capacity per user, but also towards drastically reducing latency which becomes ever more important in time-critical machine-to-machine and human-to-machine communication (e.g. autonomous cars).

While radio-based communication technologies are fiercely trying to keep up with the booming demands for wireless connectivity, this challenge is becoming more and more difficult to meet. As stated in Cooper's Law (http://www.arraycomm.com/technology/coopers-law/), the enormous growth in wireless communication capabilities with a factor of about 1 million in the past 45 years has partly (by roughly a factor of 5×) been achieved by introducing ever more powerful signal modulation techniques in order to squeeze more data capacity in a limited frequency band. Another factor of 5× has been attained by introducing frequency division multiplexing techniques in order to fit these bands into a limited spectrum. And opening up more radio spectrum yielded another factor of about 25×. The remaining factor of 1600× has been achieved by so-called spatial multiplexing: by making the radio cells covered by an antenna ever smaller, more and more antennas can serve the same area. Thus, the number of users to be served by a cell decreases and each user can get higher capacity. Within each cell, with respect to the surrounding cells, the radio spectrum can be re-used (provided that cross-talk between the cells is managed, e.g. by not allowing the same frequency band to be used in the neighbouring cells). Basically, this spatial multiplexing approach is the key solution to meet exponentially growing capacity demands; halving the cell size doubles the network's aggregate capacity. It is enhanced by introducing ever smaller pico-cells.

This article discusses how optical technologies, next to having proven their great capabilities in wired communication networks (through optical fibres), also have a great potential in addressing the challenges in wireless communication. It in particular addresses how wireless communication by means of dynamically steered narrow infrared beams can disclose huge capacities to the users while also providing high energy efficiency and enhanced privacy.

2. Optical wireless communication

Despite all the above-mentioned efforts made by radio technologies, the ongoing growth of wireless communication traffic is driving the radio spectrum into congestion. The optical spectrum can offer a huge amount of extra room, with ample opportunities for high-capacity communication as has been impressively shown by optical fibre transmission techniques.

The spectrum of visible light, roughly from 400 to 700 nm, represents no less than 320 THz of bandwidth, way beyond what the THz radio technologies being investigated can attain. Light-emitting diode (LED) illumination systems have been introduced widely; next to giving light for illumination, a LED can also convey data after being modulated (albeit with restricted bandwidth). In its networked version (known as light fidelity (LiFi) [1]), it can offer higher capacities. Also the spectrum of infrared light as used for high-speed long-reach optical fibre communication, roughly from 1500 to 1600 nm, represents a sizable 12.5 THz of bandwidth. For the latter, there is a wealth of mature high-speed optical devices available (laser diodes, photodetectors, optical modulators, etc.) already developed by the optical fibre systems domain. Optical wireless communication may advantageously build on that legacy [2–4].

Moreover, infrared light beyond 1400 nm (as used for fibre communication) is ‘eye-safe’. To be visible, light needs to reach the retina of the human eye, as illustrated in figure 1. Beyond 1400 nm, light is largely stopped at the lens, and the remainder after that by the vitreous body inside, and hence does not reach the sensitive but vulnerable retina. The human vision process commands the eye to focus visible and near-infrared light on the retina, and thus may increase the light intensity on it with a factor up to 200 000× the intensity with which the beam enters the eye (see e.g. https://en.wikipedia.org/wiki/Laser_safety). Melanin pigments in the epithelium behind the photoreceptors absorb the majority of it, which can cause permanent damage by burns in the retina. The maximum permissible exposure (MPE) of light power into the eye, expressed in W cm⁻², depends on the wavelength and the exposure time; the MPE decreases for longer exposure times. Lasers can be classified according to safety risks; the safest is a Class 1 laser, which is safe under all conditions of normal use, including looking into the laser beam with the naked eye or using typical magnifying optics such as a telescope or a microscope with a limited aperture. As shown in figure 2, for a Class 1 laser, the maximum continuous wave (CW) beam power allowed for visible light is about 0.5 mW, whereas for infrared light beyond 1400 nm, the maximum CW beam power is 10 mW. In conclusion, eye safety regulations allow infrared optical beams in free space to carry much more power than visible optical beams; the eye safety standards (such as IEC60825 and its US-oriented variant ANZI Z136) allow up to 10 mW at wavelengths beyond 1400 nm, whereas for visible light, the limit is only a few tenths of mW (see e.g. https://en.wikipedia.org/wiki/Laser_safety).

Figure 1.

Light incident onto the human eye. (Online version in colour.)

Figure 2.

Maximum allowed CW powers for the laser classes 1, 2, 3R and 3B according to the standard EN 60825-1:2007 (note: only for static, point-like laser sources, i.e. collimated or weakly divergent laser beams) (see e.g. https://en.wikipedia.org/wiki/Laser_safety). (Online version in colour.)

As Einstein already pointed out, nothing is faster than light in vacuum (299 792 458 m s⁻¹; in air, due to its refractive index of 1.0003, it is slightly lower at 299 702 547 m s⁻¹). Inside silica fibre, light is slowed down by the fibre's refractive index of about 1.5, which implies that free-space optical beams are 50% faster. Moreover, the waveguiding in the core/cladding structure of a fibre leads to waveguide dispersion, which lowers the bandwidth. Free-space optical beams do not experience waveguiding, hence next to a 50% lower latency also can offer a higher bandwidth than can silica fibre. Radio waves also travel with the speed of light; however, latency is added by comprehensive signal processing which is typically required to squeeze a high data rate signal into the limited radio link's bandwidth.

Next to providing a wealth of extra spectrum, a major benefit of optical wireless communication is its ability to be confined to narrow beams of light. By means of optical elements such as lenses and mirrors, optical beams can be collimated into small cross-sections and stay narrow over a long reach. Optical beams are being used for linking satellites in outer space which are thousands of kilometres apart. Optical beams are par excellence suited to create pico-cells, which as Cooper's Law also pointed out are the key to meet exponentially growing traffic demands. Moreover, the confinement of an optical beam implies that it can be directed such that it only reaches the user(s) for which it is meant; other users cannot get it and hence cannot tap in. Hence communication by narrow optical beams provides inherently better privacy than the wider radio beams. And an optical beam directs its power only to the intended destination; it does not spoil it elsewhere. Hence communication by optical beams is inherently more energy efficient than that by radio beams. This high directivity also means that the user receives a significantly higher fraction of the signal power emitted by the transmitter station, and therefore the signal-to-noise ratio (SNR) is much better. As Shannon's Law states, the maximum attainable capacity of a communication link is given by C = B · log₂ (1 + SNR) in bits/second, where B is the bandwidth of the channel. With the increased SNR and the high bandwidth B of the optical wireless link, the link can carry a significantly higher capacity. At the expense of comprehensive phased array antenna structures supported by complex signal processing, also radio waves can be tailored into smaller shapes and get more directive. For 5G networks, so-called massive multiple input multiple output antennas are foreseen, where e.g. no less than 1000 antenna elements are needed to provide a directivity gain of 30 dBi with respect to an omni-directional antenna. For optical wireless communication, however, collimating the light beam by simple passive optical devices such as a single lens or concave mirror can easily yield a directivity gain which is much higher than 40 dBi.

Optical wireless links need a clear line-of-sight (LoS) between the transmitter and the user's receiver, which is a disadvantage with regard to radio wireless links operating at low to moderate frequencies (such as wireless LAN systems working at 2.5 and 5 GHz). Millimetre-wave radio systems, such as those at 60 GHz, have a wider bandwidth (some 7 GHz at 60 GHz carrier frequency), but also suffer from severe attenuation if their link does not have a free LoS. On the other hand, the necessity of LoS also means that optical wireless communication cannot penetrate walls, so it provides good privacy without additional encryption measures as are needed in wireless LAN links.

Indoor wireless communication is presently predominantly done by wireless LAN technologies, notably from the IEEE802.11 standards family (WiFi). For short reach (up to a few metres), also Bluetooth (IEEE802.15) is used as cable replacement. For the upcoming Internet of Things era, where myriad small devices doing all kinds of sensing and actuating functions are used to communicate, radio technologies are well suited, in particular as they typically do not need LoS. Optical wireless communication is therefore not foreseen to fully replace radio wireless communication, in particular not for indoors. It can, however, alleviate the tasks of radio wireless LANs and take over the heavy broadband traffic load from them, thus giving them more breath for performing their low bursty data rate duties such as for the Internet-of-Things functionalities.

For broadband wireless communication, in order to position the characteristics of optical wireless communication with respect to those of radio-based wireless communication, table 1 gives a (non-exhaustive) qualitative comparison based on system parameter values as typically used. Optical wireless communication is divided into two main categories: visible light communication (VLC) which is typically aimed at deploying LED illumination systems with a wide footprint for data communication as well, and infrared light communication deploying multiple narrow footprint-steered beams. For broadband radio-based wireless communication, the IEEE802.11 suite of standards has been widely adopted already and huge amounts of products are being deployed worldwide. VLC standardization efforts have recently started and first (prototype) products are entering the market, while beam-steered infrared communication is still in the research phase. In the remainder of this article, we will focus on beam-steered communication using narrow infrared beams, which offers a range of attractive properties as given in table 1.

Table 1.

Comparison of broadband wireless communication techniques.

	visible light communication	infrared beam communication	WiFi IEEE802.11
capacity per user	shared; 9.5 Gbit s−1 over 1 m	unshared; very high (up to 112 Gbit s−1 per beam)	shared; 802.11n (2.4 GHz): <600 Mbit/s 802.11ac (5 GHz): <6.93 Gbit/s 802.11ad (60 GHz):<7 Gbit/s
reach	short–medium (few m)	medium (<10 m)- long	medium
energy consumption	high (Watts); illumination on	low (per beam <10 mW, on-demand only)	base station >100 mW
carrier frequency, available bandwidth	400–700 nm; width 320 THz	1460–1625 nm (S + C + L); width 20.9 THz	2.4 GHz band: width 83.5 MHz; channels 20 or 40 MHz 5 GHz band: width 575 MHz; channels 20 or 40 or 80 MHz 60 GHz band: width 7 GHz
safety aspects	penetrates eye; if collimated: ≪1 mW	does not penetrate eye; collimated <10 mW	EM hypersensitivity issues
privacy	good (contained by walls; not by windows)	good (contained by walls; windows may be coated)	open; needs encryption
infrastructure	share LED illumination	new indoor (fibre) infra	electrical cable (Cat5) infra
standardization	first steps made,IEEE 802.11 LC study group	not yet	extensive, mature, keeps evolvingbackwards compatible
utilization	first products on market	laboratory phase	very wide spread; more WiFi devices than people on earth; >50% of Internet traffic through WiFi

3. Beam-steered indoor optical wireless communication

Focusing primarily on indoor communication, optical wireless communication by means of steered beams may be represented by the application scenario sketched in figure 3. In this context, ‘indoor’ means inside a building. ‘Outdoor’ would imply that variable atmospheric conditions which may dynamically impact the propagation of light have to be taken into account, such as rain, fog, haze, air and turbulences. For indoor, such dynamically varying conditions are not considered. Indoor scenarios may encompass a wide variety, such as a residential home, a hospital, an office building, a conference room, a museum, a waiting lounge at an airport, an airplane cabin, a cabin in a train or bus, an exhibition hall and a shopping mall.

Figure 3.

Indoor optical wireless communication employing steered narrow infrared beams. (Online version in colour.)

Within each room of a building, narrow optical beams, so-called pencil beams, are foreseen. These beams connect the wireless devices such as laptop computers, tablets, smartphones and video monitors within each room; they are envisaged to be so narrow that each beam serves only as a single device. Thus, each device gets an unshared communication link, of which the full capacity is continuously available for it as long as needed. From the moment it has been connected, it does not need to compete for capacity with other devices. By contrast, in wireless LAN systems, a WiFi router is serving a whole room, and in most cases even several rooms, out of a single antenna site. Hence, its capacity is shared among all the wireless devices in those rooms, implying that they are in competition and do not get a guaranteed capacity.

In each room, the pencil beams are emitted from a so-called pencil-radiating antenna (PRA). As a beam may be obstructed in its LoS path towards a device, there are multiple PRAs in a room, which enables to circumvent the LoS blocking from a certain PRA by choosing another PRA. The PRAs are fed via optical fibre lines, which run from a central site, the Central Communication Controller (CCC). By means of an optical cross-connect (OXC) the required connections to the respective PRAs in the various rooms are established. The CCC hosts all the intelligence for controlling the OXC, and it acts as the interface with the broadband public access network. It translates the access network signals into signals suitable for the indoor optical wireless communication and vice versa.

4. Optical beam steering

An essential task of the indoor optical wireless network is to provide independent accurate steering in two dimensions of each optical beam. In the literature, steering by tunable little mirrors employing micro-electrical mechanical system (MEMS) techniques has been proposed [5]. Another solution proposed is based on spatial light modulator (SLM) techniques employing programmable reflection or transmission gratings [6,7]. Both solutions need a steering element per beam, with its separate control signal. Scaling to many beams hence leads to a complex steering module with comprehensive control needs. Moreover, the steering is done by active elements, so local powering and maintenance of the steering modules are required. These modules typically are mounted on the ceiling and may not be readily accessible.

Steering an infrared beam with fully passive devices may be done by tuning the wavelength of the light in combination with wavelength-diffractive functions. The optical transmitters hosted in the central site can be equipped with tunable laser diodes which readily generate the optical signals at the wavelengths needed. Laser diode tuning speed is typically much faster than the speed with which MEMS and SLM devices can be tuned. Beam steering has been reported by means of multiple tilted Bragg gratings embedded in a holographic photo-thermo-refractive glass volume [8]; each grating is designed to diffract a specific wavelength into a specific direction. The number of beams which can be steered is limited by the number of gratings.

Diffraction gratings are well-known devices which change the direction of an incident light beam depending on its wavelength, as described by the grating equation $\sin \hspace{0.17em}i+\sin \hspace{0.17em}\phi =m\cdot \lambda /(n\cdot d)$, where i is the angle of incidence, φ the angle of deflection, λ the wavelength of the light beam, m an integer number representing the order of the deflection, n the refractive index of the surrounding medium and d the groove spacing of the grating. The deflection angle φ is a periodic function of the wavelength λ, where the period is called the free spectral range (FSR). The beam is being deflected in several orders, and for the wavelength λ being within the FSR of order m, the beam does not enter the neighbouring orders (m ± 1). The FSR of order m can be calculated to be $\mathrm{\Delta }{\lambda }_{\text{FSR},m}={\lambda }_1/(m-1),$ where λ₁ is the lowest wavelength in the FSR range. The beam steering is done in the angular φ-dimension by tuning λ. The maximum angular tuning range (Δφ)_max which is achievable by wavelength tuning in order m without overlap into neighbouring orders is given by [9]

$$\cos {(\mathrm{\Delta }\phi )}_{\text{max}}=1-\frac{\lambda }{d}\cdot \frac{m}{m-1}.$$

The beam inside a room needs to be steered in two dimensions for adequate coverage. Hence, we adopted a structure deploying two diffraction gratings which are orthogonally positioned with respect to each other, as illustrated in figure 4a [9]. The first grating diffracts the beam in direction φ and has a small FSR, so it operates in a high-order m and when tuning λ over a wavelength range Δλ, the beam sweeps repetitively across the φ range. The second grating, rotated by 90° with respect to the first one and having an FSR larger than Δλ, diffracts the beam in the orthogonal direction ψ. As a result, when tuning λ over Δλ, the beam spot wanders two-dimensionally across the area to be covered, as shown by the example in figure 4b (grating 1: 13.3 grooves mm⁻¹, m = 95, i = 80.7°, FSR = 16.3 nm; grating 2: 1000 grooves mm⁻¹, i = 49.9°, m = 1; tuning from λ = 1505 to 1630 nm, yielding two-dimensional angular tuning over 5.6° × 12.7°). The grating pair may also consist of two transmissive ones or of a reflective and a transmissive one.

Figure 4.

Passive two-dimensional beam-steered module using a pair of crossed gratings. (a) Grating-based two-dimensional beam steerer. (b) Moving the beam spot by tuning its wavelength. (Online version in colour.)

Assuming that M scanning lines are needed, the wavelength tuning range Δλ should encompass M FSR ranges, so

$$\mathrm{\Delta }\lambda =\sum _{m=m_{\text{max}}-M}^{m=m_{\text{max}}}\mathrm{\Delta }{\lambda }_{\text{FSR},m}={\lambda }_{\text{min}}\cdot \frac{M}{m_{\text{max}}-M}\text{hence}{m}_{\text{max}}=M\cdot \left(1+\frac{{\lambda }_{\text{min}}}{\mathrm{\Delta }\lambda }\right).$$

For example, assume that the area in the room to be covered is 1.5 × 1.5 m² and the beam diameter is 10 cm, then M = 1.5/0.1 = 15 scanning lines are needed. The maximum order m_max in which the first grating needs to operate when the available tuning range is from 1500 to 1560 nm is then m_max = 15·(1 + 1500/100) = 900. Such high orders are difficult to achieve with conventional gratings. They can be attained with an arrayed waveguide grating as shown in figure 5, for which the grating equation becomes $d_x\cdot \sin \phi +\mathrm{\Delta }L=m\cdot \lambda$ and thus a large maximum order can be realized by introducing a large path length difference ΔL according to $m_{\text{max}}=(d_x+\mathrm{\Delta }L)/\lambda$. The maximum angular tuning range achievable (Δφ)_max is given by $\cos {(\mathrm{\Delta }\phi )}_{\text{max}}=1-(\lambda /d)\cdot m/(m-1)\approx 1-\lambda /d$; e.g. for m_max= 300, λ = 1.5 µm and d_x = 3.3 µm, we find ΔL = 447 µm and (Δφ)_max ≈ 57.0°.

Figure 5.

High-order phased array grating. (Online version in colour.)

In [10], a silicon integrated 1 × 4 phased array is reported which achieves a steering range of up to 30°. Its limited amount of only four edge-emitting couplers yielded a large beam divergence and hence limited reach (up to 1.4 m at 12.5 Gbit s⁻¹).

Also a virtually imaged phased array (VIPA) [11], based on multiple beam interference in a parallel plate, can achieve very high orders of interference m and is well suited as a small FSR diffractive element [12]; its angular range of operation, however, is only a few degrees. Also photonic integrated circuit solutions for passive two-dimensional beam steering have been proposed using an array of grating couplers fed by arrayed waveguides in silicon-on-insulator technology [13]; losses were relatively high (8.9–10 dB) and the beam was relatively wide (4°).

Adopting the pair-of-crossed gratings concept, we built a system using a 1000 grooves mm⁻¹ transmission grating in combination with a 13.3 grooves mm⁻¹ reflection grating, as shown in figure 6 [9,14]. This system achieved two-dimensional beam steering over a 6° × 12° range when tuning the beam's wavelength from 1505 to 1630 nm. The optical link losses were less than 6.15 dB, and its channel bandwidth at −3 dB was 10.35 GHz; with PAM4-modulation, we achieved 32 Gbit s⁻¹ data transmission over a reach of 3 m.

Figure 6.

System experiment: two-dimensional beam steering with reflection and transmission grating, transferring 30 Gbit s⁻¹ per beam. (Online version in colour.)

It may be noted that these grating-based solutions are fully passive and can be easily scaled to steer many beams without altering the module: many wavelengths can be fed to the module through the fibre, and each wavelength represents the direction of its beam.

Another option for implementing a two-dimensional beam steerer, deploying readily available fibre-optic devices, is shown in figure 7a [15]. It is built with a wavelength demultiplexer employing an arrayed waveguide grating router (AWGR) with a large number N of fibre output ports each carrying one of the N incoming wavelength signals. These ports are regrouped into a square two-dimensional fibre array, which in turn is put in the focal plane of a lens with a large aperture. The lateral position of a fibre with respect to the lens's axis determines the angular direction of the collimated beam after the lens. Hence, each incoming wavelength fitting on the grid of the AWGR demultiplexer maps to a specific two-dimensional angular beam direction. By putting the fibre array slightly out of focus from the lens, the beams are slightly diverging. Following the design method illustrated in figure 7b, for a given number of AWGR output fibres and given the requirement that the beam spot should have a certain diameter in order to give complete coverage of the user plane, the focal length of the lens and the pitch in the fibre array are determined as a function of the amount of defocusing. This design method yields a compact beam steerer, and due to the slight beam divergence, a coverage fill factor which is relatively independent of the distance of the user plane to the beam steerer. Figure 8 shows a system experiment with a C-band 1 × 80 ports AWGR, in which we achieved 20 Gbit s⁻¹ binary (on–off) data transmission per beam over 2.5 m within the −3 dB bandwidth of 35 GHz of the AWGR's ports, with beam steering over a two-dimensional angular range of 17° × 17° when tuning over the C-band (1530–1565 nm). The aggregate system's wireless capacity is huge, namely 80 × 20 Gbit s⁻¹ = 1.6 Tbit s⁻¹. With the same AWGR, and PAM-4 modulation aided by digital signal processing for equalization, we achieved even 112 Gbit s⁻¹ per beam, implying a potential aggregate system capacity of 80 × 112 Gbit s⁻¹ = 8.96 Tbit s⁻¹ [16].

Figure 7.

Passive two-dimensional beam steerer using an AWGR and fibre array. (a) Two-dimensional beam steerer. (b) Design of AWGR-based beam steerer. (Online version in colour.)

Figure 8.

System experiment: two-dimensional beam-steering employing AWGR and 9 × 9 fibre array, transferring 20 Gbit s⁻¹ per beam. (Online version in colour.)

In table 2, the beam-steering technologies are compared qualitatively.

Table 2.

Comparison of beam-steering technologies.

technology	scalability of number of beams	FoV coverage	steering speed	active/passive module	references
MEMS mirrors	low	20° × 20°	low	active	[5]
SLM	low	3° × 3°a	medium	active	[6,7]
two-dimensional grating	high	6° × 12°	high	passive	[9,14]
AWGR + 2D fibre array	medium	17° × 17°	high	passive	[15,16]

^a18° with lens-based angle magnifier.

When using MEMS-based mirrors or an SLM, the scalability towards many beams is limited as each beam needs a separate steering element. The field-of-view for covering the user area is relatively large when using MEMS mirrors and small when using an SLM (but can be increased with an angular magnifier, at the expense of narrowing the beam). Both for the MEMS-based and the SLM-based steering, active electronic control at the steering module is needed, which requires a separate control line, and facilitates steering of the beam at slow/medium speed (in the order of several milliseconds).

When using the diffractive two-dimensional grating-based concept, by remotely tuning the wavelength, continuous steering of a beam is feasible in one direction and stepwise steering in the orthogonal direction. The number of beams is limited only by the resolving power of the gratings, which can be high. When using the AWGR + 2D fibre array, by remotely tuning the wavelength, the beam is steered stepwise in both the directions. The number of beams is limited by the number of output ports of the AWGR, which can be in the order of 100. The beam-steering speed is dictated by the wavelength tuning speed of the laser diode. This tuning time may be only a few nanoseconds for the laser diode itself; the tuning control electronics may increase that to a few microseconds. The modules according to the two-dimensional grating concept as well as the AWGR + 2D fibre array concept are both passive, so do not require local powering nor maintenance. Their beam-steering control is done by tuning the wavelength of a remote laser diode and does not need a separate control line (as the wavelength also carries the data signal).

Given their favourable fast steering speed, easy scaling and passive maintenance-free nature, in our research we have preferred the wavelength-tuned diffractive beam-steering techniques using two-dimensional gratings or an AWGR + 2D fibre array. Moreover, the latter one is not impacted by polarization state fluctuations in the optical input signal (these may yield signal intensity fluctuations in the two-dimensional grating concept).

5. Free-space optical receiver

Next to the beam-steered optical transmitter, the receiver catching the optical beam at the user side is an essential part of a high-speed optical wireless link. This receiver should acquire as much power as possible from the optical signal as well as be able to process high data rate signals, while also being small, consuming little power (as it is to be fed from the wireless user device, which typically is battery-operated) and low cost. In order to minimize the alignment efforts with respect to the transmitted beam, it should have a wide open aperture as well as a large field-of-view (FoV). The physics law of conservation of étendue (https://en.wikipedia.org/wiki/Etendue) (also known as the A·Ω product, i.e. the product of entrance aperture A and solid angle Ω of the accepted cone of light), however, limits what optical measures such as lenses or mirrors can do to extend both the aperture and the FoV. Compound parabolic concentrator mirrors have been used to expand the aperture at the cost of a reduced FoV (https://en.wikipedia.org/wiki/Nonimaging_optics). A slab antenna structure with large receiving aperture has been proposed which uses fluorescent material in the slab waveguide to convert incoming light to a longer wavelength which is subsequently constrained inside the waveguide by internal reflections and guided to a photodetector [17]. It thus breaks the étendue constraint and achieved a gain of 50× with respect to an étendue-preserving concentrator with the same FoV. Its dimensions so far limit its speed of operation.

Alternatively, the receiver may use a relatively small aperture which is actively scanning a larger FoV to find the incoming beam. This approach can only be applied when the transmitter has already found the user's device and has successfully steered a beam to it. By deploying an SLM the concept has been explored in [18]. We achieved an FoV's half-angle of 18°, and reception of 40 Gbit s⁻¹ in on–off keying format over 0.5 m free-space reach.

At the cost of increased circuit complexity at the electronic side, the receiving aperture can be increased by deploying an array of photodiodes. Each photodiode is followed by a low-noise electrical amplifier, where the outputs of these amplifiers are subsequently combined. These amplifiers also need to have closely matched phase characteristics in order to avoid destructive signal interference. By applying a lens system on top, each of the diodes can look into a different direction, thus extending the field-of-view. Such a so-called angle diversity receiver was reported in [19] featuring 7 InGaAs MSM photodiodes of 50 µm, followed by integrated CMOS amplifiers; it achieved a −3 dB bandwidth of 3.7 GHz.

Typically a photodetector has to strike a compromise between its operating speed and its aperture, as its capacitance is increased when its active detection area is increased. A way to resolve this compromise may be found by decoupling the light detecting function from the light collecting function. Using integrated optics technology, we devised an optical receiver device where the incoming beam is captured by a surface grating coupler (SGC) which couples this light into a waveguide which subsequently feeds it to a high-speed photodiode [20]. The SGC can be scaled by apodization to a larger receiving aperture, while the photodiode stays the same, thus maintains the speed of operation. The technical feasibility has been shown by the photonic integrated circuit in figure 9. This prototype had an SGC with an optical bandwidth of more than 75 nm and an area of 10 × 10 µm². The fast photodiode was a unilateral travelling carrier (UTC) type of 3 × 10 µm², with a −3 dB bandwidth of 67 GHz at a reverse bias of 4 V. With an incident optical beam, we showed data transmission of 40 Gbit s⁻¹ on–off modulated data. The receiving aperture can be readily scaled to beyond 1000 µm² by apodization, and further extended on-chip by an array of SGC-s followed by a waveguide light combiner.

Figure 9.

Cascaded aperture optical receiver. (Online version in colour.)

6. Localization

In beam-steered optical wireless communication, the system obviously needs to collect information about the position of the users' devices. One may use triangulation algorithms in the user device which use three or more light sources at the ceiling [21]. Such techniques have been explored in several VLC systems and are operational in various application areas (such as in warehouse logistics). Localization by means of an infrared camera has been proposed in [22], where a ring of 12 LEDs mounted around the receiving aperture of the user's device and emitting at a wavelength of 890 nm was observed. The localization accuracy achieved was 0.05°, amounting to 2.5 mm at 2.5 m distance. For the device localization in our beam-steered system, we explored observation techniques deploying a camera mounted at the ceiling and four visible light LEDs arranged around the receiving aperture at the user's device [23]. Each device is identified by a unique blinking sequence of its LEDs, which is recognized by simple image processing in a Raspberry Pi module next to the camera. The low-cost camera had a resolution of 1280 × 720 pixels, and the observation area covered was 3.2 × 1.8 m at a distance of 3 m. A localization accuracy of less than 5 mm was achieved. Calibration of the coordinates found to the wavelengths corresponding with the spot cells in our beam-steered system needs to be done; with these calibrated localization data, the autonomic beam steering for tracking and tracing the user device is realized.

In order to minimize power consumption at the battery-operated user device, we also explored fully passive localization tagging at the user's device. In our beam-steered optical wireless communication system concept, the user's device may be located by scanning with a searching beam across the whole user area, and monitor at the pencil-radiating antenna any beam parts reflected from the device, as illustrated in figure 10 [24]. Corner cube reflector structures have been used, which reflect a part of the incident beam back into exactly the same direction as it came from. This reflected beam gets a slight lateral offset by the corner cube, which is proportional to the size of the corner cube. To minimize this offset while also reflecting enough light to be detected adequately at the PRA, we deployed an array of passive miniature corner cubes embedded in a foil (as commonly used for, e.g. road signage). Circular foils of 4 cm diameter were used, with a central hole of 3 cm hosting the receiver's downstream beam detector. The central hole reduced the returned power by about 3.2 dB. With a beam power of 10 mW and a beam diameter of typically 10 cm, the localization monitoring photodetector unit next to the PRA detected a returned power of −31 dBm, which was sufficient for localizing the device well within the cell resolution. We implemented this passive localization method in our laboratory demonstrator and showed its feasibility while carrying two independent video streams embedded in 10 Gbit s⁻¹ Ethernet links ( figure 10). The scanning time of the whole area is presently limited to about 15 s in our set-up, predominantly by software processes. Intrinsically, the localization time is constrained by the time needed to tune the wavelength of a laser diode over the full range, which can be done within a few milliseconds. However, the additional steps in the scanning process are stepwise tuning to each cell consecutively, deciding whether there is retro-reflected power from that cell exceeding the background noise level, and then moving to the next cell. Per cell this takes 115 ms, of which only a few ms are consumed by the intrinsic tuning time of the laser diode, and the majority is taken by the Labview software running in the laptop controller, the Arduino board which controls the laser tuning, and the read-out time of the localization power meter. Scanning the 128 cells thus takes about 15 s; more efficient algorithms are being investigated to reduce this time remarkably, preferably implemented in embedded software. Figure 11 illustrates how in the wavelength scanning process the locations of two user devices, each equipped with a 4 cm diameter corner cube reflector foil, are found. The two peaks visible in figure 11a clearly indicate the wavelengths which correspond to the positions of the two devices indicated in figure 11b, respectively.

Figure 10.

System experiment featuring fully passive device localization. (Online version in colour.)

Figure 11.

Device localization by monitoring signal power reflected from the user devices. (a) Reflected power versus wavelength. (b) Position of the two user devices in the cell-patterned user area. (Online version in colour.)

Next to being fully passive at the user side, so not drawing any power of its battery, this corner cube reflector method also offers the advantage that it is self-calibrating the location data to the beam-steering parameters: while wavelength scanning the area and monitoring the reflections at the same time, the system directly finds which wavelengths needed for the beam steering do map to the respective locations of the user devices and can store these wavelength values for the beam-steering system controller.

7. Bidirectional wireless communication systems

Most broadband services, such as downloading video streams and large files from the internet, are highly asymmetric in nature, they require much more downstream link capacity than upstream. Beam-steered optical wireless communication is well suited for such broadband asymmetric traffic, particularly in a hybrid system combination where radio links are used for upstream. The radio upstream link does not necessarily require a LoS path and hence can advantageously serve for initiating a request for an optical wireless connection and for acknowledging when the connection has been established. And it also can provide the return channel, albeit at a somewhat lower capacity than the downstream optical channel.

A hybrid bidirectional system we have built in our laboratories is shown in figure 12 [25]. In downstream, it uses the 1 × 80 ports AWGR and an f = 40 cm condenser lens, with which beams of 8.3 cm diameter over a 2.5 m reach transferred 35 Gbit s⁻¹ binary modulated data streams. In upstream, two horn antennas with 30 dBi gain were used at the user device side and a 16 dBi antenna at the PRA side, which transferred 5 Gbit s⁻¹ amplitude-shift-keyed modulated 60 GHz radio beams. For setting up the connection, the upstream two horn antennas at the user were electro-mechanically directed towards the horn antenna at the PRA.

Figure 12.

Hybrid wireless communication system featuring downstream two-dimensional beam-steered optical wireless communication and upstream 60 GHz based radio link. (Online version in colour.)

8. Laboratory system demonstrator

An elaborated set-up was built in our laboratories as shown in figure 13, which demonstrates the downstream real-time transfer of two high-definition video streams, each embedded in a 10 Gbit s⁻¹ Ethernet stream (an earlier version with 1 PRA is reported in [26]). Two monitors display the two video streams sent and two others the two streams received. The intrinsical signal transfer itself by the optical system is done with very low latency. For processing the video streams, however, the buffering in the VLC media player software adds 1.5 s delay, and the video encoding/decoding increases this to about 2 s. Reduction of the buffering and coding times is being investigated, and may reduce the total end-to-end latency of the video streams to below 0.5 s.

Figure 13.

Laboratory demonstrator. (Online version in colour.)

To increase the coverage area, the set-up features two PRAs installed at the ceiling, which each serve a user area of about 1.3 × 1.3 m². A MEMS-based OXC switch is inserted which enables dynamic switching between the PRAs to allow nomadic mobility of the user devices. Next to a PRA, also an optical power detector aiding the retro-reflector-based localization is mounted. On the laboratory table, two closely spaced optical receivers are located which each independently receive an optical beam carrying a video channel. The CCC functions including the wavelength-tunable transmitters and their control are located in the rack, from where two fibres feed the PRAs. Each cell in table addressed by a beam has a diameter of about 10 cm. The beams were designed to be slightly divergent in order to optimize the coverage when the distance to the PRA varies. With a divergence angle of only 1.2° (achieved by putting the two-dimensional fibre array at 80% of the lens' focal length), the directivity of the beam is still equivalent to an antenna gain of 40 dBi in a millimetre-wave radio system, which in such system would require a complex phased array antenna structure. The inset in figure 13 shows seven cells illuminated by a beam, as observed by an infrared camera.

The PRAs in the laboratory demonstrator were assembled by combining two separate AWGR modules which jointly handle wavelengths over an extended range (the C-band (1530–1565 nm) and lower part of the L-band (1565–1625 nm)). This composite AWGR is followed by a module which re-arranges the 128 AWGR output ports into a two-dimensional fibre array, and subsequently this array is put in front of a commercially available wide aperture F/0.95 camera lens with 50 mm focal length. In the next steps, the size and assembly costs of this modular PRA set-up can be reduced significantly by realizing the AWGR functions and the two-dimensional light emitting array in a single photonic integrated circuit in a passive photonic integration technology (silicon nitrate or silicon oxide).

9. All-optical bidirectional optical wireless communication

Although most indoor wireless traffic is asymmetric in nature, needing more downstream than upstream capacity, there may be a need for higher upstream capacity than can readily be achieved with radio technologies. To realize high-capacity communication both downstream and upstream, one may mirror the downstream beam-steered optical wireless communication system into an upstream optical system. Such an approach has been explored with MEMS mirror beam-steering modules both at the ceiling unit and at the user devices, as reported in [5]. The full-duplex operation was achieved, with a downlink at 10 Gbit s⁻¹ using a wavelength of 1551 nm and an emission power of 7 mW, and an uplink at 2 Gbit s⁻¹ using a VCSEL (vertical cavity surface-emitting laser diode) at 850 nm and 5 mW.

For a point-to-multipoint network topology, such a mirrored solution may not be the most effective one. Taking our double-grating beam-steering system concept as the basis, we explored the recovery of the optical carrier wave and re-modulation of this cleaned-up carrier with the upstream data, as shown in figure 14 [14]. For this, at the user side, part of the modulated optical downstream signal is tapped by an optical coupler and fed into a semiconductor optical amplifier which is operated in heavy saturation. Thus, the intensity modulation on the signal is largely removed, and the almost-clean carrier wave is subsequently fed into a Mach–Zehnder modulator, which impresses by intensity modulation the upstream data signal on it. Next, via the optical circulator and the lens, the upstream signal is guided back to the central headend site. The wavelength of the optical carrier is obviously not altered in the carrier recovery process, hence the upstream signal follows exactly the same route via the two gratings as it did in downstream, but now in the reverse direction, and ends via another lens at the upstream data receiver. We showed 10 Gbit s⁻¹ transmission simultaneously in both downstream and upstream with a beam of 3.3 mm diameter, over a free-space path length of 3 m (folded between two mirrors, to save space in the laboratory). In order to have enough power for upstream during the low-level time slots of the downstream signal, the extinction ratio (ER) of the binary downstream signal (the ratio of the power in a logical ‘1’ to that in a logical ‘0’ symbol) should not be too high. We used an ER = 3.6 dB, which degraded the downstream receiver sensitivity to about −15 dBm, which implies a power penalty of about 2 dB with respect to downstream only.

Figure 14.

Bidirectional full-duplex optical wireless communication. (Online version in colour.)

10. Concluding remarks

Optical wireless communication techniques offer powerful solutions to resolve the imminent congestion in radio-based communication networks, in particular in local area wireless networks. By deploying steerable narrow infrared optical beams which address each user device individually, it can provide links with ultra-high capacity where and when needed, while also offering high privacy. Their small footprint enables the creation of a pico-cell per user device and thus creates a high degree of spatial multiplexing which boosts the total wireless system's throughput. As the beams deliver communication capacity on demand only to those places where and when needed, the system achieves high efficiency in bringing the data signals to the users and thus can reduce remarkably the power consumption needed with respect to a radio-based mm-wave communication system. Moreover, as nothing travels faster than light in vacuum (and air), a minimum in latency is achieved which makes it also eminently suited for time-critical machine-to-machine interactions. Radio beams travel equally fast, but due to the lower bandwidths accessible in a radio channel, extra processing time is needed for signal conditioning, which adds to the radio link's latency. By deploying two-dimensional-steered narrow infrared beams, we demonstrated capacities up to 112 Gbit s⁻¹ per beam and demonstrated live high-definition video streaming with up to 128 beams in a hybrid optical wireless downstream and radio wireless upstream system. By means of photonic integration, the size and costs of the two-dimensional beam steerer as well as of the optical receiver circuitry at the user device can be reduced significantly; research on such integration steps is ongoing. A next goal is all-optical wireless communication in both directions. Using optical carrier recovery at the user side, all-optical full-duplex bidirectional wireless communication can be established; so far, we have realized this at 10 Gbit s⁻¹ in both directions.

Optical wireless communication by means of narrow infrared beams can provide the ultimate capabilities for delivering broadband services wirelessly in human-to-machine and machine-to-machine communications, e.g. in terms of high capacity, low power consumption, high privacy and low latency. It is not foreseen to fully replace radio-based wireless communication, among others because it needs line-of-sight. But it can off-load high-speed traffic loads from radio-based networks and thus give these the room necessary to handle the booming amounts of low-speed intermittent traffic as, e.g. generated by the emerging myriad of small internet-of-things devices.

Data accessibility

This article has no additional data.

Authors' contributions

K.M. and F.H. contributed to laboratory experiments and setting up the demonstrator. Z.C. contributed on the integrated optical wireless receiver. N.Q.P. contributed on the camera-based device localization. E.T. contributed in supervising the junior researchers and in system discussions.

Competing interests

We declare we have no competing interests

Funding

The European Research Council (ERC) is gratefully acknowledged for funding this research in the Advanced Investigator Grant project BROWSE—Beam-steered Reconfigurable Optical-Wireless System for Energy-efficient communication, and the follow-up Proof-of-Concept project BROWSE+. KPN is gratefully acknowledged for funding the work of N.Q.P.