Optical Wireless Communication

by Jean-Paul Linnartz. Follow also the latests updates on youtube and Linkedin

Every year, 4 billion WiFi chips are integrated into various devices, including smartphones, laptops, and IoT products. While WiFi has become ubiquitous, there is growing interest in alternative wireless technologies like LiFi that promise more secure and interference-free communication. Professor Fary Ghassemlooy, an expert in Optical Wireless Communication (OWC), sees OWC as a complement to radio communication, offering potential benefits in specific applications:

It eliminates exposure to RF radiation, making it suitable for use in schools
It provides more secure communication in environments like manufacturing, government offices, and hospitals. It reduces risk of eavesdropping and signal jamming compared to WiFi

The tangibility of light-based wireless communication, as opposed to invisible radio waves, could also bring a new “magic” to wireless technology.

A laser beam that carries 20 gigabit per second over a 20 kilometer distance. Free space optical wireless links can bring connectivity to remore areas. Additionally, it can ensure that the backbone infrastructure for a future generation of wireless networks, say for 6G, can deliver the promosed bit rates and low latency. Devin Brinkley, SVP and head of engineering at TAARA (Alphabet, Google) outlines which technical breakthroughs made this possible, including advanced tracking to ensure that the laser beam stays directed precisely at the receiver, algorithms that rapidly adjust the bit rate to changes in the over-the-air propagation path and photonic integration towards aggressive cost-down. Interestingly further challenges to innovate this system towards higher perfomance do not immediately target greater bit rates. More robustness, making the link more robust, more adpative and afforable all require further research.

A frequently heard critique: if you block the light, the optical link is gone. However, that probem is largely solved by distributed MIMO

A frequently heard critigue: IR light is not safe to the eye. As one of the questions in a round table debate with Prof. Fary Ghassemlooy visiting TU Eindhoven, Lev Azarkh asked about the risks of (near) infrared light and visible light. Can we guarantee eye safety? Eduward Tangdiongga mentions that deeper infrared (as used in fiber e.g. 1330 or 1550 nm) is instrinsically safer, but has different technical application fields.

This video discusses two contrasting approaches to improving optical wireless communication (OWC) systems. One approach is to focus on setting new speed records: increasing the number of gigabits per second that the links can handle. However, in many real-world scenarios, such as connecting smartphones, secure laptops or IoT devices to the internet, there are several other important performance indicators to consider. LiFi and OWC technologies excel at serving multiple devices in crowded areas. They also tend to suffer less from interference compared to radio-based systems. Additionally, LiFi can offer enhanced security over traditional wireless technologies. An important question to consider is whether LiFi is more energy-efficient than radio-based solutions. To summarize:

In a congested urban environment like Mumbai, does one truly need the a Formula 1 racing car. Similarly, as the communication systems get overloaded, do we care about the raw speed of wireless communication if we were the only user?

The radio community has spent decades developing robust reference models; is it wise for OWC engineers to bypass this crucial step?

Diving deeper into the details

What is the throughput capacity? It is remarkable to see that scientific papers have conflicting views on a factor of 2 in the formula for the throughput of an intensity modulated (IM-DD) optical wireless communication (OWC) link. This creates an opportunity to debate whether or not a correction factor of 1/2 should be applied to the available bandwidth. The expression displayed is not a “channel capacity”; rather, it represents an achievable throughput, but does not represent an upper limit on what can be accomplished. Thus is may not be called the capacity. Nonetheless, the formula can be safely used to express the number of bits per second that OFDM can achieve with a DC-offset, but the correction factor of 1/2 should not be applied. In the video, I attempt to determine why this factor nonetheless appeared in many research papers.

The claim that “LiFi uses OFDM because it is so spectrum efficient” or that “OFDM loses 50% spectrum efficiency because of a need for Hermitian Symmetry” is highly debatable, despite sometimes being mentioned casually in paper introductions. In reality, Orthogonal Frequency Division Multiplexing (OFDM) loads the exact same number of signal dimensions per second as Pulse Amplitude Modulation. OFDM cannot magically violate the laws of linear algebra to create more dimensions through a matrix FFT operation. The need for Hermitian Symmetry simply arises because complex numbers are allowed as inputs, while the channel is real-valued. So where does the notion or myth come from that OFDM fails to use 100% of the available dimensions? And if that would be the case, then why is OFDM still a popular choice for LiFi communication?

Power Consumption

How green is Optical Wireless Communication? Can LiFi use light to save energy in delivering high speed internet access? Is it more effficient than WiFi, which runs over radio waves? At Eindhoven University of Technology, we are working on models for the power needed per square meter of coverage, and we compare LiFi to Radio. But for that we first need a more reliable model of the power consumed by an LED that is modulated with data (watts/m2 per bit/s); a more reliable one than commonly used in scientific literature.

At the IEEE Sustainable Smart Lighting Conference LS24 conference, I debated suitable models. Surprisingly, many earlier publications stated that the optical power is proportional to the average intensity signal (1st moment, reasonable), while the electrical power would be proportional to variance, i.e., the average of the square of the signal (2nd moment, !?). That commonly made assumption is surprising, because that would imply that an LED violates the Law of Conservation of Energy. It would also not credit the Shockley equation or the energy needed to generate a photon. Nonetheless, many results and papers on LiFi borrow their solutions from radio, optimized as if power is proporional to the variance. An LEDs differs from a 50 Ohm antenna, in non-linear and its performance depends on biasing. That needs to lead to different system choices. So, I was really keen to expose our models to the participants of the conference LS24, formerly known as the IEEE Light Sources conference, thus where the LED lighting experts gather.

In classes and (under)grad courses on electronic design of circuits that operate at radio frequencies, we hear all about impedance matching. But should we always match impedances? Jean-Paul Linnartz and Peter Baltus review common practices, not only in antenna tuning, or in transmission lines but also in HiFi audio loudspeakers, in feeding power in to LEDS or in plugging household devices into mains outlet sockets. Extracting the highest amount of power is not the same as designing the greenest, most power efficient system. In modulating LEDs, for LiFi or Optical Wireless Systems, impedances of the LED demands special attention. Creating a good LiFi driver is not as easy as it seems: if you just connect a strong power amplifier via a bias T to an LED, things can go really wrong. It may give better reproducible results if you design your own modulator circuit with a few transistors, rather than to buy an expensive lab linear amplifier design for an RF antenna.

Communication by modulating the light intensity: There are different solutions for creating a non-negative signal that can be transmitted as a light intensity. But how different are these methods really. We found out that if you do the math, some of the most frequently cited solutions are “almost” identical. There is only a small twist. and that twist is just a slow phase rotation. A commonality that is overseen in overview papers, but that makes it much easier to choose the best modulation for LiFi. However, I am not so sure that these schemes really beat the use of a DC offset.

Directional beams

A weak radio link can be largely improved by using directional antennas. Also in Optical Wireless Communication (OWC), the bit rate, the power consumption, the security and the user density can be improved by ensuring that the modulated light is only sent in the direction of the target client device. Let’s compare the potential benefits from sectorization at the transmitter versus at the receiver. These turn out to be very different from eachother. In this regard LiFi behaves very different from radio links. It is not as simple as applying an antenna gain. A LiFi system can perferm well with many transmit sectors but only one or few receive sectors. A preview on: “Design considerations for sectorization in bidirectional LiFi links” by Jean-Paul Linnartz, Paul van Voorthuisen, and David Rojas, at the IEEE VTC conf, in Oslo, Spring 2025

In LiFi, in wireless communication via light, the design considerations for sectorization at the emitter and at the detector differ from each other. Sectorization at the transmitter can greatly boost the Signal-to-Noise ratio. At the receiver, sectorization has a number of effects, and the benefits are more complicated to understand.

To search the counterstation, is it wise to widen the beam?

To spread a message as quickly as possible, do you talk to everyone in the crowd individually? Or, would you rather stand on a chair and shout out? In this research work, we compared radio and optical systems, and we draw the conclusion that the intuition from daily life and social settings would lead to wrong design decisions for a LiFi system. To reach “everyone in a crowd”, it is better address each counter station sequentially by a narrow, powerful beam. In particular, if we take into account the light emitter and the photo diode detector with its noise-limited amplifier, we see that LiFi systems require a different optimization. Restrictions apply, you will find “ifs and buts” in our paper:
Lev Azarkh and Jean-Paul Linnartz: “Accelerated beam searching for optical wireless communication system with slow feedback channel”, 2023 IEEE PIMRC, 34th Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), Toronto, ON, Canada, 2023

How about the detector?

Scientific papers say that the size of the photo diode in a receiver for wireless optical communication is a trade-off. In fact, large photo diodes give an excellent signal-to-noise ratio, but their bandwidth is limited. Small photodiodes have high bandwidth, but may not capture enough photons to support a high data rate. Engineeers then like to know how to make that trade-off. To choose the best size of a photo detector in a LiFi system, Lev Azarkh and Xiaochen Liu and Jean-Paul Linnartz modeled the data rates achieved versus photo diode size. The trade-off seems not always to be a compromise!

Often it is just a matter of: make it big enough. In other words, don’t select your photo diode on the promised bandwidth only. OpticalWirelessCommunication (OWC) systems are limited in bandwidth by their electro-optical components. We study the receiving photo diode, its surface area and the associated capacitance. A large–area detector improves the signal–to–noise ratio, but a small size allows a high bandwidth. Does an optimum detector size exist? It depends on photodiode properties, the Trans-Impedance Amplifier noise performance and the received signal strength.

Scientifc details have been presented at SPIE Photonics West – San-Francisco, Jan 2023: Optimal Detector Size for Optical Wireless Communication Systems, Lev Azarkh, Xiaochen Liu, J.P. Linnartz, but more results and insights are coming up.

If you block the line of sight, LiFi goes on !? Distributed MIMO

LiFi, which is a form of optical wireless communication, has the potential to alleviate the increasing congestion of the radio spectrum. Additionally, since the signals do not penetrate through walls, the entire spectrum can be reused in each individual room. However, this line-of-sight propagation also comes with a drawback – an accidental blockage of the light wave can lead to a signal outage. Distributed MIMO offers a solution to this issue: the data is transmitted through multiple cooperating light sources. While MIMO technology is a known concept in radio wave communication, the behavior of LiFi links is different. At Eindhoven University of Technology, Thiago Bitencourt Cunha’s PhD research work focused on making D-MIMO work effectively for LiFi systems. And there are significant differences between the radio and light. The typical approaches used for radio communication may not work properly in the case of LiFi. But by closely examining the specific details of LiFi systems, the system can be made highly efficient and power-effective.

The performance of radio links strongly depends on the accuracy of the information that the system has about the state of the communication channel. We know this from RF radio links. Modulation and coding schemes can adapt to fades, provided that these can track the rapid and highly frequency-dependant variations in real-time. But do we really experience multipath fades in Optical Wireless Communication? In fact, a light-link behaves very differently. Mostly the frequency response is not affected by multipath fading. Instead the (low pass) frequency response op photonic-electrical converters dominate the channel limitations. Real-time channel state information, not just “average statistics” is very well possible in OWC. We had an interesting conversation with Prof. Fary Ghassemlooy to debate this.

More of our own views are in : IEEE JSAC 2025: Cascaded Lagrangian Power Allocation to Optimize D-MIMO OWC Systems, Thiago E. B. Cunha, Jean-Paul M. G. Linnartz IEEE Journal on Selected Areas in Communications, Volume 43, Issue 5, Pages 1676 – 1690