Dennis W. Prather, Professor of Electrical and Computer Engineering, University of Delaware and President, Phase Sensitive Innovations
Published: 19 Oct 2021
CTN Issue: October 2021
A note from the editor:
As the natural growth season in the northern hemisphere draws to a close, the growth of mobile data shows no sign of abating, and the mobile network industry increasingly looks to higher frequency bands (millimeter wave and Terahertz) in search for more bandwidth. The extremely high beam-bandwidth product, native to such spectra, promises throughput that is orders of magnitude higher than those of today's 5G network, yet poses daunting hardware challenges for implementation of a radio frontend that scales with growing spectral availability. Enter Radio Frequency (RF) photonics. It turns out that a phased array supporting several octaves of RF carrier frequencies and thousands of antenna elements can be efficiently built on electro-optical modulators (EOMs), photodetectors (PDs), and Fourier optical processors running at the speed of light. In this article, our colleague Dennis Prather shares with us the remarkable progress that RF-photonic device technologies have made towards becoming a viable alternative to all-RF transceiver chains and explains how thousands of such links can be brought into phase coherence in a small ensemble to form as many beams as the electrical aperture of the array allows. Indeed, the optical processor that does the spatial beamforming, being the 'brain' of the system, is RF agnostic, and can be part of a photonic integrated circuit (PIC), raising the intriguing prospect of a future-proof massive MIMO front end that keeps on giving for many generations of mobile networks to come. It is oddly fitting that it may take a RF-photonic approach to bridge the Terahertz Gap between microwave and light. As always your comments and feedback are most welcome.
Miguel Dajer, CTN Editor-in-Chief
To Infini-G and Beyond: A Case for RF-Photonics in Wireless Communication Systems
Dennis W. Prather
Professor of Electrical and Computer Engineering, University of Delaware and President, Phase Sensitive Innovations
With there being more mobile communication devices than there are people in the world, it is safe to say that the progression of mobile networks have had both a personal and widespread impact. That said, the journey from 1G to 5G has been driven by numerous innovative technical developments as well as a rapid expansion in commercial markets, with their interplay giving rise to an insatiable demand for service that drove a progression of nearly an order of magnitude increase in data capacity roughly every 10 years. Now, with the dawn of fifth generation (5G) upon us, service providers are striving to improve the user experience through expanding access, offering faster speeds (up to 10 Gbps), and providing lower latency (less than 1msec). Exactly how this will manifest, on a large scale, remains to be seen but if these technical goals are achieved it will pave the way toward the integration of artificial intelligence and machine learning to augment the reality of everyday experiences and allow for a more seamless integration between communication and sensor systems. To this end, the impact of future mobile networks will be limited by both our imagination and our ability to implement it.
This article strives to address the latter issue, i.e., how to implement future generations of mobile network technology. To put it simply, to sustain the progression of increasing data capacity, two things must happen: (1) carrier frequencies must move to higher regions of the electromagnetic spectrum and (2) each carrier frequency must offer more information bandwidth (IBW). Also, according to the Shannon-Hartley capacity theorem, this must be done while preserving a sufficient signal-to-noise (SNR) ratio. With this being said, this paper explores a technology base that can offer such sustained capabilities as they progress from 6G, to 7G, … all the way To Infini-G and Beyond, to paraphrase one of my favorite movie characters, Buzz Lightyear.
With such a broad perspective in mind, this article introduces the field of radio-frequency-photonics, i.e., RF-photonics, as part of a mobile network development cycle that may never end. To begin, consider that the capacity of information, or data, transmission is largely dependent on two primary factors: bandwidth (BW) and signal-to-noise-ratio (SNR). While the former is related to the frequency of transmission, the latter is related to the integrity of signal transmission and reception, which among other things can depend on the state of the environment, i.e., channel state. To this end, an increase in BW usually comes with a concomitant increase in carrier frequency and as the carrier increases to higher regions of the electromagnetic spectrum, there are fewer commercial technologies available and, of those that are, they tend to have reduced performance. For this reason, there is an active push to develop advanced technologies for infini-G mobile communication systems. Along these lines, the Federal Communications Commission (FCC) has enticed technology developers with the allocation of frequencies up into the THz region of the spectrum, where requisite technologies currently do not exist. This amounts to an “allocate it and they will come” mentality, which is proving to be true to its intention.
Therefore, this article introduces an emerging framework based on recent developments in the field of RF-photonics that offer a path for sustained technology development, perhaps all the way to infini-G. Historically speaking, the field of RF-photonics was limited in application due primarily to the performance of two key legacy devices, namely electro-optic modulators (EOMs) and photodetectors (PDs). While the former is responsible for up-converting the RF signal to the optical domain, the latter is responsible to bringing it back down, or down-converting it. The issue, until recently, has been that a complete up- and down-conversion cycle came with significant degradation of the SNR, or elevated noise figure (NF). Fortunately, recent advances in both EOMs and PDs have improved the up/down conversion process, which is now competitive with electronic approaches, if not superior in some cases. This is especially true when considering that as RF frequencies increase to higher regions of the spectrum, the losses and performance of their associated devices tend to become worse. While some of this is due to the lack of technical maturity other aspects are based on inherent material properties, such as conduction loss which increases with the square-root of the frequency, as shown in Fig. 1. In comparison, the propagation loss in an optical fiber is on the order of 0.2 dB/km (red line in the bottom of Fig. 1) over the entire RF spectrum. Thus, as frequencies approach the millimeter wave region of the spectrum, which is an emerging aspect of 5G, conductors have significantly more signal loss than do optical fibers. We should note that the plots shown in Fig. 1 represent the best-case scenario, i.e., lowest loss, as they only include conduction losses associated with high frequency propagation and do not include material-, connector-, and/or modal-losses.
Because conduction losses are inherent, they can only be overcome by reducing the amount of high frequency transmission within a conductor. Thus, some approaches rely on intermediate frequency (IF) transmission throughout the communication system and conversion to a high frequency RF carrier directly at the antenna, just before transmission. However, this approach can require more expensive devices and components to be included within the remote-radio-head (RRH), which can be subject to variations in weather, hot/cold temperatures, and may not be easy to access for maintenance and/or service.
Another key issue, as frequencies migrate to higher regions of the spectrum, is that they experience increased absorption by the atmosphere as well as increased free-space-path-loss (FSL), which is proportional to the square of the wavelength. While the former is inherent to the environment and must be accounted for within the link budget, the latter can be compensated by increasing the gain of the TX/RX antennas. For example, consider that a ten-fold increase in frequency, i.e., going from 3 GHz to 30 GHz, comes with 20 dB of added signal loss. To compensate for this, one can increase the gain of the antenna by 20 dB. However, an increase in spatial gain of an antenna comes with an associated reduction in spatial coverage, or field-of-regard (FoR). Thus, to regain FoR, one can use multiple beams, with each having enough beam-gain to overcome FSL. To do this simultaneously, one can use a phased array antenna system where a key figure-of-merit (FoM) is how many simultaneous beams an antenna system can transmit and receive all at their maximum IBW, which is called the beam-bandwidth-product (BBP). The goal, of course, is to simultaneously process as many beams as possible at the highest possible bandwidth.
This is where RF-photonics comes in, as it offers unique advantages in each of these areas: efficient up- and down-conversion, low loss signal transmission, and the ability to enable phased arrays to operate with significantly higher BBP. To present these advantages, we offer an overview of recent developments that begin on the device level, namely EOMs and PDs, and their integration into an RF-link for remoting RF signals. Lastly, we discuss a new technique for using RF-photonic devices in phased array RF systems based on a novel optical-upconversion technique, which offers a near unlimited BBP.
As stated above, recent years have seen the advent of a new material system referred to as thin-film-lithium-niobate (TFLN) for EOMs and modified-uni-traveling-carrier (MUTC) PDs. Together these new materials and devices have transformed the field of RF-photonics by offering unprecedented up-conversion efficiency, as denoted in a quantity called the half-wave voltage, or low-Vp, modulators and high down-conversion responsivity with high output photocurrents, respectively. This is important because as carrier frequencies continue to expand into higher regions of the spectrum, their up- and down-conversion become a critical gating factor for mobile system performance. As shown in Fig. 2, a typical receiver system captures the incoming RF signal and outputs it at a suitable IF frequency, which is then fed to a backend analog-to-digital-converter (ADC) as the first stage of a digital signal processor that serves to demodulate and route the signal within the platform. The ultimate properties of a up/down-conversion system are shown in the black box, in Fig. 2(a), where ideally the system would be frequency agnostic, unlimited BBP, multi-functional (to merge communication and sensor systems) and be able to accommodate very high IBWs. In the electronic world, this is a near impossible task, however, in the RF-photonic world this is close to a reality. Shown in Fig. 2(b) is a standard RF-photonic link that serves to up-convert an incident RF signal to the optical domain (typically ~200 THz). Once the RF signal is in the optical domain, it can be fed into an optical fiber for transmission over an ultra-low-loss medium. At the receive side of the link is a photodetector that receives the optical signal and either converts it back to an RF signal or a tunable-optical-local-oscillator (TOLO), which can be used to mix the optical signal down to a desired IF signal.
Figure 2: (a) standard down conversion process for an electronic receiver with the ideal properties of it being frequency agnostic, unlimited BBP, multi-function, and very high IBWs, (b) an RF-photonic link that embodies the ideal properties of a general RF receiver.
Figure 3: Overlayed plot of a photonic RF link based on the device performance of the EOMs and PDs used in the link. The link performance is characterized by its gain, SFDR and noise figure. They collectively depend on the combination of bias voltage ( Vπ) of the EOM, the photocurrent (Iph) of the PD, and the relative intensity noise (RIN) of the laser.
The ideal receiver properties, described in Fig. 2, are being realized with recent developments in new EOMs and PDs, as denoted in Fig. 3. From a historical view, it is easiest to see the transition from legacy device performance to that of newer photonic devices, such as thin-film-lithium-niobate (TFLN) EOMs  and modified-uni-traveling-carrier (MUTC) PDs [2,3], based on device limited performance. While it is not the easiest chart to understand, the graphical method  shows overall link performance as a collective assessment of the EOM and PD used in the link. As noted on the plot, legacy EOMs and PDs tended to have a performance range of Vp ~ 6-10V and a photocurrent of 1-10mA. With these numbers plotted in Fig. 3, the overlayed curves coincide with the red diamond on the right-hand side of the plot, which yields a NF of ~40dB, a spur-free-dynamic-range (SFDR) of ~100dB-Hz2/3 and a link Gain of ~ -32dB. With this kind of performance, it is not hard to see how the appeal of low-loss optical fibers was overtaken by unacceptable RF link performance. The good news is that recent developments in photonic device technology, most notably TFLN EOMs and MUTC PDs, have moved the red diamond to the green diamond, on the left side of the plot. This is based on device performance of a Vp ~1-2V [5,6] and photocurrents of 10-100mA, which is achievable in these new devices. The location of the green diamond shows the performance of a link to have a noise-figure (NF) of < 10dB, a SFDR of ~110dB-Hz2/3 and a link Gain of ~ +20dB.
One missing parameter that significantly affects link performance is the relative intensity noise (RIN) of the laser, which results from random fluctuations in the laser intensity, because of instabilities in the laser cavity, random variations in drive current, and the coupling of back reflections into the laser cavity, from associated devices in the link. Typically, a laser cavity can be stabilized by providing vibration isolation and a steady/controlled temperature. Back reflections can be mitigated by using a properly designed isolator. Currently, commercial lasers  offer RIN levels -175dBc/Hz that when integrated with the new EOM and PD devices have demonstrated intensity-modulated-direct-detection (IMDD) RF-photonic links that rival the performance of all electronic systems.
In this system, the Vp is 1.7V and the photocurrent was varied at 20, 40, and 80mA. The performance is plotted in Fig. 4, which shows performance as a function of both frequency and photocurrent. It is interesting to note that the NF of the system increases as the photocurrent was increased from 40 to 80mA. This is believed to result from compression of the PD response, which serves to note the importance of further developing the linearity properties of MUTC PDs. It is also important to note that the overall NF for 40 and 80mA photocurrents hovers at or below 10dB with a low point of 6dB at 5 GHz. It should be noted that there is no RF amplifier used in this link, which can ultimately operate over a BW of 0-100 GHz.
RF-Photonic Phased Arrays
Another key function that RF-photonics can offer is extreme beam-space processing, as shown in Fig. 5, which illustrates a new type of phased array based on optical up-conversion directly at each element in the RF front-end and does so while preserving the spatially coherent properties of every up-converted RF signal . Such an up-conversion amounts to mixing, as in an RF mixer, except in this case the mixing happens between the RF signal and the optical carrier to generate modulation sidebands of the RF signal in the optical domain. The up-conversion process is entirely coherent in that it preserves the amplitude, phase, and waveform of the RF wave in the optical domain with high fidelity. The novelty in this approach is that up-conversion is done at each element in the antenna array so that the output from each element is an optical fiber rather than a waveguide or coaxial cable. Notably, optical fibers are light, small, and immune to EMC/EMI issues. As shown in Fig. 5(a), the output fibers are then brought into a fiber bundle that forms an array to re-launch the up-converted RF signals back into free space where they re-form the exact beams that were incident on the front-end antenna array. Re-forming of the RF beams in the optical domain is possible due to the spatially coherent up-conversion process, which preserves not only the amplitude and phase of the RF signal at each antenna element, but also the phase relations between the elements [9,10]. Therefore, intuitively, one may think of the outputs from optical fibers in the bundle as Huygens’s wavelets, see the inset of Fig. 6, which due to the preservation of spatial coherence reconstruct the original RF wavefront, but in the optical domain. At this point, an optical lens performs a real-time, no-latency, no-power, two-dimensional (2D) spatial Fourier transformation, which focuses each incident wavefront to a unique spot, see Fig. 5. Each spot is then coupled through a lenslet to a corresponding fiber that routes the optical signal to a high-speed photodetector, see Fig. 5(a) inset and Fig. 5(b), and combines it with an optical local-oscillator for down-conversion to an intermediate frequency (IF) of choice. The IF signal is then digitized in an analog-to-digital converter (ADC) for subsequent processing in a digital signal processing (DSP) unit, as shown in Fig. 5(b).
Several key attributes of the proposed system are worth emphasizing: (1) the system outputs an IF regardless of the input RF frequency, thus no high-speed RF components are needed, (2) the entire 2D beamspace capture is non-blocking and computed at the speed of light while consuming no power, and (3) the beam forming, or imaging, process offers spatial isolation between imaged sectors, which adds multiple tens of dB to the spur-free dynamic range (SFDR) of the system and mitigates co-channel interference. Furthermore, this approach offers virtually unlimited BBP, which has been demonstrated up to 10 THz. The BBP advantage of the proposed approach compared to conventional systems is brought to a sharp focus by considering an emerging metric, BBP/Watt which is particularly relevant in space-borne phased array systems.
It should be noted that in this system each photodetector basically forms an independent receiver channel that can operate at a different band, or frequency, from those adjacent to it and it can be individually configured to implement any software-defined function via an FPGA, which enables programmable multi-functional and independent operation over each sector of the spatial-spectral environment. This is a unique capability and one that can best be realized using an antenna system that emulates the imaging process or, in the case of the proposed technology, an optically up-converted antenna system. A main reason for this is that such systems eliminate the need for the DSP to perform spatial processing on the received signals. Consequently, 100% of their computational resources can be devoted to signal-space, or temporal, processing. Moreover, due to the frequency agility of the RF-Photonic system, each channel only works at the desired IF BW, which significantly reduces cost and improves performance, e.g., spur-free-dynamic-range.
To demonstrate this technology a modest 1x8 phased array system was built, as shown in Fig. 7, where (a) is the initial 1x8 IMRX demonstration system and (b) use of the 1x8 phased array system in an interference mitigation experiment. In this system, because each signal has been spatially imaged to a unique photodetector it does not interact with any other received signal. Such signal separation mitigates co-channel and adjacent-channel interference (CCI, ACI) by virtue of eliminating signal intermixing or intermodulation, as shown in Fig. 7(b). Now, within a given spatial sector, all the RF signals are imaged (or spatially mapped) to the same photodetector, which outputs an IF, or BB, signal regardless of the input RF carrier frequency. This output is then fed to a single-channel, software-defined digital signal processor (SD-DSP) for subsequent processing, as shown in Fig. 5(b). As a result, each photodetector location forms a software-definable spatial channel that enables programmable functionality and independent operation within each sector of the spatial-spectral environment. Note also that different TOLOs can be used on each photodetector, thus enabling ultra-wide spectral agility, and one can also include multiple LO’s on the same photodetector to provide spectral channelization within a spatial sector [11,12].
By going directly from RF to IF, lower sampling rate ADCs can be used, which offer a higher effective-number-of-bits (ENOB) and consume less power in comparison to higher direct sampling-rate ADCs. The last key functional attribute in this system is latency, which is limited only by the time it takes light to propagate through the optical processor and is on the order of several nanoseconds. Note that this latency is not dependent on RF frequency, instantaneous bandwidth, or the number of beams coming into the front-end. Such performance enables near-immediate spatial-spectral processing, thereby providing a key feature of future 5G systems, which is a system latency of less than 1msec.
The future of infini-G wireless systems is boundless, however, to reach their full potential they will need a new class of technologies that enable their implementation in systems that are truly frequency agnostic, multi-band over wide regions of the spectrum, multi-function to combine the best of communication and sensing systems and offer performance that scales independent of operational frequency. While there is still much work to do, the field of RF-photonics offers many appealing attributes that offer the potential to meet these needs. In large part, the new paradigm of TFLN EOMs and MUTC PDs along with their implementation into new link and phased array architectures have demonstrated performance comparable to or beyond that of all electronic systems and it is sure to say that RF-photonics will be explored and perhaps incorporated in future mobile networks.
Figure 7. (a) illustration of a 1x8 RF-Photonic phased array system that has a beam-bandwidth-product of over 40 GHz. (b) Illustration of data captured from two different beams with the left and right columns showing individual (alternate) beam performance and the middle column showing simultaneous, non-blocking, performance, where despite both beams having the exact same carrier frequency co-channel interference is strongly mitigated.
- M. Zhang, C. Wang, P, Kharel, D. Zhu, and M. Loncar, "Integrated lithium niobate electro-optic modulators: when performance meets scalability," Optica, vol. 8, no. 5, pp. 652-667, May 2021.
- H. Ito, S. Kodama, Y. Muramoto, T. Furuta, T. Nagatsuma, and T. Ishibashi, "High-speed and high-output InP-InGaAs unitraveling-carrier photodiodes," IEEE Journal of selected topics in quantum electronics, vol. 10, pp. 709-727, 2004.
- X. Xie, K. Li, Y. Shen, Q. Li, J. Zang, A. Beling, and J. C. Campbell, “Photonic Generation of High-Power Pulsed Microwave Signals,” J. Lightwave Technol., JLT, vol. 33, no. 18, pp. 3808–3814, Sep. 2015.
- F. Bucholtz, V. Urick, M. Godinez, and K. Williams, “Graphical Approach for Evaluating Performance Limitations in Externally Modulated Analog Photonic Links,” IEEE Trans. Microwave Theory and Techniques, vol. 56, no. 1, January 2008.
- ANR Ahmed, S. Shi, S. Nelan, S. Shi, P. Yao, A. Mercante, and D. W Prather, “Subvolt electro-optical modulator on thin-film lithium niobate and silicon nitride hybrid platform,” Optics Letters, Vol. 45, No. 5, pp. 1112-1115, (2020).
- ANR Ahmed, S. Shi, A. Mercante, S. Nelan, P. Yao, and D.W. Prather, “High-Efficiency Lithium Niobate Modulator for K-Band Operation,” APL Photonics, Vol. 5, No. 9, pp. 091302, (2020).
- D. W. Prather, S. Shi, G. J. Schneider, P. Yao, C. Schuetz, J. Murakowski, et al., "Optically upconverted, spatially coherent phased-array-antenna feed networks for beam-space MIMO in 5G cellular communications," IEEE Transactions on Antennas and Propagation, vol. 65, pp. 6432-6443, 2017.
- R. D. Martin, C. A. Schuetz, D. W. Prather, T. E. Dillon, “Controlling the phase of optical carriers,” US Patent Number 8,159,737 (2012).
- C. Schuetz, R. Martin, D. W. Prather, and T. Dillon, “Method for controlling the phase of optical carriers in millimeter wave imaging systems using optical up-conversion: U.S. Patent number 7,965,435 (2011).
- G. J. Schneider, J. Murakowski, C. A. Schuetz, S. Shi, and D.W. Prather, “Radio frequency signal-generation system with over seven octaves of continuous tuning,” Nature Photonics, v 7, pp. 118-122, (2013).
- D. W. Prather, G. Schneider, J. Murakowski, “High Spectral-Purity Carrier Wave Generated by Nonlinear Optical Mixing,” U.S. Patent Number 8,848,752 (2014).
Statements and opinions given in a work published by the IEEE or the IEEE Communications Society are the expressions of the author(s). Responsibility for the content of published articles rests upon the authors(s), not IEEE nor the IEEE Communications Society.