N EXT-GENERATION (post-5G) wireless networks promise 50x more speed, 10x less latency, and 1000x more capacity than 4G/LTE [1], [2], while requiring low-latency response and high capacity access nodes for future broadband mobile fiber-wireless links. For example, the 3GPP specifies 5G new radio (NR) systems in frequency range two (FR2) to operate at least in n257, n258, and n260 bands [3] and some of band subsets are assigned differently in various countries/regions (Fig. 1(a)), necessitating both user equipment (UE) and remote radio units (RRUs) to cover multiple non-contiguous mm-Wave bands (e.g., at 24/ 28/ 37/ 39/ 43 GHz) to support multi-standard communication and international roaming services.
Phased array and multiple-input multiple-output (MIMO) transceiver architectures are extensively utilized for beamforming and null-steering interference suppression to improve the mm-Wave link performance and massive spatial diversity. Thus, multiple data streams can be adopted into two-dimensional, frequency and space divisions to support high-throughput spectral and spatial signal multiplexing (Fig. 1(b)). Due to urgent needs of future ubiquitous 5G NR links, the multi-band 5G MIMO RRUs are expected to concurrently handle multiple beams/streams signals [3]- [36]. Unlike static beamforming in conventional satellite communication, future mm-Wave links will serve as a key enabling technology for a plethora of commercial and defense applications such as drone-based networks, automotive radar, and precision positioning system for augmented reality and Internet of Industrial Things, which are required to operate in high-density, congested, fast-changing, and latency-sensitive environments (Fig. 1(c)). Therefore, fiber-wireless integrated access networks are essential to support reliable and low-latency operations for tracking multiple mm-Wave beams accurately and handling dynamic signals effectively. In practice, these multi-band 5G MIMO systems are expected to concurrently control multiple multi-Gb/s beams/streams over wide field-of-view (FoV) [3].
To mitigate intermodulation distortions during transmission and reception of concurrent multiple beams in a fiber-wireless network, the conventional radio-frequency (RF) and mm-Wave bands are employed in UEs and RRUs that require extensive baseband computations to generate phase and amplitude control signals for performing multiple sequential processes on beam-acquisition/-alignment/-forming [7]- [17]; however, the latency budget (<10 ms) specified in 5G NR cannot be met [3]. Although digital beamforming can process current multibeam operation using digital signal processing, it is susceptible to blocker signals due to the limited dynamic range of the analog-to-digital converters [37]. To mitigate this dynamic range issue and further achieve extensively spatial expansion, photonic-assisted, mm-Wave RF and intermediate frequency (IF) frontends are commonly adopted in an analog beamformer. For example, photonic-aided beamformer with array waveguide grating and dispersive fiber can achieve a wide bandwidth in the optical link [9]- [11]; however, it is difficult to operate in UE due to its bulky sizes and sensitive to mechanical vibration and temperature variation. On the other hand, mm-Wave RF beamformer requires an extremely broadband signal generation and a judicious frequency planning in UE and RRU to avoid image jamming in multi-band operations, which normally lead to a high-power consumption and low-efficiency design [20]. While, IF beamformer could be a promising candidate for fiber-wireless networks because it operates with relatively lower frequency to achieve a better phase controlling, relatively insusceptible to mismatch between traces, and less DC power consumption in a large phased array. The multi-band 5G NR operation can be easily reconfigured via applying different up/down-conversion local oscillator (LO) frequencies.
On the other hand, the aforementioned existing mm-Wave beamformers are operated with open-loop control, applying multiple tunings/calibrations to achieve precise multiple multiband signal beam-forming/-tracking. Future 5G ultra-reliable low-latency communications (URLLC) systems suffer from the increased system complexity and reconfiguration downtime.
To address these challenges in mobile data applications, we present a scalable wide FoV autonomous beamformer system with 22-40 GHz antenna array for a mm-Wave 5G NR fiberwireless access network over 10 km fiber and 56-cm wireless link. The autonomous beamformer achieves digital signal process (DSP)-free IF beamforming for multiple 5G NR modulated streams simultaneously via multiple delay-locked-loop (DLL) phase domain negative feedback [30]. The proposed beamformer also supports a bi-directional transceiver (TRX) for UE and RRU in 5G radio access networks (RAN). Without any prior information of angle-of-arrival (AoA), the proposed beamformer can dynamically and automatically sense and track wideband modulated signals at UEs and then transmit arbitrary signals back to the targeted RRUs, achieving a real-time bidirectional fiber-wireless communication. A proof-of-concept mm-Wave beamforming system has been designed and implemented; it achieves autonomous beam-tracking on multi-Gb/s signals over wide/full FoV (up to 180°) for multiple 5G NR frequency bands.
This paper is organized as follows. Section II presents the operation principle of the wide-FoV fiber-wireless system architecture with the broadband home-designed antenna array. The bidirectional multi-band and multi-stream mm-Wave self-steering TRX RRU/UE is demonstrated in Section III. Section IV shows the measurements and a performance comparison with recent reported fiber-wireless systems.
The proposed mm-Wave autonomous beamformer unit for the fiber-wireless network for uplink/downlink employs mm-Wave broadband home-designed antenna array, front-end low noise amplifiers (LNAs), wideband power amplifiers (PAs), up-/down-conversion mixers, zero-DC autonomous beamformer IC package, and electrical to optical converter (E/O converter) (Section III). The proof-of-concept autonomous beamformer unit operation principle is described in Section II, then the unit can be further combined with various ways in the proposed fiber-wireless network to achieve versatile system functionalities (Section III).
A 1 × 4 end-fire bow-tie home-designed dipole antenna array fabricated on two-layer Rogers RO4350 with 10 mil height and 1oz cooper thickness is proposed to support a mm-Wave low-loss antenna design for wide FoV and broadband coverage extension (Fig. 2). Differential feeding lines have differential impedance ∼100 Ω achieved by one of meandered feeding for creating out-of-phase (180°) difference. Then, the combined input single-end 50 Ω feeding transmission line with separate line width to the differential feeding line is applied to realize multi-section impedance matching with broadband nature; moreover, each single-end feeding line for each antenna is further meandered for equal phase distribution before the southwest connectors. In practical design, since it is impossible to create ideal infinite ground plane, the proposed design extends the PEC plane backside with optimal ∼3λ air /2 (λ air is wavelength at 30 GHz) on both sides to prevent abrupt reflective wave due to boundary condition between the air and the limited ground plane. The overall feedline matching network co-designs with antenna array achieving input matching S 11 < -10 dB and passive loss <1-2 dB over broadband 20-40 GHz.
Then, the far-field radiation performance of the 1 × 4 antenna array with a compact size 2 cm × 5.6 cm is measured with an inphase 4-to-1 power combiner at 0°incidence without any phase tuning. The antenna array is demonstrated with 6.7-10.4 dBi array gain from 22 to 40 GHz, noting that it preserves low gain variation <2 dB on individual 5G NR FR2 band subsets, i.e., n257, n258, and n260 bands, supporting multi-band mm-Wave 5G communication [3].
After the mm-Wave antenna array and front-end LNA amplification, the signals are then down-converted as IF signals and sent to the home designed IC package. It composes with the full-FoV DLL-based negative feedback for signal sensing and AoA detection and feedback voltage function according to incoming phase difference for conducting beam alignment/ forming (Fig. 3(a)). The autonomous beamformer array module [38], [39] consists of a phase detector, true-time-delay-based LC synthetic phase shifters, and progressive feedback control voltage generation as an array-based closed-loop beamformer with zero DC power consumption. For the feedback operation, the phase detector firstly senses phase difference of the adjacent channels (phase-to-voltage conversion G1) and group of progressive feedback control voltages are then generated a to a voltage-controlled phase shifters (voltage-to-phase conversion G2), completing phase difference compensation with a total large loop gain = G1G2 over wide FoV [39].
In a uniform λ/2 array, as the incoming signals are injected into the beamformer with an incident angle φ (Fig. 3(a)), the input progressive phase shift (IPPS) is θ in = πsinφ introducing to the middle central two paths. Then, this phase deviation is sensed by following power-aware phase detector and converted to voltage information, which is the corresponding feedback voltage V ctrl generation to the phase shifters for compensation phase θ FB . Because of the close-loop feedback operation, residual phase difference Δθ out is largely suppressed via subtraction between the feedback phase θ FB and the input θ in , which can be expressed as [39]:
To achieve high-accuracy beam alignment of the received signal over a wide IPPS θ in , i.e., wide FoV [39], the total loop gain G 1 G 2 is maximized for minimized Δθ out . The nonlinear conversion between the phase and voltage domain is exploited to be extremely large even at end-fire incident angle (i.e., θ in = ±180°, φ = ±90°) even with the all-passive zero-DC circuitry. The autonomous beamforming IC module can be further scalable into a large-element phased array. For example, in a uniform 1 × 4 array (Fig. 3(a)), -3V ctrl , -1V ctrl , 1V ctrl , and 3V ctrl are generated to compensate the input IPPS for the corresponding phase shifters to align the entire array towards the incoming direction. Due to the nonlinear power-aware closed-loop operation [39], the beamformer only acts on the signal power without any prior knowledge of AoA information. Moreover, the all-passive design ensures its zero DC operation power, which substantially improves the major issue under large-scaled array operation. Then, the beamformed signal is upconverted by an E/O converter to distributed unit through the dedicated fiber links for the upstream signal transmission. [39], the differential DC voltage V ctrl between rectifier outputs can be expressed as [39]
where β is the passive voltage amplification after the matching network for the rectifiers and rectification efficiency is α [39]. Note that the phase-to-voltage conversion is proportional to A 2 , which is driven by signal power even without AoA information.
The phase-to-voltage feedforward conversion is [39]
2) Voltage-to-Phase Conversion G 2 Operation: The differential DC outputs V ctrl of the rectifiers are then sent to voltagecontrolled phase shifter, generating the feedback compensation phase θ FB . Multi-section LC synthetic transmission lines are utilized for true-time delay to complete DLL-based closed loop with broadband nature and support wideband modulated 5G NR signals [39]. The V ctrl is fed back to voltage-controlled varactors in the LC network for corresponding θ F B generation and then suppressing the Δθ out . The feedback conversion is [39]
For outer path phase shifting in the array (Fig. 3(a)), the V ctrl signals are scaled via the resistive dividing loads [39] with corresponding scaling factor to provide progressive ±V ctrl , ±3V ctrl , …, in a large-scale phased array.
3) Loop Analysis: The overall closed loop is next analyzed. First, the output residual phase difference Δθ out can be expressed as [39]
The G 1 varies with different Δθ out during the feedback phase compensation instead of a fixed value, showing the negative feedback is a nonlinear loop. It can be further modified as
k is the loop gain value at incidence φ = 0°and has a quadratic relationship to the incident signals amplitude, showing that the loop response is a function of the signal input power. Moreover, the transcendental equation (6) shows that the nonlinear bifurcation behavior of the proposed closed loop can still maintain a large loop gain value during the loop operation even for full-FoV operation φ = ±90°or IPPS θ in = ±180°, as long as the k is >>1 [39], achieving effective phase error reduction and high phase alignment accuracy for reliable beamforming. The overall loop gain G 1 G 2 is explored in he proposed design to be large even though they are entirely composed by passive components and the large loop gain value can be maintained over a wide FoV [39].
To achieve a larger loop gain and minimize the residual phase difference of the four paths even at end-fire incidence (IPPS = ±180°, φ = ±90°), mm-Wave frontend LNAs and down-conversion mixers are applied to increase loop conversion gain and achieve improved system noise figure/sensitivity before the autonomous beamformer IC module. A proof-of-concept DLL-Like IF autonomous beamformer IC chip with is fabricated in a 130 nm CMOS process. It achieves fast response time <3 ms and an area of 1.63 mm × 2.53 mm [39], and then is wire-bond packaged in FR4 PCB (Fig. 3(b)). Large overall nonlinear loop gain of the system can flatten the normalized array factor over wide FoV coverage shown in Section IV. Note that this autonomous beamforming is a real-time operation by only one module instead by multiple samples facing to various angle mechanically to cover wide FoV. The signals are operated in a DLL-like loop supporting multi-Gb/s modulated signals through the following optical fiber as the proof-of-concept mm-Wave fiber-wireless networks.
IF-over-fiber technique is a promising solution on leveraging the spectral efficiency, TX DC-to-RF conversion, and hardware complexity. The autonomous beamformer unit is operated at IF frequency and is further configured in various system architectures for next-generation bi-directional multi-band and multi-stream mm-Wave fiber-wireless networks.
To achieve future dynamic multi-beam and multi-band mm-Wave mobile applications, a scalable wide FoV autonomous beamformer system for multi-band 5G NR fiber-wireless networks is proposed. After wideband antenna array and LNA frontend, multiple mm-Wave band-pass filters and down-conversion LOs with broadband mm-Wave mixer are applied into process different mm-Wave 5G NR bands for autonomous multi-band signals beamforming (Fig. 4). For each IF beamformer unit, it performs one beamforming toward one user/stream since there is only one phase shifter set. However, it could be scaled to the multi-user/-stream case if we integrate multiple beamformers via a similar fashion described in Section II, achieving calibrationfree IF beamforming on multiple 5G NR signals simultaneously with multiple DLL phase domain negative feedbacks (Fig. 4).
For an N-element uniform TRX array, if the signal beam is perfectly aligned, the signal-to-noise ratio (SNR) of the TRX array is improved by 10logN. However, their resulting pencilsharp beam drastically complicates the requirement of TX-RX alignment in fast-moving UE, which can be addressed by the proposed bi-directional self-steering TRX array operation in the fiber-wireless access shown in Fig. 5. First, the incoming signals are sensed and received by the front-end RX autonomous beamformer with calibration-and DSP-free closed-loop beamforming (Fig. 5). The feedback phase shift of the incoming signals is detected by the phase detector and generates corresponding differential DC control voltages Vctrl for the phase shifter in RX. When the loop is closed, one can note that the G1 conversion is a monotonic function (feedback DC voltage) versus different incidence, which provides and records AoA information of the incident signals according to the different Vctrl. This feedback DC voltage in RX forms a precise and autonomous signal incidence look-up table, which can be easily read out for TX. The AoA information is further fed forward to the TX with corresponding mirrored progressive phase shifts to align toward the incoming received signal direction, completing the swift bi-directional self-steering operation in the TRX array.
Both proposed mm-Wave wireless systems are measured in far field over-the-air 56 cm distance and 10 km fiber link. The wireless distance is conducted under the far-field criteria (Fraunhofer zone) as
where R f , D, and λ are the radiating far-field distance, antenna aperture diameter, and the RF signal wavelength, respectively. To demonstrate multi-band 5G NR operations, the wireless link is dominated by the lowest frequency, i.e., 24.75 GHz in the measurement, which R f is necessary to be > 47 cm so that the incoming received signals could be approximately as plane waves with the progressive phase shifts. A horn antenna mounted on a sweepable antenna arm is used to conduct the FoV coverage measurements. The distance between the horn antenna and the proposed system frontends can only be up to 56 cm due to our existing antenna chamber room size. The detailed measurement setups and results for the proposed systems are shown in follows.
First, the experimental setup of the proposed multi-beam and multi-band mm-Wave 5G mobile fronthaul is illustrated in Fig. 6. Due to the equipment restriction, two 5G NR bands with central frequencies of 24.75 GHz and 37 GHz are applied as a proof-of-concept demonstration. For the 24.75 GHz downlink path, a 16 GSa/s arbitrary waveform generator (AWG) generates a 1Gbaud IF signal at 4 GHz central frequency, which is then sent into a commercial direct modulation distributed feedback (DFB) laser with 10 GHz bandwidth. The central wavelength of the signal is 1550.76 nm and the output optical power is 3.5 dBm (Fig. 6). A 10-GHz photodetector (PD) is used after 10-km mobile fronthauling for optical-to-electrical down-conversion. At the transmitting side, the modulated IF signal is up converted by a broadband mixer with a local oscillator (LO) at 20.75 GHz. A duplicate path is implemented for 37 GHz mm-Wave signal delivery with the LO frequency of 33 GHz. Individual mm-Wave wideband PAs with 30 dB gain are applied for signal boosting. The 24.75-GHz and 37-GHz modulated signals are concurrently transmitted to the proposed system over 56 cm wireless link. At receiving side, the home-designed wideband 1 × 4 end-fire antenna is employed in the proposed system for the multi-beam and multi-band signals receiving. The corresponding receiving spectrum of the multi-band signal are inset in Fig. 6. Four wideband 24-40 GHz LNAs with 2 dB noise figure are employed to improve the received sensitivity and compensate the mm-Wave signal loss during wireless propagation. Two sets of band selectors are implemented sequentially and passive mm-Wave bandpass filters. They are mature devices, cost-effective and widely applied in commercial systems, for suppressing undesired out-of-band noise/blocker and image jamming. After frequency selection, mm-Wave signals are down converted to 5 GHz central frequency for following separate IF autonomous beamformers [30] to process multiple signal beamforming. The beamformers are with compact packages to easily deploy in the future highly condensed multi-beam 5G small cell. After analog-to-digital conversion via a 20 GSa/s real time oscilloscope (RTS), the received signals are then evaluated via Keysight Vector Signal Analysis software in terms of their EVM, SNR, and constellation diagrams.
The measured normalized array factor with and without the proposed beamformer system over 5G NR bands are displayed in Fig. 7(a) and 7(b), respectively. The main peak beamwidth is slightly narrower as the employed mm-Wave frequency increases, exacerbating the beam alignment problem in the mm-Wave systems (Fig. 7(a)). While, via the proposed beamformer system with autonomous tracking and steering toward the main beams of input signals, wide or even full FoV of multi-beam signals can be supported. The covered FoV, defined as the normalized array factor within 6 dB degradation, of 24-, 28-, 32-and 36-GHz are largely improved to 155°, 180°, 170°and 110°, respectively (Fig. 7(b)) without any external controls during the beamforming. The proposed system can support wide-/full-FoV operation (up to 180°) since the 1 × 4 end-fire antenna array element spacing is optimized around 28 GHz. When the mm-Wave frequencies deviate away from the 28 GHz, the antenna pattern is distorted and thus restricts the FoV. The relative movement of two concurrently spatial beams (1Gbaud signals of 24.75 and 37 GHz) is first investigated. As 37-GHz signal is fixed at position of 90°incidence, 24.75-GHz signal is then swept incidence from 80°to -80°, resulting in relative angles from 10°to 170°compared with 37-GHz signal source. In Fig. 7(c), the received EVM performance for 24.75-GHz signal is central-symmetric and gradually degraded towards -90°/90°i ncidence (i.e., 180°/0°relative angle refers to 37-GHz signal), which trajectory is similar to array factor plot in Fig. 7(b). The additional multi-beam signals from different incidence is with neglectable interference effect due to the band selectors and the interested 5G NR signal can be treated individually.
Fig. 8(a) shows EVM of 24.75-GHz signal versus incident angles, demonstrating similar performance with and without multibeam signal. Less than 5% EVM deviation is measured over wide FoV from 0°to 80°. After 10-km fronthauling, the bit-error-rate (BER) performance of 16QAM 24.75-GHz signal with the proposed beamformer at incidence is exhibited in Fig. 8(b). Again, the received performance is similar even with the coexisting multi-beam signal. The measured sensitivities, defined as the received power in FEC threshold criteria, are -4 and -1 dBm for 0°and 80°incidence, respectively. On the other hand, similar evaluation is also conducted for 37-GHz signal and shown in Fig. 8(c) and 8(d). Due to larger power attenuation of wireless propagation loss and cable loss at 37 GHz, the received SNR for 37-GHz signal is thus dropped by 4.86 dB and the corresponding EVM drops by 6.4% at 0°incidence compared to 24.75-GHz signal. The received sensitivities of 37-GHz 4-QAM signal, after 10-km fronthauling, achieve -3.5 and -2.5 dBm at 0°and 80°incidence, respectively, achieving wide-FoV multi-beamforming. With sharp frequency-selective bandpass filters, the measured desired signal (24.75-GHz signal in Fig. 8(a)-8(b) and 37-GHz signal in Fig. 8(c)-8(d)) is successfully demodulated with similar EVM/BER performance even with the presence of the other mm-Wave 5G signal injections, in the wireless-only/fiber-wireless transmissions, showing simultaneously multi-band autonomous beamforming without significant degradations from adjacent mm-Wave 5G channels.
Figure 9 shows the experimental setup of the proposed bidirectional self-steering TRX for mm-Wave 5G mobile fronthaul. 1Gbaud 64QAM 28-GHz signal is employed in this study and it is generated via a 16-GSa/s Keysight M8195A AWG. The employed DFB is operated at the same condition as in the multi-beam and multi-band evaluation and the maximum received optical power can achieve around 1.5 dBm after 10-km fiber transmission. After optical-to-electrical conversion via a 10-GHz bandwidth PD, the 5 GHz IF signal is then up-converted to 28 GHz via an mm-Wave mixer and a 23-GHz LO signal. In the downlink direction, a PA with 30 dB gain and a 25-dBi horn antenna are employed to improve link budget, compensating the 9 dB conversion loss of the mixer and 56.35 dB wireless propagation attenuation. At the UE side, the homemade 4 element end-fire array is applied and then the received signal is amplified by four LNAs with 25 dB gain. After down-converted to 5 GHz central frequency, the desired signals are further amplified by additional four LNAs (15 dB gain) for optimizing the input power of the power-aware beamformer. As mentioned in the Section III-B, the autonomous beamformer would firstly detect the incident angle of the mm-Wave signal, and then feedback the AoA information to the successive voltage-controlled phase shifters to steer the beam towards the input incident angle. After 4-to-1 beamforming combination, the received signal is analogto-digital conversed via a 20 GSa/s RTS for their EVM, SNR, and their corresponding constellation diagrams evaluation through the Keysight Vector Signal Analysis software. For uplink, it is a "mirror-reversed" design, the feedback AoA information is utilized to provide the information on the TX progress phase shifters to steer the TX beam back to the RRU. After amplification, up-conversion, and wireless transmission, the mm-Wave signal is received and down-converted to IF frequency and delivered back to DU via IF mobile fronthauling. Fig. 10(a) exhibits the normalized array factor with and without the receiving beamformer under the CW testing. One can note that four array factor nulls incur as the incident angle nearing ±20°and ±70°, which are slightly differed from the theoretical value ±30°and ±90°in a four-element phased array. This is because of the unequal phase addition from the four outputs and there is pattern distortion due to the limited ground plane in the home-designed antenna array. While, with the proposed beamformer, the normalized array factor is flatted with less than 3 dB fluctuation over ±85°, indicating that the proposed system effectively steers the beams to the desired signal. Fig. 10(b) shows that the EVM performance as a of incident angle for the 1Gbaud 28 GHz mm-Wave signal in over-the-air measurement, optical back-to-back (BTB), and after 10-km fiber transmission. The proposed autonomous beamformer supports the uniform received EVMs over 80°and achieves below 8% for fulfilling the 64QAM requirement. In other words, the proposed system can recover wideband modulated signals with over 160°F oV and the corresponding constellations with 56 cm wireless transmission are also inserted in Fig. 10(b). Fig. 10(c) displays that the feedback voltage increases monotonously with the input signal angle. This point-to-point mapping results can be utilized to form a control setting of the UE TX. For example, the phase detector feedbacks 0.2V in RX, meaning that the signal is injected from 30°incidence. Then, one can provide the corresponding mirror-reversed progress phase shifts in TX to transmit the beam back.
To evaluate the TX beamforming, different progress phase shifts are applied to form the beam with arbitrary transmitted angles shown in Fig. 11(a). The antenna array factor peak is gradually attenuated as it steers to ±90°since it is limited by the single-element isotropic broadside antenna pattern. The electric spectra of single element and four elements with array factor augmentation are shown in Fig. 11(b) and the corresponding EVMs are 7.86% and 6.14% respectively. Around 6 dB (10log4 array factor improvement) SNR enhancement in four-element array is achieved through beamforming signal toward the desired direction. In Fig. 11(c), under all the testing angle, the EVM deviation is under 1.5%, which is slightly larger than the RX-mode. Since the signal is amplified by the external power amplifiers without self-calibration, the TX performs more phase shifting mismatches than integrated beamformer in RX during various incidence. The EVMs reach below 64QAM requirement of 8% in wireless transmission and optical BTB cases; while, after 10-km mobile fronthauling, it can still achieve 8.9% with 75°transmitted angle. With the proposed system, the RX is selfaware device for autonomous AoA detection and feeds forward to TX for wide-FoV self-steering fiber-wireless network.
This paper demonstrates the first mm-Wave beamformer system that supports simultaneously autonomous beamtracking/-forming on multi-band, multiple wireless signals and bi-directional self-steering TRX for 5G mm-Wave mobile fronthaul over wireless links. It performs autonomous beam-tracking operation with a flat array factor enhancement and negligible SNR degradation over wide FoV (even at full FoV, ±90°). Furthermore, it supports enhanced multi-Gb/s mobile data transmission over 10-km fiber link, thus enabling future real-time dynamic high-speed multi-band, multi-beam operation for 5G fiber-wireless fronthaul applications.
Authorized licensed use limited to: Georgia Institute of Technology. Downloaded on July 08,2021 at 19:06:16 UTC from IEEE Xplore. Restrictions apply.