The fiber–wireless–fiber (FiWiFi) convergence system operating in a high-frequency band shows considerable promise for applications in disaster recovery, building interconnections, and providing seamless indoor access networks^[1,2]. The millimeter-wave (mmWave) band, recognized for its high bandwidth (BW) efficiency, low latency, and scalability, is well suited for encapsulating the fifth-generation (5G)-NR-compliant signals^[3]. As depicted in Fig. 1, this system utilizes converged wired and wireless links to connect the remote antenna units (RAUs) with the baseband units. In scenarios where the fiber cable installation is challenging or cost-prohibitive, mmWave wireless links can offer a viable alternative. The central office (CO) is emerging as a leading network architecture for the 5G/B5G systems, facilitating efficient management and coordination of numerous small cells via pooled baseband processing resources. At RAU-1, the conversion from an optical to a mmWave signal occurs, while RAU-2 undertakes the reverse process, converting wireless signals back to optical signals for transmission to the next link. Subsequently, the received wireless signal is conveyed to the receiver (Rx) (e.g.,  indoor residents or stadiums) through a secondary fiber link. Systems based on electronic receivers for radio-to-optical conversion at the antenna site are widely used^[4]. However, integrating electrical local oscillator (LO) sources at these sites complicates the system’s configuration, operation, and management. These challenges manifest as limited mobile fronthaul (MFH) capacity and unsatisfactory end-to-end user latency, underscoring the need for a simplified yet efficient transport system to accommodate the growing number of antenna sites in future radio access network (RAN) deployments.

To efficiently support high-density antennas, a simplified yet efficient MFH is crucial. Intermediate-frequency-over-fiber (IFoF) emerges as a promising solution for the high-capacity MFH, attributed to its spectral efficiency and the capability for multi-channel transmission^[5–7]. In IFoF systems, radio signals are transmitted over fiber as an intermediate frequency (IF) signals, enabling simultaneous transmission of multiple IF signals via frequency multiplexing in the IF domain. Previous studies on IFoF-based analog transmission have explored various aspects, including increasing the transmission capacity, evaluating air transmission interference, enhancing uplink signal quality tolerance, and addressing optical transmission line reflections. Demonstrating the capability of IFoF-based MFH to meet the transmission capacity required for the second fiber link in the IFoF-based FiWiFi system is paramount. Techniques like intensity modulation/direct detection (IM/DD) and optical in-phase/quadrature (IQ) modulation/coherent detection have been proposed for a second fiber link transmission to the Rx^[8]. However, these methods may limit system performance or increase receiver and digital signal processing (DSP) complexity. Consequently, developing a straightforward, cost-effective approach in the FiWiFi system is vital for the deployment of ultra-dense small cells in 5G and 5G-advanced networks. Recent research on the IFoF MFH systems includes the 5G-NR mmWave frequency ranges FR2-1 (24.25–52.60 GHz) and FR2-2 (52.6–71.0 GHz, as specified in 3GPP Release-17)^[9]. Additionally, the 5G FR1 bands including 3.3–3.6 GHz and 4.8–5.0 GHz have been identified as potential frequency spectrums by China’s Ministry of Industry and Information Technology. Due to its expansive bandwidth availability, the mmWave band offers significant potential for encapsulating 5G-NR FR1 signals. To date, no studies have investigated the encapsulation of 5G-NR-compliant signals in the FR1 band using a seamless IFoF-based FiWiFi high-frequency mmWave system.

In this Letter, we experimentally demonstrate a transparent IFOF-based FiWiFi system using DML/PM/IM-based electro-optic (EO) modulators. To simplify antenna sites, we utilize a simple optical heterodyne method using free-running external cavity lasers (ECLs) for generating mmWave signals at RAU-1. An envelope detection (ED) with a directly modulated laser (DML) method can reduce system complexity, cost, and phase noises while ensuring compatibility with existing passive optical networks for ultra-dense small cells. Furthermore, we explore optical phase modulation (PM)-/IM-based methods for transforming down-converted wireless signals into optical signals for subsequent fiber transmission. Notably, we employ a thin-film lithium niobate (TFLN) IM with linear EO performance in this work^[10]. As a proof-of-concept demonstration, we encapsulate the 5G-NR-compliant IF signals at 3.5 and 4.9 GHz over the fiber-wireless-fiber system across the 75–110 GHz band. In this work, we successfully transmit higher-order modulation and large bandwidth signals with the help of a low-noise amplifier (LNA) in the mmWave link. Satisfactory performances are experimentally confirmed for the transmission of 100, 200, 300, and 400 MHz BW 64QAM/256QAM 5G-NR signals. The proposed system can provide a simple yet cost-effective solution for high-speed bridge and uplink mobile fronthaul in ultra-high frequency bands.

The experimental setup for a proof-of-concept demonstration of the proposed IFoF-based FiWiFi system is shown in Fig. 2. The CO/baseband unit (BBU) generates and modulates the 5G-NR IF signal, and the Rx receives and demodulates this signal. The RAU-1 converts optical signals to mmWave signals, and the RAU-2 converts wireless signals to optical signals for subsequent transmission to the user end. Figure 2 also gives the corresponding photos of the experimental setup.

At the CO/BBU, two ECLs (ECL-1 and ECL-2), each with a linewidth of less than 100 kHz, are employed to generate the optical carrier. The wavelength spacing between these two carriers is maintained at 0.59–0.87 nm, corresponding to a frequency range of 75–110 GHz, as illustrated in Fig. 2(a). The two signals are combined by an optical coupler (OC), and the coupled carriers (λ1 and λ2) can be modulated via an Mach–Zehnder modulator (MZM) together. The tunability of the laser sources facilitates easy access across the full frequency spectrum. We generate the 5G-NR mobile signal using a vector signal generator (VSG). The optical carrier is then modulated with the mobile signals through a simple MZM featuring a 30 GHz bandwidth-based intensity modulation. At the RAU-1, the signal traverses a 10 km standard single-mode fiber (SSMF) with an average loss of 0.2 dB/km. A variable optical attenuator (VOA) is utilized to adjust the received optical power (ROP) into the photodetector (PD). To maximize the air interface rate, the signal is amplified by an LNA with a 35-dB gain before being input into the PD. This step converts the signal into a mmWave signal within the 75–110 GHz band, which is the aggregate of the frequency separation of the two-tone optical signal and the carrier frequency of the 5G-NR signal. The application of optical self-heterodyne technology for signal up-conversion at RAU-1 ensures relatively high-frequency stability and low phase noise of the detected signal. At the wireless link, the mmWave signals are emitted from a W-band antenna (HA) with a gain of 26 dBi. After transmission over approximately 3 m in free space, the signal is captured by another HA at RAU-2, where an LNA further amplifies it.

To streamline the network architecture, Fig. 2 presents the exploration of four simple methods, conducted to validate the feasibility of the proposed IFOF-based FiWiFi convergence system operating in the 75–110 GHz band. These approaches include: (1) Employing ED to process the amplified signal and amplifying by another LNA, subsequently analyzed by the vector signal analyzer (VSA), establishes the foundational baseline. (2) The enhanced mmWave signal is processed through an ED and modulated by a DML operating at optimal parameters. This process yields an optical signal that traverses a 5 km SSMF link. (3) A PM-based method is utilized to generate a double-sideband (DSB) signal, which is then filtered through a tunable optical filter (TOF) to yield a single-sideband (SSB) signal. After amplification by an erbium-doped fiber amplifier (EDFA), the optical signal is transmitted over a 5 km SSMF transmission. (4) A TFLN-based method replaces PM to generate a DSB signal. Finally, the IF signal is recovered through beating by a low-power PD. Figure 3(a) depicts that the PM-based method exhibits a higher signal-to-noise ratio attributable to lower insertion losses and a high-frequency fading phenomenon occurs. Figure 3(b) shows that the TFLN-based method has the effect of carrier suppression by adjusting the bias voltage of the modulator. However, the use of the TFLN-based scheme is associated with reduced robustness to noise and distortion, leading to a deterioration in system performance.

As a proof-of-concept demonstration, we transmit 5G-NR-compliant signals at 3.5 and 4.9 GHz over the IFOF-based FiWiFi system across the 75–110 GHz band, including DML/PM/TFLN-based methods. The modulation signal is set to 64 or 256QAM, and the signal bandwidth is 100, 200, 300, or 400 MHz, which is the largest order and bandwidth of a single carrier component defined by 3GPP. The performance of the employed system is measured in terms of error vector magnitude (EVM) for different ROP levels, as well as carrier frequency and bandwidth, considering 3.5% and 8% EVM threshold levels for 64QAM and 256QAM. For all IFoF-based FiWiFi cases, these measurements are carried out over a 10 km SSMF link, followed by a 3 m wireless link and concluding with a 5 km SSMF for a secondary fiber link.

In this experiment, we first evaluate the EVM performance versus the ROP values for the 5G-NR-compliant signals at 3.5 and 4.9 GHz, each with a bandwidth of 100 MHz, as shown in Figs. 4(a) and 4(b). It can be observed that both signals show similar performance trends, although the 3.5 GHz signal performs slightly better. The results also show that almost the same performance is achieved for 64QAM and 256QAM in each case. Considering a 3.5% EVM threshold level for 256QAM modulation, it can be observed that satisfactory performance is obtained for the 256QAM signal, representing the highest-order modulation defined in 3GPP. The performance is much better than the requirement of 8% for the 64QAM signal with the ROP value increasing. The performance is degraded in the lower ROP region owing to the dominance of the noise power. A simpler and more cost-efficient transmission system is demonstrated by employing an ED with a DML method. The performance of the PM-based modulation is also shown, where an ROP value of 3 dBm is required for a 3.5% level, which is 4 dB more than in the TFLN-based modulation with a lower driving voltage. Notably, the signal performance is also degraded when increasing the ROP value beyond the optimal value. Intensity modulation may enter a saturation regime, especially for the TFLN-based method. This could be caused by the nonlinear effects in lithium niobate materials, such as second-harmonic generation and third-order nonlinear Kerr effect^[11]. The optimal performance is achieved when ROP approaches 1 and 0 dBm for 256QAM signal at 3.5 and 4.9 GHz, and the constellation is represented in the illustration. The optimal ROP value of the DML/PM/TFLN-based modulation method can be obtained and used for the following validation. Moreover, the average phase error (PE), magnitude error (ME), and carrier frequency error (CPE) are quantified using the VSA. Table 1 presents the measurement results for a 100 MHz 256QAM signal at 3.5 GHz employing the DML-based method.

Figure 5 depicts the EVM performance versus variable carrier frequencies, with a specific focus on optimizing the ROP values. We achieve precise tuning of two ECLs, resulting in eight distinct carrier frequencies ranging from 75 to 110 GHz. The choice of carrier frequency plays a pivotal role in RF/mmWave transmission systems, particularly when considering bandwidth constraints. Our findings demonstrate that the DML-based FiWiFi system consistently maintains robust performance within the 3.5% EVM threshold for 256QAM signals across this frequency range. Notably, the lowest EVM value, approximately 1.85%, is recorded at a carrier frequency of 95 GHz. Comparative analysis reveals that about 1.2% relatively elevates the EVM values at these higher frequencies to the 4.9 GHz signal. This increase may be attributed to the frequency response characteristics of various components, including PD, LNA, and ED. The illustrations in this figure also give the 256QAM constellation points for both 3.5 and 4.9 GHz signals at a carrier frequency of 105 GHz. However, the PM/TFLN-based FiWiFi system can meet the 3.5% EVM level requirement within a designated carrier frequency range.

Furthermore, we explore the performance of the 100, 200, 300, and 400 MHz bandwidth 5G-NR-compliant signals for different carrier frequencies in the IFOF-based FiWiFi system, as shown in Fig. 6. As indicated in the figure, we confirm that the 5G-NR signal can meet the 3.5% EVM level for all cases, especially in the low bandwidth range. Employing the DML/PM-based method over the whole carrier frequency, satisfactory performance is confirmed for the 200 MHz 256QAM and the 400 MHz 64QAM signals at 3.5 GHz, while the signal at 4.9 GHz has a suboptimal performance. In addition, using the TFLN-based method, satisfactory performance is obtained for the 100 MHz 256QAM and 400 MHz 64QAM signals. Although the performance is degraded in the high-frequency and high-bandwidth regions due to bandwidth limitation, underdrive voltage, and insertion loss, the 3.5% EVM threshold for 256QAM signal can be met in the smaller carrier frequency range. These methods with different characteristics enhance a solution for efficient transmission of the IFOF-based FiWiFi system.

Figure 7 shows the example of the 5G-NR 64QAM/256QAM constellation and spectra of the received signals. The constellation points within the figure represent the control channel featuring 5G-NR-compliant signals. Notably, optimal EVMs of approximately 2.6 % and 2.8% are measured for the 400 MHz 256QAM signals at 3.5 and 4.9 GHz, respectively. Additionally, the 400 MHz 64QAM signals meet the required 8% level across all frequency bands. Furthermore, Table 2 provides a comprehensive overview of the observed values, including the PE, ME, and CPE across carrier frequencies and BWs. These values are measured under various conditions, adding depth to our analysis.

We have effectively demonstrated the encapsulation of 5G-NR-compliant signals in the transparent fiber-mmWave-fiber integrated system across the 75–110 GHz band. Utilizing an optical heterodyne at RAU-1 and a DML/PM/TFLN-based radio-to-optical conversion method at RAU-2 has markedly simplified antenna sites and the receiver. The system exhibits satisfactory performance, handling large bandwidth signals and broad carrier frequencies using a DML/PM/TFLN-based modulator. Moreover, we have successfully transmitted 5G-NR signals, specifically 400 MHz 64QAM and 400 MHz 256QAM, over an IFoF-based FiWiFi system comprising two fiber links and a 3-m wireless link. These results underscore the potential of our proposed system as an efficient solution for fiber-radio bridges and uplink fronthaul systems, particularly in ultra-high frequency bands pertinent to beyond-5G network contexts.