A Transceiver-Shared Photonic Integrated Broadband Multi-Beamformer Based on an Extended Blass Matrix ^†

Abstract

Multi-beam phased array antennas have become essential in modern radar and communication systems, offering high gain, superior directivity, and exceptional agility. However, traditional multi-beam phased array antennas face significant challenges in meeting the growing demand for large, instantaneous bandwidth and compatibility with transmit-and-receive multi-beamforming. To achieve these requirements, we propose a novel transceiver-shared photonic integrated broadband multi-beamforming network architecture based on an extended Blass matrix framework. Combined with wavelength division multiplexing, the architecture enables the separation and decoupling of transmit and receive channels, ensuring the independent synthesis of multiple beams for transmission and receiving. Furthermore, we design and implement a 3 × 3 transceiver-shared photonic integrated broadband multi-beamformer on a standard silicon-on-insulator platform. The proposed multi-beamformer successfully demonstrates broadband multi-beamforming across six independent directions, with transmitted beams at 15°, 30°, and 45° and received beams at 20°, 40°, and 60°, covering both the whole X and Ku bands.

1. Introduction

The escalating requirements of radar systems for high-resolution detection and multidimensional information acquisition, coupled with the pressing demands of communication systems for enhanced gain, expansive coverage, and multi-node connectivity, have positioned multi-beam phased array antennas (PAAs) as a critical enabling technology in modern radio-frequency (RF) systems and wireless applications [1,2,3,4]. Conventional beamforming networks grounded in electronic technologies face inherent limitations. First, electronic bottlenecks and substantial insertion loss (IL) caused by electrical true time delay lines (TTDLs) restrict the ability of electrical multi-beamformers to achieve wideband and wide-angle scanning in high-frequency regimes [5,6]. Furthermore, Digital Beamforming (DBF) encounters significant challenges due to the power efficiency and cost constraints of analog-to-digital converters (ADCs) and digital-to-analog converters (DACs) [7,8,9]. These limitations critically impede the advancement of DBF systems toward higher operational frequencies, extended instantaneous bandwidths, and scalable implementations.

Recent advancements in microwave photonics (MWP) and photonic integrated circuit (PIC) technologies present transformative solutions for broadband multi-beamforming networks (MBFNs), addressing critical challenges in modern radar and communication systems. MWP technology offers inherent advantages, including ultra-wide operational bandwidth spanning microwave to millimeter-wave regimes, flexible multiplexing architectures, and intrinsic immunity to electromagnetic interference (EMI) [10,11,12]. These capabilities enable seamless full-bandwidth coverage through a unified optoelectronic link, substantially simplifying system architectures while minimizing component redundancy and hardware complexity. Second, compared to electrical beamforming networks utilizing coaxial cables or other RF waveguides for true time-delay implementation, incorporating ultra-low-loss optical waveguides into optical multi-beamforming networks (OMBFNs) based on optical true time delay (OTTD) significantly reduces transmission loss, which can substantially mitigate inter-channel IL variations [13]. This advancement not only enables wideband squint-free beamforming but also ensures scalable deployment in large-aperture phased arrays. Furthermore, advancements in PIC technology effectively mitigate environmental perturbations on fiber-optic components, enhancing system reliability while reducing costs [14,15]. Several experimental validations have confirmed the feasibility of these architectures. For instance, an OMBFN employing a phase-shifted Blass matrix framework—fabricated on a silicon nitride (SiN) platform—has been optimized for scan-on-receive (SCORE) synthetic aperture radar (SAR). The OMBFN achieves a simultaneous synthesis of three independent beams with an instantaneous bandwidth of 390 MHz [16,17,18]. A separate SiN-based multi-beamformer utilizing an OTTD-enabled Butler matrix architecture has demonstrated the capabilities of the bandwidth extension, enabling eight-beam synthesis within a single spatial dimension while achieving an extended instantaneous bandwidth of 800 MHz [19,20]. In our prior work, a silicon-based integrated broadband multi-beamformer operating in receive mode, leveraging an OTTD-induced Blass matrix, has been developed. The multi-beamformer supports the concurrent synthesis of five beams at steering angles of 0°, ±20°, and ±40° across the 8–18 GHz frequency range [21].

Nevertheless, these OMBFNs are inherently constrained by their single-mode operation (either transmission or reception), imposing significant limitations on system design and application. First, single-mode beamformers necessitate dedicated hardware resources—such as independent RF circuits, control units, and synchronization links—for transmission and reception separately, thereby increasing system complexity and deployment challenges. Second, the separate design of transmit and receive paths leads to resource redundancy and inefficiencies, which are particularly detrimental in scenarios requiring high integration density and operational flexibility. Moreover, the reliance on distinct transmission and reception modules exacerbates system volume, power consumption, and cost. Finally, the inability of single-mode architectures to facilitate rapid mode switching further restricts system adaptability and overall efficiency. In contrast, transceiver-shared beamformers integrate both transmission and reception signal processing within a unified beam-control unit via a shared beamforming network. This architecture significantly streamlines hardware design by eliminating redundant components associated with separate signal transmission and reception, thereby improving resource utilization while reducing system complexity, power consumption, and cost.

To address these challenges, we propose and experimentally demonstrate a transceiver-shared (compatible with transmission mode and reception mode) photonic integrated broadband multi-beamforming network architecture based on an extended Blass matrix framework. Compared to the conventional Blass matrix, the extended framework enables multi-sided input with same-sided output, allowing for the independent separation of transmit and receive channels. By redesigning the structure of the control nodes, independent delay control for both transmitted and received signals is achieved, effectively decoupling the transmit and receive paths. This network supports both independent operation of the transmit and receive modes, as well as dual-mode operation under specific constraints. Leveraging this design, we implemented a 3 × 3 transceiver-shared photonic integrated broadband multi-beamformer on a standard silicon-on-insulator (SOI) platform and validated its functionality. The fabricated chip successfully demonstrated broadband beamforming for three transmit and three receive channels across six independent pointing directions, achieving full coverage of the X and Ku bands.

2. Chip Design

The traditional Blass matrix framework exhibits several limitations. First, the framework supports only unidirectional signal transmission. With a single set of input channels and a corresponding set of output channels positioned on adjacent sides, the framework lacks the capability for independent bidirectional transmission, making it unsuitable for transmit and receive operations. Second, the functionality of the network’s control nodes is inherently limited. The traditional control node is a four-port device including only one tunable element, restricting its ability to accommodate bidirectional signal transmission across both transmit and receive modes. Finally, the delay units within the network are constrained to constructing delay gradients along a single dimension. For instance, in our previous work [21] on multi-beamformers operating in receive mode, a series of delay gradients could be introduced between input channels, corresponding to the delay differences among antenna elements. However, the framework cannot establish independent delay gradients between different output channels.

To realize the various limitations of the conventional Blass matrix architecture, an extended Blass matrix-based transceiver-shared photonic integrated broadband multi-beamforming network architecture is proposed [22], as illustrated in Figure 1. This architecture introduces the following key improvements.

Firstly, to enable transceiver-shared multi-beamforming functionality, the number of input and output channels for both the transmit and receive paths must be expanded. Specifically, the conventional Blass matrix, which originally consists of N input channels and M output channels, is extended to accommodate 2 × N input channels and 2 × M output channels, with half allocated for transmission and the other half for reception. The transmission configuration is designed such that each input port is assigned to a specific beam direction, with the output ports directly interfacing with radiating elements in a one-to-one correspondence. In contrast, the reception configuration implements the inverse connectivity scheme. To facilitate independent control of signals in both operational modes, the transmit and receive input channels are positioned on opposite sides of the extended Blass matrix framework, while the output channels remain grouped on the same side. The corresponding signal transmission paths are illustrated in Figure 1, where red lines represent the signal paths for the transmission mode, and blue lines denote the paths for the reception mode, with arrows indicating the direction of signal propagation. The extended Blass matrix framework effectively supports bidirectional signal transmission while maintaining a separation between the two modes. Signals are processed independently and routed through distinct output channels, thereby ensuring compatibility with transmit and receive multi-beamforming.

Next, to achieve independent control of transmitted and received signals, the original four-port control node in the Blass matrix has been expanded to an eight-port configuration, with detailed structural specifications illustrated in Figure 2. This enhanced control node architecture primarily comprises two tunable optical couplers (TOCs), two optical combiners (OCs), and two optical crossings. In the proposed architecture, TOCs are employed to control the routing paths and power distribution of input signals across different channels. OCs are used to merge signals from various channels, while optical crossings are strategically incorporated to eliminate multipath effects and suppress inter-channel crosstalk. The eight ports are functionally distributed as follows: $A_T$ and $C_T$ serve as input ports for the transmission path, while $B_T$ and $D_T$ function as corresponding output ports. Conversely, $B_R$ and $C_R$ are designated as input ports for the reception path, with $A_R$ and $D_R$ serving as their respective output ports. When operating in either mode, the directional control mechanism of this node maintains functional consistency with the original four-port configuration, with the primary modification being the addition of an extra TOC that does not affect signal transmission characteristics. Specifically, for the transmission path, the input signal at $A_T$ can be routed to either $B_T$ or $D_T$ through the left TOC, while the signal entering through $C_T$ is exclusively directed to $D_T$. Similarly, in the reception path, the input signal at $B_R$ can be selectively directed to $A_R$ or $D_R$ via the right TOC, whereas the signal from $C_R$ is exclusively output through $D_R$. This architectural enhancement enables independent control and distinct channel separation for both transmission and reception signals. In the simultaneous transmit–receive operation mode, the system typically supports full beam synthesis for transmission and single beam synthesis for reception. However, the synthesis of additional reception beams requires enhanced degrees of freedom in the tunable coupler structure, specifically through increased wavelength sensitivity, which consequently imposes more stringent requirements on optical carrier wavelength selection.

Finally, to achieve independent beamforming in both transmission and reception modes, it is essential to implement distinct and independently configurable delay gradients for transmitted and received signals. The original one-dimensional delay unit array is required to be expanded into a two-dimensional architecture. Specifically, for transmitted signals, a series of delay gradients are applied across the output channels, corresponding to row-wise delays in Figure 1. Furthermore, for reception signals, another series of delay gradients is introduced across the input channels, corresponding to column-wise delays in Figure 1. This orthogonal delay configuration ensures independent control and precise tuning of both transmission and reception beams within a unified network architecture.

The integration of these three architectural enhancements has resulted in the completion of a novel broadband multi-beamforming network architecture capable of supporting transceiver-shared operations. Signal decoupling across multiple channels has been accomplished through the implementation of a multi-wavelength optical source array, where individual wavelengths function as dedicated optical carriers for separate channels. The detailed theoretical analysis of the multi-beamforming can be directed to our prior research [21].

Compared with the conventional OTTD-enabled Blass matrix, the proposed extended Blass matrix architecture introduces a two-dimensional delay unit array that enables squint-free multi-beamforming for both transmission and reception paths, without compromising bandwidth or delay precision. In terms of power consumption, although the proposed architecture integrates two sets of TOCs for transmit and receive operations—matching the total number of TOCs in two separate conventional beamformers—it requires a greater number of simultaneously tuned control signals when operating in a single-mode (transmit or receive) configuration. As a result, the average system power consumption is slightly increased.

To validate the compatibility of the proposed broadband multi-beamforming network architecture based on the extended Blass matrix for both transmission and reception modes, we design and implement a 3 × 3 integrated broadband multi-beamforming system on a standard SOI platform. The selection of the 3 × 3 configuration was primarily dictated by chip area constraints; however, the architecture itself is inherently scalable and can be readily extended to accommodate broader system requirements. The schematic diagram of this system is presented in Figure 3. For enhanced clarity in illustrating the structural details, the eight-port control nodes are represented as octagons, maintaining identical internal configurations and port definitions as shown in Figure 2. This multi-beamforming system features RF input and output interfaces, comprising six RF input channels and six RF output channels, equally distributed between transmission and reception paths. The multi-beamformer consists of three primary functional modules: an electro-optic conversion module incorporating six Mach–Zehnder modulators (MZMs), a 3 × 3 broadband multi-beamforming network based on the extended Blass matrix supporting the transceiver-shared operation, and a photoelectric conversion module containing six photodetectors (PDs). The operational bandwidth covers both X and Ku bands. Within the network’s delay unit configuration, the transmission path incorporates three distinct sets of delay units (represented by double coils in Figure 3) arranged in the row direction, establishing three different delay gradients among the transmission output channels to achieve three distinct transmission beams. Additionally, the reception path implements three sets of delay units (represented by single coils in Figure 3) in the column direction, creating three different delay gradients among the reception input channels to enable three independent reception beams.

In order to demonstrate the capability of this architecture in independently establishing distinct delay gradients for both transmission and reception paths, thereby enabling corresponding beamforming, the six beams are synthesized with angles of 15°, 30°, and 45° are for transmitting, and 20°, 40°, and 60° for receiving. The delay gradient of each beam, determined by the center wavelength of the RF signals the broadband multi-beamformer and the angle of the beam. It can be written as

$$\tau ={\displaystyle \frac{dsin\theta }{c}}$$

(1)

where d is the distance between adjacent antenna units, proportional to the operating center wavelength of the PAA systems, $\theta$ represents the angle between the beam pointing and the normal of the antenna array, and c is the velocity of the light wave in the vacuum. Following Equation (1), at the operational center frequency of 12 GHz, the designed delay gradients for the six beams are set to 10.78 ps, 20.83 ps, and 29.46 ps for the transmission path, and 14.25 ps, 26.78 ps, and 36.08 ps for the reception path.

The OCs within the control nodes employ asymmetric Mach–Zehnder interferometer (AMZI) structures to achieve wavelength division multiplexing. This design can significantly reduce the IL of the overall network and improve the IL imbalance between channels by matching the optical carriers of different input channels and the passbands of the AMZI-based wavelength division multiplexers. However, the inherent wavelength sensitivity of the AMZI structure inevitably affects the operating bandwidth of the OMBFN. To optimize the trade-off between achieving Ku-band operational bandwidth, accommodating the wavelength range of the test light source, and ensuring scalability in the number of channels, the AMZI structure parameters must be carefully designed.

The AMZI structure, as illustrated in Figure 4, is characterized by two primary parameters: the coupling coefficient ($\kappa$) of the input/output couplers and the arm length difference ($\Delta L$). The coupling coefficient determines the power splitting ratio between the two arms of the AMZI structure. To achieve an optimal extinction ratio (ER), the coupling coefficient is uniformly set to 0.5, establishing an equal power distribution (1:1 ratio) between the two arms. The arm length difference $\Delta L$ serves as the critical parameter determining the free spectral range (FSR) and operational bandwidth of the AMZI structure. The quantitative relationship is expressed by the following equation:

$$FSR=c/(n_g/\Delta L)$$

(2)

where the c is the velocity of the light wave in the vacuum and the $n_g$ is the group refractive index of optical waveguides.

To ensure the operational bandwidth of the entire network extends to the Ku band, the optical combiners in the control nodes corresponding to the three-level input channels (from top to bottom) were designed with AMZI structures featuring arm length differences ($\Delta L$) of 400 µm, 200 µm, and 100 µm, respectively. Furthermore, to enhance the network’s control flexibility, the TOCs within the control nodes were also implemented using AMZI structures with a length difference of 400 µm. The simulated optical spectra of these three AMZI configurations with different arm length differences are presented in Figure 5a, Figure 5b and Figure 5c, respectively. The simulated results demonstrate that the AMZI structures exhibit FSRs of 1.38 nm (172.5 GHz), 2.78 nm (347.5 GHz), and 5.54 nm (692.5 GHz), with corresponding bandwidths of 0.7 nm (87.5 GHz), 1.39 nm (173.75 GHz), and 2.78 nm (347.5 GHz). Notably, a halving of the arm length difference $\Delta L$ results in a doubling of both the FSR and bandwidth. As evident from Figure 5, the ILs for the three AMZI structures are 0.62 dB, 0.56 dB, and 0.53 dB, respectively, with the ER exceeding 20 dB for all configurations. Furthermore, the group delay characteristics of the AMZI structures are crucial for network design, as they influence the waveguide lengths required for constructing the delay gradients. Figure 5d presents the simulated group delay characteristics of the three AMZI configurations with different arm length differences. The results indicate group delays of 12.25 ps, 10.92 ps, and 10.25 ps for the three structures when operating within their passbands. These delay values, when converted to equivalent waveguide lengths L using Equation (3), must be compensated for in the design of the delay units to achieve the desired delay gradients.

$$L=c\tau /n_g$$

(3)

The proposed architectural framework is subsequently validated through extensive link-level simulation analyses. To minimize additional losses introduced by optical combining, the wavelength selection of optical carriers for each channel requires precise alignment with the FSR values of the corresponding AMZI-based OCs. Specifically, the wavelength spacing between optical carriers in adjacent channels should be maintained at half the FSR value of the OC in the subsequent channel. For instance, the wavelength separation between the optical carriers of Channel 1 and Channel 2 should be set to half the FSR of Channel 2’s OC, corresponding to 1.39 nm (173.75 GHz). Similarly, the wavelength of Channel 3’s optical carrier should be spaced at half the FSR of Channel 3’s OC from the central wavelength between Channels 1 and 2, with this pattern continuing for subsequent channels. The precise characterization of phase shifters integrated within all AMZI structures represents a critical factor in achieving optimal signal transmission. This optimization process involves the precise alignment of each channel’s signal with the maximum transmission point of its corresponding spectral response curve, thereby ensuring the minimization of transmission losses. Additionally, the intrinsic low IL properties inherent to the AMZI architecture substantially alleviate inter-channel IL disparities, consequently enhancing the comprehensive performance of the system.

According to the FSRs of the three AMZI structures, the wavelengths of the four optical carriers ($L{D}_0$, $L{D}_1$, $L{D}_2$, and $L{D}_3$) are precisely set to 1550.32 nm, 1549.63 nm, 1548.59 nm, and 1546.5275 nm, respectively, with each maintaining an optical power of 0 dBm. This configuration is intended to simulate channel expansion applications, for example, where one optical carrier ($L{D}_0$ at 1550.32 nm) can be externally modulated and directly coupled with the optical signal of Channel 1. Through meticulous tuning of the phase shifters in all AMZI-based OCs, the transmission curve peaks are precisely aligned with the respective channel wavelengths. The specific correspondence relationships between the transmission curves of the OCs for the subsequent three channels and the wavelengths of the four optical carriers are illustrated in Figure 6a.

Following the characterization of the control signals for each node within the OMBFN, four optical signals are simultaneously injected into their respective channels. As illustrated in Figure 6b, the output spectra, measured at a common output port, are analyzed to determine optical power levels and to calculate channel-specific IL values: 8.43 dB, 8.08 dB, 4.06 dB, and 4.68 dB. The results indicate an inter-channel IL variation of approximately 4 dB across the three channels. The integration of AMZI-based OCs has demonstrated superior performance in mitigating IL variations. Further enhancements in IL uniformity can be achieved through advanced AMZI designs, including expanded FSR, and reduced intrinsic loss.

Subsequently, the operating bandwidth of the OMBFN is also simulated. Considering the wavelength-dependent characteristics of the TOCs and OCs within the nodes, the simulation focused on the longest transmission path (for example, the path from input Channel 1 to output Channel 3 on the receiving end). As shown in Figure 6c, the simulated transmission curve demonstrates a 3 dB bandwidth of 37 GHz, indicating that the network’s overall operational bandwidth spans 37 GHz in the optical domain. When converted to the RF domain, this corresponds to 18.5 GHz, sufficient to cover the operational requirements of both X and Ku bands.

After thorough verification through link-level simulations, we implemented the design on a standard SOI platform for tape-out validation. The fabricated chip, with sizes of 3 mm × 8 mm, incorporates all necessary active and passive components (excluding the light source), comprising a grating coupler (GC) array, a MZM arrays, a 3 × 3 transceiver-shared photonic integrated broadband multi-beamforming network based on the extended Blass matrix, and a PD array. A microscopic photograph of the fabricated chip is shown in Figure 7.

3. Experiment and Results

To evaluate the performance of the integrated broadband multi-beamformer with transceiver-shared functionality and validate its feasibility, the chip testing is conducted in two distinct phases: component parameter measurement and system performance test.

3.1. On-Chip Device Testing

The key on-chip components influencing the operating bandwidth of the multi-beamformer primarily include wavelength-sensitive AMZI structures and active on-chip devices. To validate the broadband characteristics of the multi-beamformer, the bandwidths of three AMZI structures with different arm length differences, as well as the overall bandwidth of the MZM and PD, are systematically evaluated.

First, transmission curves of the three AMZI structures are measured using the test setup shown in Figure 8, with the wavelength sweep range of a tunable laser source (TLS, Santec TLS550) set to 1535–1545 nm. The variations in output power recorded by an optical power meter (OPM, Keysight 8163B) for each AMZI structure as a function of wavelength are presented in Figure 9a, Figure 9b and Figure 9c, respectively. The observed power fluctuations at peak points are primarily attributed to the wavelength-dependent loss of the on-chip GCs. By comparing and analyzing the experimental measurement data with the simulation results, it is observed that the FSRs and bandwidths of the AMZI structures exhibit a high degree of consistency with the simulated values. Furthermore, when combined with the link simulation results, these performances of AMZI structures confirm that the operating frequency range of the OMBFN can extend to the Ku band.

Subsequently, the bandwidth characterization of on-chip active components is conducted using the test setup illustrated in Figure 10a. The TSL is employed to generate a 1545 nm optical carrier with 0 dBm output power, which is amplified by an Erbium-doped fiber amplifier (EDFA, Keopsys CEFA-C) and injected into the on-chip MZM. The MZM is biased at the quadrature point, while the PD is biased at −3 V provided by a multi-channel source (MCS, NI PXIe-4162). The S21 parameters of the MZM and PD can be measured by a pre-calibration vector network analyzer (VNA, Rohde and Schwarz ZVA40), as shown in Figure 10b. The result demonstrates the 3 dB bandwidth of the on-chip active components covering 18 GHz, satisfying the design requirements for X and Ku bands in the transceiver-shared multi-beamforming system.

3.2. System Performance

System performance focuses on two critical aspects of the integrated broadband multi-beamformer: IL and delays of each channel.

To characterize the system’s IL, we selected three channels (specifically, input channels 1, 2, and 3 to output channel 1 in the receiving end) that traverse the longest optical paths and are required for synthesizing the same beam. IL measurements were also performed using the same setup as shown in Figure 10a. The control signals provided by the MCS are used to optimize the state of the AMZI structures within the test channel, thereby minimizing its loss. Subsequently, the MZMs in these channels are all biased at their orthogonal points. The S21 parameters of the channels are then acquired using the VNA to determine the final IL measurements. Figure 11 shows the IL results of the three channels at six frequency points in the 8–18 GHz range. The measurements show that the IL of the channel with the longest optical path remains approximately 40 dB throughout the X and Ku bands. In addition, the IL variation between the three channels synthesizing the same beam is approximately 4 dB, which is consistent with the simulated results shown in Figure 6b.

Next, the control signals for each channel are sequentially characterized, followed by the measurement of corresponding delays across the 8–18 GHz range using a VNA. The relative delay measurements for three transmission beams and three reception beams are presented in Figure 12a–c and Figure 12d–f, respectively. In these figures, solid lines represent the measured delay values, while dashed lines indicate the design reference values. The results demonstrate a strong agreement between the measured inter-channel delay differences and the expected reference values across the entire 8–18 GHz frequency range. Figure 12g,h further illustrate the delay deviations at different frequency points for the transmission and reception channels, respectively. The majority of delay deviations remain within 5 ps, with a maximum observed deviation of 11.31 ps. These deviations are primarily attributed to imbalances and non-uniformities in the S-parameters of on-chip active components. Such discrepancies can be effectively mitigated through the optimization of device performance parameters.

Finally, based on the measured delay results and utilizing Equation (4), the far-field radiation patterns for three transmitting beams and three receiving beams were calculated. In Equation (4) [23]

$$F\left(\theta \right)=\sum _{i=1}^N{\alpha }_i{e}^{j(i-1)({\displaystyle \frac{2\pi }{\lambda }}d sin \theta -{\varphi }_i)}$$

(4)

${\alpha }_i$ and ${\varphi }_i$ represent the amplitude and phase of the i-th antenna element, $\lambda$ is the lambda of radio frequency signal, and d is the spacing between the adjacent antenna elements, which is set to 0.0125 m i.e., a half wavelength of a 12 GHz microwave signal. The results are presented in Figure 13a,b, where solid lines represent far-field patterns calculated from measured delays, and dashed lines indicate patterns derived from reference delay values. Comparison with theoretical patterns reveals beam synthesis angle deviations of −1.40°, 0.37°, −3.82°, 1.54°, 1.37°, and −3.67° for the six beams. Furthermore, the far-field pattern for a fixed beam direction (−45°) was calculated at six different frequency points across the 8–18 GHz range, as shown in Figure 13c. These results demonstrate that the designed integrated broadband multi-beamformer chip effectively operates across the 8–18 GHz frequency range without significant beam squint, maintaining synthesis angle deviations within 3.50°. The observed angular deviations primarily stem from unequal delay variations across channels within the 8–18 GHz range, with the impact of identical delay deviations on steering angle accuracy becoming more pronounced at larger beam steering angles.

4. Discussion

In this work, a novel transceiver-shared photonic integrated broadband multi-beamforming network architecture based on an extended Blass matrix is proposed and demonstrated. While preserving the original broadband multi-beamforming capability, the proposed architecture achieves compatibility between transmission and reception modes through functional expansions in input/output channel count, control node configuration, and delay unit distribution. Based on this architecture, a 3 × 3 integrated broadband multi-beamformer supporting transceiver-shared operation has been designed and fabricated on a standard SOI platform, capable of broadband beam synthesis across six independent directions (three for transmission and three for reception) covering both the X and Ku bands. Both link-level simulations and experimental measurements confirm that the extended Blass-matrix-based integrated broadband multi-beamforming network architecture successfully supports transceiver-shared operation, offering a promising solution for compact multi-beam broadband radar and wireless communication systems.

However, the proposed OMBFN architecture still has certain limitations. Though the proposed OMBFN architecture retains the inherent scalability of the Blass matrix, allowing for theoretically unlimited expansion of both the number of input channels and synthesized beams based on application requirements, in practical implementations, two primary factors constrain such scalability. First, inter-channel IL imbalance arises due to the varying number of network nodes traversed by different input channels. As the number of channels increases, this IL variation becomes more pronounced, complicating power equalization across channels. This issue can be mitigated through the further optimization of node design and reduction in optical component transmission losses. Second, the expansion in channel count necessitates a proportionate increase in the wavelength coverage of the light sources, placing stringent demands on the operational bandwidth of on-chip components such as GCs, MZMs, and PDs. To address this challenge, hierarchical architectures such as binary-tree-based frameworks may be employed to optimize the network structure and improve scalability.

Furthermore, the proposed OMBFN architecture remains limited to fixed beam directions. To achieve independent beam steering across the entire angular range, it is necessary to incorporate tunable delay structures within each delay unit of the network. However, integrating current mainstream tunable delay lines—such as those based on optical switches with delay waveguides and microring resonators (MRRs)—into all delay units presents significant challenges in terms of network design and control complexity. For optical switches with delay waveguide structures, their relatively large footprint necessitates multi-stage cascading to support multi-angle scanning, which imposes considerable constraints on network size, layout design, and control scalability. On the other hand, MRRs, though compact, are inherently wavelength-sensitive and highly susceptible to environmental variations such as temperature fluctuations. These characteristics complicate the precise characterization and control of the system. Therefore, future work will prioritize the design of MBFNs with limited angular scanning capability, along with further optimization of the Blass matrix topology to accommodate more tunable delay units.