A 67–110 GHz Multi-Carrier Communication Front End Based on Broadband Diplexer and Filtering Waveguide-to-Microstrip Transitions Authors: Yujie Zhi, Yulong Zhong, Bo Zhang, Yibo Fan, Zuqiang Ou, Jincai Qiao Published: 2026 ABSTRACT This paper presents a 67–110 GHz millimeter-wave RF front end for 6G high-speed wireless communications. A broadband diplexer based on a staggered branch-waveguide structure is first proposed to achieve low insertion loss and high isolation simultaneously. The fabricated diplexer exhibits better than 12 dB return loss, 48.6% fractional bandwidth, and less than 1.5 dB insertion loss. A filtering waveguide-to-microstrip transition is then developed to suppress harmonics and spurious signals while maintaining efficient mode conversion. Back-to-back measurements demonstrate over 50 dB out-of-band suppression at 99.5 GHz with a center frequency of 89 GHz. Finally, the proposed components are integrated with broadband amplifiers, filters, and a local oscillator to realize a 67–110 GHz RF front end with 4–26 GHz IF output. Experimental results show harmonic suppression better than 60 dBc, demonstrating excellent spectral purity and validating the proposed architecture as a compact, broadband, and highly integrated solution for future millimeter-wave and terahertz communication systems. 1. Introduction With the rapid development of B5G/6G wireless communications, satellite Internet, intelligent transportation, integrated sensing and communications, and terahertz technologies, the 67–110 GHz W-band has emerged as a promising spectrum for multi-channel high-speed wireless communications owing to its abundant spectrum resources and wide transmission bandwidth [1,2,3]. The widespread adoption of multiple-input multiple-output (MIMO), beamforming, carrier aggregation, and multi-stream transmission technologies has imposed increasingly stringent requirements on RF front ends in terms of channel capacity, operating bandwidth, isolation, and integration density. To enable efficient routing and processing of RF, local oscillator (LO), and intermediate-frequency (IF) signals, broadband diplexers have become indispensable passive components in multi-channel millimeter-wave transceivers. Their performance directly affects channel isolation, signal integrity, spectral efficiency, and overall system capacity [4,5,6]. As operating frequencies continue to extend toward the millimeter-wave and terahertz regimes, multi-channel transceiver systems face increasing challenges, including higher transmission loss, stronger inter-channel crosstalk, enhanced parasitic modes, and the difficulty of achieving high-density integration [7,8]. Among various passive components, diplexers play a critical role in signal separation, combination, and isolation, thereby determining power transmission efficiency, channel isolation, beamforming accuracy, bit-error rate (BER), and overall communication performance [9,10]. Diplexers based on branch-waveguide directional couplers have become one of the most widely adopted solutions because of their excellent impedance matching, high power-handling capability, low transmission loss, and robust mechanical structure. Consequently, they have attracted extensive attention from both academia and industry [11,12,13]. However, due to the inherent frequency characteristics and coupling mechanism of conventional branch-waveguide structures, existing designs still struggle to simultaneously achieve low insertion loss, high return loss, high port isolation, and compact integration over an ultra-wide bandwidth, which remains a major bottleneck for next-generation broadband millimeter-wave communication systems. In millimeter-wave transceivers, active devices such as mixers generally require waveguide-to-microstrip transitions to realize efficient energy coupling and impedance matching between waveguides and planar circuits [14,15]. Nevertheless, conventional transitions often excite higher-order modes and parasitic resonances during mode conversion, generating harmonics, spurious responses, and image signals [16]. These undesired spectral components degrade channel isolation, increase adjacent-channel interference, and deteriorate signal quality, thereby reducing system capacity, communication reliability, and BER performance. Existing waveguide-to-microstrip transitions mainly focus on broadband impedance matching and low-loss transmission, while insufficient attention has been paid to out-of-band suppression and spectral purity enhancement [17]. As a result, they are unable to satisfy the stringent requirements of multi-channel millimeter-wave communication systems for high isolation, low crosstalk, and high spectral efficiency [18,19]. Therefore, integrating filtering functionality into the transition while simultaneously achieving efficient mode conversion, broadband impedance matching, and out-of-band suppression has become a critical issue in the design of high-performance millimeter-wave communication front ends. To address these challenges, this paper proposes a set of key front-end technologies for a broadband 67–110 GHz multi-channel millimeter-wave communication system. A broadband diplexer based on a staggered branch-waveguide structure is first developed to simultaneously achieve low insertion loss, low return loss, and high port isolation over a wide operating bandwidth. In addition, a filtering waveguide-to-microstrip transition is proposed to realize efficient mode conversion while effectively suppressing out-of-band spurious signals, thereby improving spectral purity and channel isolation. Based on these key passive components, the proposed front end is designed and experimentally validated, providing an effective solution for compact, highly integrated, low-crosstalk, and spectrally efficient millimeter-wave communication front ends. 2. Front-End Architecture for a 67–110 GHz Multi-Carrier Communication System To satisfy the requirements of a 67–110 GHz broadband communication system, while considering the design complexity of key components such as mixers, LO, and filters, as well as spurious suppression and implementation feasibility, a multi-carrier frequency-conversion architecture is adopted. The operating band is divided into two sub-bands, 67–88 GHz and 88–110 GHz, each driven by an independent low-side LO channel. This architecture not only reduces the design complexity associated with a single LO covering the entire ultra-wide frequency range but also effectively suppresses mixing spurious signals, improves image rejection, and downconverts both carriers to a common IF range of 4–26 GHz, thereby simplifying subsequent IF processing and signal demodulation. The overall architecture of the proposed front end is illustrated in Figure 1. Two independent LO paths are employed, and the appropriate LO signal is selected through an RF switch according to the operating sub-band. After frequency multiplication and power amplification, the LO signal drives the mixer to perform frequency conversion with the received RF signal. The resulting 4–26 GHz IF signal is then passed through an IF filter to suppress spurious components and higher-order harmonics, followed by an IF amplifier to compensate for conversion loss and provide sufficient gain. Finally, the amplified IF signal is delivered to the subsequent receiver or digital signal-processing unit for signal demodulation and analysis. The proposed front-end architecture simultaneously achieves wideband frequency coverage, low spurious responses, high dynamic range, and modular implementation, providing a practical and scalable solution for broadband 67–110 GHz multi-carrier millimeter-wave communication systems. 3. Broadband Diplexer Based on a Staggered Branch-Waveguide Structure As the input component of the diplexer, the broadband coupler must provide an ultra-wide operating bandwidth while maintaining excellent amplitude and phase flatness, both of which play a decisive role in the transmission performance of the overall diplexer. Therefore, a coupler featuring a wide bandwidth together with low amplitude imbalance and low phase deviation is highly desirable. Conventional branch-waveguide couplers, however, are unable to achieve more than 30% fractional bandwidth [20]. To enhance the coupling flatness, directivity, and isolation of traditional branch-waveguide couplers over the 67–110 GHz ultra-wide frequency range, a staggered branch-waveguide coupling structure is proposed in this work, shown in Figure 2. By introducing a center offset, denoted as m, into the conventional branch-waveguide configuration, the originally symmetric structure is transformed into an asymmetric one. This intentional asymmetry excites the higher-order TE201 mode and generates additional transmission zeros near the passband edges, thereby extending the coupling bandwidth while maintaining excellent broadband transmission characteristics. For a conventional rectangular branch waveguide with width a and height h0, the equivalent characteristic impedance can be expressed as(1)Z0=h0a⋅ZTEM1−λ02a2 where ZTEM=120π, and λ0 is the free-space wavelength at the center frequency. After introducing an offset distance m, the effective width of the staggered branch waveguide can be approximated as a+m. Accordingly, its equivalent characteristic impedance becomes(2)Z0′=ha+m⋅ZTEM1−λ02(a+m)2 where h denotes the compensated branch height. To maintain the same equivalent coupling strength as that of the conventional branch waveguide during the initial design stage, the characteristic impedances are set equal.(3)Z0′=Z0 Substituting (1) and (2) into (3) yields the compensation relationship between the staggered branch height h and the conventional branch height h0:(4)h=h0⋅(a+m)2−(λ02)2a2−λ022 Equation (4) indicates that the compensated height h increases with the offset distance m. Physically, the offset m enhances the excitation of the higher-order TE201 mode and introduces additional transmission zeros near the passband edges, whereas the increased branch height compensates for the reduced fundamental-mode coupling caused by the offset. Consequently, the proposed structure achieves a trade-off between bandwidth enhancement and coupling strength. Because the offset is constrained by the physical dimensions of the branch waveguide, its value must satisfy(5)00,(6)h=h0⋅(a+m)2−(λ02)2a2−λ022>h0 Therefore, the offset m and the conventional branch height h0 are generally of the same order of magnitude. Since h0 can be readily obtained from the classical Reed theory or its generalized formulations, the initial offset is simply chosen as m=h0. This significantly simplifies the initial design procedure. The corresponding compensated height h can then be calculated directly from (4). The branch length t is mainly determined by the guided wavelength at the center frequency. For the staggered branch waveguide, the effective guided wavelength is(7)λg,MW=λ01−λ02(a+m)2 Accordingly, the length of each staggered branch is selected as one-quarter of the guided wavelength.(8)t=λ041−λ02(a+m)2 For a multi-branch coupler, the spacing between adjacent branches is typically chosen as one-quarter of the guided wavelength in the main waveguide.(9)w=λ041−λ02a2 When the compensated branch height becomes relatively large and satisfies 2h>w, the maximum allowable offset is additionally constrained by the spacing between adjacent branches. Under this condition, (5) should be modified as(10)0