{"id":15920,"date":"2026-09-30T06:36:09","date_gmt":"2026-09-30T06:36:09","guid":{"rendered":"https:\/\/ms-telerf.com\/?p=15920"},"modified":"2026-09-30T06:37:39","modified_gmt":"2026-09-30T06:37:39","slug":"wilkinson-power-divider-miniaturization","status":"publish","type":"post","link":"https:\/\/ms-telerf.com\/fr\/wilkinson-power-divider-miniaturization\/","title":{"rendered":"Trois approches pratiques pour la miniaturisation des diviseurs de puissance de Wilkinson"},"content":{"rendered":"
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\n

In RF systems such as phased arrays, 5G base stations, and satellite communications, the Wilkinson diviseur de puissance<\/a> is one of the most widely used power distribution components. However, the core structure of a conventional Wilkinson power divider is a quarter-wavelength transmission line<\/strong>. At L-band (1\u20132 GHz), a quarter-wavelength microstrip line is physically about 30\u201350 mm long, and the entire divider may occupy tens of square centimeters. For large-scale phased arrays, the feed network area directly determines whether the system can be made small and highly integrated.<\/p>\n\n\n\n

Miniaturization has therefore become one of the core engineering challenges in Wilkinson power divider design. Three mature paths have emerged: lumped-element replacement, slow-wave structure loading, and process and architecture optimization<\/strong>. These paths are not mutually exclusive; in practice, products often combine them. Below is a breakdown of each.<\/p>\n\n\n\n

Path 1: Lumped-Element Replacement \u2014 Using LC Circuits Instead of \u03bb\/4 Transmission Lines<\/h2>\n\n\n\n

This is the most direct and classic miniaturization approach. The logic is simple: since the size of a \u03bb\/4 transmission line is inversely proportional to frequency, we can avoid the transmission line and use an equivalent circuit made of inductors (L) and capacitors (C) to achieve the same phase shift and impedance transformation.<\/p>\n\n\n\n

In implementation, a quarter-wavelength transmission line with characteristic impedance Z\u2080 can be equivalent to a pi-type or T-type LC network<\/strong>. The most common pi-type equivalent circuit consists of two shunt capacitors and one series inductor. Its parameters are derived by equating the ABCD matrix of the transmission line with that of the LC network.<\/p>\n\n\n\n

The advantage is large size reduction<\/strong>. Transmission lines are distributed elements, with physical length constrained by wavelength; LC elements are lumped elements, and their size depends on process capability rather than wavelength. In a broadband Wilkinson power divider based on thin-film integrated passive device (IPD) technology, replacing \u03bb\/4 transmission lines with pi-type LC circuits reduced the entire divider to only 1.5 mm \u00d7 3 mm \u00d7 0.65 mm<\/strong>\u2014more than 90% smaller<\/strong> than a conventional microstrip structure. The device has a center frequency of 0.7 GHz, an operating band starting from DC, a relative bandwidth as high as 200%<\/strong>, and port isolation greater than 15 dB<\/strong>.<\/p>\n\n\n\n

The lumped-element path also faces engineering challenges. First, Q factor<\/strong>: on-chip spiral inductors and MIM capacitors have limited Q, which introduces extra insertion loss. GaAs substrates have smaller parasitic capacitance and inductance effects, helping improve the Q of spiral inductors and MIM capacitors, making them a preferred substrate for IPD power dividers. Second, frequency limit<\/strong>: parasitic parameters of lumped elements become non-negligible at millimeter-wave frequencies, so this path is mainly suitable below 6 GHz. In L-band and S-band, lumped-element solutions are almost the first choice for miniaturization. Research has reported a filtering power divider based on lumped elements on a GaAs substrate with a size of only 1.3 mm \u00d7 1.4 mm<\/strong> (about 0.037\u03bb\u2080 \u00d7 0.04\u03bb\u2080).<\/p>\n\n\n\n

Path 2: Slow-Wave Structure Loading \u2014 Making Electromagnetic Waves \u201cTravel Slower\u201d<\/h2>\n\n\n\n

Slow-wave structures are the fastest-growing miniaturization technology in recent years. The physical principle is: by periodically loading capacitive or inductive structures onto a transmission line, the phase velocity<\/strong> of electromagnetic waves on the line is artificially reduced, so a greater phase is accumulated over the same physical length. Equivalently, to achieve a 90\u00b0 phase shift at a given frequency, the required physical length is shortened.<\/p>\n\n\n\n

The most common implementation is inserting metallized via holes<\/strong> into the substrate below a microstrip line. The vias form distributed capacitance with the microstrip line, producing a slow-wave effect. Another method uses periodic LC cells loaded onto the transmission line, creating a slow-wave transmission line. In LTCC processes, such slow-wave transmission lines can shorten \u03bb\/4 impedance transformers by 55%<\/strong>. Combined with high-permittivity multilayer substrates, this enables miniaturized devices. In coplanar waveguide structures, replacing conventional \u03bb\/4 lines with cross-shaped slow-wave transmission lines made of interdigital capacitors and meandered inductors can reduce power divider size to 35.7%<\/strong> of the conventional design.<\/p>\n\n\n\n

In 2025, a V-shaped Wilkinson power divider based on slow-wave microstrip transmission lines operating at 2.45 GHz used a dual-layer substrate (RO4003 + RO4403). By inserting metallized vias into the lower substrate to create a slow-wave effect, it achieved 40% size reduction<\/strong>, a surface area of 22 \u00d7 13 mm\u00b2<\/strong>, and amplitude imbalance of only 0.1 dB<\/strong>.<\/p>\n\n\n\n

Compared with lumped-element solutions, slow-wave structures have an important advantage: they remain distributed structures, so the frequency limit is higher<\/strong>. Slow-wave transmission lines can still maintain good performance at millimeter-wave frequencies, whereas lumped elements suffer significant parasitic degradation. Slow-wave structures can also be combined with other miniaturization techniques such as defected ground structures (DGS)<\/strong> to further compress size. Research has applied DGS spiral inductors to a Wilkinson power divider on a high-resistivity silicon substrate, achieving 56.6% size reduction<\/strong> compared with an ordinary meandered structure while maintaining a higher Q. A second-order power divider at 1.6 GHz occupies only 1.85 mm \u00d7 2.208 mm<\/strong>, avec 1.322 dB insertion loss<\/strong> et 61.18% bandwidth<\/strong>.<\/p>\n\n\n\n

Path 3: Process and Architecture Optimization \u2014 System-Level Compression from Materials to Topology<\/h2>\n\n\n\n

The first two paths focus on circuit-level miniaturization; the third path compresses size at the manufacturing process and circuit architecture<\/strong> levels.<\/p>\n\n\n\n

Process Level: High-Permittivity Substrates and Advanced Packaging<\/h3>\n\n\n\n

The higher the dielectric constant of the substrate, the narrower the microstrip line width for the same characteristic impedance and the shorter the wavelength, so the device area naturally decreases. Multilayer organic substrates such as LCP (liquid crystal polymer)<\/strong> are a popular choice in recent years. LCP has a dielectric constant of about 2.9\u20133.2, very low dielectric loss, and can be stacked in multiple layers. In a Wilkinson power divider based on multilayer LCP substrate, the size at 1.1 GHz center frequency is only 4.4 mm \u00d7 5.0 mm \u00d7 0.35 mm<\/strong>\u201497.5% smaller<\/strong> than a conventional microstrip design. This is among the largest size reductions reported for planar PCB processes.<\/p>\n\n\n\n

Architecture Level: Multi-Order Topology and Folded Structures<\/h3>\n\n\n\n

A conventional Wilkinson power divider is a single-section structure with limited bandwidth. Multi-section topologies<\/strong> can not only broaden bandwidth but also reduce total length through proper design of each section\u2019s characteristic impedance. A further approach is folded structures<\/strong>\u2014arranging multi-section transmission lines in a folded layout and using multilayer substrate space.<\/p>\n\n\n\n

In 2025, a team from Beijing Institute of Technology proposed an innovative multi-order magnetic coupling self-cancellation technique<\/strong> at the RFIC conference. The core idea: in folded multi-section structures, magnetic coupling between adjacent transmission lines inevitably exists, and conventional designs try to eliminate it. But the team used this coupling instead, making it interact with the circuit\u2019s distributed parameters to achieve self-cancellation. The result: the power divider area was halved<\/strong> while the relative isolation bandwidth was extended by nearly 140%<\/strong>. Based on a 65-nm CMOS process, 2-way and 8-way power dividing networks achieved core areas of only 0.069\/0.059 mm\u00b2<\/strong>, the smallest normalized area known internationally for such power distribution networks, with relative isolation bandwidths of 86.3%\/82.4%<\/strong> and crosstalk isolation up to 63 dB<\/strong>.<\/p>\n\n\n\n

The lesson: miniaturization is not simply making a device smaller, but redesigning the topology so that size and performance are optimized simultaneously.<\/strong><\/p>\n\n\n\n

Comparison of the Three Paths<\/h2>\n\n\n\n
Path<\/th>Typical Size Reduction<\/th>Suitable Frequency<\/th>Avantages<\/th>Trade-offs<\/th><\/tr><\/thead>
Lumped-element replacement<\/td>>90% (IPD)<\/td>DC\u20136 GHz<\/td>Smallest size, easy on-chip integration<\/td>Limited Q leads to higher insertion loss; parasitics at high frequency<\/td><\/tr>
Slow-wave loading<\/td>40%\u201365%<\/td>1\u201340 GHz<\/td>Retains distributed advantages; usable at mmWave<\/td>High design complexity; accurate modeling required<\/td><\/tr>
Process & architecture optimization<\/td>50%\u201397.5%<\/td>Full band<\/td>Size and performance can be optimized together<\/td>Depends on advanced processes; higher cost<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

In practice, the three paths are often combined. For example, in LTCC processes, slow-wave transmission lines and multilayer folded structures can be used together; in IPD processes, lumped elements can replace \u03bb\/4 lines, combined with multi-order topologies to broaden bandwidth.<\/p>\n\n\n\n

The selection logic can be simplified as follows:<\/p>\n\n\n\n