Bandwidth Optimization of 5G Cavity Bandpass Filters

1. Introduction

1.1 Background

The global RF filter market is valued at USD 18.8–18.9 B in 2025 and is projected to reach USD 21.8–21.9 B in 2026, growing at a CAGR of about 16.2% and potentially exceeding USD 98 B by 2036 (Fact MR / Future Market Insights, 2026). The principal growth driver is large-scale 5G macro-cell infrastructure deployment, particularly in the Asia-Pacific region (Taiwan, China, Japan and Korea). Thanks to their high Q factor, low insertion loss and high power-handling capability, cavity bandpass filters remain the core front-end element in 5G Sub-6 GHz base stations.

1.2 Motivation and Objectives

While 5G NR supports up to 100 MHz per carrier in Sub-6 GHz and 400 MHz in FR2, the RF link budget in macro-cell base stations typically requires IL ≤ 0.5 dB on the front-end filter. Classical coupled-resonator theory, however, shows that bandwidth, insertion loss and skirt steepness form an inherent trade-off — widening the bandwidth softens the skirt, while pursuing low IL necessarily narrows the bandwidth. The objectives of this study are therefore: (1) to provide a structured review of the key design variables governing bandwidth in 5G cavity bandpass filters; (2) to establish a quantitative trade-off model among BW, IL and selectivity; (3) to propose a production-oriented optimization strategy for the n78 band (3.5 GHz, BW = 100 MHz); and (4) to map the strategy onto TEMWELL's current manufacturing technology stack so that research outcomes can be transferred directly to production.

2. Theoretical Foundations

2.1 Cavity Resonators and Coupling Structures

A cavity bandpass filter is implemented as a chain of metallic cavity resonators, each supporting a fundamental TE/TM mode at the target center frequency. Adjacent cavities are coupled inductively or capacitively through irises, tuning screws or probes; the coupling coefficient k_i,i+1 defines the passband bandwidth. The input/output ports connect to 50 Ω or 75 Ω transmission lines via tap, loop or probe excitation. A representative six-pole topology is shown in Fig. 1.

2.2 Bandwidth Definitions and Specification Requirements

The following bandwidth definitions are commonly used in 5G cavity BPF design:

• 3 dB Bandwidth (BW_3dB): the frequency range over which the transmitted power drops to 50%. The most common design benchmark.
• 1 dB Bandwidth: preferred when low distortion and low EVM are required, e.g. 5G MIMO front-ends.
• Shape Factor = BW_1dB / BW_60dB: reflects the selectivity; values closer to 1 approach an ideal brick-wall response.

Table 1 summarizes the typical target specification for a Sub-6 GHz macro-cell cavity BPF.

3. Key Design Parameters for Bandwidth Optimization

3.1 Pole Number (N)

Each additional pole contributes about 20 dB/decade to the out-of-band slope but also raises the in-band IL. For Sub-6 GHz macro-cells, N = 4–6 is the industry-standard balance between selectivity and IL. Fig. 3 shows a Chebyshev approximation (0.05 dB ripple) for several N values.

3.2 Coupling Coefficient (k)

From Cohn's formula, k_i,i+1 = FBW / √(g_i · g_i+1). k directly sets the passband shape and bandwidth: larger k broadens the passband at the cost of increased group-delay ripple, and an excessive k distorts the passband. In practice, iris dimensions are extracted by eigenmode simulation in HFSS / CST.

3.3 Unloaded Cavity Q (Q_u)

Q_u sets the lower bound on in-band IL: IL_min ≈ 4.343 · (f₀ / BW_3dB) · Σ(g_i / Q_u) (dB). For fixed Q_u, IL is inversely proportional to BW. Typical aluminum cavities reach Q_u ≈ 5,000–10,000; silver-plated copper can exceed 30,000. Fig. 4 quantifies this trade-off.

3.4 Transmission Zeros (TZs)

Transmission zeros improve rejection without increasing pole count and are the most effective lever to break the BW–selectivity trade-off. Implementations include: (1) non-adjacent cross-coupling, (2) cul-de-sac topology, (3) multi-mode resonators, and (4) rectangular slots / vias in SIW structures (Nature Sci. Reports, Feb. 2026). Two symmetric TZs typically improve the shape factor by 25–40%.

3.5 S-Parameter Simulation and Passband Performance

A 6-pole Chebyshev response with symmetric cross-coupled TZs is simulated for f₀ = 3.5 GHz, BW = 100 MHz (Fig. 2). At Q_u = 8,000 the design yields IL ≈ 0.4–0.5 dB, S₁₁ ≤ −18 dB and rejection ≥ 55 dB at f₀ ± 100 MHz — meeting 5G NR n78 macro-cell requirements.

4. Recent Technological Advances (2024 – 2026)

Table 2 summarises the most relevant advances of the past 24 months in bandwidth optimisation of 5G BPFs.

4.1 AI-Assisted Multi-Objective Optimization

Conventional EM optimisation of ≥6-pole cavity BPFs requires hours of HFSS simulation per iteration. Since 2025–2026, Bayesian optimisation and deep-learning surrogates have been adopted to explore the BW / IL / S₁₁ design space efficiently, reducing design-convergence time by 50–70%. This direction maps to our planned follow-up Article E-50.

5. Design Strategy Matrix

Based on the analysis above, Table 3 presents the design-strategy matrix.

6. Conclusions & TEMWELL R&D Capabilities

6.1 Technical Conclusions

1. Bandwidth, insertion loss and skirt steepness constitute a fundamental tri-lemma; simultaneous optimization is not possible.
2. N = 4–6 remains the mainstream pole count for Sub-6 GHz macro-cells. Production-grade Q_u ≈ 8,000 meets BW = 100 MHz, IL ≤ 0.5 dB.
3. TZs are the most cost-effective lever, improving the shape factor by 25–40% without adding poles.
4. SIW and 3D-printed substrates are strategic for FR2 and miniaturization; metallic cavities remain mainstream for Sub-6.
5. AI-assisted optimization is the next differentiation lever — lowering NRE cost and shortening lead time.

6.2 Directly Deployable TEMWELL R&D Capabilities

Leveraging 25+ years of RF filter manufacturing expertise, TEMWELL can translate the above findings directly into product upgrades through five capability pillars:

6.3 12-Month R&D Roadmap

1. Deploy an in-house HFSS + Bayesian-optimization pipeline to shrink 6-pole cavity BPF iteration time from 3 days to 1 day.
2. Develop a standard n78 / n79 dual-band low-PIM cavity diplexer aligned with TW & SEA operator spectrum plans.
3. Pilot silver-plated Q_u > 15,000 premium cavities for high-power and military applications.
4. Collaborate with academia to evaluate SIW process introduction for FR2 (24.25–29.5 GHz) miniaturized lines.
5. Extend Temstron brand amplifiers / isolators / couplers to deliver turn-key RF sub-system solutions.

TEMWELL, headquartered in Taipei with 25+ years of RF-filter expertise and partners in 20+ countries, offers a full portfolio of cavity, helical, SMD ceramic and SAW filters, plus diplexers / duplexers / multiplexers, and — via the Temstron brand — RF amplifiers and passive components. The optimization strategy proposed here can be mapped directly onto TEMWELL's existing capabilities and delivered within a 7–10-day customization window.