An 8-layer RF board built entirely on RO4350B will cost enough to make volume production unrealistic. Built entirely on FR-4, the 50Ω RF traces on the surface layer will not meet spec.

Hybrid stackups are the way out. But hybrid construction is far more difficult to process than a single-material stackup—CTE mismatch causes delamination, multilayer lamination causes layer misregistration, and alternating thick and thin dielectrics make impedance hard to hold.

We recently manufactured an 8-layer RO4350B + MT40 hybrid RF board that is a typical example of this design approach. This article is not just about how the board was built. It is about a question that comes up constantly at the material selection stage: when is a hybrid stackup worth it, and when is it not.

The Board Specifications

8-layer copper construction. The top and bottom layers use 0.168mm RO4350B high-frequency material. The inner layers use 0.381mm MT40, a standard high-Tg FR-4 material, with 0.1245mm MT40 prepreg between layers. Finished outer copper is 1oz, inner copper is 0.5oz, and the laminated thickness is 1.636mm. Green solder mask, white silkscreen, palladium gold finish, 179mm × 121mm.

The board includes L1-L2 and L7-L8 blind vias. The top layer carries 6mil single-ended traces with a 50Ω target impedance.

Why RO4350B on the Surface, MT40 in the Middle

The stackup logic is clear: high-frequency performance on the surface, structural support and cost control in the middle.

The surface layers use RO4350B because that is where the RF traces are. RO4350B has a dielectric constant of about 3.48 and a dissipation factor of about 0.0037. Low Df means low dielectric loss, so microwave signals attenuate less as they travel along the surface. The 6mil single-ended 50Ω traces on the top layer are built directly on RO4350B, which gives tighter impedance control. The trade-off: RO4350B costs far more than standard FR-4.

The inner layers use MT40 because power and ground planes do not need high-frequency performance. MT40 is a standard high-Tg FR-4 material, far cheaper than RO4350B. Its job is to provide structural support and thermal stability for the multilayer stack. If the inner layers also used RO4350B, material cost would rise sharply—and the power and ground planes do not need low Df. The trade-off: MT40's electrical performance is not as good as RO4350B, but used on inner layers it does not affect RF performance.

The 0.1245mm MT40 prepreg controls interlayer thickness. Prepreg thickness directly affects the dielectric thickness between layers, which in turn affects impedance. The 0.1245mm specification was chosen to match the RO4350B and MT40 core thicknesses and hold the total laminated thickness at 1.636mm.

The core logic of this stackup: high-frequency performance only where it is needed. Resource utilization is higher, and cost is more controllable.

The Process Challenges: Three Real Problems

Hybrid dielectric stackups are far harder to process than single-material stackups. This board had three main challenges.

Challenge 1: CTE Mismatch Causes Delamination and Warpage

RO4350B is a ceramic-filled material, and its CTE differs from MT40, which is FR-4 based. During lamination, the two materials expand at different rates, generating stress at the interface that can cause delamination or board warpage.

The fix is a staged temperature ramp with slow cooling. The ramp is divided into multiple stages, each held long enough for the material to soften and flow fully. The cooling stage controls the rate of temperature drop to avoid internal stress from rapid cooling. This lets the two materials gradually accommodate each other's thermal expansion behavior during lamination, reducing interface stress.

Challenge 2: Layer Misregistration After Multilayer Lamination

A laminated thickness of 1.636mm is moderately thick for an 8-layer board. During multilayer lamination, registration accuracy between layers can drift, causing misregistration.

The fix includes additional alignment targets and optimized pressure uniformity on the lamination plates. Alignment targets provide a more accurate registration reference, while pressure uniformity ensures even force across the panel and reduces local shifting.

Challenge 3: Alternating Thick and Thin Dielectrics Make Impedance Sensitive to Thickness Variation

The stackup alternates thick and thin: 0.168mm RO4350B, 0.381mm MT40, 0.1245mm prepreg. The thickness differences between dielectric layers are significant. Impedance is highly sensitive to dielectric thickness, and variation translates directly into impedance shift.

The fix is tight incoming material thickness tolerance. RO4350B and MT40 core thickness, as well as prepreg thickness, must be strictly inspected at incoming inspection to ensure batch-to-batch consistency. Post-lamination dielectric thickness also needs to be verified by cross-section to confirm impedance stays within design range.

The Cost Advantage Over an All-RO4350B Stackup

The core value of a hybrid stackup is reducing cost while preserving high-frequency performance. This board's cost advantage shows up in three ways.

First, material cost drops significantly. Replacing RO4350B with MT40 on the inner layers cuts material cost sharply. RO4350B is priced far above standard FR-4, and an 8-layer board built entirely on RO4350B would be very expensive. A hybrid stackup keeps RO4350B only on the surface layers, making it suitable for volume production of medium-to-large RF modules.

Second, MT40's heat resistance compensates for the thermal reliability weakness of pure ceramic materials. MT40 is a high-Tg FR-4 material with good heat resistance and can survive multiple SMT reflow cycles. RO4350B offers excellent electrical performance, but its thermal reliability under repeated reflow is not as good as FR-4. Using MT40 in the inner layers improves the board's overall thermal reliability.

Third, resource utilization is higher. High-frequency performance is only needed on the signal surface layers. Power and ground layers do not need low-Df material. Putting RO4350B where it counts and MT40 in the inner layers is a more rational allocation of resources.

When You Should Not Use a Hybrid Stackup

This is the part most stackup write-ups skip, and it is the most useful part for engineers at the material selection stage.

If the entire board needs high-frequency performance. If inner layers also carry RF traces, or if power layers have strict impedance requirements, the hybrid advantage disappears. In that case, an all-RO4350B or other high-frequency stackup is more suitable.

If the board is small. When material cost is a small share of total cost, the savings from hybrid construction are limited while process complexity increases. A single-material stackup may be simpler and more reliable for small boards.

If multiple reflow cycles are not required. MT40's thermal reliability advantage only matters in multi-reflow assembly. If the assembly process is simple, that advantage goes unused, and the hybrid complexity is unnecessary.

If cost is not the first constraint. If the project prioritizes performance far above cost, an all-high-frequency stackup avoids the process risk that hybrid construction introduces. The value of hybrid is balance—if you do not need balance, you do not need hybrid.

The Bottom Line

An 8-layer RO4350B + MT40 hybrid stackup is not a general-purpose solution. It is a targeted solution: it earns its process complexity when high-frequency signals are confined to the surface layers, the board is large, cost is sensitive, and multiple reflow cycles are required.

The core logic is: high-frequency performance on the surface, structural support and cost control in the middle. The process challenges—CTE mismatch, layer misregistration, impedance variation—all have countermeasures, but those countermeasures add process complexity.

If your design does not meet those conditions, a single-material stackup may be the more sensible choice.

If you are working on an RF module stackup and need to balance performance against cost, we would be glad to review your requirements. Stackup decisions should start from your frequency, impedance targets, and cost goals—not from which material sounds best.

Have you built hybrid stackups? What process issues did you run into?