An RF board usually runs down the same line as a digital board, and that is where its performance is quietly lost. The stack-up is a mix of PTFE or hydrocarbon laminate and FR-4, the copper is heavy in places, and the ground pad under a power transistor needs a paste volume that a standard stencil does not give. The board still passes a DC check and still looks right under AOI, and the insertion loss has already moved by the time it reaches a chamber.
So RF assembly is a build problem before it is a measurement problem. What happens at print, placement, reflow and depanel decides how much of the stack-up you paid for actually arrives at the connector.
Where an RF build loses performance
Four places account for most of the loss we see when a customer sends back a board that worked as a hand built sample.
The first is paste under the ground paddle. A large thermal pad with partial coverage leaves voids, and a void under a source pad is both a local hot spot and a change in the ground path. The second is the connector attach. An SMA or a board edge launch depends on the ground vias around it and on a fillet that reaches the barrel, and a touch up after reflow rarely gets both. The third is the reflow profile on a mixed stack, because PTFE and FR-4 do not expand at the same rate and the laminate side wants a slower ramp. The fourth is depanel stress, where a routed tab or a V-cut close to a microstrip bends the trace beside it.
Where assembly decisions move the RF result
| Build step | What goes wrong | How it shows up |
|---|---|---|
| Paste print on a thermal pad | Coverage too low or paste slump before placement | Voiding under the die paddle, hot spot, drift in gain |
| Connector attach | Fillet short of the barrel, ground vias starved | Return loss at the launch, sensitivity to cable torque |
| Reflow on a mixed stack | Ramp set for FR-4 only | Laminate blister, pad craters, shift in impedance on the soft side |
| Depanel | Tab or V-cut too close to a trace | Micro crack in the trace, opens after thermal cycling |
| Coating over tuning structures | Coating thickness varies across the board | Frequency shift that appears only after coating |
What the build package has to state
A quote on an RF board is only as good as the package behind it. Five items decide the process, and when one is missing the build either stops or runs on an assumption that nobody wrote down.
| Item | What we need | Why it decides the build |
|---|---|---|
| Stack-up | Laminate and prepreg type per layer, copper weight, finished thickness | Sets the reflow profile, the drill schedule and the plasma step on PTFE |
| Impedance targets | Single ended or differential, tolerance, reference layer per net class | Decides the coupon, the TDR check and which traces are held to a tighter window |
| Paste requirement on thermal pads | Aperture pattern and expected coverage | Decides whether the pad voids under a large ground plane |
| Connector and shield attach | Iron, hot bar or reflow with a fixture | Decides the fillet that carries the ground and the shield bond |
| Test spec | DC, functional and RF, with limits and the fixture | Decides what we sign at outgoing and what goes to a lab |
How SUPERB PCBA runs an RF build
On the bare board side, the range covers Rogers RO4003C, RO4350B and RO4835, Taconic RF-35, TLY-5 and CER-10, and PTFE with plasma treatment before plating, in RF only or mixed with FR-4 in one stack. Boards run to 68 layers, 0.05 mm line and space, impedance held to plus or minus five percent with two percent available on critical nets, back drill stub of 0.25 mm or less, thickness from 0.2 to 12 mm and up to seven lamination cycles, across 500 MHz to 77 GHz.
On the assembly side the work runs on seven SMT lines with 01005 placement, 3D SPI after print, 3D AOI after reflow and X-ray read to IPC-7095 with voiding held under five percent. Reflow is a twelve zone nitrogen oven under 500 ppm oxygen with SAC305, followed by selective coating where the build calls for it. The flow is seventeen steps with ten QC stations, held to IPC-A-610 Class 2 and Class 3, under ISO 9001 and ISO 13485 for medical work. Every build is made to your released package, and the component list is bought to the approved maker and part number you name.
Stack-ups, drill rules and impedance coupons are described on the RF and microwave PCB capability page, and the module level builds we support are listed under RF and microwave PCBA.
What we test, and what we do not claim
This is the part buyers ask about last and should ask about first. Our line tests DC continuity and isolation, runs functional test on a fixture built to your spec, and inspects visually and by X-ray. Gain, return loss, isolation and phase are measured in your own lab or in a partner RF laboratory, not on our bench, so we do not sign an RF number we did not take.
What we can do is keep the build inside the process window that produced your measured sample: the same paste pattern on the ground pad, the same profile, the same depanel method, and a record per lot that says so. When a lot is compared against a sample in a lab, the process is the variable we can hold.
RF PCB assembly questions
Can you assemble a board that mixes PTFE and FR-4?
Yes, and it is one of the common RF builds we see. The constraint is the profile and the press cycle, not the placement. PTFE needs plasma treatment before plating and a slower ramp, and the FR-4 side has to stay inside its own window at the same time. Send the stack-up with the build and the profile is set from the laminate data, not from a default.
Do you measure RF performance before shipping?
No, not on our own equipment. We test DC, run functional test on your fixture and inspect by X-ray. RF parameters belong to your lab or a partner laboratory, and the correlation between the two is agreed before the first lot. The reflow side of the build is covered under nitrogen reflow soldering.
What if the approved part is short and an alternative is offered?
The alternative is proposed with the datasheet comparison and never fitted without your written approval. On an RF board the substitution rule is stricter than on a digital one, because a change in package parasitics moves the match. How the sourcing side runs is described under RF component sourcing.
Next step
Send the Gerber set, the stack-up, the impedance table and the test spec, or send the BOM with the approved maker and part number. The reply returns the process route, the paste and stencil call for the thermal pads, the profile window for the laminate, and a price with the tooling listed as a separate line.