High Frequency PCB: From Prototype to Production

A high frequency board rarely goes straight from a sketch to a production line. It is designed once to prove the concept, then redesigned to be manufacturable, then optimised until the loss and the impedance tolerance meet the specification, and only then released. Each of those stages has a different objective, and treating them as one continuous process is how projects end up with a prototype that cannot be built in volume or a production board that does not perform like the prototype.

Stage One: Proving the Electrical Concept

The first board exists to answer one question: does the circuit work at the intended frequency? The answer depends on the transmission lines, the transitions and the components, so the prototype is built on the material that will be used in production rather than on whatever is available, because substituting an FR-4 board at this stage produces a result that says nothing about the final design.

The prototype is usually single sided or two layers with generous geometry. Impedance is verified with a coupon or with a time domain reflectometer measurement, and the transitions are measured rather than simulated, because a connector launch or a via structure is where the simulation and the reality diverge most. Our high frequency laminate article describes how the material for that stage is chosen.

Stage Two: Making It Manufacturable

Once the electrical behaviour is confirmed, the geometry is tightened. Minimum feature sizes are brought to the fabricator’s standard capability rather than left at the prototype’s comfortable values, panel utilisation is improved, and the stack-up is adjusted to use standard material thicknesses rather than special orders. This is also the stage where the assembly process is decided, because the placement accuracy and the stencil design depend on the pitch of the smallest components.

Design for manufacture has an electrical side as well. Vias are added where they are needed for routing but not where they would disturb a controlled impedance line, test points are placed so that a flying probe or a fixture can reach every net, and fiducials are added for the placement machine. Each of those additions changes the geometry, which is why the impedance calculation has to be repeated after them rather than before. Our design release checklist places those checks in the review sequence.

Stage Three: Optimising Loss and Tolerance

With a manufacturable board, the remaining work is performance. Conductor loss is reduced by widening traces where the impedance calculation permits and by choosing a lower roughness copper foil. Dielectric loss is reduced only by changing the material, and that decision has to be made on the basis of a measured link budget rather than a hunch. Via stubs are removed by back drilling where the data rate justifies it.

Impedance tolerance is the other lever. As tolerances tighten, the fabricator has to control the dielectric thickness more closely, which usually means a higher grade of material and a more expensive lamination cycle. The specification should therefore say what the system actually needs rather than what would be nice, because a tolerance quoted without justification is paid for on every board. Our component tolerance and reliability notes describe how the tolerance stack is assessed.

Stage Four: Qualification and Volume

Production introduces its own problems. A stack-up that worked on one panel has to work on every panel, and the fabricator’s process capability is what determines whether it does. Impedance coupons are measured on each lot, microsections confirm the plating, and the electrical performance is verified on production samples rather than only on the first article.

Assembly yields follow the same logic. The first production run should be watched closely, because the interaction between the stencil aperture, the paste volume and the pad geometry is what determines the joint quality, and it is easier to adjust before the process is fixed. Where a design uses a fine pitch package or a ball grid array, the x-ray inspection data from the first run is the most useful information the project will get.

Managing the Transitions

The riskiest moment in the whole sequence is the transition between stages, because it is where a change is made that has not been verified. A material substitution, a stack-up adjustment, a change of fabricator or a change of assembly process all invalidate part of what was measured before, and each should be followed by a measurement rather than an assumption.

The practical rule is to change one thing at a time and to measure the result. A project that changes the laminate, the stack-up and the via structure in a single revision learns nothing when the performance changes, and has to repeat the exercise to find out which change was responsible. Our cost reduction notes describe how those changes are evaluated against the price of the finished board.

Working With the Fabricator

The relationship with the fabricator changes as a high frequency design matures. At the prototype stage the questions are about availability and about whether the stack-up can be built at all. At the production stage they become questions about tolerance, about lot to lot consistency and about what the process capability actually is rather than what the data sheet claims. Asking for the capability figures before the layout is frozen is far cheaper than discovering them afterwards.

The most useful document to agree early is the stack-up drawing. It names each layer, its material, its thickness and its copper weight, and it is the reference against which the impedance is calculated and the coupons are measured. Once it is agreed, any change to it is a change to the electrical design, and it should be treated as a revision rather than as a production detail. Our layer assignment notes describe how that drawing is constructed.

Volume, Testing and Yield

The economics of a high frequency board are different from those of an ordinary one, because the material is expensive and the yield is lower. That makes panel utilisation worth attention: fitting more boards on a panel reduces the cost per unit more than any other single measure, and it is achieved by arranging the panel around the board dimensions rather than by shrinking the board itself.

Testing has to be scaled to match. Full electrical test on every board is normal, but the radio frequency performance is usually verified on a sample, with the impedance coupons measured per panel and the full parameter set measured periodically. Where the product is safety critical or is supplied to a demanding customer, the sample rate rises and the documentation becomes part of the deliverable, which is a cost that should be planned rather than discovered at the first audit.

FAQ

Can the prototype and the production board use different materials? They can, but the electrical result will differ and the difference has to be understood before the change is made. For radio frequency work the two usually stay the same.

Why is impedance verified on a coupon rather than on the board? Because the coupon is built on the same panel with the same process, so it measures what the fabricator actually produced rather than what was designed. It is also easier to measure.

When should a low loss material be introduced? When the measured link budget at the intended frequency and trace length no longer closes on the cheaper material. Introducing it earlier pays for performance the product does not use.