Multilayer PCB Prototyping: What to Verify Before You Order

A prototype is supposed to answer questions. Too often it answers the wrong one, because the board was ordered before the stackup, the finish or the drill structure was confirmed, and the result is a build that cannot distinguish a design problem from a fabrication problem.

Multilayer PCB prototyping is cheap compared with a production respin, and it is expensive compared with a careful review. The purpose of the review is not to slow the order down; it is to make sure that whatever the prototype shows is caused by the design.

What Makes a Multilayer Prototype Different

The first difference is that the stackup has become an electrical parameter. On a two-layer board, the dielectric thickness matters mainly for impedance on a few traces. On a multilayer board it defines the characteristic impedance of every signal referenced to a plane, and it is fixed during lamination.

The second difference is that the board cannot easily be modified. A missed connection on a two-layer prototype can sometimes be patched. On a dense multilayer board, a rework that crosses a plane or a fine-pitch device usually means a new build.

The third is time. Because the prototype is usually followed immediately by prototype PCB assembly, the assembly package has to be right at the same moment as the fabrication package. A footprint error discovered on the day the boards arrive costs a full cycle.

Verify the Stackup Before the Order

Stackup verification is the step that most often earns its cost. It means confirming that the material, the dielectric thickness on each layer, the copper weight and the finished thickness produce the impedance the design expects, and that the fabricator can hold those values in routine production.

Two questions make the review concrete. Which layers are the reference planes for each critical signal, and what happens to the impedance if the dielectric thickness drifts within the fabrication tolerance? A stackup that only works at nominal thickness is a stackup that will produce boards outside the impedance tolerance.

Where the design needs controlled impedance, ask for the coupon design and the measurement method at the quotation stage. It is much easier to agree a test structure before the panel is built than to argue about which value was intended afterwards.

Impedance: Specify the Tolerance and the Method

Stating a target impedance without a tolerance and a measurement method leaves the outcome undefined. A practical specification names the net or net class, the reference layer, the target value, the acceptable range and the test coupon that will be measured.

It is also worth deciding which signals truly need controlled impedance and which do not. Every controlled net adds review and test effort, and applying the requirement indiscriminately increases cost without improving the design.

The other half of impedance tolerance is consistency. A prototype that measures at the edge of its tolerance and a second build that measures at the other edge can behave noticeably differently, and a product that will be produced more than once should be specified with the variation in mind.

Drill, Finish and Solder Mask Choices

Hole size and annular ring decide whether the design can be drilled reliably, and the drill schedule should be checked against the smallest via in the design. A via that is theoretically possible but outside the routine process window will be the first thing to fail in a slightly different build.

Surface finish is frequently chosen for convenience and should be chosen for purpose. A finish that is excellent for fine-pitch assembly may be unnecessary for a board that will be hand-soldered, and a finish that is incompatible with a coating or a test probe creates problems downstream.

Solder mask clearance and silkscreen rules also deserve a check. A mask dam that is too small between pads is a yield risk, and a silkscreen printed over a pad or a fiducial is a defect that assembly will have to work around.

Design File Review: The Procedural Checks

A design file review is deliberate and checkable, and it is worth running it as a list rather than as a feeling.

• Every footprint checked against the datasheet land pattern, especially for fine-pitch and thermal pad devices.
• Reference designators unique and in sequence, with no duplicated or missing numbers.
• Board outline closed, with tooling and fiducial features that match the panel and assembly requirements.
• Net classes assigned and applied, including the high-speed, power and analog groups.
• Copper clearance and thermal relief checked against the fabricator process window.
• Silkscreen readable after assembly and not covering any pad, test point or fiducial.
• Drill and via schedule consistent with the stackup and the impedance plan.

Fabrication data should be generated from the reviewed files, with the layer set and the drill file matched to the stackup. Most respins that are blamed on the fabricator begin with a data set that was generated before the design was final.

Preparing Assembly in the Same Cycle

Prototypes are almost always assembled as soon as they arrive, which makes it sensible to prepare both packages together. The placement data has to correspond to the coordinates in the current revision, and the assembly drawing needs to state the polarity and orientation conventions for the parts that can be placed in more than one way.

Working with a partner that runs SMT assembly as well as PCB manufacturing also means the panel and the stencil can be designed together, which avoids the common situation where the panel that suits the board is awkward for the printer.

The practical meaning of a prototype to production flow is that this prep work is not repeated. The stackup, the footprint decisions and the assembly data are carried into the next build instead of being recreated by a different team.

Ordering the Prototype Efficiently

Speed in prototyping comes from removing questions, not from rushing the shop. Send the complete package: fabrication data generated from the frozen revision, the stackup with material designations, the impedance specification with a tolerance and a coupon, the drill schedule, and the assembly data if the boards will be placed immediately.

Then state the acceptance criteria. If the prototype is intended to confirm impedance, say which coupon is measured and what the acceptable range is. If it is intended to confirm that a fine-pitch device can be assembled, say which inspection method will be used. A prototype with a defined question produces a useful result even when the answer is negative.

Finally, allow for the fact that the first build of a new stackup is itself a process trial. Building a few spare boards is cheaper than a second order, and keeping one unassembled board as a reference makes later comparisons possible when the design moves into production.

FAQ

How much does a stackup review add to the schedule? Usually a day or less, and it is frequently shorter than the delay caused by a single impedance or footprint problem.

Should the prototype use the production stackup? Yes, wherever possible. A prototype built on a different stackup does not prove that the production board will behave the same way.

What is the most common prototype error? A footprint that does not match the purchased part, followed closely by data generated before the design was frozen.

How many prototypes should be ordered? Enough to allow one or two to be damaged during bring-up, and enough to keep one as a reference build.

Summary

Multilayer prototyping pays back when the review is treated as part of the process. Confirm the design file review and stackup, specify impedance with a tolerance and a method, choose the drill and finish deliberately, and generate the fabrication and assembly data from the same frozen revision. Then the prototype answers the question you asked, and the stackup verification record becomes the first page of the production documentation.