Choosing a PCB Fabricator for Prototypes and Pilot Runs

Choosing a PCB fabricator for a prototype is usually treated as a purchasing decision, made by comparing the price on two websites. For a simple double-sided board that is the right approach, because the fabrication is standard and the differences are in service rather than in capability. For anything unusual the decision should be made by the engineering team, because the parameters that determine whether the boards can be built are the ones the designer chose, and a shop that accepts the order without reviewing them is not necessarily a shop that can build it.

Start with the Parameters That Decide the Answer

The list of parameters that separates a routine order from one needing an engineering review is short and worth running through before any quotation is requested.

The layer count and the stack-up come first, because the material between the layers, the dielectric thickness and the copper weight all have to be available in the constructions that shop builds. The board thickness and the finished copper weight follow, since a heavy copper layer changes the minimum trace width the process can hold. Impedance control is the fourth: a board with an impedance requirement needs test coupons, a controlled dielectric thickness and a measurement report, and not every shop provides them as standard. The via structure is the fifth, because a blind or buried via changes the number of lamination cycles and therefore the process. Finally, the quantity matters, because a prototype run of five boards and a pilot run of five hundred are different orders even when the data is identical.

Where all of those are within the standard capability of the shop, an online route is efficient: the parameters are configured on the order form, the price is immediate, and the risk is low. Where any of them is at the edge of the capability, the order should go through a person before it is placed.

What to Keep on Record

Whatever route is used, the order parameters should be recorded rather than remembered. The set that matters includes the board thickness, the copper weight, the surface finish, the solder mask colour and specification, the legend colour, the quantity, the test method and the promised lead time.

The reason is repeatability. A second prototype built three months later against a different set of parameters is a different board, and if the difference is not written down the comparison between the two builds is meaningless. It is also the record that settles a dispute about a reorder: a board that arrives with a surface finish other than the one specified was specified correctly only if the specification was recorded.

Fabrication and Assembly as One Handover

Where the boards are assembled immediately after fabrication, the interface between the two steps deserves as much attention as the fabrication itself.

The finish has to be one that the assembly process expects, and it should be stated in the fabrication data rather than left to the shop’s default. The panel or array drawing should match what the assembly line can handle. The data package that the fabricator works from and the package that the assembler works from should be the same revision, because a board built to one revision and placed to another is a defect that appears in the first build and is difficult to attribute afterwards.

Our PCB manufacturing and SMT assembly operations work from a single released data set for that reason, and the review that confirms the two agree happens before either starts.

What a Quotation Should Include

Comparing quotations is only useful when they describe the same scope. The items that are frequently omitted from a low quotation are the ones that matter to a prototype.

Electrical test on every board is the first. Optical inspection of the artwork and of the finished board is the second. A test coupon for impedance, where the design has an impedance requirement, is the third. Packaging that protects the boards in transit, and a delivery date that is held, are the fourth and fifth. Where a shop quotes a price without those, the price is for a different order.

The capability list is the other document worth requesting. A shop that publishes the layer counts, materials, minimum trace and spacing, minimum drill and tolerances it works to has given the designer the information needed to check the design before the order, which is cheaper than discovering the limit in a data query. Our PCB capabilities page exists for that purpose.

Using the Prototype Run as a Trial

The first order is also an assessment of the supplier, and the information that matters comes from how they behave rather than from the boards themselves.

Did the data package produce a query, and was the query specific? A shop that asks about a real ambiguity is reading the data. Did the boards arrive on the promised date, or was the date revised after the order was placed? Did the delivery include the electrical test data, the coupon measurements and the panel drawing, or only the boards? Was a discrepancy between the order and the delivery reported before shipment or discovered on receipt?

The answers to those questions predict how the relationship will behave at volume far better than the price of the prototype batch. Where the product is expected to scale, the same assessment is done once and used for every subsequent order, which is the arrangement our high volume PCB assembly programmes use with their customers.

Panel Utilisation and the Real Cost

Two boards of the same area can cost different amounts to make, because what the shop buys is panel area and the price follows the utilisation rather than the outline. An outline that tiles badly leaves material unused on every order, and the effect is larger on small quantities where the setup cost is spread over fewer boards.

Asking about the preferred panel size and adapting the outline or the array to it is therefore a cost decision, and it is one of the few in this area that can be made once and used for the life of the product. It also reduces the number of surprises later, because a design that fits the shop’s standard panel is easier to schedule at volume.

Lead Time and Communication

Two suppliers with the same quoted lead time can behave differently in practice. One holds the date because it schedules the engineering review at the start and asks its questions immediately; the other accepts the order, asks the same questions a week later and pushes the date once the answers arrive. The difference is visible in how quickly the first data query comes back, which is why a short prototype order tests a supplier regardless of the order size. A shop that reads the data on the day it arrives is a shop that will schedule a production order competently, and the same behaviour is what a customer sees in our order process.

FAQ

Is the cheapest online quotation a risk? For a standard board with complete data, no. For anything outside the standard capability, the risk is that the order is accepted and built at the edge of the process.

When should an engineering review be requested? Whenever the design touches a limit of the shop capability, or whenever a failure of the order would delay a build.

What is most often forgotten when ordering? The surface finish, which affects assembly, and the record of the parameters, which affects every reorder.