From Pilot Build to Volume SMT: What Must Be Frozen
The first twenty boards are finished, the customer has adjusted the firmware, and one driver chip was replaced because it went out of stock. The order is now five hundred boards. The natural question is whether the same production files can simply be reused, and the answer is usually no. The purpose of a sample build is to prove the design; the purpose of a volume build is to repeat an identical set of conditions, and the conditions are not yet identical.
Review One: What the Sample Data Still Contains
A small build is carried out with adjustments that never reach the documents. Wires are added to test a hypothesis, parts are substituted by hand, resistor values are changed temporarily, and a different firmware version may be loaded on each unit.
Before the quantity increases, those decisions have to be written back into the formal data. The customer should confirm that the Gerber data, the bill of materials, the placement coordinates and the assembly drawing belong to the same revision, and should supply a change list relative to the sample batch. Where a driver, a wireless module, a power device or a connector has changed, the list should name the original part, the new part and the affected designators.
The incoming data is then re-checked: package, polarity, coordinates and stencil information. An important point is that an alternative component approved for a sample is not carried into a volume order by default. Every substitution has to be approved again against the quantity and the batch it will be used in, because a substitute that was acceptable for twenty boards is a design decision when it applies to five hundred.
<img src="https://www.gopcba.com/wp-content/uploads/2020/12/project_image_3.jpg" alt="pilot build records compared with volume production data” />
Review Two: The Stencil and the Machine Programme
The pads may not have changed, and the stencil may still need to change. If the sample build produced insufficient paste, tombstoning or local bridging that was repaired by hand before delivery, the stencil aperture and the printing conditions should be reconsidered before the batch runs.
SPI records from the printing stage are the evidence for that decision, because they show the paste volume, the covered area and the offset for each position. On a control board the positions that matter are different from each other: the thermal pad under a driver, the small passives around it and the area beside a wireless module each need their own paste decision. A single volume target applied to all of them is a compromise that shows up as a defect on one of the three.
The machine programme needs the same scrutiny. The feeder table, the nozzle selection, the placement orientation and the first article checklist all have to match the frozen revision, because a programme left over from an earlier version carries obsolete designators and rotation angles that no longer correspond to the data.
Review Three: First Article Release Is Not Only Solder Joints
Optical inspection after the first article covers missing components, offset, reversal, bridging and the joints that can be seen. On a functional product, the release should extend past the joints to the programmed state of the board.
An irrigation valve controller will have valve outputs, sensor inputs, buttons or a communication interface. For the first article to be released as a product rather than as an assembly, the customer has to provide the program file, the programming method, the supply parameters, the harness definition, the load conditions and the pass criteria.
Testing can then proceed in a defined order: the static current and the main rails first, then the program version, then the output channels and the specified feedback. Where the sample build was tested with a real valve and the batch will be tested with a simulated load, the difference between the two methods has to be stated in advance. Measuring a voltage at an interface is not the same as confirming that the function performed correctly.
<img src="https://www.gopcba.com/wp-content/uploads/2026/09/218-1.jpg" alt="SPI report from the printing stage of a batch” />
Review Four: Bounding the Risk Across the Batch
In a run of several hundred boards, something will change while the line is running: a reel is replaced, a stencil is cleaned, a parameter is adjusted. Each of those events is a point in the batch, and recording the point is what makes the batch traceable.
If one designator keeps generating alarms, the affected range can be narrowed using the SPI records, the material batch and the production interval, which turns a decision about the whole order into a decision about a few panels. Boards held for review, boards already reworked and boards that passed test should be kept separate, because mixing them removes the ability to answer that question later.
Where a rework touches a supply or a driver circuit, the relevant load test should be repeated rather than only the joint inspected. A reworked joint that looks correct on a board whose output was never re-tested is a known unknown.
The Documents the Order Should Carry
Four documents make a batch reproducible. A change list against the pilot build, naming every part, value and programme that differs. A released data set with a revision number that appears in the file names and on the drawing. A process sheet recording the stencil revision, the print parameters, the reflow profile and the inspection level agreed for this order. And an acceptance definition stating what is tested, with what equipment, against which limits, and what happens to a unit that fails. Each of those is a page at most, and together they replace the conversation that otherwise takes place halfway through a production run, when the line has already been set up from an assumption. The value of the set is that it survives a change of shift, a change of operator and a repeat order placed eighteen months later.
What Is Worth Keeping
The most valuable output of a pilot build is not a golden sample. It is a set of checkable records: the released data, the parameters used, the first article results and the test records for the pilot units. Those are what allow the batch to be specified, and they are what allow a defect in the batch to be compared against a known state.
Our high volume PCB assembly programme starts by reconciling exactly those records with the data for the new order, and the data package that a pilot should leave behind is the same one our order process asks for at the start of any build. Where the transition also involves a change to the board itself, the design side of the change is handled with the same discipline in our design review.
FAQ
Can a pilot build be approved as a production reference? Only if the documents were frozen at the same time. A pilot that was completed with hand adjustments is a design validation, not a production baseline.
Does a substitute part approved for a sample carry over? No. It has to be re-approved against the batch it will be used in.
What single record makes the batch easiest to control? The SPI data from the printing stage, because it shows the state of the process at a point in the run rather than only the result at the end.
