First Article Inspection for PCB Assembly: What to Measure
A first article inspection exists to answer one question before a production run begins: does the assembled board match the assembly drawing and the process that was qualified? If that question is answered honestly, the run follows; if it is answered by a signature, the run produces a batch of defects that all trace back to the same unchecked assumption.
What FAI Is For
The first article is a build of one to five assemblies, taken through the real process rather than a special one, and inspected against the drawing with the same criteria that production will face. Its purpose is to catch the mismatches that only appear once the physical parts meet the physical stencil, the real profile and the actual placement program.
Because the sample is small, the FAI is a check of the process rather than of the parts. A measurement that is in tolerance on the first article tells you the setup can produce good hardware; it says nothing about whether it will keep doing so, which is what process monitoring afterwards is for. The form that carries the results should be the customer standard or the shop form the customer approved, because a shop-specific checklist tends to quietly drop the characteristics that are hardest to measure.
Documentation Package and Revision Control
The FAI starts with the paperwork: assembly drawing revision, bill of materials revision, approved vendor list, and the process documentation actually used on the line. The most common FAI finding is a mismatch between the revision the customer released and the revision the shop built, and that mismatch invalidates every other check on the form.
The document set should be assembled and reviewed before the build, not after. Deviations, approved alternates and any concession granted in a previous run belong in the package, so the first article is inspected against what was agreed rather than against what the drawing literally says. Where the customer supplies a first article report form, its fields define the minimum record, and anything the shop adds belongs in an appendix rather than merged into the form itself.
Placement and Polarity Verification
Verify component reference designator, value and polarity for every part on the assembly. Visual checks catch the obvious reversed electrolytic and the rotated diode, while an automated check against the placement program catches the systematically wrong part that a human reviewer would accept as plausible. Polarity is where the greatest number of FAI escapes originates.
Placement position itself is verified against the drawing for fine-pitch and connector parts, where a 0.1 mm offset is visible and matters. The measurement should be recorded with the placement accuracy data from the same run, so an offset is attributed to the machine rather than to the operator who happened to place the part.
Solder Joint and Fillet Assessment
Every solder joint class criterion that applies to the product is exercised on the first article: wetting angle, fillet shape, hole fill percentage for through-hole, and the absence of voids, bridges or disturbed joints. Class 2 and Class 3 differ mainly in the tightness of those criteria, and the FAI is where the intended class becomes visible in a decision.
Judging on a photograph of a section is the reliable method: criteria such as 75 percent hole fill or a specific fillet contour are measurable on a section and remain arguable on a visual image. Where a criterion is borderline, the first article is the right time to raise it, because once the run is released the same joint becomes a reject.
Measurement: What to Record and How
Record the measurement method alongside the result. A dimension taken with a calibrated caliper to plus or minus 0.02 mm supports a different conclusion from one taken from an image with no scale, and the record has to make that clear. Note the instrument, its calibration status and the location measured, so a later dispute can be resolved from the file.
Choose sampling that fits the risk: dimensional checks on the parts that carry the tolerance, and a full check of at least one assembly. Where a dimension is controlled by the process rather than by the design, such as paste deposit volume, record the SPI result from the same boards instead of a hand measurement that cannot reproduce it.
Process Data: Profiles, Paste, Stencil
Attach the process data that produced the article: reflow profile with the peak and time above liquidus, paste lot and date code, stencil thickness and tension, and the printer settings used. This is a process record rather than an inspection item, and it is what allows a later defect to be traced to the conditions of the first build.
The profile deserves particular attention because it is reproducible only from a dated recording. A recorded peak of 240 degrees Celsius with 60 seconds above liquidus is a fact; a note saying the profile was nominal is not. Store the raw file with the FAI package rather than a summary. The oven log is stronger evidence than a hand-written reading, because it carries a timestamp and the zone temperatures actually delivered to that panel.
X-ray and Hidden Joint Verification
Area array packages, bottom-terminated components and any joint that cannot be seen require X-ray as part of the first article. Look at voiding percentage, ball collapse and bridging, and compare against the drawing’s acceptance criteria rather than against an impression. The same inspection should also confirm that no part is missing under a package.
Where X-ray resolution is marginal, an oblique view or a cross-section adds the information. Combining X-ray with the optical inspection already performed also catches the case where the visible joints are perfect and the hidden ones are not, which is exactly the case a visual-only FAI misses.
Disposition of the First Article
The article is either accepted, accepted with documented deviation, or rejected with a corrective action. An accepted-with-deviation result must include the disposition of the parts already built under the same condition, because those units exist and cannot be left undecided while the process is adjusted.
Rejected articles should be treated as evidence rather than scrap until the cause is understood. Keeping the article with the FAI package allows the defect to be re-examined if the same symptom appears later, and it makes the corrective action verifiable rather than a claim.
Records, Repeat FAI and Change Triggers
The FAI record is retained for the life stated by the customer or the standard, commonly several years, and it is reused as the baseline for comparison. A repeat FAI is required when a change could affect form, fit or function: a new component source, a revised stencil, a new line, a different finish or a moved manufacturing site.
The practical rule is that anything not covered by the existing record triggers a new one. Where the change is small, a partial FAI covering the affected characteristics is usually enough, as long as the decision to do a partial rather than a full check is documented and justified. A trigger list kept with the record means the shop does not have to reinterpret the requirement every time a change is proposed.
FAQ
How many assemblies should a first article cover? One to five, built through the real process. Where the process is marginal, build more and inspect them so the decision rests on data rather than on a single board.
Does a first article need X-ray? Yes, whenever the assembly has area array packages, bottom-terminated components or any joint that cannot be seen optically. Hidden joints are the ones a visual FAI misses.
When is a repeat first article required? When a change can affect form, fit or function: new component source, revised stencil or profile, new line, different surface finish or a moved site.
