Most assembly programmes inherit their test list from the previous product. Inspection after reflow, in circuit test, functional test at the end, and a signature on the traveller. The list feels complete because it is long, and it stays unchanged for years because nobody can point to the fault it let through.
Test coverage is not the number of stations a board passes through. It is the set of fault classes the sequence can actually detect, measured against the fault classes that matter for that board. A line can run four stations and still ship a lifted pin, and a line can run two and catch everything that has ever failed.
What an inherited test list misses
Coverage gaps form where the tests overlap and where they do not reach. Visual inspection and AOI both look at joints from above, so a void under a ground pad passes both. In circuit test proves continuity and value, so a correct resistor with a cold joint on one end can still pass if the probe contacts the pad rather than the lead. Functional test proves the board works at the moment it is tested, which is exactly the condition a marginal joint survives.
The gaps are rarely random. They cluster on area array packages, thermal pads, connectors, and anything with a shield can over it.
Which station catches which fault
| Station | Fault classes it detects | What it cannot see |
|---|---|---|
| 3D solder paste inspection | Volume, height, offset, bridging in paste | Anything that happens after reflow |
| Post reflow AOI | Missing part, wrong part, shift, tombstone, visible solder defects | Joints under a package body or shield |
| DIP AOI | Through hole fill, polarity, pin in hole | Barrel fill below the visible surface |
| X-ray, 2D and CT | Voids, open joints, ball collapse, bridging under area array | Electrical value, and faults outside the field of view |
| In circuit test, needle bed | Opens, shorts, wrong value, missing device, orientation | Nodes with no probe access, and marginal joints that still conduct |
| Flying probe | The same electrical faults without a fixture | Speed, so coverage per hour is lower |
| Functional test | Whole board behaviour against a specification | Any fault that does not change the measured output |
Read across a row, not down a column. A board is covered when every credible fault class appears in at least one row that can see it.
Sizing coverage to volume and risk
Coverage has a cost per board, so the strategy has to change with quantity and consequence of failure.
| Situation | Coverage that usually pays back |
|---|---|
| Prototype, under 50 boards | Paste inspection, AOI, X-ray on area array, functional test at the bench |
| Mid volume, 50 to 1,000 | Add a needle bed fixture, keep X-ray sampling on the risk packages |
| High volume, above 1,000 | Full ICT plus functional test, X-ray sampling plan tied to process capability |
| Medical and safety related | Full traceability per unit, 100 percent functional test, recorded parameter limits |
| Radio frequency boards | DC and functional coverage on the line, RF parameters measured in a calibrated laboratory |
How SUPERB PCBA sets coverage
Coverage is agreed before the fixture is built, from the fault classes on the board rather than from a standard list. Where a node has no probe access, that limitation is written into the plan rather than discovered at first test.
Inspection runs from paste through to final: 3D paste inspection on a S8080, post reflow AOI, DIP AOI for through hole work, and X-ray with 2D and CT modes on a View-x1800, judged to IPC-7095 with void acceptance under 5 percent. Electrical test uses a needle bed in circuit tester and flying probe where a fixture is not justified. Functional test fixtures are built at the Shenzhen assembly site against the customer specification, and each one is proven on known good and known bad boards before release. First article is compared to the file on a dedicated station, and the whole route sits inside 17 process steps and 10 quality stations, under ISO 9001 and ISO 13485 for medical work, to IPC-A-610 Class 2 or Class 3.
Radio frequency parameters are a separate case. The production line proves DC behaviour, power rails and digital function, and runs the customer fixture where one exists. Gain, noise figure, return loss and phase are measured by the customer laboratory or an accredited partner laboratory, and the results are attached to the lot record. Claiming otherwise would be easy and wrong.
The cost side of the decision is covered in which test plan pays back at 500 boards, and the inspection comparison in AOI versus X-ray.
Common questions
Is 100 percent AOI the same as full coverage?
No. It is full coverage of the faults visible from above. Joints under a package, barrel fill and electrical value stay outside that field, which is why a plan built only on AOI leaves the same gaps on every lot.
When does in circuit test stop being worth the fixture?
When the fixture cost is larger than the escapes it prevents over the programme. That normally happens below a few hundred boards, where flying probe or functional test with a targeted probe set carries the same fault classes at lower fixed cost.
How is coverage recorded?
As a matrix: fault class against station, with the result per lot. That record is what makes the second production run cheaper, because the sampling plan can be defended with data. The equipment behind each station is listed under functional test and ICT testing.
Next step
Send the Gerber, the BOM and the test specification you hold, with the annual quantity and the acceptance standard for the product. The reply returns a coverage matrix for that board, the stations it needs, and the fixture cost against each volume band so the decision is made on numbers rather than habit.