- Inspection checks how a board was built; electrical test checks whether it works. A sound strategy uses both, because each catches defects the other can't.
- AOI sees what's visible (presence, placement, polarity, visible solder joints) but can't see under BGAs or other hidden joints.
- X-ray sees hidden joints but mostly not component bodies, so it can't confirm polarity or catch a wrong part that looks the same.
- Neither camera nor X-ray can confirm the value of an unmarked chip resistor or capacitor. That takes an electrical measurement.
- Flying probe needs no fixture, which suits prototypes, NPI and low volume. Bed-of-nails ICT costs more up front but tests each board in seconds, which suits stable, higher-volume designs.
Inspection vs. electrical test
PCB assembly verification splits into two families:
- Inspection, meaning solder paste inspection (SPI), automated optical inspection (AOI) and automated X-ray inspection (AXI), looks at the physical build: is the right part in the right place, the right way round, with a good solder joint?
- Electrical test, meaning flying probe, in-circuit test (ICT), boundary scan and functional test (FCT), checks electrical behavior: opens, shorts, component values, and whether the board does its job when powered.
Neither family covers everything. Inspection can't tell you a capacitor has the wrong value. Electrical test can pass a solder joint that's mechanically marginal and will fail later. That's why production lines layer methods instead of picking one.
Method-by-method comparison
| Method | Catches well | Misses / limits | Setup | Best fit |
|---|---|---|---|---|
| SPI Solder paste inspection | Problems in printed solder paste (too little, too much, misaligned) before parts are placed | Says nothing about components; no solder joint exists yet | Inline machine, programmed from paste-layer data | Any SMT line; catches print problems before they become solder defects |
| AOI Automated optical | Missing, misplaced and skewed parts; reversed polarity; visible solder defects such as bridges; a wrong part when it looks different (marking, size, color) | Hidden joints (BGA, J-lead, many odd-form connectors); the value of unmarked passives | Program per board, built from CAD/centroid data | Every build, prototype to production |
| AXI / X-ray | Hidden-joint problems: shorts and bridges under BGAs, insufficient solder, lifted leads, voids | Polarity (except parts with asymmetric internals, such as tantalum capacitors); a look-alike wrong part; hidden opens can be difficult; slower than AOI | Programming; often targeted at BGAs and hidden joints instead of the whole board | BGA, QFN-style and other hidden-joint builds; high-reliability work |
| Flying probe | Opens, shorts, and component values (including unmarked passives) on nets it can reach | Nets without physical access; slower because probes contact nodes one after another; probing fine-pitch leads can damage parts | No fixture: programmed from CAD data | Prototypes, NPI, low volume, frequent design changes |
| ICT Bed-of-nails in-circuit | Similar faults to flying probe, very fast, with good coverage including power rails and unpowered analog | Needs a test point on each net; nets without access are missed | Custom fixture: significant up-front cost (NRE), usually specific to one board revision | Stable designs at production volume |
| Boundary scan JTAG, IEEE 1149.1 | Opens and shorts on interconnects between JTAG-capable devices, including BGA-to-BGA nets no probe can reach | Nets that don't touch a scan-capable device (most analog circuits and passives) | Connection through the board's test access port (TAP) header; no bed-of-nails | Dense digital boards with limited probe access; pairs well with flying probe |
| Functional test FCT | Whether the powered board actually performs its function; some wrong values show up indirectly | Often can't pinpoint which part failed; latent defects (such as head-in-pillow) can pass | Product-specific fixture and test software | Final gate on most production builds |
Not sure which mix your board needs? We'll recommend an inspection and test plan during DFM review.
Schedule a DFM review →AOI and X-ray work together
AOI and X-ray aren't competing choices; each covers the other's blind spot. AOI can't see joints hidden under a BGA or a J-lead; at best it confirms that the part is present and roughly positioned. X-ray sees through the package to the joints but mostly can't see the component body, so it usually can't tell that a part is reversed or that a look-alike wrong part was placed.
A published trial makes the split clear. In 2002, Solectron ran a server board (about 1,855 components, 52 boards) with deliberately induced faults through 2D AOI, laminography X-ray and manual visual inspection:
| Defect type | AOI detected | X-ray detected |
|---|---|---|
| Reversed polarity | 100% | 0% |
| Insufficient solder | 37% | 79% |
| Lifted leads | 43% | 100% |
| All induced faults | 69% | 72% |
The lesson still holds: AOI and X-ray find different things, and manual visual inspection alone misses most of them.
The unmarked-passive blind spot
Small chip resistors and capacitors (0402 and smaller) usually carry no markings. If the wrong reel is loaded, a 10 kΩ resistor where a 1 kΩ belongs looks identical to a camera and to X-ray. Only an electrical measurement, such as flying probe or ICT, confirms the value. Functional test may catch the result indirectly, but often without pointing to the part at fault.
Good lines also prevent the error at the source, with checks such as verifying reels at setup and barcode-verifying the feeder setup.
Head-in-pillow BGA defects
In a head-in-pillow defect, a BGA ball rests in the solder paste but never fully fuses with it. Because the ball still touches the pad, the joint can conduct during in-circuit and functional test, so the board passes, then fails outright or intermittently in the field.
That's the main argument for X-ray on BGA-heavy builds, together with tight reflow process control. Be careful with claims that X-ray finds every head-in-pillow defect: they're hard to see in 2D X-ray, and angled (2.5D) or 3D/CT X-ray does better.
Flying probe vs. ICT vs. boundary scan
| Factor | Flying probe | Bed-of-nails ICT | Boundary scan |
|---|---|---|---|
| Fixture | None | Custom, per board revision | None (TAP header connection) |
| Up-front cost | Low (programming) | High (fixture NRE) | Software and test development |
| Test time per board | Minutes | Seconds | Fast, for digital interconnect |
| Access needed | Physical contact with each net | A test point on each net | Scan-capable devices on the net |
| When the design changes | Reprogram | Fixture rework, or a new fixture | Update the test model |
| Best fit | Prototype, NPI, low volume | Stable design, higher volume | Dense digital and BGA boards |
For low-volume boards with limited access, boundary scan for the digital interconnect plus flying probe for passives and analog nets can be a cost-effective combination. Some ICT platforms also combine techniques for boards where test-point access is limited.
Choosing by build stage
Prototype and NPI
The design is still changing, so avoid fixtures. A typical plan is AOI on every board, X-ray on BGAs and other hidden joints, flying probe for electrical verification, and functional test once the product team can supply a procedure.
Pre-production
Lock the design, add test points where coverage was weak, and decide whether production volume will pay back an ICT fixture. This is also when to settle the functional-test fixture and software.
Production
Keep AOI on every board and X-ray where hidden joints matter. Add ICT if the volume justifies the fixture, or stay with flying probe if it doesn't, and use functional test as the final gate.
Design for test (DFT)
- Give every net you want tested a reachable test point, kept clear of component bodies.
- If the board has JTAG-capable devices, bring the test access port out to a header so boundary scan is possible.
- Decide early which nets functional test will cover, so electrical test can focus on the rest.
- Flag BGAs and other hidden-joint parts so X-ray coverage is planned from the first build.
At Surface Art, every build gets automated optical inspection. We add X-ray for hidden joints such as BGAs, flying probe for fixtureless electrical test, and functional and in-circuit test fixtures for production programs.
Our DFT review checks test-point access and coverage before the first part is placed, so the test plan is settled before the build, not after.
Sources
- Solectron, AOI vs. AXI trial (detection-rate tables), IPC APEX 2002. ipc.org (PDF)
- GÖPEL electronic, comparison of inspection and test strategies, IPC-hosted technical paper E7&S13_01. ipc.org (PDF)
- JTAG Technologies, test strategy and passive-value verification, IPC APEX 2005 paper E17&S10. electronics.org (PDF)
- XJTAG, "Choosing a PCB test method" FAQ. xjtag.com
- Keysight, i3070 in-circuit test: Cover-Extend Technology flyer. keysight.com (PDF)
- I-Connect007, "A look at Saki's approach to 2D, 3D and X-ray technology" (2015). iconnect007.com