Test & inspection
4 flying-probe heads, 3,072 in-circuit test points on the Agilent i3070, custom functional fixtures and burn-in, plus 100% 3D AOI on every SMT line and X-ray on a minimum of 5 boards per batch.
Choose the test method
Flying probe asks whether the board is connected. In-circuit test asks whether every component is the right one, the right way round, and within value. Functional test asks whether the product works. Pick the question you need answered at the batch size you are building.
Four probe heads move and measure simultaneously against the netlist, so no fixture has to be designed, built or stored. Choose it for prototypes and small batches, where fixture cost and fixture lead time cannot be spread across the order. It finds opens, shorts and broken net continuity.
3,072 test points on a bed-of-nails fixture check opens, shorts, resistor and capacitor values, diode orientation and IC connectivity, and leave a per-board pass/fail record. Choose it for series volume, where one fixture is amortised across the whole batch and every board is tested in seconds.
A purpose-built jig powers the board up, exercises its interfaces, verifies signal integrity and loads firmware, the same operations the product performs in the field. Choose it when a fault would only appear once the board is powered, not when it is merely measured.
Flying probe vs fixture ICT
The two electrical test routes trade fixture cost against throughput. The comparison below is the one we walk through with a customer before a test plan is written, because it decides the unit cost of every board in the batch.
| Parameter | Flying probe, Takaya APT-1500F | Fixture ICT, Agilent i3070 |
|---|---|---|
| Probing | 4 heads move to each net in turn, contacting the board directly | 3,072 test points contact the board at once through a bed-of-nails fixture |
| Fixture cost | None, nothing has to be designed or stored | A custom fixture has to be designed, built and debugged before the first board is tested |
| Program generation | Generated from the netlist, so testing can start within the same project window | Program is written and debugged against the physical fixture |
| Throughput per board | Slower, the heads travel point to point across the board | Faster. All points are contacted simultaneously once the board is clamped |
| Best batch fit | Prototypes, engineering builds and small batches | Series runs and repeat batches where the fixture is reused |
| What it proves | Net continuity: opens, shorts and mis-routed connections | Continuity plus component values, diode orientation, IC connectivity and a per-board pass/fail record |
Functional test
A functional fixture is not a catalogue item. It is built around your board, your connectors and the sequence your firmware expects, which means the test script is co-developed with your engineering team and archived with the project so a later batch is tested the same way as the first.
Test chain
Each link catches a class of defect the previous one cannot see, which is why a board does not skip a link to save a day. The chain ends at outgoing inspection, where the sample size is set by AQL Z1.4 sample rather than by how the batch feels on the day.
Fixture development
A fixture programme is a small project with its own deliverables. Each of the five steps below ends in a document or a piece of hardware that you receive, which is what keeps fixture lead time visible instead of hidden inside a production schedule.
Your netlist, functional specification and connector list define the test coverage we commit to. Deliverable: a written test spec and coverage list.
Probe layout, clamping and connector access are drawn against 3,072 possible contact points. Deliverable: the mechanical and probe layout drawing.
The fixture is machined, wired and aligned, then debugged against a known-good board. Deliverable: a working fixture with its probe alignment record.
A first article runs the full script with your limits in place. Deliverable: a signed first-article record and the test log, held with the project file.
Every board in the batch runs the released script and gets its own pass/fail record. Deliverable: the test report that ships with the goods.
QA data & reporting
Coverage is defined by basis, not by a single percentage: some gates inspect every board, one gate inspects a batch sample, and one gate inspects the shipment. The matrix below is the whole sampling rule set on one screen, followed by the one industry benchmark that qualifies it.
The test floor
Electrical test, functional test, burn-in and outgoing inspection each produce a different record. Between them they cover the three independent checks every product passes before it is packed: an in-line inspection, an electrical measurement and a final visual gate.
The Takaya APT-1500F runs straight from the netlist, which is why an engineering build can be tested before any fixture exists.
Every board in a series batch is clamped into the i3070 fixture and measured against the released program, one pass/fail record each.
Boards run powered in the ageing rack, with duration and cycling set by the project requirement rather than by a default.
Packing, labelling, quantity and paperwork are checked on a sample sized by AQL, then the goods are sealed with the certificate of conformance.
Test plan review
Tell us the batch size and what the board has to prove. An engineer answers with the test route, the fixture proposal and the report package, engineering review and quotation within 24 hours.
Send your Gerber, BOM or a plain description. An engineer, not a sales script, reviews it and replies within one working day. NDA on request.