How to Choose an SMT Assembly Factory: Equipment, Inspection and Support

By the time a hardware project reaches the SMT stage, the design is finished and the components are on order, and the yield of every unit after that point depends on the factory. Cold joints, bridges, shifted parts and starved pads are all process outcomes, and any one of them can turn a batch into rework. Choosing an SMT assembly factory is therefore a decision about process control rather than a search for the lowest quotation, and the differences between factories that look similar on a price sheet show up in the yield data a few months later.

The Hard Metrics: Equipment, Inspection and Capacity

Three metrics decide what a factory can actually build, and they are worth checking before anything else is discussed.

The first is placement capability, and an automated SMT line is only as good as the accuracy it holds at production speed. The placement machine sets the smallest component the line can handle reliably and the accuracy it can hold at speed. Products have been moving toward 0201 and even 01005 passives for years, and a line that cannot place them consistently will force a change of factory later, at the worst possible time. Ask for the machine model, the placement accuracy and the smallest package the factory runs in normal production rather than in a demonstration.

The second is the inspection chain. Solder paste inspection, automated optical inspection and X-ray inspection each catch a different class of defect, and a factory that runs all three is inspecting the process rather than the finished product. Paste volume problems are caught before they become joints, placement and visible solder defects are caught after reflow, and the hidden joints under a BGA or a QFN are only visible to X-ray.

The third is capacity and production planning. A factory with meaningful spare capacity can absorb a schedule change, run a small batch without deprioritising it, and hold a promised date during a busy season. Capacity is also about people: a line running three shifts needs engineers on all three.

What Each Inspection Step Actually Catches

The value of the chain is easiest to see in what a single stage misses. Solder paste inspection measures the volume and position of every printed deposit and compares it to the stencil design. It catches insufficient paste, excess paste, bridging and misalignment while the board can still be washed and reprinted, which is the cheapest possible moment to correct a problem.

Automated optical inspection covers the placement and the visible solder after reflow: missing parts, wrong parts, polarity errors, tombstones, bridges and poor fillets. Its value depends on how the programs are built and how false calls are handled. A line that tunes AOI to pass everything is worse than a line without it, because it produces inspection records that mean nothing.

X-ray inspection sees what optical systems cannot: the solder joints under area array packages and the voiding in a thermal pad. For boards with BGA, QFN, LGA or press-fit connectors, X-ray is not optional equipment, it is the only way to verify the joint. Together, the three stages are usually described as SPI AOI X-ray inspection coverage, and their real output is a defect rate the factory can quote with confidence.

The Details That Decide Yield

Stencil management is the first detail most buyers ignore. The stencil controls paste volume, and paste volume controls everything after it. A factory that stores stencils properly, checks tension, cleans them on a defined schedule and replaces them when they wear will hold a fine-pitch process stable. A factory that treats stencils as consumables of unknown age will drift.

Changeover efficiency matters as soon as a product line has several variants. If every changeover takes hours, small and medium batches get pushed to the back of the queue, and delivery promises become unreliable. Feeders set up offline, standardised programs and a structured line setup routine all shorten the changeover and give the planner room to be flexible.

Moisture control is another quiet variable. Components in moisture-sensitive packages absorb water and crack or delaminate during reflow unless they are baked or stored correctly. A factory with documented MSD handling, dry storage and logged bake cycles is managing a risk that will otherwise appear as random failures months later.

Engineering Support: The Difference Between Build and Solve

The most valuable thing an SMT PCB assembly partner provides is not the machine time. It is the review that happens before production starts.

A capable factory checks the Gerber data, the bill of materials and the placement files against each other, then reports what it finds: a footprint on the board that does not match the part in the BOM, a polarity marking that will be ambiguous on the production floor, a panel design that leaves no room for the conveyor, a fine-pitch part that needs a different stencil aperture. Every one of those items is a few minutes of engineering time before the build and a day of rework after it.

That review is also how a components procurement problem is caught early. If a part is hard to source, or a second source is electrically different, the right time to find out is before the kits are assembled, not during the build.

A prototype to production handover is where this pays off. When the same organisation handles prototype PCB assembly and then the volume order, the process knowledge transfers: the profile, the stencil and the placement program that were proven on the prototype are the ones used in production, and any change is a deliberate decision rather than an accident.

Questions to Ask Before Placing an Order

• What is the smallest passive the line places in normal production, and at what accuracy?
• Which inspection stages run in line, and what defect rates do they report?
• How are stencils stored, cleaned and retired?
• What is the typical changeover time, and what is the minimum batch size the planner will run?
• How is moisture-sensitive component handling documented?
• Who reviews the design before production, and what does that review produce?
• What happens to a batch that fails inspection: rework, scrap, or a root cause investigation?

Reading the Quotation

A PCB assembly supplier that cannot explain what is inside its quotation is quoting machine time rather than a service. Two quotations for the same board can differ by a factor without either being dishonest, because they assume different levels of service. A low quotation may assume the data is perfect, the parts arrive on time, the stencil is existing and the first article is the customer responsibility. A complete quotation includes engineering review, stencil fabrication, inspection, first article documentation and rework, and it usually turns out to be the cheaper option once the project is measured end to end.

For a product moving from high-volume PCB assembly back into engineering change orders, the same principle applies in reverse: the factory that can absorb an ECO without rewriting every program is the one that keeps the schedule honest.

FAQ

Do I need X-ray inspection for every board? No, but any board with area array packages, press-fit connectors or a thermal pad under a power device needs it. The joint quality there cannot be judged visually.

Is SPI worth it for small batches? Yes. Paste defects are the cheapest class of defect to correct, and SPI is the only way to see them before reflow.

How do I compare two factories fairly? Send the same data package and the same volume assumptions to both, and compare what is included: engineering review, stencils, inspection, first article and rework policy.

What is a reasonable first article process? The first board of a run should be inspected against the drawing, and the results documented and approved, before the rest of the batch proceeds.

Does an automated SMT line guarantee quality? No. Automation makes the process repeatable, but the inspection strategy, the stencil condition and the engineering review decide what the repeatability is worth.

Summary

An SMT assembly factory is chosen for the process it can hold, not the price it can quote. Check placement capability against the smallest components in the design, confirm the inspection chain reaches the hidden joints, look at how stencils and moisture-sensitive parts are managed, and test the engineering support by sending a data package and reading what comes back. The factory that finds the problems before a build starts is the one that will keep the quality management system meaningful once production is running.