Solder paste printing, not the placement head, sets the real yield ceiling on a PCB line.

The line runs, the printer reports good boards, and the first article still fails at reflow because nobody re-checked paste height after a squeegee swap. Yield losses usually start in that gap between the report and the panel.

A PCB printer is mechanical work dressed as an office task: squeegee, stencil, paste and a camera judging registration. When one of those four drifts, the defect surfaces two stations later at reflow or AOI, far from the cause.

That distance is why printers get blamed for placement faults and cleared for defects they caused. Anyone reading specs from the PCB printers market should ask how registration is measured, because brochure numbers rarely match production panels.

Stencil apertures set paste volume first

Paste volume comes from aperture area, foil thickness and wall finish, not from the pressure dial. A laser-cut aperture with a rough wall releases less paste than a smoother one, and that shows as height variation across the panel.

Area ratio is the number that decides whether paste transfers at all. Narrow apertures on a thick foil release poorly, so fine-pitch work needs a thinner stencil or a stepped area rather than more pressure, which only smears the deposit.

Aperture geometry also matters relative to print direction: paste rolls ahead of the blade and needs somewhere to go. Apertures aligned with squeegee travel fill more evenly than those across it, which is why rotation is worth testing.

Pressure and separation decide release

Too much pressure scoops paste out of the aperture and leaves a thin, dished deposit; too little leaves paste on the foil. The setting is the lowest pressure that clears the stencil, found by measuring deposits, not watching the blade.

Separation speed is the quieter variable. Pull the board away too fast and paste strings or lifts from the pad; slow it and release is clean but cycle time grows. Most bridging complaints trace back to separation, not paste.

Blade condition belongs in the same conversation. A worn edge with a rounded profile behaves like a wider squeegee and changes the pressure actually delivered to the foil, so a setting validated on a new blade can drift within weeks.

Mask registration follows panel growth

Mask registration is judged against copper, but the panel has already moved by the time the mask is exposed. Lamination and baking shrink or stretch the substrate, so artwork that fits a bare panel sits off centre afterwards.

Drill position compounds it. If the drill program is aligned to the panel edge rather than the datum used for artwork, small scale changes push vias toward the pad edge and thin the mask dam on one side.

Shops that drill vias notice this fastest, which is why tooling decisions in the micro drill for PCB segment affect registration as much as hole quality. Spindle runout and hit count feed back into positional error.

What a first article should measure

Most first articles check that the board prints, places and reflows without a stop. That is a process check, not an approval. The sheet should list paste height, volume, area and offset per aperture class, plus mask dam width at tight pitch.

Height on its own hides volume errors. A tall, narrow deposit and a short, wide one can both pass a height window while delivering different amounts of paste, which is why volume and area belong beside height on the form.

Offset deserves its own line on the sheet. Registration offset pushing one way on one side of the panel points to stretch or stencil tension, while random scatter points to vision or clamping, and the two need different fixes.

Where SPI data stops being trustworthy

An SPI machine reports what it can see: height, area, volume and offset on the pads it recognised. Shiny or contaminated pads produce grey readings, and those readings still travel downstream carrying the same confidence as good ones.

The practical check is correlation. Pull boards that SPI flagged and boards it passed, run both through reflow and AOI, then compare. If passed boards keep failing at the same aperture, the limits are wrong, not the process.

Limits copied from a machine default rarely fit a real product. Volume windows should come from the deposit the stencil actually produces on that board, with the reflow result acting as the arbiter rather than a vendor's suggested range.

Batch drift looks like paste buildup

Paste builds on the stencil underside, so the deposit degrades slowly. Boards at hour one and hour six both pass, while those between sit at the edge of the window and fail at reflow for reasons nobody links to cleaning.

Wipe frequency, solvent choice and vacuum assist change how fast that buildup reaches the panel. A dry wipe every few prints with a solvent wipe at intervals is a starting point, not a rule that suits every paste chemistry.

Room humidity moves the same needle. Dry air thickens paste on the stencil and slows transfer, humid air encourages slumping after separation, and both effects surface as defects the printer gets blamed for. The fix usually sits with the HVAC.

Thawing and mixing still get skipped

Cold paste is thick paste, always. A cartridge pulled from the fridge and printed immediately transfers poorly, so the operator quietly compensates with pressure and then over-compresses the deposit once the paste warms up halfway through a shift.

Thaw time depends on package size and belongs on the container label, not in somebody's estimate. Mixing matters too: paste separates in the jar, and a few slow, gentle turns restore consistency without folding air into the deposit.

Time on stencil is the other clock. Paste left rolling for hours loses solvent and gains viscosity, so a line pausing for a long meeting should scrape and reload rather than resume with what is already on the foil.

The second reflow pass changes the profile

Double-sided boards get cooked twice, and the second pass is where bottom-side paste and heavy parts show their limits. Paste that printed cleanly on the first side can slump on the second, especially near a large thermal mass.

Panel sag matters as much as paste. Thin boards supported only at the edges bow noticeably inside the oven, and that bow changes deposit geometry while the paste is melting, distorting joints that looked perfectly fine straight after printing.

Support pins, pallets and a reviewed profile for the second pass usually cost less than reworking a whole batch. The profile should be validated with thermocouples taped to the real assembly, not simply copied from the first side's recipe.

Changeover is a tooling problem

The clock starts when the last good board leaves and stops when the next first article is approved, not when the stencil is clamped. Most of that time goes to finding tooling, verifying it and re-establishing paste settings.

Dedicated stencils with labelled frames, a premixed paste kit and a saved printer recipe per product cut changeover far more than a faster machine does. Buying speed you cannot use between setups is money spent against the wrong specification.

Quick-release frames help only when tension is repeatable. A frame that clamps at a different tension each time changes paste release, and the operator ends up quietly re-tuning pressure for a stencil that was printing well an hour earlier.

Acceptance checks that still hold later

Acceptance should test the machine against your board, stencil and paste, with your operators running it. A demo on a generous test coupon proves the printer can print, not that it can hold your fine-pitch device at cycle time.

Ask for the registration method, the vision resolution and how the machine compensates for panel stretch on a production panel. Ask what happens when a fiducial is marginal, because that is where an unstable printer starts scrapping boards quietly.

Put paste height, volume and offset limits in the document with the measurement method named, and require a repeatability run at the start and end of a shift, so drift is caught while the supplier is still on site.

FAQ: questions buyers ask

How often should the stencil be cleaned?

Depends on paste chemistry and board density, but the trigger should be measured rather than timed. Print a test coupon every set number of boards, check deposit volume, and clean when it drifts outside the window instead of waiting for a scheduled wipe.

Does a thinner stencil help fine pitch?

It improves release for small apertures and reduces paste volume for large pads elsewhere on the same board. Stepped stencils exist for that reason, so the decision should follow the aperture list rather than a general preference for thinner foil.

Why do first articles pass and production fail?

Because the first article is usually printed on a rested machine with fresh paste and a clean stencil. Production adds heat, buildup and operator variation, so approval should include a run at the end of a shift under normal conditions.

Can SPI replace inspection after reflow?

No. SPI sees the deposit before reflow and cannot really predict how it will wet, slump or void. It narrows what reaches AOI, and the two data sets should always be compared rather than treated as substitutes for each other.

What should buyers ask about mixing?

Ask for the thaw and mix procedure in writing, the time limit for paste on stencil, and how the line handles a paused shift. Vague answers usually mean the procedure lives in one operator's memory and leaves with that operator.