Triage first: support error or stencil error
Before removing the stencil, look at how the deviation is distributed. The two failure classes leave very different fingerprints.
Stencil error follows the aperture. Every aperture of a given type — all 0402 pads, all QFN thermal pads — behaves the same way regardless of where it sits on the board. Change aperture type and you get a different behaviour. Aperture geometry, area ratio and wall condition are properties of the stencil, and all apertures of one type on the board share them.
Support error follows the position. The same aperture type behaves differently in different board regions: correct near a support point, thin or heavy away from it. Conversely, different aperture types in the same region drift together, because what they share is one patch of board that moved.
Two additional discriminators:
- Fit a fresh stencil and the symptom persists → the deviation is not in the stencil.
- Rotate the board 180° in the fixture and reprint. Does the affected area travel with the board or stay with the machine? Staying with the machine (always thin toward the exit side) points to support or clamping geometry. Travelling with the board (always the same board corner) points to that board's own flatness or to components already mounted on it.
This step is worth ten minutes. Get the triage wrong and you will spend the next several hours on a stencil that was never the problem.
Why stencil gasketing is really a question of board stiffness
A stencil is a thin foil tensioned in a frame. It has tension but effectively no bending stiffness of its own, so its ability to seal against the board depends on the board presenting a stable plane.
As the squeegee travels it does two things at once: it drives paste into the apertures, and it applies a downward force to the stencil that is transmitted through to the board. That force is a local load sweeping across the board. How the board responds to that moving load determines the print:
- Adequate stiffness and correct support → the board does not move, the stencil stays sealed, paste can only go down through apertures.
- The board yields locally → the stencil sinks with it, a wedge-shaped gap opens around the aperture, paste bleeds laterally, and the deposit ends up oversized in outline and off-target in height.
- The board yields then rebounds → the relative motion during separation becomes uncontrolled, and deposits get dragged or peaked.
Thinner and larger boards show this sooner. SKPCB's thickness capability spans 0.2–6.0 mm (6.0 mm maximum), and the two ends of that range behave like different workpieces at the printer. Near the lower bound the board is essentially a flexible membrane and needs continuous or dense support to hold a plane; near the upper bound it is stiff enough on its own, and the job of support shifts from flattening to locating.
One coupling that is easy to miss: inadequate support narrows the whole process window. Squeegee pressure and separation speed normally have a usable range, but once the board can move, raising pressure amplifies deflection and lowering it fails to fill the apertures. Both directions are blocked. So when no combination of parameters seems right, go back and look at support rather than continuing down the parameter table.
Method 1: universal support pins
The common approach: movable pins placed on the printer's support platform under the board's clear areas, holding it at a number of discrete points.
Suits small to medium panels with enough clear area on the bottom side, boards that are not too thin, and production that changes over frequently. Pins can be rearranged in minutes, which is their real value — high-mix low-volume lines almost inevitably run pins as the primary method.
Limits: pins are discrete support, so the board still spans unsupported between adjacent pins. Deflection across that span grows as the span increases and as the board gets thinner. The method's effectiveness therefore depends completely on whether pin density and placement match this board — in particular whether pins sit directly under the high-print-load areas (large pads, dense fine-pitch devices), not just at the four convenient corners.
Practical points:
- Pins must land in clear areas, not on mounted components, test points, plated holes or solder mask edges.
- Pin heights must match. Mismatched heights actively bend a flat board into a curve, which is worse than no support at all.
- The pin tip type (flat, ball, vacuum cup) determines whether a pin only supports or also locates; these get mixed up during changeover, so verification needs to be a step in the procedure.
Method 2: dedicated support blocks and machined fixtures
A plate machined to this product's bottom-side topography: pockets milled where components already sit, and a solid supporting surface everywhere else.
Suits stable production, boards with many bottom-side components, thin or large-format boards, and products with demanding print requirements (fine pitch and large pads in the same design). It provides near-continuous area support, which removes the span deflection inherent to pins.
Limits: one fixture serves one product, sometimes one side of one product. A design revision, a package change, or a bottom-side layout change can scrap it. Its precondition is a frozen design.
Practical points: the fixture's machining datum must match the board's outline datum, or the pockets will land off the components. Surface ESD properties need consideration. The fixture's own flatness becomes the board's flatness, so incoming inspection of the fixture is not optional.
Method 3: vacuum platforms, magnetic platforms and edge clamping
This family does not push up; it pulls down or holds sideways.
Vacuum platforms draw air through ports in the support surface and pull the board onto a plane. They address something neither pins nor blocks can: a board that arrives already warped. Discrete support can only hold a bent board in its bent state; vacuum can pull some of that warp out before printing. The limit is that the bottom surface needs enough sealable area — many bottom-side components or a dense through-hole field reduces the achievable hold.
Magnetic platforms fix blocks and edge stops by magnetic force. Fast to change over, usually combined with pins or blocks. They are a fixing method, not an area support method.
Edge clamping and stops set the board's in-plane position and do not carry vertical load. That division of labour has to be explicit — a lot of shop-floor trouble comes from using clamping to compensate for missing support, which just bends the board harder.
Practical point: vacuum pulls the board toward the platform, so if the platform is not flat, the board takes the platform's shape. Flatness requirements for a vacuum platform are therefore higher than for pins, not lower.
Method 4: local clearance for the already-populated side
When printing side two of a double-sided board, side one is already placed and reflowed. The bottom surface is no longer a plane; it is a component field with varying height.
Local clearance means designing the support to make room for those components and taking support only from the remaining clear areas. With pins that means re-laying out pin positions to miss every mounted part; with a fixture it means milling pockets.
Limit: when side-one components are dense enough that almost no continuous clear area remains, the number of usable support points can fall below what is needed to hold a plane. At that point you either accept a process downgrade (lower squeegee pressure, more conservative aperture design) or reserve support areas for side-two printing back at the design stage. The latter is a DFM input — by the time the board reaches the printer, process engineering can no longer create it.
Side two: component height turns flatness into a 3D problem
The difficulty of side two is not only that clearance is required, but that what remains after clearance is usually an unbalanced support distribution.
A typical failure chain: side-one large components cluster on one side of the board (connectors and bulk capacitors together along one edge), so during side-two printing that side can only be supported from a few edge points while the opposite side takes a dense pin field. Under the squeegee the board develops a slight tilt or twist about its centreline, and paste thickness on the sparsely supported side drifts systematically. The deviation is not random; it maps onto side one's component layout — which is itself a clue that you are looking at a support problem.
Two further points:
- Large component height sets pocket depth and decides whether the board can lie flat at all. SKPCB's PCBA capability covers large devices up to 74 mm diagonal (BGA/QFP); once parts of that size are on side one, side-two support almost always needs a dedicated fixture rather than universal pins.
- Fine pitch combined with large devices leaves the least process margin on side two. Our assembly capability reaches 0.2 mm pitch for BGA and 0.3 mm for QFN/QFP, with components down to 01005. Those pads tolerate very little thickness deviation, and if they happen to fall in a sparsely supported region the print error converts directly into bridging or insufficient solder.
Arrays, board thickness and warp
How an array is connected determines whole-panel stiffness. A V-cut array is weakened along the scored lines, and the scored line is exactly where bending prefers to happen. A tab-routed or breakaway-tab array has stiffness set by the number, width and placement of its tabs. An array held together by a few narrow tabs is really several small boards joined at weak points, free to move relative to one another — such arrays need support under each individual board, not support that treats the panel as one rigid plate.
Board thickness sets how dense support must be. This is a qualitative relationship, not a numeric threshold: for the same support span, a thinner board deflects more. Thin boards push toward continuous area support or vacuum; thicker boards tolerate discrete pins.
Initial warp decides whether support can solve the problem at all. Discrete support under an already-bent board simply holds it bent. That warp comes from the board's lamination and copper distribution history, and by the printing step it is a fait accompli; the options are partial flattening by vacuum, or accepting a wider process deviation. The real fix is upstream, in copper balance and lamination control — see the related reading below.
Small boards: our minimum board size goes down to 5×5 mm. Boards that small are normally delivered in arrays, so the support discussion applies to the array and its tabs, not to the individual board.
What insufficient support looks like on SPI
SKPCB's PCBA test capability includes SPI, inline 2D AOI, post-reflow 3D AOI, X-Ray, ICT and FCT, and post-print SPI is the most direct tool for diagnosing support. The point is knowing what to look at.
Look at the distribution, not individual pads. SPI reports volume, height and area deviation per pad. The signature of a support problem is spatial correlation: plot deviation against board coordinates and you see continuous regional trends — one side high, the centre low, a whole quadrant offset — rather than scattered points. A stencil problem clusters by aperture type and is indifferent to coordinates.
Look at repeatability. Print a run of boards: a support problem produces a stable, repeating deviation map, because the support geometry is identical every time. Deviation that varies randomly board to board is more likely paste condition, squeegee wear, or scatter in the boards' own warp.
Look at how it carries into reflow. Print deviation on large pads shows up after reflow as a shift in void tendency. Our soldering void specification is <10% under nitrogen reflow — a conditional process capability, not an unconditional promise for any pad geometry and any print condition. Staying reliably at that level requires the printed paste volume to be controlled first. Post-reflow 3D AOI and X-Ray confirm whether the print deviation actually produced joint defects, judged to IPC-A-610 Class 3.
What to specify when you order
Support is a process-side activity, but most of the information it needs comes from the design side. Supplying it up front avoids a round of trial and error on the line.
1. Bottom-side layout information. For double-sided boards, state which side is placed first, how components are distributed on the populated side, and the maximum component height. Without it, side-two support cannot be designed in advance and has to be worked out after the line is set up.
2. Permitted support areas. Mark where pins may contact the bottom side and where they may not — components, test points, sensitive structures, surfaces that must not show witness marks. It is a simple drawing that saves an iteration.
3. Array information. Array type (V-cut, breakaway tabs, routed slots with tabs), tab positions and count, individual and overall dimensions. Tab position determines directly where support belongs.
4. Board thickness and any known warp history. Thickness is mandatory. If this product has had warp issues before, saying so is cheaper than having the line rediscover it.
5. High-print-load regions. Which areas carry large pads and which carry the finest pitch. Those get support priority ahead of the board corners.
6. Which deviations are unacceptable. State the acceptance class tied to paste consistency on critical devices (fine-pitch BGA, shield frames, power devices) rather than only writing "per IPC". Our PCB acceptance classes are IPC-6012 Class II and Class III, and assembly is IPC-A-610 Class 3; which one you choose changes how much process preparation is justified.
7. Volume expectation. One-off prototypes and sustained production lead to different support choices (pins versus dedicated fixture). BOM size — ours supports up to 1000 lines — and changeover frequency both feed that decision.
This information is separate from the stencil data package. The stencil file answers how apertures are cut; support information answers how the board is held. Supply only the first and half the variables in your print result remain unconstrained.
Related reading
- Aperture and paste release fundamentals: SMT stencil aperture area ratio and paste release
- Thickness strategy for mixed assemblies: Step stencil multilevel aperture design
- Choosing between printing and dispensing: Stencil printing vs jet dispensing
- Where thickness and flatness originate: Finished board thickness tolerance control
- The upstream cause of warp: Multilayer copper balance and warpage control
- Stencil service and capability: SMT stencil manufacturing
- Assembly and printing capability: PCBA assembly services
FAQ
Q1: My stencil is within spec but the print is uneven. Is it always support?
Not always, but support is one of the first things to check. Judge by the distribution: deviation that follows aperture type is a stencil issue; deviation that follows board position is a support issue. A fresh stencil that changes nothing, plus continuous regional trends, points to support.
Q2: Pins or a dedicated fixture?
It depends on whether the design is frozen, how crowded the bottom side is, board thickness and volume. Frequent changeover, enough clear area, not too thin → pins. Frozen design, crowded bottom side, thin or large format, demanding print → dedicated fixture. Both can coexist: fixtures for runners, pins for everything else.
Q3: Can vacuum hold-down fix board warp?
Partly, not fundamentally. Vacuum can pull some warp out before printing, but it needs enough sealable bottom-side area and loses effect with many bottom components or a dense through-hole field. And if the platform itself is not flat, the board takes the platform's shape. Warp originates in lamination and copper distribution upstream.
Q4: Why is side-two printing harder?
Because the bottom surface is no longer a plane. Side-one components occupy support area, the remaining support points are usually unbalanced, and the board tilts or twists under the squeegee, with deviation mapping onto side one's layout. When side one carries large devices — we place up to 74 mm diagonal — side two generally needs a dedicated fixture.
Q5: What single piece of information matters most at order time?
For double-sided boards: the placement sequence and the permitted bottom-side support areas. Those two decide whether side-two support can be designed in advance, and they are the information most often missing and hardest to guess.
Q6: Can print parameters compensate for weak support?
Usually not, and trying narrows the window further. If the board can move, more squeegee pressure amplifies deflection and less pressure under-fills apertures; both directions are blocked. When no parameter set works, go back to support instead of continuing to tune.