Which pads count as "large", and why they are a different problem
"Large" here is not an absolute dimension. It is a relationship: once the aperture opening is much greater than the foil thickness, the wall stops being the dominant influence on the paste.
On one and the same stencil, the two small apertures for an 0402 resistor and the single large aperture for a QFN centre thermal pad have identical foil thickness, yet the paste inside them lives in completely different worlds. In the small aperture the paste is gripped on all sides. In the large one, only a thin perimeter touches metal while the bulk of the deposit is unsupported, held in shape by its own viscosity alone.
In practice large apertures cluster around a few pad types:
- Shield frames — long or rectangular continuous pads with large perimeter and high aspect ratio.
- Thermal pads — the large centre pads under QFN, DFN, LGA and most power-management ICs.
- Power device base pads — the large drain/collector pads of MOSFETs, IGBTs and power modules.
- High-current bus pads and large ground copper pads, common in heavy-copper and high-current designs.
- Hold-down pads of large connectors and mechanical attachment areas.
What they share is large area plus, usually, a requirement for uniform joint thickness and controlled voiding — precisely the two things a single large aperture struggles to deliver.
The mechanism reverses: scooping, and slump on separation
None of the three characteristic large-aperture failures can be explained by insufficient release.
One: squeegee scooping. The squeegee sweeps the foil at an angle. Over a small aperture it merely packs the paste and passes on. Over a large aperture its leading edge dips into the opening and carves a trough out of the middle of the deposit, like a spoon. What prints is not a flat brick of paste but a deposit that is thin in the centre and thick at both ends. Soft rubber squeegees do this more readily than metal blades, and the longer the aperture is along the direction of travel, the worse it gets.
Two: slump on separation. The moment the stencil lifts, a large block of paste loses its side-wall support and its own weight spreads it outward. The printed outline ends up visibly larger than the designed aperture, eating into the clearance to neighbouring pads.
Three: trapped volatiles. During reflow, flux volatiles in the centre of a large continuous paste block have the furthest to travel to escape, so voids concentrate in the middle rather than at the edges. SKPCB's PCBA line states solder voiding as below 10% under nitrogen reflow — a conditional process capability, not an unconditional promise for any pad geometry. Reaching that level requires the paste on a large pad to be divided into blocks with escape paths, not left as one slab.
Line the three up and the conclusion is plain: every large-aperture problem comes from one block of paste being too big and too unconstrained. The fix is to make it smaller and give the constraint back. That is what webs do. A web is not primarily there to reduce paste volume (although it does reduce it somewhat); it is there to restore side walls and to open channels for volatiles.
The decision: when splitting is mandatory and when it is not
Decide with three questions, none of which starts with a dimension.
First, does this pad need controlled joint thickness and voiding? Thermal pads and power-device base pads do, because the joint is both the heat path and the mechanical attachment, and voids drive thermal resistance directly. These are split almost without exception. Conversely, purely mechanical hold-down pads and large copper areas that carry no electrical or thermal duty are insensitive to joint uniformity and may be left whole or treated only with perimeter reduction.
Second, is there a device on top, and can it float or shift? Devices with an exposed base pad (QFN, power parts) sit down onto molten paste during reflow. Uneven paste distribution tilts them and can pull nearby fine leads open. SKPCB's assembly capability covers components down to 01005, BGA at 0.2 mm minimum pitch and QFN/QFP at 0.3 mm minimum pitch — the finer the pitch, the more sensitive the neighbourhood is to paste spreading off an adjacent large pad, and the stronger the case for splitting.
Third, how long is the aperture along the squeegee's direction of travel? An aperture that is long in the travel direction — a shield-frame rail placed perpendicular to the blade is the classic case — scoops far more readily than a square aperture of the same area. Sometimes getting the split direction right is the entire fix.
One criterion that often gets overlooked: clearance to neighbouring pads. If a large pad sits right next to fine-pitch leads, the consequence of paste spreading is not a slightly thicker joint, it is a bridge. In that situation the pad needs splitting and perimeter reduction to bound the paste, even if the pad itself has no joint-thickness requirement.
Method one: bar webs (single-direction split)
The simplest split: cut one large aperture into several parallel strips along one axis, leaving foil webs between them.
What it solves — mainly scooping. With webs arranged across the squeegee path, the blade finds foil again after every strip and its leading edge cannot keep diving into the deposit. Each strip of paste is also narrower, so slump on separation is reduced.
When to use it — elongated pads (the straight rails of a shield frame), narrow high-current connection areas, and pads where scooping is the issue and outgassing is not a concern.
Watch out for — web direction has to be decided together with squeegee direction, not simply cut along the pad's long axis. In real production the squeegee direction can change with panel orientation, so a split scheme must be fixed together with the panelisation, never on the single board alone. Also remember that bar webs constrain only one axis: the paste is still long in the other direction and the escape path for volatiles is one-directional.
Method two: grid webs (two-direction split)
Cut the large aperture with webs in both axes, producing a chequerboard array of small apertures.
What it solves — this is the one method that addresses all three failure modes. Each small aperture has metal on four sides, so the paste regains full side-wall constraint; slump is confined within a single cell; and most importantly the webs become escape channels for flux volatiles during reflow, letting gas out between coalescing paste blocks instead of trapping it in the middle of a slab.
When to use it — thermal pads, power-device base pads, and any large solder area where voiding must be controlled. This is the standard treatment for a QFN centre pad.
Watch out for — grid splitting reduces total paste deposit, since open area equals pad area minus the web area. You need to know how much solder that pad actually requires, because over-splitting leaves the joint too thin, raising thermal resistance or weakening the attachment. Grid density is therefore not a "denser is better" parameter but a balance between constraint-and-outgassing on one side and total paste volume on the other. Where that balance sits depends on device type, pad size and foil thickness, and must be settled jointly by the stencil supplier and the assembly side for the specific pad. There is no universal ratio to copy.
Grid symmetry matters too. An asymmetric grid biases paste to one side and tilts the device during reflow. The grid centre should align with the pad centre; if the leftover cells at the edge come out too narrow, merge them into the adjacent cells or handle the edge with uniform perimeter reduction instead.
Method three: perimeter reduction and outline inset
The third approach does not cut the interior at all. It pulls the whole aperture outline inside the pad outline, or keeps an unopened ring of foil around the pad.
What it solves — paste spreading and bridging risk. With the aperture smaller than the pad, even paste that slumps still lands within the pad area.
When to use it — when a large pad sits next to fine-pitch leads; when the relationship between solder-mask opening and copper edge is awkward and paste tends to climb the mask dam; and when you simply need less paste without changing how it is distributed.
Watch out for — the inset amount is a joint decision between stencil supplier and assembly process, depending on paste type, foil thickness, mask opening style and device requirements. There is no value that transfers across products. This article gives no inset figure, and any claim of a single universal inset should be treated with suspicion, because it ignores differences in paste rheology and mask opening style.
In practice the three methods coexist on one stencil: a shield frame may use bar webs on its straight rails and outline inset at the corners; a large thermal pad may use a grid and also sit inset from the pad outline overall.
Splitting vs thinner foil vs step stencil
When a large pad carries too much paste, there are three directions available, and they are not interchangeable.
Splitting (webs/grid) changes the paste's planar distribution: at the same thickness, one block becomes several, with some volume reduction as a by-product. It is the only method that creates outgassing channels, and the only one that addresses scooping and slump at the same time.
Thinning the whole foil changes paste thickness for every pad on the board. It works for the large pad, at the cost of thinning fine-pitch and small components too — and those apertures may already be near their release limit. When a board carries both large pads and fine-pitch devices, uniform thinning is usually not the answer.
A step stencil creates regions of different thickness on one foil, thin where the large pads are and thick where fine pitch is. It resolves the "one stencil, two requirements" conflict, but introduces printing problems at the step boundary: the squeegee briefly loses contact crossing the step, print quality for pads near the step needs separate assessment, and step-region layout is constrained by squeegee direction and step spacing. Step stencil process detail is covered separately (see links); here we only set the selection boundary.
The practical order is: use splitting first to fix distribution and outgassing; consider local thinning only when splitting cannot bring total volume down far enough; and reach for a step stencil only when the board genuinely has two widely separated thickness requirements that splitting cannot reconcile. Reverse that order — step stencil first, splitting later — and you usually end up with a stencil that is both expensive and hard to print.
What to monitor once the split is in production
The split scheme is decided on a drawing, but whether it is right shows up on the line in three places.
Post-print paste shape and volume (SPI). Does every small aperture produce a formed deposit, is there merging or dropout, does the total volume per pad land where expected. The first run of a new split scheme must be reviewed against SPI data, not visual inspection alone.
Post-reflow voiding (X-Ray). Whether the grid really opened escape channels is an X-Ray question: look at void distribution and proportion under the large pad. If voids still concentrate at the centre, the grid is too coarse or its orientation is unfavourable for outgassing. SKPCB's PCBA line runs SPI, inline 2D AOI, post-reflow 3D AOI, X-Ray, ICT and FCT; verification of a large-pad split scheme leans mainly on the SPI and X-Ray stages.
Device position and coplanarity. Does the device tilt or shift after reflow, do adjacent fine leads bridge or starve because of paste spreading off the large pad. The larger the part, the more this matters — SKPCB's assembly capability covers large BGA/QFP up to 74 mm diagonal, and at that scale uneven base-pad paste magnifies into visible coplanarity problems.
Acceptance follows IPC-A-610 Class 3, with the bare board accepted to IPC-6012 Class II & III, and the flow carries 10+ quality gates. One qualification matters though: these standards define joint acceptability, not the correctness of a stencil split scheme. A split can pass visual acceptance and still leave a thermal-resistance or long-term-reliability issue behind. That is exactly why large-pad splits need SPI and X-Ray evidence rather than a single visual sign-off.
What your order must include
If the stencil supplier receives a paste layer that is just a copy of the pad layer, they are left guessing how your large apertures should be split. To avoid that, make the following explicit when you order:
- Which pads are functional large pads — identify thermal pads, power pads and shield frames, and state what each one does (heat, current, shielding, mechanical). Function decides whether and how far it can be split.
- The device part numbers and base-pad construction — exposed pad or not, joint thickness requirement, voiding requirement.
- Any existing split scheme — draw the webs and grid into the paste layer rather than describing them in text. Text notes get lost easily in multi-party handoffs.
- If you have no scheme, authorise the stencil supplier to propose one — and give them the function and constraints of each large pad so the proposal has a basis.
- Panel orientation and squeegee direction agreement — split direction depends on squeegee direction, and rotating the panel can turn a correct bar-web scheme into the worst possible one.
- Location of adjacent fine-pitch devices — which large pads sit next to 0.2 mm pitch BGA or 0.3 mm pitch QFN/QFP. Paste boundaries at those locations must be tightly bounded.
- Inspection requirements on the assembly side — whether full SPI is required, whether large pads need X-Ray sampling, and the source of the void acceptance criterion (for example IPC-A-610 Class 3).
Paste-layer delivery format, origin definition and the completeness checklist for stencil data are a separate topic (see links). The single point here is this: the split intent has to travel with the files to the stencil supplier, not stay in the designer's head.
Related reading
- SMT stencil service
- Step stencil multi-level aperture design
- Stencil aperture area ratio and paste release
- Stencil aperture position accuracy and fiducial registration
- PCBA and assembly capability
- Quality system and inspection capability
Image alt: A stainless steel SMT stencil foil lying flat on a machining table with a processing/inspection head above it; the foil already shows cut pad arrays and component outlines, with solid metal webs left between several large square apertures.
---
FAQ
Does every large pad have to be split?
No. The criterion is function, not size. Pads that carry heat or current, have a device on top, or sit next to fine-pitch leads need splitting. Purely mechanical hold-down pads and non-soldered copper areas usually do not, or need only outline inset.
Is a denser grid always better?
No. The denser the grid, the more area the webs consume and the less paste is deposited. Over-splitting leaves the joint too thin, hurting both heat conduction and mechanical strength. Density is a balance between constraint-and-outgassing and total volume, set jointly by the stencil supplier and assembly for the specific pad.
How do I choose between bar webs and a grid?
Bar webs when the issue is scooping, the pad is elongated, and outgassing is not critical. A grid when voiding must be controlled, the pad is roughly square, and a device sits on top. Thermal pads and power-device base pads normally take a grid.
Does splitting weaken the joint?
Splitting reduces paste volume, not solder area. The deposits coalesce during reflow and no gap remains where the webs were. The strength risk comes from over-splitting leaving the joint too thin, not from splitting as such — which is why you first establish how much solder the pad needs.
A stencil is already made and the large apertures were not split. Can it be rescued?
Stencil manufacture is subtractive: foil can be removed, not added back. So webs cannot be added to an existing stencil afterwards. In the short term the printing process can mitigate (squeegee type and parameters, panel orientation), but that is a stopgap. Splitting is a stencil structural matter and ultimately requires a new stencil. This is why large-pad splits must be confirmed before the order goes out.
Can a split scheme be reused across projects?
With the same device, the same foil thickness and the same paste system it serves as a starting point, but it still has to be re-checked against this board's panel orientation and neighbouring devices. Do not treat one project's grid parameters as a general rule.