The EQ does not start with a solder bridge on the assembled board. It starts on the solder mask layer, between neighboring SMT, fine-pitch IC, or BGA pads.
Designers often call that strip a solder mask bridge. In review it is the solder mask dam: the intact web of mask left between adjacent solder mask openings. The land pattern usually assumes one opening per pad and a dam between them so reflow solder stays on each land. A CAD polyline is not evidence that the dam will still be there after imaging. Copper and solder mask have to be overlaid, and the remaining web after the openings are applied is what fabrication can attempt to build.
If that remaining width is only a sliver, or the openings have already merged, the dam may not survive imaging and develop. That geometry is the finding. It is not yet a solder-bridge defect. It is a mask feature that may be absent after process. The usual review notes describe the same object in three places: SMT or fine-pitch IC pad spacing too tight to form a dam; pad-to-pad spacing too small for a dam to be made; a BGA field where the dam cannot be retained. Closing the EQ is a build choice on this part or region: keep a manufacturable dam, or release a gang solder mask opening.
When the Drawn Solder Mask Dam Will Not Survive Imaging
The object under review is the solder mask dam between neighboring SMT pads, fine-pitch IC pins, or BGA lands.
The incoming package contains copper lands plus solder mask openings. A dam drawn in the footprint library is not a held dam in fabrication. That strip has to image and develop onto laminate as a continuous web. A sliver can break, lift, or fail to form after registration. Once that happens, the board no longer matches the individual openings the land pattern assumed.
Two states on the same land pattern trigger the question:
- Separate openings still exist in the customer solder mask layer, but the remaining web is too thin to treat as a real dam.
- Openings have already merged in the released mask layer, so there is no dam left to keep.
The EQ is raised on that geometry. It does not claim that solder will short on the first article. It claims that the mask feature between those pads may not exist after imaging.
Pad-to-Pad Spacing Still There, Solder Mask Web Already Gone
Designer intent lives in the land pattern and the library solder mask expansion. Pitch is not the starting measurement.
The comparison on each adjacent pair runs in this order:
- Pad-to-pad spacing — copper edge to copper edge on adjacent lands.
- Remaining solder mask web — the gap between adjacent solder mask openings.
- What the mask layer actually contains — individual openings with a drawn dam, a sliver, or one gang opening.
Those two gaps diverge because openings are usually larger than the lands. Solder mask expansion, flash size, and any later mask adjustment consume the web while the layout viewer still shows copper spacing. Pitch is center-to-center. This check is not pitch. Production builds the remaining dam, not the copper pitch.
The review trigger is straightforward: copper still has spacing, but the solder mask web is already gone; or the mask still draws a sliver that is unlikely to hold. That is the condition behind "pad spacing too small, solder mask dam cannot be made."
The copper layer is compared with the solder mask layer on the same land pattern. The 3D viewer and the assembly drawing are not the source of this finding. No published minimum dam table belongs in the review. The question is whether a real web remains on this geometry.
| What the land pattern still shows | What the overlay measures on the same pair | Why it becomes an EQ |
|---|---|---|
| Pad-to-pad copper spacing still present | Remaining solder mask web already gone or reduced to a sliver | Imaging would have to form a dam the openings no longer support |
| Individual solder mask openings drawn per pin | Openings already touching or merged in the mask Gerber | The drawn dam is not in the manufacturing data |
| Fabrication note: keep solder mask dams | Mask layer already ganged, or only a sliver remains | Note and artwork conflict; neither can be assumed as the ship condition |
| Pitch looks acceptable in the library viewer | Pitch is center-to-center; web is opening-to-opening | The wrong gap is being used as the process object |
What CAM Looks At in Fine-Pitch Land Patterns
Dense land patterns are windowed first: SMT connector rows, fine-pitch IC / QFN / SOP rows, and BGA arrays.
The sequence is consistent:
- Solder mask is overlaid on copper.
- Missing, touching, or sliver webs after the openings are flagged.
- Mixed treatment on one part is caught — some pairs still dammed, neighbors already ganged.
- Fabrication notes that say "keep solder mask dams" are checked against a Gerber that has already merged openings, or the reverse.
A dam that the openings have already deleted cannot be restored from a note. A sliver is a poor candidate for a held web. The files also do not show whether SMT will accept a gang opening on that part, whether the stencil assumed a dam as a gasket, or whether a particular pin pair is critical. These are the kinds of PCB solder mask and via design issues that CAM review commonly finds, and the missing decision is the EQ rather than a silent mask edit.
Released mask polarity and expansion are part of the same window, along with whether a dam is present in the customer solder mask layer or only implied by the land pattern. All three EQ locations — SMT/IC rows that cannot form a dam, spacing too small to make a dam, and a BGA field that cannot retain a dam — are the same overlay on different packages.
| Overlay check | What is being compared | What is not used as the check |
|---|---|---|
| Solder mask on copper | Remaining web between openings vs copper pad-to-pad spacing | Layout 3D view or rendered assembly preview |
| One part, all adjacent pairs | Same land pattern treated the same, both sides if the part appears twice | A single representative pin pair |
| Customer mask layer vs note | Dam actually drawn vs a note that says keep dams | An implied dam from the footprint library alone |
| Expansion already in the file | Openings as released, including any library or later mask adjustment | A dam expected to reappear after develop |
When an SMT or Fine-Pitch IC Row Cannot Hold the Dam It Draws
On these rows, SMT means surface-mount component pads, not solder-mask-defined lands. The land pattern draws one solder mask opening per pin, so a dam is assumed.
Pad-to-pad spacing on the tight pairs is often small enough that after expansion there is no web, or only a sliver. Two file states show up on the same class of part:
- Spacing cannot form a dam. The usual wording: solder mask dam cannot be made; a gang solder mask opening is proposed.
- The file still draws a dam that fabrication is unlikely to produce.
If that second state is built as a nominal dam, the result is often incomplete ink, lifted slivers, or an irregular gang opening after develop. The board then matches neither the customer plot nor a clean gang opening.
The tightest adjacent pair on that part is the review location, both sides if the same land pattern appears twice. This type of early design-for-manufacturing review can help prevent SMT shorts from narrow mask bridges before the board reaches assembly. This is not an IPC-7351 land-pattern redesign exercise. It is a mask-web decision on the released pair.
Those two SMT/IC notes are the same finding: pad spacing insufficient to form a solder mask dam, with a gang opening proposed; and pad spacing too small for the dam to be made.
BGA Arrays Where the Dam Cannot Be Retained After Expansion
Many BGA land patterns are NSMD: the opening is larger than the copper land, with an intended dam between balls. Pad-to-pad spacing already consumes most of the copper gap. Released or adjusted openings then collapse the web across the array, or leave dams only on the outer ring.
One global expansion hits the whole matrix. If a dam is required, the web still has to exist after the openings are large enough that misregistered mask does not encroach on copper. If that combination is not in the data, the field becomes a gang opening whether the drawing wanted dams or not.
The mismatch is specific: the drawing still shows dams; the Gerber after mask adjustment does not. That is the BGA case of a dam that cannot be retained. The densest BGA overlay is the window, including whether conductors between balls would be exposed if the dams are dropped. Pad cratering and via-in-pad belong to a different review.
What Happens If the Dam Is Left to the Process
When the finding is not closed in writing, fabrication still has to image something.
A sliver may image poorly, break in develop, or be removed during mask prep. The result is a ragged web or an undeclared gang solder mask opening. Assembly then sees laminate between pads. Stencil apertures often assumed a dam as gasket. With no dam, solder can run. The later defect is a solder bridge; the cause in this cluster is a missing dam, not a separate SMT mystery. Inspection has the same problem: bare-board mask will not match the customer solder mask plot if openings were merged without a written exception.
Four views have to stay separate:
- The designer wanted individual openings.
- Only dams that can be held are a reliable ship condition.
- The factory ships the released mask layer.
- The customer has to accept that condition.
If a paste layer is present, per-pad paste against a ganged mask opening may be mentioned in the EQ. That is a consistency note, not a second process change made on its own. All three source findings, left unanswered, are the same uncontrolled loss of the dam.
Retain the Solder Mask Dam or Release a Gang Opening
Typical reviewer wording is direct: the solder mask dam between these SMT, fine-pitch IC, or BGA pads cannot be formed or retained; confirm a gang solder mask opening on this part or region, or revise land and mask data to keep the dam. Similar 8-layer blind-via PCB DFM cases involving solder mask and NSMD BGA issues also require the final decision to align the released artwork with the actual manufacturing requirements.
The question is not a published capability number. It is a build decision on this geometry.
What needs to come back:
- Is the dam required for SMT on this part?
- If a gang opening is accepted: the whole part, one row, or only the failing pairs?
- Do notes that said "keep solder mask dams" get superseded here?
Normal close paths:
- Accept a gang opening on the named parts. Those openings are merged, the exception is noted, and the job is released.
- Require the dam. Solder mask expansion is reduced, or the land and opening are trimmed, until a real web remains. The overlay is repeated.
- Mixed. Dams stay where the web is real; a gang opening is used only on failing pairs or the BGA field. That mix has to be written.
"Open window" in shop shorthand maps to gang solder mask opening. It is not a new process. All three EQs close the same way. The written choice is then checked against the edited solder mask and copper on those exact locations.
| Written close | What happens to the mask layer | When this close fits |
|---|---|---|
| Accept gang opening on the named part or row | Merge those openings, record the exception, release | Dam is not required for SMT on that geometry |
| Require the dam | Reduce expansion or trim land and opening until a real web remains; overlay again | Dam is required and the data will change |
| Mixed: keep real webs, gang only failing pairs or the BGA field | Edit only the named pairs or region; leave the rest | Must be explicit; an unspoken mix is an undeclared gang opening |
| Note "keep dams" left standing with no geometry change | No safe edit | Artwork and note still conflict; the EQ stays open |
Solder Mask Overlay Checks Before the Files Leave CAD
Open the released solder mask Gerber over copper. The layout 3D view is not a substitute.
On every fine-pitch SMT row, IC row, and BGA:
- Openings should still be individual if a dam is intended.
- Measure the remaining solder mask web, not only pitch and not only pad-to-pad copper.
- If the web is already gone in CAD, fabrication will not grow a dam back.
After a global library expansion, the tightest land pattern is the one that will be windowed in review. If a dam is needed, the land pattern or the expansion has to change before release. If a gang opening is acceptable, that should be stated in the fabrication notes so the EQ is not the first time the decision appears.
FAQ
Q1: The note says the solder mask dam cannot be made. Is that the same as a solder bridge?
A1: No. The EQ is a mask-feature finding: the web between openings may not exist after imaging. A solder bridge is an assembly defect that can follow if that dam is missing and solder runs. The mask condition has to be frozen first.
Q2: The layout still shows pad-to-pad spacing. Why is the dam gone?
A2: The remaining solder mask web between openings is the process object, not copper edge-to-edge, and not pitch. Expansion and mask adjustment consume the web while copper spacing is still visible.
Q3: What does "open window" mean on this EQ?
A3: Gang solder mask opening on the named part, row, or BGA field. It is an accepted merge of adjacent openings, not a separate fabrication process.
Q4: Can the drawn dam be kept if the note says "keep solder mask dams"?
A4: Only if a real web still exists after the openings. A note does not hold a sliver, and it does not restore openings that have already merged in the Gerber.
Q5: Is an IPC-7351 footprint redesign required to close this?
A5: Not for this finding. The close is retain a manufacturable dam on this pair, or release a gang opening. Land-pattern redesign is a separate design task.
Q6: If a gang opening is accepted, does that apply to every pin on the part?
A6: Only if the reply says so. Scope has to be stated: whole part, one row, failing pairs only, or the BGA field. Mixed treatment without a written scope is how an undeclared gang opening ships.
Q7: The BGA drawing shows dams between balls. Why is the field called out?
A7: One expansion is applied across the array. If the remaining web does not survive that opening size, the drawing and the mask data no longer match. Outer-ring dams are not treated as proof for the whole matrix.
Q8: Should paste apertures change if the mask is ganged?
A8: Per-pad paste against a ganged mask opening may be flagged in the EQ. Whether stencil data changes is an SMT decision that has to come back with the mask decision. Paste is not silently edited to match an undeclared gang opening.
Q9: What is re-checked after the reply?
A9: The written choice, then solder mask versus copper on the exact SMT row, IC row, or BGA overlay that was raised. The EQ is closed when the released mask layer matches that choice.
