In production we treat HDI solder mask expansion as a hard process window, not a design preference. CAM engineers normally set the opening expansion to 0.025–0.05 mm per side for most HDI builds and reject any value that falls outside the factory's registration capability for that layer stack. When the incoming Gerber shows expansion smaller than 0.02 mm or larger than 0.075 mm on fine-pitch pads, the file is immediately flagged for clarification or automatic compensation before panelization.
The expansion number itself is simply the radial clearance between the copper pad edge and the solder-mask opening edge. On HDI boards this single dimension controls whether the mask will stay clear of the pad under real registration shift or whether it will leave an unprotected copper ring that invites bridging.
Where the expansion number starts fighting the registration budget
HDI panels run through laser direct imaging or traditional film exposure with residual registration error of ±0.025 to ±0.04 mm after lamination and scale compensation. The solder-mask process adds its own contribution from coating thickness variation and development undercut. When the designer calls for an expansion of only 0.015 mm on a 0.25 mm BGA pad, the remaining process window is already smaller than the expected shift. The opposite extreme—expansion of 0.1 mm or more—eats into the solder-mask dam between pads that are already spaced at 0.15–0.2 mm. In both cases the CAM station sees a geometry that cannot survive the real panel movement that occurs after the artwork is fixed.
Material behavior compounds the problem. Thin core and high-Tg prepreg used in sequential HDI build-up expand and contract differently from the base laminate, so the scale factor calculated at the start of the lot is never perfect. The result is a systematic offset that the expansion value must absorb.
What the line sees when expansion is left uncontrolled
Too little expansion produces solder-mask residue on the pad surface after development. On HDI this residue is often only a few microns thick and hard to detect optically, yet it prevents proper wetting during assembly and shows up as open joints or high void rates. Too much expansion removes the dam between adjacent pads. When the remaining mask bridge drops below 0.05–0.06 mm, the liquid solder mask can flow and create a short after reflow, or the AOI simply rejects the panel for insufficient dam width.
Yield impact is immediate. A single 0.4 mm pitch BGA area with incorrect expansion can scrap 8–12 % of the panels in a lot because the defects are concentrated and cannot be reworked once the mask is cured. Shipping delays follow because the lot has to be stripped and re-coated or, in the worst case, the entire panel is scrapped and a new lot started.
How CAM and process actually keep the expansion inside the window
CAM first measures every pad diameter and the nearest neighbor spacing on the critical layers. The expansion value is then forced into the factory's process capability table: for 0.2–0.3 mm pads the opening is set to pad size plus 0.05 mm total (0.025 mm per side); for pads below 0.2 mm the value is tightened to 0.04 mm total and the registration tolerance is tightened by using higher-resolution LDI and tighter scale control. If the designer's original expansion is outside this band, CAM either overrides it with a note to the customer or generates a compensated Gerber that restores the dam while still clearing the pad under worst-case shift.
Panelization adds another layer of control. Critical HDI areas are kept near the panel center where registration is tightest; coupon strips with the same pad geometry are placed at the edges so that the first-article measurement of actual expansion can be fed back before the full lot is released. Development parameters are locked to a fixed undercut target of 0.01–0.015 mm so that the final opening size stays predictable.
Inspection closes the loop. Automated optical measurement of pad-to-mask clearance is run on every panel; any reading outside ±0.02 mm of the programmed expansion triggers a hold. This is why the CAM review always includes a "minimum dam after expansion" check—once the expansion is applied, the remaining mask bridge must still exceed 0.075 mm for most HDI processes.
When the factory will accept a non-standard expansion
Relaxation is allowed only when the design uses pads larger than 0.4 mm or when the customer explicitly accepts a lower registration yield and signs off on the risk. In those cases CAM can open the expansion to 0.08 mm per side, but the panel must still carry a process note that the lot will be run under tighter scale control and 100 % AOI for dam width. For true HDI fine-pitch areas the window remains closed; the process simply cannot guarantee both pad clearance and dam integrity if the expansion is left outside the normal 0.025–0.05 mm band.
The practical rule on the floor is simple: if the expansion value forces the remaining dam below the process minimum or leaves the pad clearance smaller than the expected registration error, the file is returned. That single check prevents the majority of solder-mask related scrap on HDI boards.
