Bottom line first: mask failures around a BGA come from two independent sources — a dam too narrow to survive development, and a plugged via that voids and bursts in reflow. The first is governed by ink thickness and side-etch, the second by vacuum level and cure profile. They need separate acceptance criteria. Folding both into a single "mask appearance acceptable" sign-off is the most common starting point for lot-level rework.
The dam-width limit is set by ink thickness and development side-etch
Take a BGA on 0.4 mm pitch with 0.25 mm pads: only 0.15 mm remains between adjacent pads. Subtract the 0.05 mm per-edge opening compensation on both sides — 0.1 mm in total — and the nominal dam width left is 0.05 mm. That number survives on a drawing; on the line it is already below the process floor.
Green photoimageable solder mask runs 15–25 μm thick, and development is wet, with 8–12 μm of side-etch per edge. A 0.05 mm dam therefore loses 0.01 mm on each side and ends up at 0.03 mm, with micro-cracking across the crest. The industry generally treats 0.075 mm as the lower limit; below that the design has to move: shave 0.02 mm off the pad, switch to an NSMD pad definition, or accept a bridged opening without a dam, together with a revised stencil and flux strategy.
A second, quieter killer is uneven ink deposit. When the screen printer lays down 5 μm more at the panel edge than at the centre, the same panel shows intact dams in the middle and broken dams at the edge. AOI judges the opening outline; a break on the inner side of the opening is routinely missed. The reliable check is 50× microscopy or a cross-section of the dam crest on first article.
Opening compensation and registration tolerance must be budgeted together
- Set opening compensation at 0.05 mm per edge, larger for thicker ink. Below 0.03 mm the risk of resist residue rises, and a thin mask film left on a BGA pad causes outright non-wetting.
- Hold opening registration at ±0.025 mm. Compensation and registration are independent error directions; in the worst case they add to 0.075 mm, which consumes the dam entirely.
- Below 0.4 mm pitch, protect dam width rather than maximise the opening: 0.02 mm of extra opening buys very little in soldering, while a broken dam scraps the lot.
- Insufficient mask pre-treatment — brushing or chemical cleaning — lowers adhesion and lets the dam lift at reflow. Check adhesion by cross-hatch or tape pull, not by eye.
Three sources of voids in plugged vias
The first is trapped air. When a 0.2–0.3 mm via is capped at one end by ink during screen printing, the air has nowhere to go and becomes a chamber after cure. Vacuum plugging, pulling the cavity below -80 kPa before the ink is applied, is the main countermeasure.
The second is solvent. If the ink surface skins over during cure, high-boiling solvent is sealed inside and forms a fine bubble cluster. Ramp the cure in stages: 20 minutes at 75 °C to let solvent escape, then 100 °C, then 150 °C to complete cross-linking. Driving straight to 150 °C gives the highest void rate.
The third is side sequence. Plug side A, pre-cure, then plug side B; this prevents the second pass from pushing uncured ink out of the first. Simultaneous two-side plugging is tolerable above 0.3 mm and almost always voids on 0.2 mm microvias.
For acceptance, work as a fraction of hole depth: total void length in a single via no more than 10% of depth, and never crossing the pad centre. A void directly under a BGA pad expands at reflow and blows through the joint; on X-ray it appears as a regular round void inside the ball, easily misread as flux volatiles.
Who checks what, and what to change tomorrow
- Settle dam width at design review: put pitch, pad, compensation and registration into one calculation, list every device below 0.075 mm, and change the design device by device instead of asking the shop to force it.
- Do three things on first article: 50× microscopy of the dam crest, cross-section for via fill (target ≥ 90%), and dimensional measurement of openings against the drawing.
- Use AOI in volume for openings and shift, plus a separate X-ray sampling of plugged vias. They judge different things and cannot substitute for each other.
- Put dam width and via voids on one first-article report but in two columns, so that "cosmetically acceptable" cannot mask a functional risk.
- Repeat the first article after any ink lot change. Thixotropy and solvent system vary enough between lots to move dam width and void rate at the same time.
Solder mask is the only organic layer covering the whole board, and it carries four jobs at once: insulation, solder dam, protection and appearance. A failure in any of them only surfaces at assembly. Dam width is a design margin; via fill is a vacuum and cure-profile problem; neither can be sorted out afterwards. AIPCB.cc supports up to 44 layers, a maximum panel size of 610×1100 mm, heavy copper up to 13 OZ, line-width accuracy of ±8 μm, impedance tolerance of ±8% (±10% standard), and aspect ratio up to 20:1 (25:1 for backplanes), with 24-hour rush prototypes. For engineering and quotations, contact info@ipcba.com.cn or +86 138 2310 9230.
Send your Gerber files and stackup requirements; a 24-hour expedite channel is available.