Bottom line: CO2 laser blind vias seldom fail because the beam cannot punch through. They fail at the via bottom immediately afterwards. When any one of three variables — single-pulse energy, pulse count, or focal plane — drifts by more than 15%, a 5–15 μm resin residue is left behind. AOI does not see it, copper plating covers it instead of removing it, and the joint separates at the via bottom after three thermal excursions. Locking dielectric thickness, energy settings and cross-section criteria into one interlocked condition set is the first gate on HDI stacked-via reliability.
1. The energy window is bounded on three sides
The usable range is not a single point but a band clamped by three boundaries. Break any one and the via geometry starts to distort.
- Lower bound — under-energy: the dielectric is not fully penetrated and a resin film of 10 μm or more remains at the bottom. On a cross-section this reads as a flat via bottom with no reflection from the inner-layer copper; after plating it shows up as high resistance or an open.
- Upper bound — over-energy: the beam passes through the dielectric and keeps going into the inner-layer copper, burning a crater. Once the crater exceeds 30% of copper thickness, the current-carrying cross-section of that trace is already reduced, and the recast copper at the crater rim becomes a stress riser.
- Side bound — focal plane drift: a 50 μm variation in panel flatness moves the focal plane off optimum. Hole diameter can then differ by 15 μm across one panel and taper falls from 0.85 to 0.70.
The practical method is an energy ladder. Hold pulse width at 8–20 μs, step single-pulse energy from 3 mJ to 10 mJ in 1 mJ increments, fire 2–4 pulses at each step, and cross-section every step to measure bottom residue and copper crater depth. Take the band where residue is ≤ 5 μm and the crater is ≤ 10% of copper thickness, then move 20% toward the middle as the production setpoint. The margin that remains absorbs panel thickness variation and machine ageing.
2. Two distinct residue sources need opposite fixes
The same smear on a cross-section can have two causes, and the corrections point in opposite directions. Misdiagnosis is the most common reason a fix makes things worse.
- Resin residue: energy is low or pulse count is short, so resin never fully vaporises. It appears as a uniform film over the via bottom with clean edges. The fix is more single-pulse energy or one extra pulse — not an added desmear step.
- Glass bundle protrusion: glass filaments in the dielectric are not fully cut, leaving point or strip features. The residue concentrates at isolated spots on the via bottom with irregular shape. The fix is higher pulse overlap or a UV laser opening step; adding energy alone only drives the crater deeper into the copper.
Diagnosis relies on cross-section plus SEM: resin residue forms a continuous film, glass protrusion shows fibre fracture. If either exceeds 5 μm, go back to the energy ladder rather than relying on downstream plasma. Plasma removes organic residue; it cannot remove glass pushed into the via bottom.
3. Taper and aspect ratio decide whether stacking is feasible
Higher blind-via aspect ratio is not automatically better. CO2 is most stable near 0.8:1; beyond 1:1 the energy density at the via bottom drops and residue control degrades quickly. UV laser reaches 1.5:1, but per-via time rises and wall roughness increases.
- Dielectric of 60–100 μm pairs with via diameter of 75–150 μm. That is the normal CO2 working zone; outside it the yield curve steepens sharply.
- Hold taper between 0.75 and 0.95. Too small a taper means an oversized opening that eats the pad; too large a taper means the bottom closes faster than plating can fill it.
- Stacked vias require the two layers to align within 25 μm centre-to-centre. That is close to the limit of the registration system, so both imaging steps must use the same datum set and the same exposure tool.
Where stacking tolerance genuinely cannot be held, staggered vias are the better trade: same layer count, better reliability, at the cost of a little routing space.
4. Acceptance needs three tests, not one
Blind-via failures are mostly delayed, so a first-article cross-section alone is not enough.
- Cross-section: three vias from the densest area and three from the sparsest area per lot, measuring bottom residue, copper crater, taper and wall roughness.
- Pull test: measure the bond between via copper and inner-layer copper, judged against the IPC-6012 Class 3 minimum wall copper and adhesion requirements.
- Thermal excursion: -55 °C to +125 °C, 100 passes, then re-section the via bottom. This is the only test that exposes a residue problem before volume.
AIPCB.cc is operated by Shenzhen LingXin Technology. Capability covers up to 44 layers, a maximum panel size of 610×1100 mm, an aspect ratio of up to 20:1 (25:1 for backplanes), impedance tolerance of ±8% (±10% standard), line width accuracy of ±8 μm, and copper weight up to 13 OZ, with 24-hour expedited prototyping. 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.