Project Logs

Collapse

• CAM Adjustments That Prevent Class 3 Yield Loss

  Ryan O'Connor • 13 hours ago • 0 comments

  This update adds a brief look at how CAM engineers adjust layout data before fabrication.

  Pad growth, mask expansion, trace width compensation, and drill-offset budgeting are applied to prevent annular ring loss, copper exposure, and breakout in Class 3 builds.

  A simple before and after diagram is included to illustrate how these corrections stabilize yield.

• The Real Tolerances Behind Class 3 PCB Fabrication

  Ryan O'Connor • 4 days ago • 0 comments

  Five Tolerances That Matter Most in Class 3 Fabrication

  PCB CAD tools show fixed numbers and perfect geometry. Real Class 3 fabrication does not work that way. These are the five tolerance behaviors that most often decide whether a high reliability board builds cleanly or turns into scrap.

  1. Drill Wander

  Drill bits do not travel in perfect straight lines. They can walk or bend slightly depending on:

  • total stack height
  • glass weave pattern
  • resin content
  • entry material
  • drill wear and sharpness
  • via aspect ratio

  On a pad that is 0.010 inches, even 0.0015 inches of drift can be enough to break a Class 3 annular ring requirement.

  If a stacked microvia sits on top of that through hole, the allowed window gets even smaller. Now several features need to land inside one narrow target at the same time.

  In CAD, the hole always appears exactly centered. In fabrication, the hole is centered within a distribution that has a real width.

  2. Plating Growth

  After drilling, every hole grows copper during plating. The growth amount depends on:

  • barrel plating rate
  • copper balance across the panel
  • tank loading
  • current density and field variations
  • via density and local pattern
  • panel position inside the rack

  All barrel plating makes the drilled hole smaller in the final state. That means you do not really get the drilled size. You get the drilled size minus the copper added during plating.

  In Class 3 work, where minimum finished hole sizes are tight, this affects:

  • press fit pins
  • stacked or buried via structures
  • component leads that must meet contact rules
  • test coupons
  • impedance and reference vias

  A finished hole that is only half a mil smaller than intended can push a connector or a critical via out of specification.

  3. Lamination Movement

  During lamination cycles, several things are happening at once:

  • resin flows and fills gaps
  • copper expands and contracts with temperature
  • prepreg softens and then sets
  • layers can shift relative to tooling pins
  • glass weave can steer movement in preferred directions

  On a simple two layer board, total movement may stay small. On a Class 3 rigid flex or complex HDI stackup that goes through several lamination cycles, movement can accumulate.

  Registration rarely drifts in a random way. It tends to move more in one direction or along one axis due to how the materials and press setup behave. That is why alignment that looks perfect in CAD does not always align that way on a real panel.

  4. Etch Factor and Copper Taper

  Etching does not produce vertical copper walls. When you look at a cross section from a Class 3 build, you see that sidewalls are angled. A trace might measure:

  • about 5 mils wide at the top
  • closer to 3.6 mils at the base

  The numbers depend on copper thickness, resist, and etch chemistry.

  This matters when:

  • impedance must stay inside a tight window
  • differential pairs must track closely
  • copper balance is uneven across the panel
  • long runs sit in regions that etch slightly differently

  In high reliability work, this sidewall shape is one of the hidden reasons why impedance shifts away from the value that was simulated.

  5. Material Reality

  Material datasheets show clean values for dielectric constant, loss, and thickness. In actual Class 3 production, you see:

  • dielectric constant variation across a panel
  • resin rich and glass rich regions
  • weave skew that pulls fields off the ideal path
  • local thickness differences between inner layers
  • behavior that changes with temperature and process conditions

  A trace that was designed to be 50 ohms with perfect modeling might come out around 46 ohms in one area and 52 ohms in another. You cannot model all of this in the design tool, but you can design with a realistic guard band.

  Why Passing DRC Does Not Guarantee Class 3 Success

  DRC checks for rule violations in the digital description of the board. Fabrication checks for violations...

  Read more »

View all 2 project logs