Close

The Real Tolerances Behind Class 3 PCB Fabrication

A project log for Five Tolerances That Matter in Class 3 Fabrication

Class 3 PCBs fail when real fabrication tolerances stack up. Drill wander, plating growth, and lamination movement decide yield.

ryan-oconnorRyan O'Connor 11/20/2025 at 16:310 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:

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:

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:

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:

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:

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

This matters when:

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:

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 of actual physics and process limits.

Standard layout tools do not model:

Class 3 reliability comes from designs that respect these real process limits.

A Practical Checklist for Class 3 Designers

If you want a Class 3 design to move through fabrication with high yield and predictable results, these habits help a lot:

Most Class 3 issues are not pure schematic or layout mistakes. They are tolerance budget mistakes that only show up once material and process variation are added to the design.

Closing Thought

High reliability boards do not live inside the layout tool. They live in heat, pressure, chemistry, mechanical force, and real materials that move inside real limits.

A clean DRC run is still important, but it is not the finish line. It is the starting point for a chain of steps where small process variations can accumulate.

Understanding those tolerances, especially the ones that never appear on the designer screen, is what separates a Class 3 design that only passes a rule check from a board that also survives fabrication and testing with predictable yield.

If you have seen different behavior in your own builds or want to discuss specific cases, add a comment below. I am always interested in how these effects show up in other environments.

Discussions