High-Speed PCB Quality Control: Standards and Acceptance

Quality control on a conventional board is built around visual and electrical criteria: no shorts, no opens, no exposed copper and a plating thickness within the specification. On a high speed board those criteria are still necessary but they are no longer sufficient, because a board can meet all of them and still fail to pass a signal.

The additional requirement is that the electrical properties the design assumed are the ones the finished board actually has. That moves the emphasis from inspecting the product to controlling the process, and it changes what has to be measured and recorded.

What Changes at High Speed

At low frequency the only electrical properties that matter are continuity and isolation. As the frequency rises, the impedance of every line, the loss of the material, the skew between the two halves of a differential pair and the behaviour of every via transition begin to matter.

Those properties are not visible to a visual inspection and they are not measured by a continuity test. They are controlled by the stackup, the artwork compensation and the fabrication tolerances, and they are verified by a coupon that is built alongside the product.

<img src="https://www.gopcba.com/wp-content/uploads/2020/05/steel_product1.jpg" alt="Impedance coupon measured on a high speed panel” />

Impedance Control and Coupons

The impedance coupon is the primary control. It carries traces of the same width and on the same layer as the product, and it is measured with a time domain reflectometer after fabrication. The result is compared with the specification and with the tolerance the design allows.

The coupon has to be representative. A trace on a coupon placed in a sparse area of the panel is etched and laminated under different conditions from a trace in a dense product area, so the coupon is either placed beside the product or designed with a similar copper density.

Material Verification

The laminate properties that the design assumed, particularly the dielectric constant and the dissipation factor, are verified from the supplier certificate and, where the requirement is critical, by a coupon measurement. A substitution of one grade for another changes the impedance and the loss, and it has to be agreed rather than discovered.

The copper foil is part of the same verification. A foil with a different surface roughness changes the conductor loss, so the foil type specified in the stackup is confirmed rather than left to the fabricator to substitute.

Registration and Layer Alignment

Layer to layer registration affects the impedance of a reference plane structure and the reliability of every via. The measurement is made on a test coupon that includes features on several layers, so that the offset between them can be quantified.

As the layer count rises, the cumulative error grows and the registration budget becomes tighter. High layer counts therefore need more accurate lamination tooling and more frequent measurement, which is one of the reasons a high layer count board costs more to control.

Surface Finish and Plating Quality

The barrel plating thickness affects the impedance of a via slightly and its reliability significantly. The specification gives a minimum thickness at the centre of the barrel, and it is measured by cross section rather than by an indirect method.

The surface finish affects the loss of a trace that runs on the outer layer, because the finish is part of the conductor. A thick finish or a rough one adds a small amount of loss, which matters only in the tightest designs but is nevertheless part of the electrical specification.

Inspection at Each Stage

High speed boards are inspected at more stages than a conventional board. The inner layers are checked for line width and spacing against the compensated artwork, the panel is checked after lamination for thickness and registration, the drilled holes are checked for position and size, and the finished board is checked for impedance and for the final dimensions.

Each of those checks produces a record. The value of the record is that it allows a drift to be detected before it produces a reject, and it allows a later failure to be traced to the operation that caused it.

Acceptance Criteria and Sampling

The acceptance criteria should be written in terms of measurable quantities: an impedance target with a tolerance, a minimum barrel thickness, a maximum registration offset and a defined coupon location. Where the criteria are qualitative, two inspectors will reach different conclusions.

Sampling follows the criticality. The impedance coupon is measured on every panel or on a defined frequency, the barrel thickness on a sample per lot, and the dimensional checks on the finished board. The frequency should be justified by the process capability rather than by habit.

Documentation and Process Capability

The documentation package for a high speed board includes the stackup, the impedance targets and tolerances, the coupon design and the measurement results. Where a customer requires traceability, the panel and lot identifiers are recorded with the measurements.

gopcb builds high speed boards with coupon based impedance verification, measured barrel plating, controlled material lot tracking and the inspection records that a qualification programme requires.

Coupon Design and Placement

The coupon is only useful if it describes the product. It should be built on the same layers as the critical traces, with the same copper weight, the same dielectric build and a similar copper density around it, and it should be placed on the panel where the process conditions match the product area rather than in the quiet corner that is convenient for the fabricator. Where a panel carries several different products, each with its own impedance target, the coupon set is designed accordingly rather than shared.

The coupon also has to be readable. A time domain reflectometer measurement needs a launch that does not disturb the line, enough length for the instrument to resolve the impedance and an unambiguous reference plane under the whole structure. A coupon that is too short, that lacks a ground reference or that places the launch pads asymmetrically produces a measurement that describes the coupon rather than the board, and the resulting argument costs more time than the coupon saves.

Trending and Continuous Improvement

A single measurement confirms that a panel is inside its tolerance; a series of measurements over time shows whether the process is drifting. The etch rate, the lamination pressure and the plating current all move slowly, and the trend is visible on a chart long before a panel is rejected.

That trend data is what allows the tolerance to be tightened or relaxed with evidence. Where a process has run in the middle of a narrow band for months, the design can rely on a tighter tolerance, and where it wanders through the band, the design has to assume the worst case. The data therefore feeds back into the design rules rather than only into the production record, which is how a fabricator and a customer improve a product together rather than arguing about individual lots.

FAQ

Why does a board pass electrical test and fail a signal test? Because electrical test verifies continuity and isolation, while a signal test is affected by impedance, loss and skew. A board can be perfectly connected and still have the wrong impedance on a critical line.

How often should the impedance coupon be measured? On every panel for a critical design, or at a defined interval where the process has been shown to be stable. The decision should follow the process capability rather than a fixed rule.

Does the surface finish affect impedance? It affects the loss of an outer layer trace, and it changes the thickness of the conductor slightly. The effect is small but it is part of the electrical specification on a tight design.

Related reading: impedance discontinuity analysis, PCB yield and quality control, PCB manufacturing tolerances, and high frequency laminate selection.