Inner Layer Registration: Scaling, Tooling and Annular Ring

Every layer in a multilayer board is imaged on its own sheet of copper, and every sheet changes size during the processes that follow. The artwork is therefore scaled before imaging, so that the pattern arrives at the right size after the material has moved. Getting that prediction wrong is the most common cause of a reduced annular ring.

Where the Movement Comes From

Copper clad laminate absorbs moisture and expands. It also expands when heated during lamination and contracts as it cools. The film artwork used to expose the panels moves too, in response to the humidity and temperature of the imaging room.

The net result is that a panel printed at exactly one hundred per cent scale will not register after etching and lamination. The fabricator compensates by applying a scale factor to the artwork, and the factor is derived from measurements of previous panels rather than from a material data sheet.

The compensation is not uniform. Movement along the glass warp direction differs from movement along the weft, so the scale factor is usually applied separately in X and Y rather than as a single number.

Why the Tooling Holes Set the Reference

The scaling is meaningless without a reference, and the reference is the tooling system. Our tooling holes notes describe how the holes are located and how the layers are pinned together through the process.

If the tooling holes themselves move relative to the pattern, the registration measured from them is wrong even though the pattern is correct. This is why the hole positions are established early and checked against the pattern rather than assumed to be fixed.

A three point tooling system, with two holes on one edge and one on the opposite edge, locates the panel without over-constraining it. A four hole system can fight the thermal expansion of the material if the holes are tight, and the panel then distorts rather than sliding as it grows.

Measuring Registration on the Production Panel

The practical measurement is the distance from the edge of the drilled hole to the edge of the inner layer pad, evaluated on the layers with the smallest pads and at the panel corners. Modern fabrication uses automated optical measurement of these targets across the panel, producing a map rather than a single number.

The map is more informative than the average because registration errors are largest at the corners of the panel and smallest in the centre. An average that looks acceptable can hide corners where the annular ring has already been consumed.

The measurement also separates translation, rotation and scale. A pattern that is displaced uniformly indicates a tooling shift, one that is rotated indicates an alignment error, and one that is larger or smaller than nominal indicates that the scale factor needs to be revised.

Design Rules That Accommodate the Process

The design controls how much tolerance the process can consume. A larger inner layer pad relative to the finished hole size leaves more annular ring for registration error, and the trade is simply that the pad takes up routing space.

The pad is normally sized from the finished hole diameter plus a growth allowance for plating plus the required annular ring. Where the design compresses that ring to gain routing density, the margin for registration disappears.

Vias on a coarse pitch are therefore more forgiving than vias on a fine pitch, and a design that places the smallest pads on the layers farthest from the reference has the least margin of all. Our layer count notes explain how these choices interact with stack-up planning.

Scaling and the Etch Factor

Etching removes copper laterally as well as vertically, so the width of a trace after etching depends on the copper thickness and the etch process as well as on the artwork width. Artwork is compensated for this by biasing the line width, which is a separate correction from the scale factor.

The two corrections interact in the sense that both are derived from process measurements and both drift if the process drifts. Our etching process notes describe how the compensation is applied and verified.

Where a design specifies controlled impedance, the etch bias matters directly, because it changes the conductor width that the impedance calculation assumed. A coupon on the panel is the way to confirm that the width actually achieved matches the width designed.

Material Behaviour Across the Panel

Material movement is not identical everywhere on a panel. The resin flows during lamination, and the flow is influenced by the copper distribution, because copper constrains the resin beneath it more than bare laminate does. A panel with heavy copper on one side and none on the other moves asymmetrically.

This is one reason why copper balance is a fabrication concern and not merely an aesthetic one. Balancing the copper across a layer reduces the differential movement and makes the scale factor more stable from panel to panel.

Thin laminates and high layer counts accentuate the effect, because there is less material to resist the movement and more interfaces at which it accumulates.

What the Designer Should Specify

The designer’s contribution is the annular ring requirement, stated explicitly rather than implied by a pad size. A minimum ring value on the fabrication drawing gives the fabricator a target and gives the inspector something to measure.

Where the design is dense and the ring is small, it is worth saying so, because the fabricator may then build a registration test coupon or measure the corner targets more frequently for that order.

It is also worth specifying which layers are critical. In most designs only one or two layers carry fine pitch routing, and concentrating the registration effort there is cheaper than tightening the whole stack.

Diagnosing a Registration Failure

When the ring is reduced, the question is whether the artwork was scaled incorrectly, the tooling drifted, or the material moved further than expected. The three have different signatures.

A scale error appears as a ring that is thin on one side of the panel and thick on the opposite side, symmetrically. A tooling shift appears as a uniform offset with the same thin side everywhere. Material movement appears as variation that follows the panel geometry rather than the pattern.

Comparing the measurement map against the previous lots for the same design distinguishes a change in the process from an error in the first build. Our board outline notes describe the related tolerance that positions the pattern relative to the finished edge.

FAQ

Does the designer need to apply a scale factor? No. Scaling is applied by the fabricator to the supplied artwork, using their own process data. The designer supplies dimensions at nominal size.

How much registration error is normal? It varies with layer count, material and panel size, and it is larger at the corners than in the centre. The correct figure comes from the fabricator’s process capability for the specific stack.

What does gopcb provide for a tight registration build? We provide measured registration maps for each lot, scale factors derived from that data, tooling systems matched to the panel size, and registration coupons where the design is dense. Where a design has no margin, we say so before the panel is built rather than after.