Heat Sink Inspection: Fin Pitch, Thickness and Coplanarity

A heat sink's fin row carries the heat-transfer area and the convective surface. Fin thickness, pitch, height, tip straightness, the root transition and the consistency of the whole row across the base together decide how the sink performs at a given airflow. Yet in routine quality control the fin row is often the least covered region on the part: calipers will not enter a gap of one or two millimeters, micrometers load thin fins and deflect them, height gauges touch a handful of sections, and neither cumulative pitch drift nor continuous tip form is captured in a single setup. Measuring the whole row does not mean digitizing every point of every fin. It means describing the key geometry of the row completely, repeatably and traceably from one datum. That is the task this article addresses.

Why the fin row resists contact measurement

Four characteristics compound here. The row is dense: pitch on higher-end heat sinks commonly sits in the one to two millimeter range, and fin thickness can fall to a few tenths of a millimeter, so the contact tip of a caliper or probe is close to the size of the gap itself and interferes with the neighboring fin before it reaches the true contour. The fins are thin: aluminum fins have limited stiffness, and even a light measuring force deflects them elastically, so the thickness that comes back is not the thickness in the free state. The surface is reflective: untreated aluminum behaves almost like a mirror, anodized aluminum scatters, and the two can appear on the same part, which makes edge extraction swing between the bright band and the shadow under a single lighting condition. And the row accumulates: pitch is not an isolated per-fin value, small deviations add up along the direction of the row and end as an uneven distribution across the whole array, which single-point sampling cannot reveal.

Together these effects kept fin row inspection at the level of a few fins and a few sections, leaving a real gap between measurement coverage and the geometric information the row actually contains.

Where optical dimensional measurement enters

A vision measuring machine builds on non-contact optical imaging, which suits dense, thin, reflective fins by nature. A SIG vision measuring machine, for example, captures the fin contour through a high-resolution industrial camera and a telecentric lens, and the parallel projection of the telecentric path suppresses the parallax caused by height differences, so tips and roots at different heights keep the same magnification in the image.

Programmable multi-channel ring light and coaxial light can be set in segments along the direction of the row: frontal light lifts the tip edges, low-angle side light strengthens the fin flank and the shadow between fins, and coaxial light suppresses the specular band on aluminum. Under these combinations a sub-pixel edge algorithm extracts both flanks of each fin, and the software fits a centerline fin by fin along the row and returns thickness, pitch and tip position together.

One image capture covers dozens of fins inside the field of view, so coverage stops being limited by how fast an operator can place a gauge on each fin.

Height needs a third dimension

For height-direction features such as fin height and tip form, a two-dimensional image only gives a projected position, and three-dimensional data has to fill the gap. A line-scan laser head is available as an option: once the optical system has located the row in plane, the laser scans continuously along the tip line and collects point clouds from the tips and the root transition.

Scanning and imaging run on the same motion platform and share one coordinate frame, so the in-plane position and the height of the row are evaluated together, without the datum drift that comes from moving the part between two machines. Tip straightness, overall row flatness and the height trend along the row are fitted from that point cloud instead of being inferred from a few isolated sections.

Thickness, pitch and cumulative deviation

The two most informative outputs on a fin row are the thickness distribution and the cumulative pitch deviation. Thickness fixes the conductive cross-section of each fin; pitch fixes the width of the channel between fins; the uniformity of both along the row drives the consistency of the heat-transfer surface. The traditional sequence, measuring each thickness with a caliper and a few pitches with a steel rule, recording by hand and calculating afterwards, is slow and handles accumulation badly.

After extracting the edges of the whole row, the software generates the center coordinate of every fin. Thickness comes from the distance between the two fitted flank lines, and pitch from the difference between adjacent centers. On that basis the cumulative deviation required by the drawing can be computed: the difference between the actual accumulated distance from the datum end to the n-th fin and the theoretical one, plotted as a curve along the row.

That curve says more than a set of point readings. A uniform offset, a progressive drift or a local jump each produce a different shape, and the shape points at a different cause: cutter wear, repeatability of the indexing mechanism, or uneven clamping in a welding fixture. For skived or welded rows, this trend information feeds process adjustment directly.

Tips, roots and whole-row form

Tip straightness and the root transition are the two features most easily left out, and both affect assembly and airflow. A tip with a local collapse or a wavy profile opens a gap where the sink meets a fan shroud or duct, part of the airflow escapes through the bypass and the effective air volume falls. The root fillet controls stress at the junction between fin and base; a radius that is too small or a transition that is discontinuous becomes a crack origin under thermal cycling and vibration.

The high-magnification optical channel evaluates tip straightness at sub-pixel level, while the laser samples the root transition along its length and the software fits the fillet radius and checks continuity. For the row as a whole, tip heights are fitted into a trend line along the direction of the array, from which overall flatness and tip coplanarity are reported. Results are presented as values and as deviation maps, so the form of the row moves from visible but unquantified to measurable and comparable.

Fixturing, programming and batch use

Repeatability on a fin row depends heavily on clamping and datum consistency. Aluminum fins are compliant: too much clamping force distorts the row, too little lets the part creep during measurement. In fin row work the primary datum is normally the underside of the base or a specified mounting face, combined with low-stress flexible supports, so the row is imaged in its free state. Once the coordinate frame is built from that datum, thickness, pitch and tip height are all evaluated against the same reference, and the datum shift caused by different clamping methods disappears.

In production, the program can be generated from the heat sink CAD outline: row direction, fin count, nominal thickness and pitch are imported from the drawing, the operator confirms the datum and the lighting scheme, and the software scans segment by segment and evaluates automatically. When the same model returns to the line, the program is recalled directly; on changeover, the scan region and row parameters are adjusted, and no dedicated gauge has to be built for each fin pattern.

Results are stored in a database against part number and timestamp. Thickness distribution, cumulative pitch deviation and tip straightness trends by batch are available to process engineering, which gives the skiving index, the welding location or the straightening step data measured on the floor rather than estimated from experience.

Where optical measurement stops

The role of a vision measuring machine in fin row work is to provide a non-contact, programmable and well-covered optical evaluation, not to replace every other metrology method. Internal material defects, the very bottom of deep narrow slots, or micro-profile analysis that needs an extremely high point density remain the territory of contact profilometers, industrial CT and ultrasonic testing. The relationship is a division of labour: optics evaluates the macro geometry of the whole row quickly, and the other methods go deeper into specific regions when the question calls for it.

Moving fin row inspection from a few fins to a whole row that can be evaluated is not about pushing a single index further than conventional tools. It is about letting thickness, pitch, height and tip form form one connected dataset under a single datum, so that the manufacturing quality of the row has a geometric basis that can be controlled batch by batch. For more application notes on heat sinks and liquid cooling parts, visit www.sinowon.com.