XRF: Preparation, Placement and Process Control

XRF is the standard way to check coating thickness on a finished board, and it is also easy to use badly. The instrument reports a number to three decimal places whether or not the calibration, the spot size and the substrate behind the coating all match the assumption built into the measurement, so a reading is only as trustworthy as the setup behind it.

What XRF Measures and What It Assumes

X-ray fluorescence excites the atoms in a coating and reads the characteristic X-rays that come back. The intensity depends on the thickness and the density of the layer, so the instrument converts intensity into thickness using an internal model of the coating and of the layers beneath it. If the real stack differs from that model, the number is wrong even though the instrument is functioning perfectly.

The practical consequence is that the same coating measures differently over nickel and over copper, and differently again over a thin versus a thick underlying layer. Knowing the layer stack, and selecting the matching measurement program, is the first requirement for a usable reading. The instrument cannot know whether the layer beneath the coating is the one the program expects, so the operator has to establish the stack and select the matching program before a number means anything.

Calibration and Reference Standards

Calibration is verified against a certified reference standard of known thickness on a known substrate, and the check is done at the start of a measurement session rather than once a year. A reading that drifts by more than a few percent from the standard value means the instrument or the program needs attention before any production number is recorded.

Standards themselves must be appropriate to the range being measured. A standard of 1 micron gold does not validate a measurement of 0.05 microns, because the fluorescence behaviour and the counting statistics differ. Where a specification calls for a thin immersion gold layer, the reference should be in the same range.

Spot Size and Feature Size

The measurement spot is typically 0.1 to 0.5 mm across, and everything inside the spot contributes to the result. Measuring an edge, a small pad or a plated hole wall mixes in the substrate or the surrounding mask, which biases the reading. The safest practice is to measure on a flat area larger than the spot, ideally on a dedicated coupon.

Where the feature is smaller than the spot, a collimated instrument or a smaller aperture is required, and the reading should be reported with the aperture size so a second laboratory can reproduce it. Reporting a thickness without stating the measurement area is a common cause of disagreement between a shop and its customer.

Measuring Thin Gold Over Nickel

Immersion gold over <a href="https://www.gopcba.com/gold-finger-plating-edge-connectors/” title=”electroless nickel”>electroless nickel is the classic thin-layer measurement. The nickel is usually 3 to 6 microns and the gold 0.05 to 0.1 microns, so the gold reading is a small number that depends strongly on the nickel beneath it. The instrument must be programmed for the gold-on-nickel stack, and the count time extended until the uncertainty is small relative to the specification.

Two errors are common. Measuring over a copper area with no nickel produces a gold reading that is meaningless. Measuring a gold layer thicker than the immersion process can produce, such as a reading of 0.4 microns where the process targets 0.08, usually indicates that the spot included a gold-plated feature or that the program is set for a different stack.

Measuring Tin, Silver and OSP Layers

Hot air solder leveled tin is thick and irregular, varying from a micron to tens of microns across a pad, so a single XRF reading represents only the point measured. Silver and tin immersion layers of roughly 0.1 to 1 micron are measured more reliably, and organic surface preservative layers are usually assessed indirectly because XRF does not see the organic film directly.

For a rough coating, take several readings across the pad and report the range rather than the mean alone. A mean of 15 microns with a range of 2 to 30 describes a different process condition from a mean of 15 with a range of 13 to 17, even though the averages agree.

Substrate and Interaction Effects

Thicker underlying copper changes the fluorescence background and can shift a thin coating reading, particularly when the coating and the substrate share elements. Plated through-hole walls, gold fingers over different base layers, and coatings over alloy 42 all need their own measurement program if the numbers are to be comparable.

Where the layering is complex, a correlation against a cross-section is the only way to establish confidence. Measuring the same coupon by XRF and then sectioning it gives a direct comparison, and the difference between the two methods becomes the known uncertainty of the non-destructive method. That figure belongs in the finish specification, because it tells an inspector how much disagreement between a supplier certificate and an incoming check is normal rather than suspicious.

Sampling a Production Lot

Define how many readings, on how many coupons, from which positions on the panel. A workable plan takes five to ten readings per coupon and coupons from across the panel, because plating thickness varies with position in the bath and with current density. A single reading from the middle of a panel supports almost no conclusion about the lot.

The sampling plan should also state what happens on a marginal result: re-measure at more points, section the coupon, or hold the lot. Deciding this in advance prevents a marginal reading from being rounded into acceptance, which is how a finish problem reaches the assembly line. A workable default is to hold any lot whose reading falls below the specification limit at a single point, and to release it only after a second coupon has been measured and sectioned where the result is marginal.

Correlation with Microsection

XRF and microsection measure different things: the first measures a mass per unit area converted to thickness, the second measures a geometric distance on a cut face. They agree only when the coating density matches the assumption and the section is cut through a representative point. Establishing the offset between the two methods on a reference coupon is a good investment for any finish that is tightly specified.

Repeat the correlation whenever the process or the instrument changes, and keep the result with the finish specification. A supplier and a customer who have agreed the correlation in advance can resolve a dispute with a measurement rather than a negotiation.

Records and Supplier Verification

Record the instrument, the program, the aperture, the coupon, the reading and the reference standard check with each lot. That record is what allows a later problem to be traced to a specific measurement rather than to a general claim about the finish.

Where the surface finish is purchased with the boards, verify with incoming XRF on a sample and compare against the certificate. A supplier whose readings correlate with yours is a supplier whose certificates can be relied on for the lots you do not measure, which is the practical basis for reduced inspection.

FAQ

How thick is immersion gold on an ENIG finish? Typically 0.05 to 0.1 microns over 3 to 6 microns of electroless nickel. The gold reading depends strongly on the nickel layer, so the instrument must be set for that stack.

Why do XRF and microsection disagree? XRF converts mass per unit area to thickness using an assumed density, while a section measures geometric distance. The two agree only when the coating and the cut point match those assumptions.

How many XRF readings does a finish lot need? Five to ten per coupon with coupons taken across the panel, since plating thickness varies with position in the bath and with current density.