Solder Paste Inspection: What SPI Measures and Why

Solder paste inspection measures the deposit that printing leaves on the pads, before any component is placed. It is the one process in surface mount assembly where the input can be measured directly, and a printer that is drifting will show up in the paste volume long before it produces a visible defect on a finished board.

The measurement is not an end in itself. What matters is whether the numbers predict the joints that will be produced, whether the limits are set where they catch real problems, and whether the data reaches the person who can adjust the printer. A system that reports volume without any of that is an expensive camera.

What the System Actually Measures

A paste inspection machine projects a structured light pattern onto the pad and reconstructs the shape of the deposit from the distortion. From that it derives height, area and the volume calculated from them, along with the print offset of the deposit relative to the pad and the presence of bridging or smearing between adjacent pads.

The measurement is a comparison against a reference derived from the stencil aperture, so the numbers are reported as a percentage of the nominal volume rather than in absolute units. That normalisation is useful, but it hides the fact that the nominal itself is an assumption, and the aperture on the stencil is not always what the design intended.

Volume, Area and Height Tell Different Stories

Paste volume is the headline number, and it is the one most closely tied to joint quality. A deposit with the correct volume but the wrong shape can still fail: the solder that forms the joint comes from the volume that is in contact with the pad and the lead, and a tall column of paste that collapses sideways contributes less to the fillet than the same volume spread evenly.

Height and area are therefore worth watching separately. A low area with a normal height usually means the aperture is partially clogged. A large area with a low height suggests the paste has slumped or that the stencil was not released cleanly, and both point at different parts of the printing process.

Where the Variation Comes From

Paste printing has more variables than most people expect. Squeegee pressure and speed, the separation speed of the board from the stencil, the condition of the apertures, the temperature and humidity of the room, the age of the paste on the machine and the flatness of the board and its support all change the deposit.

The variation between pads on the same board is usually smaller than the variation between boards, and both are smaller than the drift over a shift. That structure matters, because it determines what a sample can tell you: measuring a few pads on one board shows the within board variation and says little about the process as a whole.

Limits That Catch Something

Specification limits should be derived from the process capability, not copied from a default. The relevant question is how much volume variation the assembly can tolerate, which depends on the pad size, the component, the paste and the reflow profile, and it is different for a fine pitch quad flat pack and for a large thermal pad.

A practical approach is to set the limit where the defect rate begins to rise, using data from the process rather than a rule of thumb; the reasoning mirrors the tolerance discussion in the notes on manufacturing tolerances. Where a product has a history, the correlation between paste volume and defect rate can be measured, and the limits chosen at the point where the relationship becomes steep.

Closing the Loop to the Printer

The value of the data depends on how quickly it reaches the printer. A closed loop system adjusts the printer parameters automatically when a trend is detected, while a manual system relies on an operator reading a report. Both work if the response time is fast enough, and neither works if the inspection is done at the end of the shift.

The most useful output is a trend rather than a pass or fail at a single board. A slow drift in average volume over a few hundred boards tells the operator that the paste is drying or that the stencil is loading up, and the correction can be made before the first defective print appears.

Correlating with the Finished Joint

Paste inspection predicts defects, but the correlation has to be established on the actual product. Insufficient paste on a fine pitch device produces opens and weak joints, while excess paste on the same device produces bridging. The same volume deviation in opposite directions can therefore produce two different defects, which is why limits must be two sided.

The finished joint quality is the reference, and the technique used to judge it matters. X-ray inspection of the first article and the routine checks of yield and quality control should be tied back to the paste data, so that the inspection limits reflect what happens in the oven rather than what happens on the pad.

False Calls and Their Cost

A high false call rate destroys confidence in the system. Operators begin to accept boards without looking, and the data is ignored. The usual causes are a reference that does not match the actual stencil, a board that is not held flat under the camera, and measurement settings that have not been tuned to the paste and pad combination.

Gauge studies are the way to separate real variation from measurement noise, and the same discipline is applied to pad design standards when a new footprint is introduced. Measuring the same board several times, and having different operators measure it, shows how much of the reported variation belongs to the machine, and that number sets the smallest change that can be trusted.

Where to Apply It

Full inspection on every board is expensive, and it is not always necessary. The cases where it pays are products with fine pitch devices, ball grid arrays and any assembly that cannot be reworked, because the cost of a missed defect is higher than the cost of the measurement.

A sampled regime works where the process is stable and the product is not critical, provided the sample is taken in a way that reflects the whole shift. Sampling pads from one board once a day will not catch the drift that happens after lunch, and the schedule should be designed around the sources of variation rather than around the convenience of the operator. The paste itself is part of that variation, and its handling is described alongside the other requirements in the notes on manufacturing tolerances.

FAQ

Can paste inspection replace visual inspection after reflow? No. It measures the input to the process, not the joint that results. The two checks find different defects and complement each other.

What volume tolerance is normal? It depends on the pad and the device, but a band of roughly plus or minus twenty per cent around the nominal is a common starting point for a stable process on a fine pitch product.

Does a good paste reading guarantee a good joint? It removes one category of risk. Placement, reflow and the component itself still have to be correct.