SPI and AOI Data Correlation for Process Control on SMT

SPI measures the paste before placement and AOI inspects the joints after reflow, and the two systems together describe the process from end to end. Used separately they answer single questions; used together they allow a defect to be traced back to the deposit that produced it. Data correlation between the two is what turns a pair of inspection machines into a process control system.

What Correlation Means in Practice

Correlation means that the result of an AOI inspection can be compared, board by board and pad by pad, with the SPI measurement that preceded it. That requires the two systems to share a reference, so that a defect at a given position on the board can be matched with the deposit at the same position.

Once the link exists, patterns become visible. A rise in insufficient solder at one package can be matched with a fall in deposit volume at the same pads, and the cause is then a printing problem rather than a placement or reflow problem. Without the link, the same evidence is spread across two reports that never meet.

SPI Data and What It Measures

SPI reports volume, area and height for every deposit, together with the position error and the shape of the paste. Volume is the parameter most closely linked to the final joint, and it is the one usually used for control, but area and height add information about slumping and about the release from the aperture.

The technique and its limits are described in the guide to solder paste inspection. It is a fast measurement taken on a wet deposit, so it describes the input to the process rather than the result, which is exactly what makes it valuable for diagnosis.

AOI Data and Its Limitations

AOI inspects the finished joint and reports the defects that the machine has been programmed to detect. Its output is affected by the programming, by the lighting and by the thresholds, and a defect that is outside the programmed criteria will not be reported however obvious it appears to a person.

Its strength is coverage and consistency: every board is examined, and the criteria do not change between shifts. Its weakness is that it measures the result of the whole process, so a defect found by AOI has already been produced by the printer, the placer and the oven. That is why the AOI result needs the SPI data to be actionable.

False Calls and Escape Rate

AOI generates two kinds of error: a false call, where a good joint is reported as defective, and an escape, where a defective joint is passed. The false calls are visible because they consume operator time, while the escapes are invisible until a customer finds them, which makes the escape rate harder to measure and more important.

Correlating with SPI gives an independent way to check both. A joint that SPI predicted would be short of paste and that AOI passed is a candidate escape, while a joint that SPI measured as well formed and AOI rejected is a candidate false call. Both are worth investigating, and neither can be found from one data set alone.

Linking SPI and AOI Results

The link is made through a common coordinate system, a common board identifier and a common unit of analysis. The systems are usually different machines from different suppliers, so the interface is normally a file or a database exchange rather than a direct connection, and the accuracy of the mapping has to be verified.

A practical approach is to define a small number of analysis zones, such as a package or a group of pads, and to compare the SPI volume distribution with the AOI defect rate in each zone. Zone level comparison is robust to small coordinate differences and is usually enough to identify the process step that is responsible.

Using the Data for Process Feedback

The purpose of correlation is feedback. If the AOI defect rate in a zone rises while the SPI volume falls, the printer is the place to look, and the adjustment is made where the problem is. If the volume is stable and the defects rise, the placement or the reflow becomes the suspect, and the investigation moves downstream.

Feedback is most effective when it is fast. Data that is reviewed at the end of the week can describe what happened but cannot prevent it, while a live display of the two data sets allows the operator to see a drift during the shift. The interval between measurement and action is what determines the value of the system.

Sampling and Coverage Strategy

SPI is often run in sampling mode because a full inspection at the printer can limit throughput, while AOI usually inspects every board. The sampling plan therefore determines how much of the correlation is possible, and a sparse sample limits the analysis to trends rather than to individual joints.

The plan should be set from the stability of the process. A stable line can sample less, while a line that is being changed, or one that is running a new product, benefits from full inspection for a defined period. The choice should be a decision with a reason rather than a default setting.

Verification of the Correlation

The correlation itself has to be verified before it is trusted. A simple check is to introduce a known defect, such as a deliberate under print at one pad, and to confirm that both systems report it in the same place and with the same identity. If the mapping is wrong, the analysis will produce confident conclusions from mismatched data.

Verification should be repeated after any change to the programming, the conveyor, the board orientation or the software version on either machine. The comparison of AOI with X-ray and with other methods is described in the guide to X-ray and AOI inspection.

Reporting and Line Discipline

Reporting should be simple enough to be used. A chart of deposit volume against defect rate for the critical zones, updated every shift, is more useful than a database that nobody opens. The report should also state the sample size, so that a change based on a handful of boards is not over-interpreted.

Discipline means acting on the information. A correlation that is collected but never used is a cost without a benefit, and the value only appears when the data changes what the line does. The criteria used to judge the result belong to the same family as those described in the guide to judging PCB quality.

FAQ

Do SPI and AOI need to be from the same supplier? No. Correlation works through a shared board identifier and a coordinate mapping, which can be exchanged between systems from different vendors. What matters is that the mapping is verified, not that the machines match.

What is the most useful single comparison? Deposit volume per zone against the defect rate in the same zone. It links the input to the output at a resolution that is fine enough to be useful and coarse enough to be robust against small coordinate differences.

Can correlation replace either machine? No. SPI describes the input and AOI describes the output, and each answers a question the other cannot. The value of the pair is in the comparison, which is lost if either is removed.