First Pass Yield Improvement on the SMT Line

First pass yield is the fraction of units that complete the process without rework, repair or retest. It is the most honest single number a line produces, because it does not care how much effort went into recovering a unit; it only records whether the process got it right the first time. Improving it is usually the fastest route to lower cost and shorter lead time, and the work is systematic rather than heroic: define the number, find where the losses are, remove the largest cause, then hold the gain.

Defining First Pass Yield Precisely

The definition has to settle two questions: which operations are inside the yield, and what counts as a failure. A unit corrected on the line by an operator, without any formal repair record, must count as a failure of the process, otherwise the number flatters itself. A unit that passes functional test but needed rework earlier still fails first pass yield, even though it ships. Writing these rules down and applying them consistently matters more than the exact formula chosen, because a number that changes meaning between shifts cannot be improved.

Throughput yield, calculated by multiplying the yields of the individual steps, is a useful companion figure because it shows where losses multiply. A line with six steps each at ninety-nine percent has a throughput yield near ninety-four percent, and that loss is invisible in any single step’s report. Publishing both numbers, first pass yield for the line and throughput yield by step, tells the improvement team where to look and stops the argument about which department is responsible for the loss.

Building the Defect Pareto

Improvement starts with data that is actually collected. Every defect found at inspection, at test or in the field should carry a code, a location and the operation where it was detected. With that, a quality checklist becomes a measurement instrument, and a Pareto chart ordered by frequency usually shows that three or four codes account for most of the loss. Chasing the long tail before the head is the most common way to spend effort without moving the number.

The Pareto must be built on the right denominator. Comparing defect counts between products with very different volumes is misleading, so the ranking should use defects per million opportunities or per thousand units. It is also worth separating defects by detection point, because a defect found at the end of the line has already passed several opportunities to be caught and corrected, and each of those missed opportunities is a separate process weakness and, in a mature line, a separate improvement project.

Attacking the Top Contributor

The top contributor usually has a physical cause rather than a behavioural one. Solder bridging from a marginal stencil aperture, tombstoning from an asymmetric pad or profile, or placement offsets from a feeder that is drifting all have traceable mechanisms. Root cause analysis should end with a change to a process parameter, a tool or a design, and that change should be recorded with the data that justified it rather than applied quietly and hoped for. The record of that change becomes the evidence that the improvement was caused by the action.

Where the cause is genuinely behavioural, such as an operator working around an awkward fixture, the fix is still physical: change the fixture. Discipline problems created by process design cannot be trained away for long. This is why the improvement team should include the people who actually run the line, whose understanding of the workaround is usually more accurate than the documentation they were given at handover.

Process Control versus Inspection

Inspection finds defects; process control prevents them. Adding a visual check after a step raises detection and leaves the defect rate untouched, so first pass yield may look better while the underlying capability has not changed at all. The sustainable route is to control the inputs, which means monitoring paste volume, placement accuracy, profile and cleanliness at intervals rather than waiting for the output to be graded by an automated inspection system.

That distinction also decides where the money goes. An extra inspector is a recurring cost that scales with volume, while a control chart on the printer is a one-time effort that reduces the number of units needing attention. Where both are present, the aim is to shrink the inspection burden as the process becomes capable, and to keep inspection where it provides independent evidence rather than acting as the primary screen. Inspection that is genuinely independent still catches the surprises that control charts never predicted.

Rework and Its Effect on the Number

Rework and first pass yield are linked by definition, and pretending otherwise creates a reporting problem rather than a quality improvement. Every rework action should be recorded, charged to the step that caused it and counted as a failure of first pass. When the number is reported honestly, the line usually finds that one station contributes most of the rework, and that station usually has a fixable process problem rather than an operator problem, which is good news for everyone except the equipment vendor.

Rework also has a hidden second cost: the reworked unit may fail later. Handling damage, thermal exposure and probe marks all accumulate, so a high-rework line produces a field failure rate that its final test results do not predict. Tracking the rework rate alongside the field return rate and the staged inspection data exposes that relationship before the customer discovers it, and it justifies the effort spent on the original process rather than on recovery.

Sustaining the Gain

Improvements decay unless they are built into the standard. The parameter that fixed the bridging must become part of the printer recipe, the fixture change must be documented, and the visual aid that transmitted the lesson must stay on the line. A short audit at intervals, checking that the standard is still followed, keeps a solved problem solved; without it the defect returns at the next changeover or when the original operator moves to another line. Document control is what makes those standards survive staff changes.

Reporting should continue after the project closes, because the number is the evidence that the gain is real. A quarterly review of first pass yield, throughput yield and the top three defect codes, with an owner assigned to each, provides the continuity that keeps the effort alive. Programmes that stop measuring after the celebration usually find the yield back where it started within a year, and the second improvement is always harder than the first one.

FAQ

Should reworked units count as failures of first pass yield? Yes. The unit did not pass the first time, and counting it as a pass hides both the cost and the cause.

What is a realistic first pass yield target? Set the target from the capability of the controlled processes and the complexity of the product, then improve step by step. A target copied from another product’s line is rarely meaningful.

How do I know an improvement is real? Measure the same way for at least a month after the change, and confirm that the defect code you attacked has moved while no new code has appeared in its place.