Via Sampling Plans: Separate Holes, Circuits and Panels

Measurement purpose

Ensure ceramic via inspection counts and lot decisions use the correct hole, circuit and panel units with a traceable selection design.

Specimens and conditions

Population identity

Preserve lot, panel, circuit and via identifiers through inspection and any singulation.

Requirement mapping

State whether a nonconformity applies to one hole, a connected group or the delivered circuit.

Equipment and records required

• Inspection method: Demonstrate observability of the electrical or visual unit claimed in the count.
• Traceability system: Retain coordinates, first results, retests and missing records without duplicating inspected opportunities.

Method sequence

1. Define units

   Agree lot scope and separate sampling, inspection and nonconforming units.

   Record: Sampling definition and disposition authority.

2. Select and inspect

   Apply the approved selection and record each result at all hierarchy levels.

   Record: Linked hole, circuit and panel register.

3. Summarize

   Calculate each denominator and disclose exclusions before applying the agreed rule.

   Record: Hierarchy counts and lot disposition.

Decision and uncertainty

Apply the acceptance rule only to its specified unit and population; never substitute a lower-level count to enlarge the sample.

Clustering, targeted selection, incomplete observability and missing results limit inference beyond the inspected population.

The customer and quality owner agree acceptance units and contractual sampling requirements before results are reviewed.

Traceable outputs

Define the four units in the sampling statement

Start with the lot boundary: which material, processing history and time window the disposition will cover. Next define the sampling unit selected from that lot, such as a panel or individual circuit. Then define the inspection opportunity, such as a particular via or a specified via-chain test. Finally define the nonconforming unit under the drawing requirement.

These units need not be identical. A circuit can contain several failed vias yet count as one nonconforming delivered circuit. Conversely, a continuous redundant via bank may contain an individual open that violates a per-hole requirement without failing the group continuity criterion. Agree the requirement before deciding which count belongs in an acceptance calculation.

Report each denominator beside its numerator

Consider a hypothetical inspection of ten panels, each containing twenty circuits with twelve relevant vias per circuit. The record contains 2,400 hole opportunities and 200 circuits. Suppose six nonconforming holes occur in three circuits on two panels. The observed fractions are 0.25 percent of holes, 1.5 percent of circuits and 20 percent of inspected panels containing at least one affected circuit.

All three statements can be correct, but they answer different questions. None alone estimates the quality of an uninspected population without the selection design and uncertainty. Reporting only the smallest percentage would conceal concentration; reporting only the panel fraction could exaggerate the number of individual affected circuits. Preserve the hierarchy instead of choosing the most attractive denominator.

Show why many holes on one panel share exposure

Vias on one panel may share paste preparation, printing conditions, drying, firing and handling. A local region may also share a common geometry or loading condition. These relationships can make outcomes correlated. Counting more holes on the same panel improves knowledge of that panel, but does not necessarily add equivalent information about variation between panels or lots.

If all selected circuits come from a convenient panel corner, the plan may miss a different panel position even with a large nominal hole count. Record the selection route and coordinate distribution. This page does not prescribe a variance model; its immediate task is to prevent a clustered inspection count from being described as an independently selected production sample.

Use the independent-hole calculation only as an explicit model

For a circuit with m required vias, if every via independently has the same nonconformity probability p and any nonconforming via rejects the circuit, the model circuit probability is one minus the quantity one minus p raised to m. Both independence and the any-hole rejection rule are strong assumptions. A redundant network or common panel event changes the interpretation.

For an illustrative p of 0.001 and m of twenty, the independent model gives about 1.98 percent of circuits with at least one nonconforming via. This is not a conversion formula for the factory dataset unless the assumptions are supported. With completely correlated via outcomes, the same marginal hole probability could produce a very different distribution of affected circuits.

Design selection around the population claim

If the decision concerns delivered circuits across a lot, distribute the selection across the identified production population using an agreed random or appropriately stratified plan. Retain selected panel identities even after singulation. A visually convenient tray of circuits may contain only adjacent positions from one panel and therefore provide narrower coverage than expected.

Use targeted inspection separately when investigating a suspected edge effect or common defect. Such targeting can be highly useful for finding a mechanism, but its result should not be presented as a random lot estimate. Label the purpose of each group: acceptance sample, process investigation, retained specimen or destructive analysis. Combining them after seeing the result changes the meaning of the count.

Check whether the test observes the unit being counted

An end-to-end measurement of a parallel via bank may not detect one bypassed open. If the requirement is per-hole integrity, the plan must establish individual fault observability or use an agreed alternative evidence path. A photograph of every hole and one group continuity reading cannot automatically be counted as a complete electrical test of every hole.

Similarly, a series witness coupon observes its own route, not every via on neighboring delivered circuits. Keep the relation between witness and product explicit. A coupon may support process monitoring while product acceptance relies on another method. Do not inflate the sample count by adding witness holes to delivered-unit holes as though the test boundaries were identical.

Preserve missing results and first observations

Identify inaccessible holes, damaged specimens, aborted tests and unreadable records before calculating the denominator. A missing result is neither a conforming opportunity nor automatically a product defect. The inspection plan should define whether it requires replacement sampling, further investigation or a held decision, with reasons preserved.

Keep original and repeat results linked. Reprobing can distinguish a fixture contact issue, but the initial observation must remain available. Repeated measurements of the same hole do not create additional inspected holes. When a circuit is repaired or reworked, preserve the original disposition and the post-action verification as separate states rather than rewriting the historical count.

Freeze the disposition rule before examining the outcome

Specify the lot, sampling unit, nonconformity classes, test boundary and permitted acceptance rule with the responsible quality owner. If a contractual sampling standard is required, identify its applicable edition and parameters through that agreement. Do not select an acceptance number from an unrelated table or infer a universal plan from the number of vias available.

The final report should allow another engineer to reproduce all three hierarchy counts and the actual lot decision. Include coordinates, selection method, missing observations and any departures from the approved plan. A larger inspection dataset is useful only when its units and scope remain intelligible; more recorded hole values alone do not establish broader production confidence.