On this page
A resistor value is the combined outcome of resistor paste, fired geometry, termination interaction, print-field position, substrate history, and measurement boundary. If every witness uses one body and one overlap at one convenient location, a shift cannot be assigned responsibly. A deliberately structured matrix can separate common paste movement from edge effects and termination-dependent behaviour while preserving the identifiers needed for corrective action.
Key design decisions
- Repeat identical resistor bodies across the print field before comparing geometry.
- Pair short- and long-overlap terminations within each position group.
- Keep paste lot, conductor lot, firing load, position, and measurement identity in the data model.
Map the contributors before choosing resistor shapes
Start with the product resistor families: aspect ratio, nominal value region, conductor termination, overlap, nearby glass, trim access, and electrical test points. Mark which families share paste and firing but differ in geometry. The witness plan should answer a limited question about these contributors rather than reproduce every value in the product catalogue.
Distinguish three levels. Lot factors travel with material or incoming substrate. Position factors repeat at print-field, carrier, or furnace locations. Geometry factors belong to resistor body and termination layout. A valid analysis preserves all three labels. If a specimen loses one label, it may remain useful for screening but cannot support the intended separation.
Arrange paired geometries at repeated field positions
Choose a reference resistor body and repeat it at centre, edges, and other positions justified by the screen and loading route. At each position, place paired termination-overlap variants while holding body dimensions constant. Add a limited aspect-ratio contrast only when the product contains that geometry and when artwork space permits. Pairing within position makes the termination comparison less vulnerable to broad spatial gradients.
Avoid placing every short-overlap body on one side and every long-overlap body on the other. That layout confounds termination with position. Rotate or mirror selected structures only when orientation is an explicit factor. Give each resistor a machine-readable identifier containing panel, position, geometry, overlap group, paste lot, conductor lot, and firing-load reference.
Calculate a termination contrast within position
Use the median or another preselected robust centre when occasional measurement anomalies would unduly affect a mean. Compute the short-minus-long overlap contrast inside each position and firing group. Then compare those contrasts across positions. A stable contrast across the field suggests a repeatable termination-layout relationship; a contrast that changes only at an edge indicates interaction with position and requires a more detailed model.
For illustration, short-overlap witnesses at one position have a 12.8 kΩ median and long-overlap witnesses have a 12.1 kΩ median. The 0.7 kΩ difference describes that group only. It does not prove termination physics or establish tolerance. Repeat positions, lots, and firing loads are necessary before the contrast can guide artwork.
ΔR_t,p = median(R_short,p) - median(R_long,p)
- ΔR_t,p: termination-overlap contrast at position p
- R_short,p: resistances from short-overlap bodies at p
- R_long,p: resistances from long-overlap bodies at p
- median: chosen centre applied consistently
Body geometry, material route, measurement boundary, and firing group are matched within position. The contrast remains an observed association until validated.
Keep lot and spatial terms visible in the analysis
Plot individual resistance by paste lot and physical coordinate before fitting a model. A mixed histogram can hide a coherent edge band or a shifted lot. Use a blocked comparison with lot, panel, position, body geometry, and overlap as explicit fields. Where sample size supports it, estimate interaction terms; where it does not, describe the observed strata without claiming statistical separation.
Do not treat consecutive prints as independent lots. Record the actual paste issue, working-container history, screen state, substrate batch, and firing load. These variables may explain a chronological pattern that would otherwise be assigned to paste. Retests remain linked to the original measurement and should not replace it silently.
Read witness patterns as competing hypotheses
If every geometry and position moves together, investigate shared material, thickness, firing, and measurement factors. If both overlap variants move only along one field edge, investigate printing and position. If short-versus-long contrast repeats at several positions while the common centre stays stable, termination interaction becomes a stronger hypothesis. A single extreme witness does not establish any of these patterns.
Inspect resistor edges, fired dimensions, termination coverage, cracks, contamination, and probe contact by coordinate. Trimmed and untrimmed resistors must not be pooled because trimming changes geometry and history. When destructive analysis is justified, select specimens from contrasting groups rather than only the worst value, and retain controls from the same lot.
Assign each witness comparison a specific decision
The table keeps common movement, spatial movement, and overlap response from collapsing into one generic resistor result. Each comparison should lead to a known action: material review, print-field investigation, termination-layout review, or measurement check. If the outcome cannot change a decision, reconsider the witness before consuming panel area.
Set analysis rules before seeing the data. Define grouping, treatment of censored readings, retest handling, and minimum traceability. Report raw values and group centres together. A fitted model never repairs missing identifiers.
| Observed pattern | Matched comparison | Primary next question | Required confirmation |
|---|---|---|---|
| All groups shift together | Same geometries across lots | Material, thickness, firing, or measurement? | Independent lot and process evidence |
| One field region shifts | Paired geometries by coordinate | Print or load position effect? | Mapped dimension and process data |
| Overlap contrast repeats | Short versus long within position | Termination interaction? | Additional overlap levels or sections |
| Only one body shape shifts | Aspect ratios within position | Geometry sensitivity? | Fired dimensions and repeat load |
Validate the matrix with blind confirmation material
Use development runs to confirm that the matrix detects a deliberately introduced position or geometry contrast without misclassifying general measurement noise. Establish measurement repeatability first. Then execute independent prints and firing loads. Do not tune group boundaries until every unusual result disappears; preserve the anomalies and determine whether they expose an uncontrolled factor.
Reserve confirmation material whose results were not used to select the analysis rule. Require traceable lot and position identifiers, adequate representation of both overlap groups, and coordinate-linked geometry inspection. If the same nominal condition produces a different pattern, narrow the inference and investigate screen, paste history, substrate, or furnace loading.
Release the witness matrix as a controlled experiment
Control the artwork, body dimensions, termination overlaps, conductor construction, resistor paste, substrate, print orientation, field map, firing-load positions, measurement current, stabilization, grouping method, calculation, exception codes, and data retention. State which product resistor families the matrix represents and which remain outside its geometry.
Reopen the design when paste or conductor changes, overlap artwork changes, screen orientation moves, resistor aspect ratio changes, firing loading changes, trimming is introduced, or the substrate lot structure changes. RFQ input should include resistor targets, body and termination geometry, trim requirement, field constraints, and functional consequence. All acceptance remains by drawing and application review.
Provide resistor geometry and factor structure
Send the information needed to design a lot-, position-, and termination-resolving witness matrix.
- Resistor paste and conductor systems, substrate, body dimensions, aspect ratios, termination overlap, glass boundaries, and trim access.
- Screen field, print orientation, panel coordinates, substrate lots, paste issue history, and firing-load mapping.
- Measurement circuit, current, timing, probe geometry, retest rule, target grouping, and data format.
- Known shifts or failures, product consequences, candidate hypotheses, sample availability, and review owner.
The drawing-upload form loads as you reach this section.