4.1 SPI Recap & Cp/Cpk
Solder pastePaste Inspection (SPI) transforms the stencil printing is the first real gatekeeper of assembly quality, and SPI turns that step into measurable, actionable data. By linking paste volume, height, and area to known downstream failure modes, manufacturers can predict defects before they ever appear at AOI, AXI, or test. Capability analysis adds another layer of insight, showing not just whether prints are in spec, but whether the process itself is stable and centered enough to stay there. With well-chosen limits, dashboards, and disciplined habits, SPI evolves from ana inspectionmanual toolart into a predictive, data-driven system. By measuring the three-dimensional geometry of every paste deposit — Volume, Height, and Area — SPI provides the earliest warning sign for downstream reflow defects (Chapter 3.5). The translation of this raw measurement data into Process Capability Metrics (Cp and Cpk) is essential for achieving a stable, high-yield manufacturing process control engine that drivesminimizes yieldthe andCost predictability.of Quality (CoQ).
4.1.1 Why SPI isMetrics and Defect Prediction
SPI measures the best early-warning system
SPI (solderprinted paste inspection)volume measuresagainst the theoretical volume defined by the stencil aperture (Chapter 1.4).
SPI Metric | Definition | Defect Signal | Upstream Process Check |
Volume (% of Target) | The measured mass of paste. | Low: Opens, Tombstones, HIP risk. | Stencil |
Height (µm) | The peak elevation of the deposit. |
| Squeegee Pressure ( |
Area (% of Pad) | The footprint of the deposit. | High: Bridging, Smear, Solder Balls. | Separation Speed (Chapter 1.5), |
Transfer Efficiency (TE) | Measured | The | Paste Age ( |
Process Note:Think The goal of SPI asis yourto “predict a reflow defect (Chapter 3.5) based on a print health”abnormality dashboard;(Chapter Cp/Cpk1.5). thenFor tellexample, youlow whetherVolume thatTE healthon passive chips is the primary precursor to the capableTombstoning defect.
4.1.2 Capability Analysis: Cp vs. Cpk
Process Capability Indices quantify how well a stable process can meet a set of engineering tolerance limits. This analysis must be performed by feature family (e.g., 0402 chips, 0.5 mm BGA pads) because each has different tolerance requirements.
A) Process Potential (Cp )
Cp measures the maximum potential capability of the process, assuming the output is perfectly centered between the Upper Specification Limit (USL) and Lower Specification Limit (LSL). It only reflects the width of stayingthe insidedistribution your(σ).
Cp = ( USL - LSL ) / 6σ
B) Process Performance (Cpk)
Cpk measures the actual performance by factoring in whether the process output (µ, the mean) is centered within the specification limits. If the average print volume is too high or too low, the Cpk drops, even whenif the lineCp wiggles.(process width) is acceptable.
=
4.1.2 Map SPI signals to real defectsmin (cause →( effect)
USL - µ ) / 3σ , ( µ - LSL ) / 3σ )
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
|
If you keep TE inside sane bands (from Chapter 7.6) and area/height stable, most of the reflow Pareto simply… doesn’t happen.
4.1.3 SPI Limits and Capability 101 (what Cp/Cpk actually say)Targets
CpLimits
checksmustthebe set tight enough to predict defects but wide enough to be achievable. The process should targetwidthCof your process vs thewidthpkof your spec window:Cp = (USL − LSL) ÷ (6σ)(USL/LSL are your TE limits for that feature family; σ is the TE standard deviation.)Cpkaddscentering:Cpk = min((USL − μ) ÷ (3σ), (μ − LSL) ÷ (3σ))(μ is the mean TE. If you’re off-center, Cpk drops.)
Rules of thumb for SPI:
Cpk≥ 1.33→forcomfyvolume(stable,oncenteredallprints).1.00–1.33high-risk→ watch list (tighten cleaning/recipe).< 1.00→ not capable (expect defects unless you change the process).
Always compute by feature family (chips vs QFN edges vs BGA pads vs thermal pads). Mixing families hides trouble.
4.1.4 Setting sensible SPI limits (specs you can live with)
Start with Chapter 7.6’s defaults, then tighten per product:features.
Feature Family |
|
|
|
| Volume Cpk |
Fine- | 85% – 115% | Area Cpk |
| 75% – 125% |
|
QFN | 50% – 65% | Volume Cpk |
4.1.4 Closed-Loop Feedback and Continuous Improvement
The true value of SPI is not the inspection itself, but the use of the Cpk trend to drive permanent improvement.
- Baseline and Audit: The First Article (FA) (Chapter 2.5) must establish the initial Cpk baseline. This baseline proves the capability of the specific paste, stencil, and recipe combination.
- Center the Mean (µ): If Cpk is low due to poor centering (e.g., µ = 90% when the target is 100%), the primary action is to adjust the squeegee pressure or speed (Chapter 1.5) to bring the mean into the target range.
- Reduce Variation (σ): If Cp is low, the process is too wide. The action is to improve fundamental stability via stencil cleaning cadence (Chapter 1.5), paste hygiene (Chapter 1.2), or switching to a higher-quality Electroformed stencil (Chapter 1.3).
- Design Feedback: When a single aperture geometry (e.g., a specific QFN pad) consistently prevents the process from achieving Cpk ≥ 1.33, the failure lies in the Stencil Design (Chapter 1.4). The high σ data must be fed back to the design team to revise the aperture shape for the next stencil iteration.
Final Checklist: Capability Control
Item | Status | Action/Metric |
Capability | Cpk achieved for Volume ≥ 1.33 for all high-risk features. | Dashboard Tracking of Cpk by family. |
|
| Adjust Squeegee Parameters (Chapter 1.5). |
Process Width | Cp is high (≥ 1.33). | Ensure Feeder/Paste Hygiene and minimize print variation (σ). |
Golden Limits | SPI limits are Feature-Specific and locked in the Golden Recipe. | Prevents high-risk features from being run with loose tolerances. |
Pair TE specs with height floors (e.g., ≥60–70% of stencil thickness) to catch “paste on mask.”
4.1.5 A tiny worked example (numbers that talk)
Feature: 0402 chips Spec: TE 75–125% (LSL=75, USL=125) Pilot data (1000 pads): μ = 96%, σ = 6%
CpUSL − LSL = 125 − 75 =506σ = 6 × 6 =36Cp = 50 ÷ 36 =1.39(capable width)CpkUSL − μ = 125 − 96 =29→ 29 ÷ (3×6=18) =1.61μ − LSL = 96 − 75 =21→ 21 ÷ 18 =1.17Cpk = min(1.61, 1.17) = 1.17(a bit off-center low)
Read it: your prints are tight, but biased low. Expect tombstone risk under stress. Fix by nudging blade pressure/speed or aperture reduction (if over-reduced), then recheck.
4.1.6 How to use capability (not just calculate it)
Baseline per familyon the FA lot (+ first real lot).Center first(shift μ toward target), thentighten spread(reduce σ) via:Printer recipe: angle/pressure/speed, separation, cleaning cadence.Stencil tweaks: nano-coating, step maps, aperture edits (7.3–7.4).Paste hygiene: bead size, open time, fresh jar policy (7.2).
Promote families withCpk ≥1.33to “green”; focus ops/engineering on theorange/redfamilies.
4.1.7 Dashboards that prevent defects (what to plot)
TE Cpk by family(chips, QFN edge, QFN thermal total, BGA pads).First-Print Pass %andWipes/Board(printer stability).Reprint rate(local recovery doing work?)Bridging/Tombstone hitsat AOI overlaid on SPI heat-maps for the same panels.Before/aftermarkers when you change stencil or recipe.
If AOI calls drop where Cpk rose, you’ve proven cause→effect.
4.1.8 Small habits that make stats trustworthy
UseI-MRcharts per family; ignore the very first panel after a wipe when setting control limits.Keep SPIillumination profilesper mask color/finish; bad lighting fakes area/height drift.Run aweekly GR&Ron a golden board (3 ops × 3 repeats) so σ is real, not a metrology artifact.Don’t average families together;worst familysets your risk.
4.1.9 Pocket checklist (print-side)
TEUSL/LSLset per family;height floorsenabledCp/Cpkcomputed on FA + first lot; orange/red families identifiedActions assigned:center μ(recipe/aperture),shrink σ(cleaning/foil/paste)SPI ↔ AOI overlay reviewed after changes; defect Pareto moved the right wayGolden limits/recipes updated if the fix sticks