Skip to main content

1.6 SPI Metrics & Closed Loop

SPI turns printing from an art into a controlled process by making paste volume measurable, trackable, and correctable in real time. With transfer efficiency as the anchor metric, it links aperture design, stencil performance, and printer stability into a single feedback loop. Automated limits, SPC charts, and closed-loop adjustments catch drift before it becomes scrap, while Pareto analysis points back to stencil fixes for lasting improvement. When trusted and maintained, SPI makes solder paste deposition as predictable as any other engineered process step.

1.6.1 SPI in one minute (what it’s for)

SPI (solder paste inspection) is your speedometer for printing. It measures volume, height, and area at every pad, compares results to limits, and—when you let it—tunes the printer (cleaning, pressure, speed, alignment) before small drifts become big defects.



1.6.2 The three numbers (and one that ties them together)

  • Volume (% of target) — the main quality knob.
  • Height (µm) — catches scooped/short prints and “paste on mask.”
  • Area (% of pad) — great for bridging and edge smear.
  • Transfer Efficiency (TE) = (Measured Volume ÷ Theoretical Volume) × 100%.
    • Theoretical Volume = aperture area × stencil thickness.

Pro tip: watch TE stability by feature family (chips vs QFN edges vs BGA). It’s the cleanest way to see process drift.



1.6.3 Limits by feature family (start here, tighten with data)

Use two tiers of limits: Yellow (warn & log) and Red (action). Tune per product after a few lots.

Feature family

TE target

Yellow band

Red band

Height floor

What trips “Bridge?”

Chips 0201–0603

100%

±15%

±25%

≥ 60–70% of stencil

Area > pad by +10–15% or contiguous blobs

Fine-pitch gull-wing (≤0.5 mm)

95–105%

±10%

±20%

≥ 60–70%

Any area growth into gap on 2+ neighbors

QFN/DFN edge pads

90–110%

±10%

±20%

≥ 60–70%

Same as above

QFN thermal pad (total)

50–65% coverage

45–70%

<45% or >70%

—

N/A (use total coverage)

BGA/CSP balls

90–110%

±10%

±20%

≥ 50–60%

N/A (use neighbor rules)

Large power pads / shields

80–100%

±15%

±25%

≥ 60–70%

Area “lips” beyond land outline

Why the fuss on QFN thermal? Too much paste ⇒ float/tilt; too little ⇒ voids/thermal trouble.



1.6.4 SPC without pain (how to chart and react)

  • Chart type: most lines use Individuals (I-MR) per panel or subgroups by quadrant.
  • Warm-up: ignore the first panel after a wipe or stencil change for control limits; then include it for history.
  • Western-Electric-lite rules:
    • Any Red point = stop the line and act (below).
    • 2 of 3 consecutive in Yellow on the same feature family = tighten cleaning or blade settings.
    • Trend of 6 up/down = check temperature/paste age and gasket.

What you plot on dashboards

  • % First-Print Pass (no reprint, no wipe)
  • Reprint Rate and Wipes per Board
  • TE Cpk by family (chips/QFN/BGA)
  • Defect Pareto (low volume, high volume, bridge, smear)



1.6.5 Closed-loop corrections (the simple decision tree)

Let SPI push small buttons automatically; reserve big ones for humans.

  1. Bridging alerts at a local region
    → Add understencil clean now (wet+vac if already did dry) → slow separation 10–20% on next prints in that region.
  2. Low TE on fine features across panel
    → Ease blade pressure slightly (−0.1 to −0.2 kg) → slow blade speed one step → confirm bead size (keep it narrow).
  3. High TE / smear
    → Increase angle or speed slightly → ensure pressure is not excessive (no gray film behind blade).
  4. Random lows in one quadrant
    → Local alignment correction on that quadrant → check board support pins under that area.
  5. Volume drift with time
    → Shorten cleaning interval; if still drifting, refresh bead; swap to a fresh jar (7.2).

Only change one knob at a time and record deltas in the job recipe. SPI should tag which auto-tweak was applied and whether it fixed the next panel.



1.6.6 Reprint & wipe policies (make them predictable)

  • Auto-reprint a region if Volume Red or Bridge is flagged and the board hasn’t moved.
  • Escalate to wipe if two reprints fail or if Bridging Red hits on adjacent apertures.
  • Panel reject only when three regions fail in Red after a wipe—otherwise you’ll kill throughput.



1.6.7 Trust but verify (measurement health)

  • Daily 3D calibration on the reference plate; weekly GR&R spot-checks on a golden panel (three ops × three repeats).
  • Disable “smart smoothing” that hides real variation; better to see the wobble and fix it.
  • Keep illumination profiles per mask color/finish—AOI lighting lessons apply to SPI as well.



1.6.8 Turning charts into better stencils (close the loop to design)

When a single aperture geometry dominates the SPI Pareto, fix the shape—not the printer:

  • Chips tombstoning → move to home-plate/inverted (7.4.2).
  • QFN thermal out of band → adjust window coverage or add chimneys (7.4.3).
  • BGA variation → confirm powder size and stencil reduction (7.1, 7.4.4).
    Document the tweak, spin the stencil, and watch the next lot’s TE Cpk climb.



1.6.9 First-Article & recipe lock

  • Run one FA panel, review SPI heat-maps, tweak one variable, reprint, then freeze angle/pressure/speed, separation, and cleaning cadence in the job file.
  • Capture golden SPI screenshots of tight features and store with the lot record so nights/weekends know what “good” looks like.



1.6.10 Release checklist (stick this on the printer)

  • TE/height/area limits set by family (table above) with Yellow/Red bands.
  • Closed-loop enabled: cleaning, pressure, speed, local alignment.
  • Reprint/wipe policy in the recipe (counts and escalation).
  • Calibration & GR&R schedule posted; golden panel available.
  • Dashboards up: First-Print Pass, Wipes/Board, TE Cpk, Pareto by family.
  • Design feedback path ready (tag risky apertures for review in 7.4).



Conclusion: Define clear SPI metrics, enforce closed-loop corrections, and feed findings into stencil design revisions. This keeps paste prints stable across shifts, reduces firefighting, and steadily drives yield upward.


Back to top