Joint Quality, Inspection and Failure
What a good joint looks like, defects and causes, inspection, voids, whiskers, thermal fatigue and cleanliness.
What a good joint looks like
A good joint is judged by its wetting (a smooth, shallow angle where the solder meets pad and lead), its shape (a concave fillet that follows the lead) and its surface (smooth and shiny for leaded solder; satin or slightly grainy for lead-free). Under the surface it has a continuous thin intermetallic layer and no cracks or large voids.
- Fillet: concave, covers the pad and rises up the lead, with the outline of the lead visible.
- Contact angle: under 30° is excellent; over 90° means the solder has not wetted.
- Surface: bright and smooth for tin-lead; lead-free is acceptable if smooth and free of cracks, though duller.
- Barrel fill for through-hole: at least 75 % of the hole filled, with solder visible on both sides.
- No damage: no burnt board or insulation, lifted pad, scratched track or cracked part.
Defects, causes and cures
| Defect | Appearance | Cause | Cure |
|---|---|---|---|
| Cold joint | Dull, grainy, lumpy | Not hot enough, or the joint moved while freezing | Reflow with flux and hold still |
| Dry (insufficient) joint | Too little solder, thin fillet | Too little solder or heat; poor wetting | Add solder and flux |
| Excess solder | Convex ball, the lead hidden | Too much solder | Wick off the excess |
| Bridge | Solder between adjacent pins or pads | Too much solder, dragged tip | Flux and wick or drag |
| Non-wetting | Solder balls on the pad, base metal visible | Oxide, contamination, no flux | Clean, flux, reheat |
| De-wetting | Solder first covers, then retreats leaving islands | Contaminated or poorly plated surface | Clean and re-plate/tin |
| Solder ball / splatter | Tiny balls around the joint | Wet paste or flux; fast heating | Dry boards; slower ramp |
| Tombstone | Chip standing on one end | Uneven heating (see Surface-Mount Soldering) | Even heat; same paste on both pads |
| Insufficient barrel fill | Hole not full, solder only on one side | Not hot enough through the board; ground plane | Preheat; more heat time; flux |
| Lifted pad / burnt board | Pad detached, brown board | Too much heat or time | Repair (see Rework, Desoldering and Repair) |
| Cracked joint | Hairline crack around the lead or pad | Mechanical stress, vibration, thermal fatigue | Reflow; add strain relief |
| Disturbed joint | Rough, rippled, frosty | Moved while freezing | Reflow and hold still |
| Head-in-pillow (BGA) | Ball and paste have touched but not merged | Warped part or board, poor wetting | Better profile; flux; X-ray to find |
Acceptance standards: IPC classes
The IPC-A-610 standard (the visual criteria) and J-STD-001 (the process requirements) divide products into three classes, and the same joint may be acceptable in class 1 but not in class 3.
| Class | Product | Barrel fill (through-hole) | Typical wetting criterion |
|---|---|---|---|
| 1 | General electronics; function matters, appearance less so | 50 % | Evidence of wetting |
| 2 | Dedicated service electronics (computers, instruments, industrial) | 75 % | Lead and pad wetted, concave fillet |
| 3 | High performance and harsh environment (medical, aerospace, automotive safety) | 75 % (100 % of the barrel on the primary side in some cases) | Full wetting, defined fillet, no exposed base metal |
Worked example: checking barrel fill
- A 1.6 mm board with a 75 % barrel fill needs solder at least 1.2 mm up the hole from the bottom (for class 2). A joint that has solder to only 0.8 mm (50 %) passes class 1 but fails class 2.
- In volume terms (see Through-Hole Soldering) the 1.0 mm hole around a 0.8 mm lead holds 0.45 mm³ when full; 75 % is 0.34 mm³.
The numbers shown are typical readings of the standard; buy and read the current IPC documents for contract work.
Inspection: eyes, magnification, X-ray and tests
| Method | What it finds | Notes |
|---|---|---|
| Visual with magnification (3 to 10×, a microscope for fine pitch) | Wetting, shape, bridges, tombstones, damage | The first and cheapest; IPC allows 1.5× to 20× depending on the pad size |
| Automated optical inspection (AOI) | Presence, position, polarity, bridges, fillets | Production; sees only what is visible |
| X-ray (2D and 3D) | BGA, QFN voids, hidden bridges, barrel fill | Needed for hidden joints |
| Electrical test (in-circuit, flying probe, functional) | Opens, shorts, wrong values | Finds faults but not marginal joints |
| Resistance measurement | High-resistance joints | Four-wire; a few milliohms |
| Shear and pull tests | Joint strength | Sample basis, destructive |
| Cross-section (micro-section) | Intermetallic layer, cracks, voids, fill | Destructive; for qualification |
| Dye-and-pry, thermal imaging | Cracked BGA joints; hot joints | Failure analysis |
- Inspect under good, angled light, at the right magnification: a hairline crack hides at 2×.
- Wiggle-test suspect parts with a plastic probe while the circuit runs; intermittent faults often live in a cold joint.
- Look for the pattern: every joint on a row defective points to the process; one bad joint points to the operator or the pad.
Voids
Voids are gas bubbles trapped in the solder as the flux outgasses. A few small voids do little harm, but large ones reduce the conducting area, the strength and, most importantly, the heat conduction of thermal pads. BGAs usually allow up to 25 % void area by IPC-7095; QFN thermal pads often 50 % or less; the customer’s specification decides.
Worked example: voids and chip temperature
- A thermal pad has a solder thermal resistance of 2.0 K/W with no voids. A void fraction v blocks the heat path, so R = R0 ÷ (1 − v).
- The chip dissipates 3.0 W. With 30 % voids R = 2.86 K/W and the temperature rise across the joint is 8.6 K instead of 6.0 K: 2.6 K hotter (a simplification: voids at the edge matter less than at the center).
| Void area | Thermal resistance | Temperature rise at 3 W | Extra temperature |
|---|---|---|---|
| 0 % | 2.00 K/W | 6.0 K | 0.0 K |
| 10 % | 2.22 K/W | 6.7 K | 0.7 K |
| 25 % | 2.67 K/W | 8.0 K | 2.0 K |
| 30 % | 2.86 K/W | 8.6 K | 2.6 K |
| 50 % | 4.00 K/W | 12.0 K | 6.0 K |
Reduce voids with a longer soak (so the flux outgasses before the solder melts), less paste, a window-pane stencil on large pads, fresh dry paste and vacuum reflow for critical products.
Tin whiskers
Pure tin and high-tin plating can grow hair-thin crystals (whiskers) of a few micrometers in diameter and up to some millimeters long, driven by compressive stress in the plating. They can bridge adjacent conductors and cause short circuits, usually small, intermittent and self-fusing at a few tens of milliamps, but dangerous on fine pitch and in high-reliability systems. Tin-lead alloys with over about 3 % lead nearly stop the effect, which is one reason why lead-free brought the problem back.
Worked example: can a whisker bridge the gap?
- The gap between adjacent pins is the pitch minus the pad width (see Surface-Mount Soldering). A whisker 0.3 mm long can bridge a gap shorter than that.
| Package | Gap | Result |
|---|---|---|
| 1.27 mm pitch SOIC | 0.67 mm | whisker of 0.3 mm cannot reach |
| 0.65 mm pitch | 0.25 mm | whisker of 0.3 mm can bridge |
| 0.5 mm pitch QFP | 0.20 mm | whisker of 0.3 mm can bridge |
| 0.4 mm pitch QFP | 0.17 mm | whisker of 0.3 mm can bridge |
- Mitigation: matte (not bright) tin over a nickel underlayer; annealing after plating; conformal coating; keep the pitch large; or use tin-lead where it is permitted (for example in the military and aerospace exemptions).
- Whiskers are seen on the plating of leads and connectors, not usually on reflowed solder.
Thermal fatigue and why joints fail
A solder joint joins two materials that expand differently: a ceramic chip (about 6 ppm/°C) on an FR-4 board (about 15 ppm/°C), or a leaded part on a rigid board. Each time the temperature cycles, the joint is stretched and compressed through a small shear strain, and the solder creeps and fatigues. After enough cycles it cracks. The failures to expect are:
| Failure | Cause | Prevention |
|---|---|---|
| Thermal fatigue | Temperature cycles with mismatched expansion | Compliant leads, smaller parts, underfill, low-CTE boards |
| Vibration and shock | Heavy parts on small joints | Glue or clip heavy parts; boards stiff and supported |
| Creep | Steady stress at elevated temperature | Reduce stress; lower temperature |
| Brittle intermetallic fracture | Thick intermetallic after long or repeated heating | Correct profile; limit rework (see Rework, Desoldering and Repair) |
| Electromigration, corrosion | Flux residue, moisture, bias voltage | Clean; coat; correct flux |
| Mechanical overload | Connector pulled; board flexed | Strain relief; board support |
Nfield ÷ Ntest = (ΔTtest ÷ ΔTfield)1.9 × (ffield ÷ ftest)1/3 × exp[1414 × (1/Tmax,field − 1/Tmax,test)]
Worked example: what does a 1 000-cycle test prove?
- Test: −40 to +125 °C (ΔT = 165 K), 24 cycles per day, peak 398 K. Field: 20 to 60 °C (ΔT = 40 K) once a day, peak 333 K.
- (165 ÷ 40)1.9 = 14.8; (1 ÷ 24)1/3 = 0.35; exp[1414 × (1/333.15 − 1/398.15)] = 2.00. The acceleration factor is 14.8 × 0.35 × 2.00 = 10.2.
- So 1 000 test cycles stand for about 10 237 field cycles, which at one a day is 28 years. The model is a guide for SnPb, not a guarantee: always test the real assembly.
Cleanliness and corrosion
Flux residue, fingerprints and ionic salts on a board attract moisture and become conductive, causing leakage, dendrite growth and corrosion. The traditional limit for ionic contamination by the ROSE test is about 1.56 µg of sodium chloride equivalent per square centimeter; modern practice (J-STD-001) is to set limits for the process by measuring the surface insulation resistance (SIR).
Worked example: contamination budget for a board
- A 100 × 80 mm board (both sides) has about 160 cm² of surface; at 1.56 µg/cm² the allowed contamination is about 250 µg of NaCl equivalent in total.
- No-clean flux residue left unactivated, or a thumbprint, can exceed that on its own, so handle boards by the edges, with gloves, and clean any board that goes into humid or high-impedance service.
Counting defects
A board with hundreds of joints can have a defect rate that sounds tiny but still matters. The industry measures it in defects per million opportunities (DPMO).
Worked example: yield from the defect rate
- At 200 DPMO and 400 joints, the chance that a board has no defect is (1 − 200×10−6)400 = 92.3 %; 7.7 % of boards have at least one defect.
- At 50 DPMO the good fraction rises to 98.0 %.
- To be 95 % confident that the defect rate is under 1 %, you must inspect 299 joints with none bad; for 0.1 % you need 2 995.
Troubleshooting
| Symptom | Likely cause | What to do |
|---|---|---|
| Intermittent fault that comes and goes with tapping | Cold or cracked joint | Wiggle test; reflow suspect joints with flux |
| Fault appears when warm | Cracked joint opening with expansion | Heat/cool the area with a gun and freeze spray; reflow |
| Board fails after a year in service | Fatigue, corrosion or flux residue | Cross-section; clean; add strain relief |
| Many joints dull and grainy | Cooling or movement problem in the process | Check the profile and handling |
| Shorts on fine-pitch pins | Bridges or tin whiskers | Inspect under magnification; clean; coat |
| Chip runs hot | Voided thermal pad | X-ray; adjust stencil and profile |
Related
- Soldering: the overview of this guide.
- How Soldering Works: wetting and the intermetallic layer.
- Rework, Desoldering and Repair: fixing defects.
- Design, Testing and Troubleshooting: finding the faulty joint.
FAQ
How can I tell if a joint is cold?
It looks dull, grainy, rough or lumpy, with a high contact angle; a good leaded joint is smooth and shiny. A cold joint may still look fine and fail on a wiggle test.
Are lead-free joints always dull?
They are usually less shiny than tin-lead but should still be smooth with no cracks. A rough, frosty or crumbly appearance is a fault.
What is a good barrel fill?
At least 75 % of the barrel height for class 2 and 3 products (IPC-A-610), with solder visible on both sides.
Do I need to X-ray my boards?
Only for hidden joints (BGA, QFN pads) and critical products. Visual inspection is enough for through-hole and leaded SMD parts.
Why do solder joints crack after years in service?
Thermal cycling and vibration fatigue the solder where parts of different expansion are joined. Strain relief and compliant leads delay it.
References
- IPC-A-610, Acceptability of Electronic Assemblies; IPC J-STD-001; IPC-7095, Design and Assembly Process Implementation for BGAs.
- Norris, K. C. and Landzberg, A. H., “Reliability of Controlled Collapse Interconnections”, IBM Journal of Research and Development 13 (1969).
- JEDEC JESD201, Environmental Acceptance Requirements for Tin Whisker Susceptibility; NASA tin whisker website (nepp.nasa.gov/whisker).
Last reviewed: October 2026. Acceptance limits, void limits and fatigue constants are typical: use the standard and data that apply to your product.