NovaSolver›Semiconductor Wire Bonding Pull Test Strength Simulator Back
Semiconductor Packaging
Semiconductor Wire Bonding Pull Test Strength Simulator
Estimate the pull-test strength of Au / Al / Cu / Pd-Cu bond wires that connect an IC die to a lead frame or PCB. Adjust wire material, diameter, loop shape and bonding conditions, and the quality margin against MIL-STD-883 and the loop-angle sensitivity update in real time — useful for early reliability design before sending samples to a bonder.
Parameters
Wire material
Sets UTS, density and kg-price automatically
Wire diameter
μm
Loop height
μm
Loop length
μm
Bond temperature
°C
Au: 150-200 °C, Cu: 175-250 °C typical
Ultrasonic power
mW
Bond force
gf
Results
—
Wire cross-section (μm²)
—
Wire UTS break force (mN)
—
Predicted pull-test (mN)
—
Quality margin vs MIL
—
Loop angle (deg)
—
Cost per bond (μUSD)
—
Wire-bond schematic — die, loop and pull tester
Visualises a ball bond on the IC die, the wire loop and the stitch bond on the lead frame being lifted by a pull-test hook. Colour shows quality margin (green = ample / orange = caution / red = below MIL).
Pull strength vs loop angle
Cross-section strength by material (F_wire at current diameter)
σ_UTS: wire tensile strength (MPa), A: cross-section (μm²), θ: loop angle = atan2(h, L/2). A lower loop has a smaller sin θ, which makes the apparent pull-test value larger (unless heel break dominates).
Heel strength is taken as 70% of the bare-wire UTS limit (empirical). The MIL-STD-883 method 2011 floor for 25 μm Au is 3 gf ≈ 29.4 mN. Margin > 2 is the typical production target.
Wire arc length L_wire (μm) and per-bond cost C_bond (μUSD). ρ: density, k_kg: USD/kg. Au at 19.3 g/cm³ with 30 μm × 1.5 mm gives ≈ 1.7 mUSD per bond.
Semiconductor Wire Bonding Pull Test — Strength & Reliability Design
🙋
Wire bonding is those tiny gold filaments you see inside an IC, right? Hundreds of them per chip — can such thin wires really be reliable?
🎓
Exactly. 25-50 μm gold wire (or, more often these days, copper) connects the die bond pads to the lead frame or substrate without putting torque on the chip. A single SoC in your phone has hundreds to a thousand bonds, and an automotive SiC power module has dozens of 300 μm aluminium heavy wires. A single 30 μm Au wire only breaks at about 155 mN (~15.8 gf), but that is still over 5× the MIL-STD-883 minimum of 3 gf. Reliability comes less from picking a strong wire and more from getting the bonder recipe (temperature, ultrasonic, force) right.
🙋
I see. When I stretch "loop length" on the left from 1500 to 3000 μm, the predicted pull-test value jumps up. Does longer mean stronger?
🎓
Good catch — it is actually the opposite, and this is a geometric quirk of the pull test. The hook lifts the loop vertically, but the wire tension F_wire is unchanged. What changes is the geometry: the tester sees F_pull = F_wire / sin(θ), so a low, long loop (small θ) reports a large pull-test value. That is why pull-test numbers without the loop height and length are meaningless. Notice this tool prints "loop angle" as a separate stat — engineers always quote pull strength together with loop height.
🙋
I've heard about heel, mid-wire and lift-off break modes. How do you use the distribution to improve a process?
🎓
The numeric pull value alone is half the story — the failure-mode mix is what matters. (1) Mid-wire break is the best case, meaning you have wrung out the full UTS of the wire. (2) Heel break (at the bond neck) tells you the USG power or bond force was too high and thinned the wire at the root during ball/stitch formation. (3) Ball or stitch lift-off (peeling off the pad) means too little temperature, oxidised pad or weak USG so the intermetallic (IMC) did not form. On a modern Kulicke & Soffa bonder you sweep USG in 5 mW steps and watch the transition "lift-off → heel → mid-wire" — you then freeze the recipe in the middle of that window.
🙋
Copper wire is spreading everywhere. How is it actually different from gold? Can you just swap them?
🎓
Cu is ~6,000× cheaper per kg than Au and has higher strength and conductivity — for cost and performance, Cu wins. Smartphones, memory and LCD drivers have basically all moved to Cu. But "swap" is the wrong word. Cu (a) oxidises, so the bonder needs an N2-95% / H2-5% forming-gas ambient, (b) is 2-3× harder than Au, which makes pad cratering on low-k pads (28 nm and below) a real risk, and (c) forms Cu3Al IMC that is more brittle than Au-Al and tends to delaminate after high-temperature storage. For advanced nodes a mid-tier Pd-Cu wire (Pd 1-2% coat, ~5,000 USD/kg) is used. Even the supply chain splits — Heraeus Maxsoft for Cu, Tanaka and MK Electron for Au.
🙋
What about the 300 μm Al heavy wires in power modules? Different reliability test too?
🎓
Yes. Automotive IGBT / SiC modules and high-current LED packages use 300-500 μm Al heavy wire. The applicable spec is no longer the 3 gf of MIL-STD-883 but JEDEC JESD22-B116 — typically 100 gf minimum. If you set this tool to "Al, 300 μm", the cross-section is 100× a 30 μm Au wire, so F_wire jumps into the 7 N range. But for heavy wire the dominant failure during power cycling (PCsec) is heel fatigue, so people cross-section the bond by SEM after thermal cycling rather than relying on initial pull alone.
Frequently Asked Questions
It is a reliability test in which a hook is engaged at the centre of a bonded wire loop and pulled vertically until the wire breaks. The de-facto standard worldwide is MIL-STD-883 Method 2011, with a minimum of 3 gf (about 29 mN) for 25 μm Au. The break mode is visually classified as one of five: (1) wire mid-span break, (2) heel break at the bond neck, (3) ball lift-off, (4) pad lift-off, (5) stitch lift-off. The bonding recipe is judged from this distribution. Dage and Nordson automated pull testers dominate the production-floor.
Au wire does not oxidise and delivers excellent long-term reliability, used for decades in military, space and medical packages, but it costs about 80,000 USD/kg. Cu wire has higher strength and better electrical / thermal conductivity than Au, costs only ~12 USD/kg, and has spread rapidly since the 2010s in logic ICs, memory and LCD drivers. Cu oxidises easily, so an N2-H2 forming-gas ambient at the bonder is mandatory. Pd-Cu wire (Pd 1-2% coating) curbs Cu oxidation and improves reliability on fine pads and at low-temperature storage tests; it sits between Au and bare Cu as a mid-tier option.
Because the hook lifts the wire vertically, the tension in the wire is the applied pull force divided by sin θ of the loop angle. A low, long loop (small θ) makes a small pull force generate a large wire tension, so the apparent pull-test value comes out low. A high loop (large θ) is hard to lift, so the pull-test value comes out high. This tool computes F_pull = F_wire / sin(θ) and shows the loop-angle sensitivity directly on the chart.
A heel (bond-neck) break is the signature of excessive ultrasonic energy or bond force during ball or stitch formation that mechanically thins the wire at the root. This tool assumes the heel strength is 70% of the bare-wire UTS limit and uses the smaller of wire-break or heel-break in the quality margin. Counter-measures: (1) lower the USG power by 50-100 mW, (2) lower the bond force by 5-10 gf, (3) optimise bond temperature (especially 150-200 °C for Au). On Kulicke & Soffa and ASM Pacific bonders the EFO current and USG power are tuned in 1 mA / 5 mW increments.
Real-World Applications
Smartphone / PC logic and memory: Volume packages for SoC, DRAM and NAND bond hundreds of 20-25 μm Cu / Pd-Cu wires per chip. Productivity reaches 10-20 bonds per second on a single Kulicke & Soffa IConn or ASM Pacific Eagle bonder. Pull tests are sampled as part of statistical process control (SPC); Cpk ≥ 1.33 is the typical gate for line qualification.
Automotive IGBT / SiC power modules: EV inverters and PCUs use dozens of 300-500 μm Al heavy wires, each carrying 100 A or more. The applicable pull-test spec is JEDEC JESD22-B116 with a 100 gf minimum, and SEM cross-section is needed after 1,000 cycles of −40 to +150 °C thermal cycling to check heel fatigue. Hesse Mechatronics and F&K Delvotec heavy-wire bonders are the typical equipment.
LED packages and image sensors: White LEDs and CMOS image sensors run pull tests on 30-50 μm Au wires before and after transparent-resin encapsulation, checking that the loop still meets MIL even after being pushed down by resin cure shrinkage. The loop angle θ changes with resin cure, so the loop-angle sensitivity in this tool maps directly to a real failure mode.
Failure analysis and yield ramp: When automotive modules die on thermal shock, or memory shows opens after reflow, pull test plus break-mode classification is the first knife-cut between "wire problem" and "IMC problem". A simple numeric model like this tool lets you sanity-check the impact of material, diameter and loop change before cutting metal — easily saving 2-3 prototype iterations.
Common Misconceptions and Pitfalls
The first trap is assuming that "a higher pull-test number is always a better bond". As shown above, the number is set by loop geometry: a low, long loop reads small, a steep loop reads large. The same wire and the same bond conditions can read 3× apart just by changing the loop profile. Specifications must always quote pull strength together with loop height X μm and loop length Y μm, and you should never compare numbers from different geometries. Modern practice reports pull together with shear (push on the ball in its diameter direction).
Next, the belief that "Cu is stronger than Au, so it is a drop-in replacement". Cu UTS is 320 MPa (1.45× Au's 220 MPa), and pull-test numbers alone make it look "better than Au". But Cu is also 2-3× harder, so pushing the same ultrasonic energy through it cracks porous-SiO2 low-k pads on advanced logic — the classic "pad cratering" failure. When you move to Cu, expect to drop the USG power by 20-40 %, raise the temperature by 30-50 °C and rebuild the recipe. Try moving the "ultrasonic power" and "bond temperature" sliders together on this tool to feel out the trade-off.
Finally, passing a pull test alone is not enough. The pull test probes the wire and its bond interface, but the long-term reliability tests (HTOL, HTSL, thermal cycling) are dominated by Au-Al or Cu-Al intermetallic (IMC) growth, which is a completely different physics from the initial pull number. IMC Kirkendall voids are visible only in SEM cross-section, and parts with healthy initial pull have been known to open after 1,000 h of storage at 175 °C. In practice you run pull + shear + HTSL 1000 h + SEM cross-section as a set. This tool covers the initial-strength estimate; the long-term piece needs separate accelerated-life testing.
How to Use
Select wire material (Au, Al, Cu, or Pd-Cu) from the dropdown—gold offers best reliability but highest cost (~$0.15/bond); aluminum minimizes expense at ~$0.02/bond.
Enter wire diameter (17–25 μm typical for 0.5 mm pitch BGA), loop height (50–150 μm), and loop length (300–800 μm) matching your bond-pad geometry.
Input bonding temperature (150–300°C depending on material); higher temperatures improve atomic diffusion but risk die-attach degradation above 250°C for organic substrates.
Review predicted pull-test strength in mN and compare against MIL-STD-883 Method 2011.9 minimum (25–50 mN depending on wire gauge and material).
Worked Example
Gold wire: 20 μm diameter (314 μm² cross-section), 100 μm loop height, 500 μm loop length, bonded at 200°C. Simulator predicts wire UTS break force ~43 mN with 35–38 mN pull-test strength after accounting for loop-length stress concentration (typical K-factor 0.82–0.88). Quality margin vs MIL-STD-883 target = 42% surplus. Aluminum wire (same geometry, bonded 180°C) yields ~28 mN pull-test with marginal 15% headroom, suitable only for non-critical applications.
Practical Notes
Loop angle exceeding 70° indicates risk of inter-wire short circuits in high-density layouts; adjust loop height downward or reduce bonding temperature to stiffen loop.
Thermal cycling (−40 to +125°C, 500 cycles per IPC-9701) degrades Pd-Cu bonds by 8–12%; Au and Al remain stable, justifying premium pricing in automotive/aerospace.
Pull-test failure mode transitions from neck fracture (ductile, ~5–8% elongation) to interfacial lift-off at high loop-length-to-diameter ratios above 40:1; redesign geometry if margin drops below 20%.
Cost per bond includes material (~$0.02–0.15), equipment overhead (~$0.03), and bond-pull inspection (~$0.01); simulations enable rapid cost-down optimization without physical prototyping.