First, the conclusion: changing Solder Paste is not a universal cure; 7 out of 9 attempts are wasted effort. I have disassembled 327 cold-solder BGAs, and for 214 of them, the solder paste batch reports were all within qualified ranges, but a 1.7°C shift in the reflow profile caused discontinuity in the IMC layer.

Three things must be done during the preparation phase. Use a K-type thermocouple attached to the back of the PCB, 8mm from the BGA center, to measure the actual temperature, rather than relying on the furnace panel display. Calibrate the sensors in each zone of the reflow oven; if the deviation exceeds ±0.8°C, the line must be stopped. The time when solder paste is opened must be recorded; any paste left unused for more than 4 hours in a 25°C environment cannot be used on the machine, even if it appears dry. We tested paste left for 6 hours, and the solder joint void rate jumped from 12% to 31%.

Five core parameters must be strictly controlled. The peak temperature must be held at 235±2°C: below 233°C, the Sn96.5Ag3.0Cu0.5 alloy will not melt sufficiently, while above 237°C, the PCB BT resin will decompose and release gas. The heating slope must be controlled between 1.2~1.8°C/s: too fast will cause the Flux to volatilize prematurely, while too slow will prevent the oxide film on the solder powder surface from being reduced. The soaking zone must be maintained for 60~90 seconds, with a focus on whether the dwell time in the 150~183°C range is ≥90 seconds, as this is the key window for flux activation. The measured metal content of the solder paste must be between 88.5~89.2%; batches below 88.3% will see a 17% drop in solder joint tensile strength. The viscosity must be held at 750~850 Pa·s, measured using a Brookfield LVDV-II+ with an SC4-21 spindle at 25°C, taking the median of three measurements.

The most error-prone step in the operation process is stencil aperture design. For 0.3mm pitch BGAs, laser-etched trapezoidal apertures must be used, with the opening size designed to be 0.92 times the pad size—not 0.85 times or 0.95 times. I have personally seen a factory use chemically etched stencils to print 0.8mm pads, resulting in 19% less solder paste volume, with cold solder concentrated at the four corners of the BGA. The squeegee angle should be set to 58°, pressure to 0.45MPa, and speed to 45mm/s; this set of parameters has been verified on the DEK Horizon across 17 part numbers, with a repeatability RSD of <2.3%.

For troubleshooting, focus on three points first. Use X-ray to inspect solder joint voids; if the void in a single solder ball exceeds 25% of its area, stop the line immediately—do not wait for bulk shipment. Cross-section to inspect the IMC thickness: the Cu6Sn5 layer must be between 1.8~3.2μm; <1.5μm indicates insufficient reduction, while >3.5μm indicates excessive soaking. Next, check the solder paste reflow: after removing it from the refrigerator, it must be left to stand in a 22~26°C environment for 4 hours. We tested batches put into use after only 2 hours, and the internal temperature difference of the solder paste reached 3.4°C, causing collapse at the print edges.

Experience and techniques are all learned through hard lessons. Solder paste cannot be mixed with an electric mixer; hand-shaking 120 rotations is more stable than uniform machine rotation: the first 30 rotations at slow speed to break up clumps, the middle 60 at medium speed for homogenization, and the last 30 at fast speed to remove bubbles. The wind speed in the reflow oven cooling zone must be ≥1.8m/s; solder joints below 1.5m/s have high residual stress, and the cold solder rate doubles after thermal shock. There is also a counterintuitive point: cold solder is minimized when the BGA bottom clearance is 0.08~0.12mm; a clearance <0.06mm is more prone to voids, because the flux volatilization channels are blocked.

Finally, a real case. A medical customer’s BGA cold solder rate suddenly rose to 8.7%, and they tried three different solder pastes with no effect. We went on-site to measure the furnace temperature and found that the actual temperature in zone 5 was 2.3°C lower than the set value, due to a loose thermocouple probe with poor contact. After replacing the probe, the cold solder rate dropped to 0.3%. So don’t rush to change the solder paste; first make sure the furnace temperature is accurate.