First, look at the void location. If voids at the center of the BGA bottom exceed 15% and voids on the QFN thermal pad exceed 25%, you can basically rule out PCB pad oxidation or stencil aperture issues, and directly dig into Solder Paste and reflow oven parameters. Don't rush to change the solder paste; first use an infrared thermometer attached to the board to measure the real-time temperature profile of three boards. You must use a K-type thermocouple fixed with high-temperature resistant adhesive, and cannot rely on the oven's built-in thermocouple readings, which are 1.2mm away from the conveyor surface, with actual deviations often reaching ±5.3°C.

Solder paste selection is not based on the manufacturer's advertised "low void" label, but on three hard parameters: metal content must be 88.5±0.3wt%, Flux softening point must be between 87°C and 92°C, and viscosity at 25°C must be 720±50Pa·s. Last week, the production line switched to a batch of Japanese solder paste with a nominal metal content of 89.2%, but the actual measurement was only 87.6%. As a result, the BGA void rate soared from 6.3% to 13.7%, and the solder joints removed during rework were gray, indicating obvious insufficient reduction.

The most misleading thing in reflow oven parameters is the set value. Put the oven temperature profiler under the board and run it to record the measured peak temperature of each zone. The heating slope of preheat zones 1 and 2 must be controlled between 1.5°C/s and 2.0°C/s. If the slope exceeds 2.2°C/s, the flux will boil violently and expel gas, but if it is less than 1.3°C/s, the solvent will not evaporate completely. The ten-zone oven used in our factory often has zone 3 measured 4.1°C higher than the set value, while zone 7 is 3.8°C lower. If this deviation is not calibrated, adjusting parameters is like catching sparrows blindfolded.

The time above liquidus (TAL) must be strictly controlled at 65 to 85 seconds. When using Sn96.5Ag3.0Cu0.5 alloy, the measured melting point is 217.2°C. The oven display of 217°C is not reliable; you must use a thermocouple to measure the actual solder joint temperature on the board surface. Once we found that the TAL was only 52 seconds, and it was found that the chain speed was increased from 70cm/min to 85cm/min without re-running the temperature profile. The peak temperature was still sufficient at 238°C, but the holding time was insufficient, the copper-tin intermetallic compound (IMC) did not grow enough, and voids accumulated at the interface layer.

The peak temperature cannot only refer to the broad range of 235°C to 245°C. For 0.4mm pitch BGA, it must be stable at 239°C±1°C; for large-area thermal pads on thick copper boards, it must be raised to 242.5°C and maintained for ≥12 seconds, otherwise the flux residue cannot push open the nitrogen protection layer. Last month, we tried 243.8°C, but the void rate increased instead, because the rosin in the solder paste cracked too fast, and carbonized particles blocked the exhaust channels.

Don't blindly believe in 99.99% nitrogen concentration. When the measured oxygen content exceeds 100ppm, the void rate starts to rise linearly. After exceeding 350ppm, for every 50ppm increase, the void area increases by about 2.1%. We switched to an online oxygen analyzer for monitoring and found that the sealing ring inside the pressure reducing valve of the gas storage tank was aging, causing air to mix into the incoming nitrogen. The measured oxygen content reached 620ppm. After stopping the machine and replacing the sealing ring, the void rate fell back to 5.4% on the same day.

Stencil thickness and aperture size will affect the solder paste release amount, but the main cause of voids is not here. For a 0.12mm stencil printing on a 0.3mm pad, if the solder paste volume after demolding is less than 0.012mm³, no matter how good the solder paste is, it cannot save the voids. However, this situation is very rare. In our data for the past six months, there were only 2 cases, both caused by the squeegee pressure being mistakenly adjusted from 4.5kg to 6.8kg, resulting in the solder paste being flattened.

Finally, check the Solder Paste Storage. After taking it out of the refrigerator, it must be warmed up for more than 4 hours. If the unopened solder paste is placed in a 25°C environment for more than 12 hours, the flux activity will decay by about 17%. A batch of solder paste was put into production after only 2 hours of warming up, and voids were concentrated on the solder joints of Chip 0402 components. Under the microscope, the flux residue was flocculent, and the oxide film was not completely dissolved.

Before changing the solder paste, first do a small batch verification. Take 20 boards, keep the oven temperature profile unchanged, only change the solder paste, and measure the void rate of 5 typical devices. The acceptance criterion is not an average value of less than 8%, but a maximum void rate of ≤11% on a single board, and no abnormal points exceeding 9% in 10 consecutive boards. Last time, we eliminated three suppliers according to this method, and the final selected one had a measured metal content of 88.47wt%, and the TAL fluctuation was controlled within ±3.2 seconds.