Process-Level Problem Solving on the Factory Floor

"After observing this soldering work for a few days, I felt it was not a very good technique, because the soldering tool’s temperature was very high and the operators’ experience and skill levels varied. Some new operators had to keep the soldering tool in contact with the electrode for quite a long time before completing the soldering, and the high temperature of the soldering tool was very likely to affect the chemical structure inside the transistor as a result."

"I dug out the heat transfer textbook I had studied at MIT and did some rough calculations, and found that my concern was correct. So over the next few days, I tried an indirect heating method: not letting the soldering tool directly contact the electrode, but only letting it contact the copper wire, using copper’s high thermal conductivity to melt part of the electrode and complete the soldering. My method was slower than the original, but the likelihood of disrupting the transistor’s internal chemistry should be lower than with the original method, so the final yield should be higher. After I myself became proficient in operating my soldering method, I began training the two most experienced operators. After one or two days, their soldering speed using the new method had reached 80–90% of the original method. We accumulated several hundred transistors soldered using the new method and compared the yield with another group of transistors soldered using the original method. Sure enough, the yield of the new method was noticeably higher than that of the original method. My supervisor came over to take a look, and the production manager also came to see it, and even sat down and asked me to teach him the new soldering method. A few days later, the entire production line switched to my method."

"Finally one day, I was talking with the foreman when I suddenly heard a tester shouting from twenty feet away. The foreman and I immediately rushed over. While we were hurrying over, the tester had already stood up and was waving his arms excitedly. Operators from the same line also gathered around the tester. We guessed there must be good news. Sure enough, that batch of products actually had a 40% pass rate, and the tester was so excited he couldn’t even speak. A few minutes later, Luce came running over with a big smile—he had already heard the good news. A few minutes after that, my supervisor also came running over with a big smile. The question they cared about most was: “Do you remember how this batch was done?” Of course I remembered. Not only did I remember, I had also written it down in my notebook. That day the full-day yield was 25%, several times higher than any previous day. Everyone probably remembers the happiest moment of their life; that day was the happiest day in my twenty-seven years of life. It was September 1958, four months after I joined TI."

"That night I did not sleep at all—partly from excitement, partly from fear. Anyone familiar with semiconductors knows that semiconductor yield is not very stable. Reaching 25% on one day does not necessarily mean reaching 25% every day afterward. Fortunately, our process control was quite rigorous. After that, yields went up and down, but the weekly and monthly averages were always above 20%. A few months later, we made further improvements to the design and process, and the yield rose a few more percentage points. One year later, after continuous ongoing improvement, the yield was very stable at above 30%."