To make matters worse, some cloned parts exhibited better performance than authentic parts. SMT engineers tested ten samples of an authentic shift register and ten cloned samples of the same device. Figure 2 shows test results indicating a better high-voltage variation among the cloned parts. "The counterfeiter actually took the time to dial in this spec better that the OCM." That statement drew a chuckle from the audience. Sharpe said the cloned parts were likely made with newer technology than the older authentic parts, hence the better test result. Sharpe gave another example of a Schmitt Tigger where the cloned part outperformed the authentic part by having a wider hysteresis. Other test such as performance over temperature may show that the cloned parts are inferior, though Sharpe didn’t discuss that. A significant difference between early counterfeit parts and today's clones is that these counterfeit parts are made from scratch in clean rooms.
Figure 2. A sampling of ten cloned and ten authentic shift registers revealed better performance by the clone, at least for this parameter. Click image to enlarge.
Cloned parts are not the only problem. Sharpe noted a situation where an authentic part had been tampered. It was an obsolete part purchased on the open market for a defense contractor. SMT engineers identified the part as a possible fake because of the packaging quality. Digging deeper, they found that the counterfeiter needed to gain access to the die to erase the programming. Sharpe believes that the intent here was to sell the part as a new blank.
Early suspected cloned parts could be identified by their inferior packaging, but that's no longer the case, explained Sharpe. In Figure 3, you could argue that the cloned MM74C923 (bottom) looked better than the authentic Fairchild part. The counterfeiter even copied the "drip" in the center of the "MM" in the part number.
Figure 3. By 2016, Cloned Parts were looking better packaged than authentic parts. Click image to enlarge.
In 2017, SMT was asked to test 20,000 pieces of a supposed Altera device used to program FPGAs on power-up. After parts had passed electrical tests, Engineers looked at the packages, which showed no rework and every mechanical dimension was within manufacturer's specs. These were new parts.