Issue link: https://iconnect007.uberflip.com/i/1440051
94 SMT007 MAGAZINE I JANUARY 2022 the component. e only tests that are look- ing for external evidence are the chemical tests (FTIR, chemical) which can detect black-top- ping and clones and are also covered by most of the other internal examinations and the EVI. e EVI is an optical examination at 3X to 100X. eoretically, a high-potency EVI is suf- ficient for detecting all counterfeit types. It raises the question: Why are all the oth- er methods part of the counterfeit mitigation standard? Unfortunately, this is because the conventional EVI alone is not potent enough to detect many of the counterfeit evidences there are in counterfeit components. Solderability Estimation Electronic component solderability is an es- sential capability in the electronics manufac- turing domain. It stands for the ability of an electronic component to be reliably and re- peatably soldered onto a circuit board in an automated production environment 1 . ere are several available methods to inspect elec- tronic components 2,4 . To obtain a reliable and repeatable soldering process, all the produc- tion parameters are utilized and optimized by the electronics manufacturer. is process in- cludes managing the board preparations, sol- der paste dispensing, component placement, and reflow parameters. However, the conditions and the solderabil- ity of the soldering leads in the electronic com- ponents are rarely evaluated before assembly. is is because of the assumption that the pro- ficiency of the process is sufficient to mitigate the variations within the acceptable parame- ters of the component leads. In the defense in- dustry, it is required that samples from compo- nent batch are inspected for solderability ac- cording to MIL-STD-202, Method 208 7,8 , and in other industries such as the automotive ISO- 26262 standard 9 . Typically, samples of the test- ed components are selected and tested under specified conditions to gauge the solder wet- ting ability on the component leads. e eval- uation process is by visual examination of the solder coverage on the leads aer the solder dip process. is conventional process is man- ual, labor-intensive, expensive, and performed on samples only. Soldering quality assessment aer the as- sembly as a part of the automatic optical in- spection (AOI) was presented 10 while ma- chine learning convolutional neural network (CNN) was performed 11,12 . e average true positive accuracy of manually detecting the soldering faults with the aid of a magnifying glass is almost 90% and the faults detection accuracy using CNN, in this case, was 84%. e visual processing of the soldering quality based on CNN can be powerful in achieving human-level accuracy if they are trained on a large dataset with an equal number of diverse examples. Unfortunately, in many cases, this is not suf- ficient to ensure a reliable bond for all the as- sembled components. is is because the un- derlying assumption is that sampling one com- ponent out of a batch represents the entire population. At the same time, there are many cases where this assumption yields to the re- Theoretically, a high-potency EVI is sufficient for detecting all counterfeit types. Figure 4: An illustration of the effect of surface roughness on emissivity.