Rubbish datalogs may not be rubbish afterall
![image](http://a.storyblok.com/f/90669/1200x800/86859032bf/untitled-design-13.png/m/filters:quality\(60\))
We often get requests to “clean up” the database and remove the rubbish datalogs. These datalogs are those that are trial runs at the subcon test floor or retest attempts of QA units that fail.
While a “clean” database may be easier on the eyes during search and during analysis, the product engineer may be missing out a very important source of information on what happened during testing at the subcon if the rubbish datalogs are removed from the database. Why is this so?
It is because repeated setup attempts at production are revealed when the product engineer sees many small datalog files with a few units each, mostly failing one or more tests. And this may indicate issues with testers, handler interface board, test programs and other components of the test setup. It is important for product engineers and test engineers to understand what is causing setup problems at the subcons because such problems may result to low yield or low production rate or units per hour (UPH).
More serious would be several attempts to retest QA failures. If such attempts occur frequently in several lots then there may be a limits or spec issue. Maybe the test limits are too wide and not accounting for setup variation. Or maybe there is a test program bug causing too much variation resulting to QA failures.
We have a very extensive search filtering capability which allows the user to filter out the “rubbish” datalogs in order to perform yield analysis without them. The search filter also allows the user to show only the rubbish datalogs to analyze possible setup issues.
yieldHUB recommends that semiconductor companies store the rubbish datalogs in the yield management database if the yield management system, like yieldHUB, is capable of filtering them out during search if needed. Manufacturing issues at the subcon may be inferred from these datalogs. These also allow the user to estimate the duration of downtimes using the lot process analysis tool of yieldHUB and thus reveal major manufacturing inefficiencies.