Design For Testing - Diagnostics

Diagnostics

Especially for advanced semiconductor technologies, it is expected some of the chips on each manufactured wafer contain defects that render them non-functional. The primary objective of testing is to find and separate those non-functional chips from the fully functional ones, meaning that one or more responses captured by the tester from a non-functional chip under test differ from the expected response. The percentage of chips that fail test, hence, should be closely related to the expected functional yield for that chip type. In reality, however, it is not uncommon that all chips of a new chip type arriving at the test floor for the first time fail (so called zero-yield situation). In that case, the chips have to go through a debug process that tries to identify the reason for the zero-yield situation. In other cases, the test fall-out (percentage of test fails) may be higher than expected/acceptable or fluctuate suddenly. Again, the chips have to be subjected to an analysis process to identify the reason for the excessive test fall-out.

In both cases, vital information about the nature of the underlying problem may be hidden in the way the chips fail during test. To facilitate better analysis, additional fail information beyond a simple pass/fail is collected into a fail log. The fail log typically contains information about when (e.g., tester cycle), where (e.g., at what tester channel), and how (e.g., logic value) the test failed. Diagnostics attempt to derive from the fail log at which logical/physical location inside the chip the problem most likely started. By running a large number of failures through the diagnostics process, called volume diagnostics, systematic failures can be identified.

In some cases (e.g., Printed circuit boards, Multi-Chip Modules (MCMs), embedded or stand-alone memories) it may be possible to repair a failing circuit under test. For that purpose diagnostics must quickly find the failing unit and create a work-order for repairing/replacing the failing unit.

DFT approaches can be more or less diagnostics-friendly. The related objectives of DFT are to facilitate/simplify fail data collection and diagnostics to an extent that can enable intelligent failure analysis (FA) sample selection, as well as improve the cost, accuracy, speed, and throughput of diagnostics and FA.