Scan based two-pattern tests: should they target opens instead of TDFs?

We make the case that TDF timing tests, even when aggressively applied at-speed, uniquely detect mostly open defects within standard cells. The majority of these defects can also be detected at somewhat slower test speeds without the risk of unnecessary yield loss from test noise. Meanwhile, many other opens that can cause operational failures remain undetected by current LOC, and even LOS, TDF tests. These can significantly degrade product quality. We therefore argue that it may be better to explicitly target all open faults in the circuit, with the tests being applied at the highest possible speed consistent with limiting yield loss and supporting high test compression efficiency. TDFs will implicitly be covered by such an approach, which can potentially reduce test costs and improve test quality.