Cypress
Semiconductor
Texas
2001-2008
A simplier solution is simply to shift the probe card over and retest questionable die results with another probe card site/needle for confirmation.

If it turns out that too many die are failing with one probe card needle but are passing with another, the operators have to weigh the time it would take to halt testing and repair the probe card, against allowing die retest to continue - albeit taking time to shift the probe card and retest die, but recovering good die none the less.

The purpose of this project was to perform the above function: to shift the probe card over and retest suspicious die. The section on die retest describes the main plan file parameters available to engineering to (statistically identify questionable failures and) automatically turn further die retesting on and off.
Learn More about Falsely Failing Die
As a probe card containing a collection of needles (or sites) gets shifted a cross a wafer to test groups of die, sometimes a very fine piece of dust, or even a slight oxidation of the metal pad surface, can interfere with the electrical connection needed for testing.

When a probe card needle or site just drops out, it is obvious. There will be a consistent pattern of functional die failures across the wafer, most often with the same bincode (a number, usually 0-15, which indicates the failing functional test). At other times however, the effect is temporary, resulting in a higher percentage of failures for that probe card site - even though the die would have tested out as being functionally good, had they been probed with one of the other needles on the probe card.

Some ways of getting around this problem are to polish or automatically clean the probe card needles after some number of die are probed. Another way is to clean and realign the probe card neeedles more often (which is time consuming).