Semiconductor Technology from A to Z

After all front-end process steps are complete, the wafer contains a large number of finished but still unseparated dies. Before the wafer is diced, every single die is electrically tested while still on the wafer – this step is known as wafer test or probe test.

The test serves two purposes: first, it identifies defective dies so they are not unnecessarily processed further in the subsequent assembly steps. Second, it provides early electrical characterization data, such as threshold voltages or leakage currents, which can be used for process control.

A test system essentially consists of three components:

• Wafer prober: a precision tool that positions the wafer and moves it die by die under the test needles
• Probe card: the actual contacting unit with fine needles that land on each die contact pads
• Tester: the electronic test equipment that generates test patterns and evaluates the circuit response

Since modern chips can have several hundred to several thousand contact pads, the probe card must feature a correspondingly large number of precisely positioned needles that reliably make contact without damaging the sensitive chip surface.

The probe card forms the mechanical and electrical interface between tester and wafer. Older designs use free-standing needles made of tungsten or a beryllium-copper alloy, mounted radially on a ring. For very fine pad pitches, MEMS-based probe cards are increasingly used today, whose contact tips are manufactured using lithographic processes from silicon or metal – a method that allows significantly higher positioning accuracy.

When the needles land on the pads, a small but deliberate force (overdrive) is applied that breaks through the native oxide layer on the pad surface, ensuring a low-resistance electrical contact. However, excessive overdrive damages the pads and can harm underlying layers – tuning this force is therefore a critical process parameter.

To increase throughput, modern test systems frequently contact multiple dies simultaneously (multi-site testing). This allows several chips to be tested in parallel, substantially reducing the test time per wafer.

The result of each individual test is stored with its exact position in a so-called wafer map – a digital map that records, for every die on the wafer, whether it passed the test (pass) or not (fail). This map is referred to as a yield map and accompanies the wafer through all subsequent stages of production.

The yield map serves several important functions:

• During the subsequent dicing and assembly processes, only functional dies are further processed based on the map, while defective ones are discarded
• Spatial patterns in the yield map (such as clustered failures near the wafer edge) provide clues to systematic process issues
• For memory devices (DRAM, NAND flash), the map enables the targeted use of redundant circuit blocks: if a die contains individual defective memory cells, these can be electrically replaced by built-in spare circuits (redundancy), allowing the die to remain usable overall

The average yield of a wafer – that is, the proportion of functional dies relative to the total number – is one of the most important metrics in semiconductor manufacturing, as it directly determines the production cost per chip.