Access Official Technical PDF

Share Summary:LinkedInEmail

In fan-out wafer-level packaging and fan-out panel-level packaging, lithography must account for die that are not positioned exactly where the design expects. Placement and molding introduce translation and rotation errors, while subsequent processing can affect the molded substrate. Optimized stepping addresses this challenge by combining external die-placement measurements with overlay correction and yield prediction, allowing exposure-field size to adapt to measured conditions.

Rather than treating every die as an isolated exposure, the method evaluates groups of dies and selects field sizes intended to meet a specified registration target. For packaging engineers and operations teams, that creates a way to consider lithography throughput and overlay yield together.

Why die placement errors matter in fan-out packaging

Fan-out processes place known-good dies on a reconstituted wafer or panel. The extra space around each die allows input/output connections to extend beyond the die footprint through redistribution layers (RDLs). These layers are patterned using photolithography, so their features must register accurately with the underlying die and previously formed structures.

Die placement errors can originate during pick and place. Molding and changes in the mold compound through repeated process cycles can add further variation. If the lithography pattern does not align with the actual die position, overlay may fall outside the process’s acceptable limits. The challenge becomes more demanding as RDL features and registration requirements tighten.

One option is to measure die displacement on the lithography stepper before exposure. However, the measurement can take as long as the exposure itself, making metrology a significant part of the cycle. Measuring on a separate system and sending corrections to the stepper can reduce that bottleneck.

How optimized stepping works

The process starts with an external metrology system that measures die positions across a wafer or panel. Software uses those measurements to calculate corrections, such as translation, rotation and magnification adjustments, and to predict how different exposure-field sizes are likely to perform against a defined overlay limit.

The lithography tool then applies the selected corrections during exposure. The measurement, analysis and correction cycle can be repeated for each layer, so the exposure plan reflects the latest placement data rather than relying on a single set of assumptions.

Optimized stepping depends on coordination among three elements:

  • Metrology: Measures die displacement accurately and quickly enough to support the production cycle.
  • Lithography equipment: Accepts externally calculated corrections and can expose different field sizes, including selecting the appropriate reticle area.
  • Control software: Analyzes displacement, predicts yield for candidate exposure configurations and manages process data and run-to-run adjustments.

The metrology step must be fast enough that it does not become the limiting operation after exposure is accelerated. In the reported implementation, a system measured placement errors for more than 5,000 dies on a 510 mm by 515 mm panel in less than 10 minutes.

Balancing overlay requirements and throughput

Smaller exposure fields provide more opportunities to align patterns to individual die, but exposing dies individually can reduce throughput. Larger fields can cover more dies per exposure, though placement variation within the field may make it harder to satisfy a strict overlay limit.

To manage this trade-off, a reticle can contain multiple field configurations, ranging from a single-die field to arrays containing many dies. Software evaluates the measured displacement data against the user’s overlay and yield requirements. If a large field is predicted not to meet the target, the analysis can divide it into smaller fields and evaluate those configurations. The resulting exposure plan aims to use the largest suitable fields without exceeding the specified failure allowance.

This is not simply a choice between maximum speed and maximum accuracy. The process gives the operator a way to set yield requirements for available field sizes and tune how aggressively fields are divided. That flexibility can be useful when process priorities differ across products or production conditions.

What the test results demonstrate

A test reticle containing fan-out-style pads, vias and RDL structures was used to evaluate the measurement and correction approach. Some structures included predefined offsets for feed-forward testing. Applying corrections calculated from measured die placement errors produced overlay errors below ±3 µm in the reported evaluation.

A panel example illustrates the potential cycle-time effect. For a panel containing approximately 4,500 dies, a conventional serial approach took a little over six hours, with about three hours allocated to measurement and three hours to exposure. Moving measurements outside the stepper and running them in parallel with exposure reduced the example cycle time to about three hours. Optimized stepping could further reduce exposure time by selecting larger fields where the predicted yield allowed it; the example presented a productivity-focused case with cycle time below 10 minutes, alongside a more yield-focused configuration with a longer—but still substantially reduced—cycle time.

These figures describe the specific test and process scenarios, not a guaranteed result for every packaging line. Actual outcomes depend on placement variation, overlay limits, available equipment and the chosen yield criteria.

Implementation considerations for packaging teams

Before adopting optimized stepping, engineering and operations teams need to confirm that measurement speed and accuracy suit the panel or wafer format and the required overlay budget. The stepper must be capable of using external corrections and changing exposure-field size. The control system must connect measured displacement to yield prediction and exposure selection.

Process teams should also define the acceptable registration error and yield requirements for each field size. These settings determine how the system trades exposure efficiency against the risk of fields failing the overlay target. Statistical process control and process genealogy can support monitoring of the measurement, correction and exposure sequence.

Conclusion

Optimized stepping provides a structured approach to die-placement variation in fan-out packaging lithography. By moving metrology outside the stepper, applying measured corrections and predicting yield across multiple field sizes, teams can reduce exposure bottlenecks while retaining control over overlay requirements. Its practical value depends on integrated equipment and software, reliable measurements and clearly defined yield targets—but it offers packaging manufacturers a flexible way to balance lithography productivity with registration performance.

Access Official Technical PDF

About The Author