Details

Inspection planning based on excursion risk management

Inspection Planner 1.0 (IP1.0) is the first inspection planning software offered to the semiconductor industry. Using process, defect, and economic data, IP1.0 allows fabs to identify the defect inspection capacity that minimizes excursion risk and maximizes fab profitability.

Importance

Analysis with fab data estimated up to 2% fab yield increase due to sample plan optimization*. There was no increase in tool cost, utilization, false alarm risk, or cycle time.

Introduction

The key to identifying the optimal inspection capacity is to calculate and balance the competing inspection and excursion costs.
IP1.0 captures inspection costs with a cost-of-ownership model and the excursion costs by calculating excursion detection delays. The delays are calculated with state of the art stochastic formulations based on process tools’ in-&-out-of-control cycles.

The excursion detection model

Process tools are modeled to experience out-of-control (OOC) states at random points in time. In the OOC state, a defect count of some type has increased. These events, also knows as defect excursions, need to be detected quickly to minimize yield impact.
The below figure depicts a single process tools’ in-&-out-of control cycle. Root cause analysis (RCA) commences after detection, followed by the fixing of the problem. After the fix, the cycle starts over.

Variables affecting the time to detection

The time to excursion detection is a function of cycle-times & detection probabilities. IP1.0 takes a vast number of inputs into account to calculate these values, examples include:
  • inspection tool throughput
  • inspection tool MTBF/MTTR
  • cycle-time to inspection and/or review station
  • sampling (% of lots sampled, % of wafers sampled, % of wafer area inspected)
  • lot-to-lot and wafer-to-wafer defect count variance
  • capture rates for individual defect types
  • review sample size, and
  • classification accuracy & purity for individual defect types

The complexity addressed by IP1.0

The above in-&-out-of-control diagram is for a single process tool & excursion. A high dimensional probability problem arises when multiple process tools with multiple excursion types are considered simultaneously. Add to this multiple inspection tools receiving excursion signals from various stages of the process. Recent algorithm advances allow IP1.0 to model this problem fast and accurately. Many other factors are considered. Please refer to our services and literature pages for more information on the many intricacies of inspection capacity planning. IP1.0 is also a fab capacity model. In addition to standard tool count and cycle-time trade-offs, it captures how process control can affect process tool availability. Therefore, IP1.0s dynamic queueing model captures the full fab throughput and cycle-time impact associated with sampling plans. These are cost impacts current fab capacity models omit.
* ISMI Tech Transfer Report #08124977A-ENG