Developed to help customers lower their DPPM rates to the single digit range, the Optimal+ Outlier Detection solution delivers the industry’s only end-to-end solution for the detection of marginal outlier units.  Using automated, rules-based analytics to identify outliers in good die populations, Outlier Detection automatically re-bins suspect devices with no manual intervention required.

 

Highlights

  • One of the complementary solutions that comprise the Optimal+ Semiconductor Operations Platform
  • Reduces customer returns (RMA) by as much as 50% by detecting marginal parts
  • Increases quality and reliability by killing or downgrading outlier units
  • Saves downstream costs by catching outliers at wafer sort
50% RMA Reduction

Outlier Detection applies advanced product analytics to detect marginal parts that impact quality for real-time and offline decision making. Integrating easily into any manufacturing environment, it saves downstream costs by catching outliers at wafer sort and cuts customer returns (RMA) by as much as 50%

Increases Quality & Reliability

A critical filter in keeping devices of questionable quality from reaching the customer, Outlier Detection enables engineering and product teams to properly assign every device using a quality index while killing or downgrading questionable dice during test

Supports Automatic Bin Switching

When suspect dice are identified, they are automatically re-binned with no manual intervention required. Users can also define their own outlier detection rules and automatically publish them to the entire test fleet.

Automated Rules Publication

Employing advanced PAT algorithms, Optimal+ Outlier Detection leverages deep product data analytics to capture quality issues within the outlier process and controls the publication of outlier rules to test floors in a closed-loop process, enabling the detection of a wide variety of quality issues which could lead to a test escape.

The Optimal+ Semiconductor Operations Platform Workflow

Our solutions are installed in 90% of the foundries and subcons serving the global semiconductor industry, enabling IDM and fabless teams to seamlessly collect, clean and collate their data sets directly from the source of their creation in preparation for extreme analytics and time-sensitive action. The data then goes through a multi-stage process that enables teams to manufacture actionable intelligence that drives every quantifiable performance metric, as described in the diagram below:

Examples

Using GPAT to Filter Die for Defined Markets

The Issue
A key market segment where returns are not an option must receive only the highest quality die.

Performing the Analysis
Optimal+ provides the industry’s most proven algorithms for escape prevention including parametric, geographical and cross-operational outlier detection. Using our GDBN algorithm, device manufacturers can dynamically re-bin suspect parts (yellow 5s in the diagram below) and prevent them from being shipped to a particular market segment. However, these are still functional dice, so while they are not suitable for Segment “A” based on our escape prevention algorithms, they can still be used for other market segments where their quality level is not a concern for mission-critical reliability.

image0044