Semiconductor engineers today face an increasing technical dilemma with the evolution of Advanced Packaging technologies. While components previously operated largely independently, modern systems require close collaboration between suppliers, posing a challenge of how to share technical models that describe component behavior without revealing secret "recipes" and proprietary manufacturing processes.
The Data Gap in Integrated Supply Chains
Previously, engineers relied on datasheets and design rules to transfer information between companies, while manufacturing secrets remained confined to the supplier. However, in the era of advanced packaging, this method is no longer sufficient; what one company does in its manufacturing process directly affects the performance of components that another company will design within the same technical package.
This interdependence creates a need for something resembling a "technical membrane" that allows the passage of data necessary for performance prediction, while blocking details that constitute intellectual property (IP). The difficulty lies in defining the boundaries of this membrane: concealing too little means leaking manufacturing secrets, while concealing too much may render the technical model unable to provide accurate or reliable predictions.
Risks of Surrogate Models and Encryption
The industry currently uses several methods to attempt to solve this dilemma, but each has its technical limitations:
- Surrogate Models: These are models that simplify complex physical representations to accelerate the analysis process. However, these models remain reliable only within the range of the data they were trained on, and they lose accuracy once the design moves outside this range.
- Encryption: Companies such as Synopsys support encrypted electromagnetic models that allow sharing simulations without revealing internal geometry or materials. Although encryption brilliantly protects intellectual property, it does not tell the receiving engineer whether the model is still valid for use in new operating conditions.
The fundamental problem is that the model may be completely secure and protected, but at the same time "technically incorrect" if used outside the scope of the assumptions upon which it was built.
What is Changing in Practice?
The fundamental change lies in the transformation of the technical model from a mere simulation tool into something resembling an "engineering contract" between companies. It is no longer enough for the supplier to send static numbers; they must send the model along with its full context, including:
- Validity Range: Where does the accuracy of this model begin and end?
- Calibration Conditions: How was this data measured and what are the assumptions upon which it was built?
- Real-world Correlation: How does material behavior (such as thermal expansion or viscosity) change with changing environmental conditions?
The companies that will succeed in this era are those that will be able to transform their secret manufacturing knowledge into models that accurately illustrate "results and effects" without the need to reveal the "process" itself. In short, the recipe must remain secret, but its results must be clear and understandable to the engineer who will build the next step upon them.