Semiconductor manufacturing stands apart as a unique process when you compare it to other manufacturing processes. From a systems perspective, production in most manufacturing industries is either discrete or process manufacturing – semiconductor manufacturing can be defined as process manufacturing with a discrete outcome. Here’s what I mean by that.
The semiconductor manufacturing process is like process manufacturing in that most of what happens is adding value to the flow of materials through the process. The wafer is fabricated, tested, sawed/separated, packaged, and tested again. The output of the preceding step is assumed to be the input to the next step. There can be options and variability within the flow. The build instructions focus more on how to process each step, with little information in most steps about material components to add. In discrete manufacturing you can design sub-assemblies that come together to make a product. In semiconductor manufacturing you may build to strategic bank inventory points, and the process can have multiple outcomes at various stages of production.
The semiconductor production process is also like discrete manufacturing since it does create defined part numbers in the end that are sellable parts. While many process manufacturing solutions focus on process industries like paint or chemicals where the process is a solution, in the end semiconductor products are components of someone else’s bill of materials.
The semiconductor process flow, and the variability of that flow, require a “one-to-many” process which means that the Bill of Materials (BOM) is inverted, compared to a discrete manufacturing BOM.
To illustrate, here’s an example of process manufacturing in the semiconductor industry:
It starts with the wafer fabrication process. Your company may provide the raw wafer or you may have your foundry provide it. The fabrication process then uses a wafer mask (mask set) and a multi-step process to create multiple individual die on the wafer.
After the wafer fab process is completed, the wafer can be put into inventory (wafer bank) or it can be moved into a wafer test process – commonly referred to as a sort or probe process. This testing process may simply mark die as good or bad, or it may categorize products more finely by bins (performance grading). These performance grades – and the overall yield for each wafer – drive multiple outputs from a single input process step.
When the wafer sorting process completes, the sorted die are often stored in a die bank – an inventory point of unassembled raw product. Often this is a build to forecast point for semiconductor companies – the place where they hold product pending specific customer demand for finished goods.
To complete a finished product the raw die would be moved through an assembly process (often more than one option per die) that puts the die into a package, moves the assembled die into a test process (where binning may occur again, and yield is tracked), and finally into some type of pack process for distribution (options include tube, tray, or tape and reel packaging). Again we have multiple outcomes from single step processes, and multiple paths for products to take. Programmability (and sometimes re-programmability), unique product marking (or remarking) for specific end users, downgrading, and rework (moving backwards in the production process to create a different product) all combine specific unique requirements for process and discrete manufacturing. Even after manufacturing a chip may be combined with other chips to produce a multi-chip IC or added to a module as a raw material in an entirely new assembly. Semiconductor manufacturing companies may require support for module production post IC manufacturing through contract manufacturing and outsourced fulfillment.
So, why is this a problem? Because business software solutions are typically designed to handle either discrete manufacturing or process manufacturing. In order to handle the needs of their unique industry, companies may resort to an ERP system with supply chain management (SCM) and manufacturing execution (MES) systems layered on top, and supplemented by various homegrown and/or manual systems. This can be expensive to maintain, difficult to integrate effectively, and ultimately ineffective.
Tensoft SemiOps supports the unique manufacturing, supply chain management, and financial accounting needs of the semiconductor business, without all the mess. There is no need to customize the system or work around it – it’s an industry-specific solution right from the start. It’s designed to handle the semiconductor industry’s inverted BOM (one-to-many production), yield (and cycle time and cost) management at every step of the production process, deep genealogy and product attribute tracking, and the ability to manage internal and external manufacturing execution.
If you’re interested in learning more about Tensoft SemiOps, please visit our product pages or contact us directly. We’d be happy to answer any questions!
Find out more: