Mask shops play a pivotal role in the semiconductor industry. Manufacturing reticles is a low-volume, almost artisanal endeavor where each reticle ordered is unique and has its own distinct manufacturing challenges. Unlike high-volume semiconductor manufacturing, where millions of identical computer chips are mass-manufactured, the wafer fab’s economy of scale simply doesn’t exist in the mask shop, nor is there the same degree of process learning that is the inevitable result of manufacturing the same thing over and over. This combination of uniqueness and customer demand for consistently perfect results shapes the mask shop’s business model in unusual and unexpected ways. (Read our previous article to learn about key differences between mask and wafer manufacturing.)
The cost of perfection
Expensive tools and the resulting depreciation expense are a primary cost driver in both mask shops and wafer fabs. However, high-volume manufacturing affords some mitigations to the wafer fab that don’t exist for the mask shop; these permit statistical sample plans that allow most lots to avoid metrology and inspection steps while still assuring a high degree of product quality. In the mask shop, the need to measure and inspect every reticle – driven by the requirement to consistently deliver a perfect product – amplifies capital and depreciation expenses, as more tools are required to achieve the necessary factory output.
In some situations, defective reticles can be repaired during manufacturing, but this too requires more specialized tools and results in higher costs and increased cycle time.
Reticles that cannot be repaired must be scrapped and a new “attempt” started because the mask shop must deliver a perfect reticle to its customer. This means higher material costs, because an additional mask blank must be consumed for each new attempt – an expense that ranges from hundreds of dollars for a commodity mask to tens of thousands of dollars for a high-end extreme ultraviolet (EUV) reticle.
Cycle time
Manufacturing cycle time can also be a factor driving mask shop costs. Cycle time can vary significantly – even for reticles of similar technology – because of the unique nature of each reticle order. The mask shop can quote an expected cycle time when a reticle is ordered based on the processing history of masks using a similar process technology, but the actual cycle time can vary significantly depending on how many attempts it takes to build a perfect reticle. If manufacturing goes beyond the committed ship date, the mask shop may have to pay for expedited shipping to minimize the delay to the customer.
In cases where a prime customer has an urgent need, the mask shop may take a calculated risk and start multiple attempts in parallel, betting that one of them will finish quickly enough to meet the customer’s deadline. Needless to say, this also incurs additional material costs for the mask shop and reduces overall capacity, as manufacturing the additional reticle attempts consumes valuable tool processing time.
Cycle time is also influenced by numerous other variables, many of which are beyond the mask shop’s control. This includes hidden factors like mask data transfer time and data preparation time before manufacturing begins, and random shipping delays to the customer after manufacturing is complete. These also can significantly increase cycle time.
Product mix
Capacity is a complex function of yield, number of process tools, processing time, the number of steps in a process flow and, in wafer fabs, the number of die per wafer. Optimizing these parameters can increase capacity which can, in turn, increase revenue. This makes sense if you define capacity as the number of things you can manufacture in a given amount of time – a good bet for high-volume wafer fabs with low product mix.
But mask shop capacity is less about the number of reticles manufactured and more about the mix of reticles manufactured. In general, there are two types of reticles: binary and phase shift. Binary reticles are simpler to build, require far fewer process steps (a third or less than the phase shift reticles), and have consistently higher yield (that is, less need for multiple attempts to manufacture a perfect reticle). If capacity is all about manufacturing the greatest number of reticles, you’d want all reticles ordered to be binary reticles.
Unfortunately, the price of a binary reticle is a fraction of the price of a phase shift reticle. Phase shift reticles can be sold for as much as 20 times that of binary reticles. A little arithmetic tells us that despite more process steps, longer processing times, and lower yield, a mask shop would likely have far greater revenue if they only manufactured phase shift reticles!
Unpredictable demand
However, the order mix of relatively simple binary reticles and much more complex phase shift reticles is essentially random, impacting both mask shop capacity and revenue. This random order mix can also increase cycle time variability, making lead-time prediction unreliable. Lack of predictability can make managing capacity and cycle time a delicate balancing act within the mask shop.
How an MES can help
A Manufacturing Executions System (MES) is crucial for achieving the mask shop’s critical performance and efficiency objectives. An MES provides real-time data and insights, enabling better decision-making and efficient management of production processes. By effectively utilizing the capabilities of an advanced MES, mask shops can optimize scheduling and tool utilization, reduce cycle times, and improve overall yield and efficiency, improving their ability to meet high customer standards for quality and timeliness.
