Finally, Analyzing All Test And Manufacturing Data Automatically

Product quality and yield, operational efficiency, and time-to-market continue to be dominant drivers in the semiconductor industry. Adding to this complexity is a diverse manufacturing and test supply-chain of independent providers all continuously generating enormous amounts of different types of chip-related data in various formats. The knowledge contained within this data is critical to properly managing these key drivers. However, current solutions in the market used for collecting and storing data, normalizing and monitoring the data for quality and completeness, and automatically analyzing and providing meaningful and actionable insights on all of this data are falling well short of expectations… until now.

It has been well chronicled that maybe there is just too much data out there and we wouldn’t have the time or ability to analyze all of it even if we wanted to. A better question is, what if you had no apparent limitations on consuming all data at your disposal and it was all cleaned, normalized and aligned for you? Would you choose to analyze it? “Maybe,” you might say. What if all of this data was cleaned, normalized, aligned and analyzed automatically for you with traceability of each chip’s history across the full manufacturing and test data continuum captured? Also, if analytical correlation was made possible across every test ever run on that chip at the push of a button?

Further, what if a summary of all points of interest and critical findings of potential issues of your chip were automatically shown to you right out of the gate with no special querying or manual manipulation of datasets required? To top it off, what if then it only took a few clicks to find the root cause? You might be saying, “Well, now you have my attention.” I’m happy to say these are all true statements and the solution that provides these amazing capabilities resides in a revolutionary technology called SiliconDash.

SiliconDash is the next-generation, high-volume semiconductor big data analytics solution for fabless companies, IDMs, OSATs and foundries. It provides comprehensive yield management, quality management and throughput management of your IC and multi-chip module (MCM) products throughout the manufacturing and test process across the entire manufacturing supply chain. It delivers comprehensive end-to-end real-time intelligence and control of manufacturing and test operations for executives, managers, product engineers, test engineers, quality engineers, sustaining engineers, device engineers, yield engineers and test operators.

The data challenge

In a standard manufacturing process, your chip goes through a series of manufacturing and test stages where any stage of the process could reveal valuable information. If caught in time, the data would prevent any compromised chip from going through the entire costly process and, more importantly, prevent it from being shipped to your customer and potentially causing catastrophic failures while in use. In each of these stages, within the huge amounts of data hides the answers to most of your serious chip issues. However, gathering and storing all possible data is a task in and of itself but also making sure that all of the data is accurate, normalized and aligned adds another level of complexity altogether that most companies are not prepared to address.

Even if they can address it, they typically limit the task to gather only specific test data types and not the full spectrum of possible test data. While some solutions may support some test data types, SiliconDash automatically handles the complex management of all of your manufacturing test data. SiliconDash provides the total preparation of your data including the collection and storage either as a cloud service or on-premises at your site, monitoring the data for quality and completeness, and ensuring the data is normalized and aligned with no data escapes.

Applying robust algorithms and analytics
You’ve now managed to collect more data from as many different test domains as possible. You’ve added the latest and greatest test methods to your test program. You’ve staffed up your yield team. You’ve brought in big data and machine learning expertise. But you’re still not getting the full value of your data. Engineers are still spending 80% of their time searching, aligning and filtering datasets while only 10% – 20% of your data is ever used. You need to turn data into meaningful insights and corrective actions, but you haven’t been able to.

SiliconDash applies robust algorithms and analytics onto the data as the data is being streamed in real-time 24/7. Valuable dashboards and reports known as ‘Insights’ are automatically generated and made available instantly out-of-the-box. These Insights require no user setup, configuration or training required. Insights provide completely automated, unsupervised analysis and labeling of test data results. They highlight key points of interest and enable you to analyze quickly, triage and identify root causes of issues within minutes.

The figure above shows just a small sampling of what powerful Insights are made available automatically and instantly when opening the tool. These are classic examples of what you would expect to have in a solution. The difference here is that these are provided without any input or query from the user, providing an order of magnitude greater productivity savings.

SiliconDash also leverages embedded Electronic Chip IDs (ECIDs) to perform automated part-level traceability and analytics across the entire supply chain. This enables powerful correlation of every test ever performed on your chip which is a critical capability if you find yourself having systematic failures later in the manufacturing process such as during Final Test (FT) for instance. You need the ability to correlate test results across earlier stages of manufacturing to see if there is a way to predict these FT failures earlier such as during Wafer Acceptance Testing (WAT) or Wafer Sort (WS) testing. For example, one possible solution to your systematic failures at FT may be to adjust certain limits for a specific WS test parameter. You then catch these issues earlier in the process and bin out those suspected parts thus saving you significant test and manufacturing costs further downstream.

Beyond analysis to root cause and production control
SiliconDash goes beyond analysis, turning Insights into corrective actions by automatically providing full production control in the supply-chain and in real-time. With the use of preconfigured libraries of recipes, algorithms and scripting support, you can configure and deploy rules running 24/7 on the manufacturing test floor to intercept test operations when triggered by specific issues. These rules can also be configured to automatically send out alerts or emails and if required, halt specific test operations altogether in real-time until the issue is remedied.

The figure above shows a lower than acceptable yielding part at FT. Notice that the left-most top-level screenshot contains a summary section highlighting key points of interest of where you should start to investigate. In this example, yield limiters were identified for you automatically. It shows a significant yield issue in hard bin 11. Then, with the use of automatic correlation of historical tests of this device, a specific WAT test parameter was identified as a predictor to this FT issue. Drilldown links were provided and highlighted to get you to an automated correlation report. In this example, the WAT parameter “Test157” predicts FT fail bin 11. In only a few clicks with no special querying needed by you the user, you are automatically brought to this finding in a matter of seconds. The probable next step here is that the fabless company should share these findings with their foundry for a resolution at or before WAT to improve downstream yield at FT.

The power in this solution is that all this reporting and analysis was done in advance where you only needed to click a few links to see the problem. Factoring together that all the data was also preprocessed in advance and made available and then analyzed automatically and presented to you in easy-to-use format saves you conservatively many man-weeks of effort.

Future considerations – yes, even more data on the horizon
The notion of debugging and optimizing your chip not just in manufacturing but across the entire lifecycle of the chip from the pre-silicon design stage on through manufacturing and its life in the field as part of an end product is not just a vision; it is actively becoming a reality.

Today, there are already links between the pre-silicon design and manufacturing test data stages with the ability to do cell-aware sub-die analysis that goes beyond identifying systematic issues at the die level. We can now go within the die and tell you which components, cells or nets are the source of the issues. Over time, the ultimate goal is to go further into your chip’s actual operation while it is in use in the field for basic maintenance checking and to do real-time optimizations within the chip while it is in use, if allowed.

Commonly referred to as Silicon Lifecycle Management (SLM), SLM is a relatively new process associated with the monitoring, analysis and optimization of semiconductor devices as they are designed, manufactured, tested and deployed in end-user systems. SLM is based on two underlying principles: gather as much useful data about each chip as possible and analyze that data throughout the chip’s entire lifecycle to gain actionable insights to improve chip and system-related activities. The Synopsys SLM Platform provides insight into critical performance, functionality, reliability, safety and security issues for the entirety of a chip’s lifespan. This enables the optimization of operational activities for all participants throughout the life of the SoC. Value from SLM will be realized by multiple teams including design, bring-up, test and end users of systems incorporating these SoCs.

The figure above depicts the SLM product lifecycle continuum from design through manufacturing to in-field. Embedded monitors and sensors provide the common thread to gather the necessary data throughout the lifecycle of your chip that is needed while your chip will be in use in the field. More exciting developments and new technologies are coming online soon and will make this vision of end-to-end chip optimization throughout the life of the chip – a reality.

For more information on SLM and SiliconDash at Synopsys, please visit us at Silicon Lifecycle Management Platform. Please join us for our upcoming webinar on SiliconDash on:

• March 25, 2021 in North America and Europe
• March 31, 2021 in Asia and Japan

Learn how companies like Marvell are able to maximize the benefits of a big data analytics solution that was finally architected correctly.