The advancement of ICT is bringing change and enhancing efficiency through systems that support decision-making leveraging data across all areas, including corporate R&D, manufacturing, and sales. Now, it has become an era where companies must move beyond the previous ways that depended on people’s experience and capabilities and instead, be able to use automated information analysis systems to ensure business competitiveness.
The area of DRAM, which is one of SK hynix’s core business areas, has continued to develop by combining various core technologies and technology migration to meet the increasing demands of customers seeking high performance and density. The problem is that this makes product development extremely challenging and complex. It requires more resources and development time, making it difficult to respond to the needs with the previously existing development systems and processes. If the trend continues, it is expected that we will meet our limit or lose product competitiveness in the future unless we innovate the way we work.
To this end, in DRAM design, many efforts are made to change the way we work to effectively accumulate and utilize various data. The direction of that effort is enabling the system to process simple and repetitive tasks, while empowering people to focus more on strategic and value-adding tasks to enhance work efficiency and improve performance and quality of products.
Building “Quality Design” System through Quantifying and Systemizing Criteria
DRAM design process includes designing circuit aligned with product requirements. This process concurrently takes place with developing a verification environment to test the designed circuit and testing whether the product operates well in line with its specifications and actual usage environment.
Test results can be categorized into items that can clearly be validated whether it is a pass or fail via system, and the other items that are impossible to validate via system because the boundary that defines Pass or Fail is ambiguous. The latter, such as signal power, signal deviation, or signal interference, are items that require validations based on personal experience or capability. These items have a trade-off1 relationship with performance, quality, and production cost, and due to such nature the subjective judgement of engineers has been pointed out as the causes that create fluctuation in product quality.
This year, DRAM Design implemented “quality design”, innovating the way we work to increase the minimum threshold of such ‘impossible-to-validate’ items and manage them in advance. The goal of “quality design” is for all relevant members to agree on the detailed validation criteria for items that were impossible to validate via system, clarify and quantify them in order to enable them to be verified whether they are Pass or Fail via the system.
First, we established quality criteria for impossible-to-validate items and developed a validation standard for such criteria. We then pursued to systemize the agreed criteria.
The quality criteria of impossible-to-validate items are comprised of criteria on removing potential defects, optimizing product performance, and removing design quality deviations. DRAM Design department came up with a total of 80 quality design criteria items by the first half of this year and plans to continue to identify new criteria in the second half of this year to shape DRAM Design’s culture around pursuing better quality.
Over the 80 quality criteria, we defined minimum standard for validation threshold, the appropriate level of dispersion2, and the standard for anomalies, and quantified reference points determining Pass or Fail. By doing so, we objectified tasks that were being conducted with personal judgement based on experience and continued to increase the defined minimum criteria as we plan to also pursue initiatives on ensuring product robustness.
DRAM Design DAM completion ceremony held on June 24th.
In addition, we established DAM3, a system that gathers and stores data required in development, during the first half of this year to secure ‘a system that provides data’ for anyone involved in development to access. We included the before-mentioned validation standard for criteria in the validation system and transformed the fundamentals of development process, including identifying defects and analyzing signals, which as a result is enhancing our design quality.
From the second half of this year, we plan to manage the automation rate of quality validation processes while adding achievement rate of validation criteria fulfillment into “design completion confirmation process (sign-off)”. We plan to first apply it to pilot products and expand to entire product lines after an adjustment progress. We also aim to increase the execution of “quality design” by continuing such efforts.
“Quality Design” Bringing Improvement to Product Quality and Development Efficiency
We expect that once the quality design system is established, it will not only help optimize product performance by removing potential defects, but also remove deviations in design quality. As a result, we will be able to ultimately enhance both product quality and development efficiency.
The Key Factor for Success (KFS) in building the quality design system is DAM. DAM is a system that accumulates and stores all data, including circuit, layout, and signal data needed in design for product development. Data stored in DAM is mostly used to build the quality design system, enhance signal quality, and ensure robustness of multiple signals.
DRAM Design aims to continue to update new quality criteria by leveraging data stored in DAM and conduct multi-dimensional analysis. In addition, with regards to improving signal quality and ensuring robustness of multiple signals, our goal is to quantify the quality of all signal data of products and optimize internal margins between multiple signals and characteristics, in order to secure product robustness.
There are three changes that we expect to achieve with DAM. First is the change in reference point for validation, which can be done by using accumulated data related to quality design. Second is enhancement of the level of completeness, which can be achieved by expanding the coverage of predictability based on accumulated data. And third is improvement of development efficiency, which can be achieved by resolving previous challenges in the development process as we resolve such challenges one by one based on data science (DS).
DRAM Design aims to align DAM with New, Unique, Difficult, Different (NUDD)4 system so that DAM can become a core system in the DRAM development process and share changes in data via system. It also aims to improve development efficiency as well as enhance the overall product quality of all SK hynix products by expanding Single Design5 system.
Transforming Development Culture and Fundamentals
Capabilities required in product development are becoming more advanced and the development process is becoming segmented and detailed as well. To respond to such changes, it is important to build a base system that efficiently connects specialized organizations. To bring together different organizations organically and to enable them to function as one team, it should not stop at simply changing the composition of an organization or enlarging the combined size of the organization. Instead, this requires integrating and sharing data among different organizations. That is how we build an actual network that adjusts the activities among organizations and integrates interactions.
In addition, it is predicted that we will be living in an era where the ability to analyze data will directly impact product competitiveness. Our product development processes and the way we work will change accordingly to adapt to such changes.
We achieved a shift from a simple checklist type of method into criteria that can be shared and improved by anyone involved in the development process through “quality design” system. We successfully applied it to our product development process, and also established a system that enables decision making based on quantified, objective data rather than personal experience. Furthermore, we alleviated the polarization of design quality and quantity of data individuals obtained depending on needs or capabilities, and at the same time allowed all developers to fairly obtain necessary data they need.
Quality design enables us to secure competitive advantage by ensuring product robustness through making design quality internal criteria, and embedding them in our systems, which is so much more than simply systemizing the Pass/Fail validation. Eventually, by establishing the quality design system, we will be able to minimize repeated confusions in the decision-making process during early development phase, and by focusing on overall optimization we will transform into an organization which is capable of resolving company-wide issues and building a positive cycle in product development.
We will continue to build a new product development culture through quality design system, and make efforts to maintain this culture in a systematic and stable manner.
1Trade-off: A relationship, where gaining one of the two objectives will delay or lose in achieving the other.
2As dispersion widens, possibility of product defect increases, and therefore, it is necessary to manage dispersion to be kept within a certain level.
3) DAM: Like dams built along rivers that stop or restrict the flow of streams to create added value such as power generation or tourism, DAM refers to a ‘Data DAM’ that gathers and manages data generated in different parts of the development process in one place, opening it to developers to enable them to create added value.
4NUDD: New, Unique, Difficult, Different. Refers to a risk assessment methodology for new products due to changed data in new products compared to previous products.
5Single Design: a collaboration system among design teams across different divisions of SK hynix. It creates synergy effect by benchmarking, IP, and horizontal rollout of methodologies.
ByJoohwan Cho
Head of DRAM Design
