Analog foundry processing demands robust PDK
Lou N. Hutter,
Dongbu HiTek Co., Ltd., Seoul, S. Korea
As an increasing number of IDMs accelerate their move toward fab-lite and fabless business models, analog foundry chip processing promises to be transformational for our industry. Foundries moving in this direction must comprehend the vast differences between analog and digital if they are to succeed. For starters, the steady advances of digital processing technology have been largely guided by the ITRS (International Technology Roadmap for Semiconductors. By contrast, there is no ITRS outlook for analog processing technologies, nor is there likely to be one. The analog space is just far too broad, diverse and fragmented. That lack of a roadmap, however, creates opportunities for differentiation.
Although the analog space can be more complex than digital, analog chip development and processing presents a compelling business opportunity for nimble foundries that can fully comprehend and manage the complexities. Since analog chips are less driven by feature size than their digital counterparts, the capital equipment needed to manufacture them is less expensive. In fact, much of the older fab equipment designed to process 200mm wafers in the 0.18–0.35μm range is adequate for the most advanced analog chips in play today. Moreover, the price erosion rate for analog chips is generally much slower than for digital chips – and analog chips generally have a longer useful life than digital chips.
As attractive as analog foundry processing may be as a new business opportunity, established digital foundries will not succeed in the analog space by simply having world class fabs. Although these are required to play in the analog space, they are not sufficient. Perhaps the most critical success driver is a robust analog PDK that has been tested and proven to provide accurate simulations and smooth transitions to silicon. In designing a digital logic function, for example, you may need 20 components in your process, but in designing an analog function, you may need 150 or more components…so the PDK becomes very critical…and there is a lot more to be worried about.
Think Like An Analog IDM! To succeed as a player in the analog space, foundries must think like analog IDMs that have demonstrated time and again how to design, develop and manufacture best-in-class analog products. IDMs have spent considerable time and effort over the years to develop robust capabilities to aid their designers. Good role models are ADI, Linear Technology, STM, TI and Maxim. Indeed, to succeed as an analog foundry, you can't just be as good as the next foundry – you have to be as good as the successful analog IDMs since your customers will compete with those IDMs in the marketplace.
This brings us back to the importance of using a well tested analog PDK during the design phase. As far as the analog chip designer is concerned, the PDK is the all important tool within which accurate chip designs must be simulated and verified before being implemented in silicon. Accordingly, analog PDKs should be routinely test driven by foundries to ensure that they will deliver accurate chip models.
In test driving an analog PDK, emphasis must be on quality control as numerous metrics are continuously tracked and all 2nd and 3rd order effects must be addressed. This is the only way to ensure that the simulation-to-silicon transition will be smooth. Being good at the analog PDK level enables development of best-in-class analog chips that rival those from analog IDMs. It can also be a clear differentiator among foundries that are expanding their repertoire of services.
Lou N. Hutter, is SVP and GM of the Analog Foundry Business Unit, at Dongbu HiTek, 222-1, Dodang-Dong, Wonmi-Gu, Bucheon, Gyeonggi-Do, 420-712 Korea; ph.: +82-32-680-4140; email firstname.lastname@example.org.