The Time Is Now For A Common Model Interface

By Ahmed Ramadan and Greg Curtis

Driven by consumer demand for “cheaper, faster, and better,” the semiconductor industry is continually pushing the migration to smaller process geometries. This continued scaling of complex designs into advanced process nodes is critical for applications ranging from high-performance computing to low-power mobile devices.

In the past, products like smartphones were replaced every few years with the latest technology, negating the need for long-term reliability. However, recent data suggests that consumers are delaying smartphone replacement, shifting focus instead on longer-term reliability.

Long-term reliability is even more critical for automotive applications. Semiconductor devices and sensors must be designed to last the life of the vehicle, or at least exceed the life of the warranty. Given the fact that the average age of a car is 12 years¹, advanced devices need to operate reliably and safely for longer than ever before. In fact, in 2015, almost 14 million vehicles on the road were over 25 years old². That is almost double the 8 million on the road in 2002. This will likely accelerate with an increasing number of electric vehicles hitting the road. Beyond this, warranties on electric vehicles will push the lifetime of a vehicle up into many hundreds of thousands of miles. Moreover, autonomous and self-driving vehicles will push the reliability and safety measures even further, as there is not only concern about the safety of passengers but also the safety of everything in the vehicle’s proximity. As a result, accurate simulation of device aging over time becomes a serious challenge across many technologies. This makes technology-specific, best-in-class aging simulation a verification requirement today.

For designs utilizing mature technology nodes, aging effects are often addressed by simply overdesigning, leaving valuable margins on the table. In today’s extremely competitive market, designers have to optimize each and every design parameter to gain margins. These slim, cost-saving margins cannot be achieved without also considering, by accurately estimating, the long-term impact of aging.

Increased reliability problems at the device level are a direct result of degradation of the gate dielectric and of the interface between gate dielectric and silicon over time. Two important mechanisms that contribute to such degradation are the Hot Carrier Injection (HCI) and the Positive/Negative Bias Temperature Instability (PBTI/NBTI). These mechanisms are more prominent at smaller geometries because the gate dielectric is scaled to only a few atoms in equivalent thickness. Over time, device aging increases the threshold voltage and decreases the channel carrier mobility, which degrades circuit performance, shortens circuit lifetime, and introduces potential failures in the field.

Until recently, foundries had to support a variety of model interfaces as a common, industry-standard interface solution for access to aging models did not exist. Likewise, simulation suppliers were also required to support their own, unique interface. This non-standard approach added complexity and increased support costs for the supplier and end-user alike.

Why the Open Model Interface (OMI) is the standard solution
Originally developed by Taiwan Semiconductor Manufacturing Company (TSMC), OMI is built on top of the TSMC Model Interface (TMI) that was licensed to the Silicon Integration Initiative (Si2) in 2013. The Si2 Compact Model Coalition (CMC) voted in June 2014 to support one interface for both the model interface built on TMI and for reliability and appropriately named the new interface the Open Model Interface (OMI). The first version of the OMI standard was officially released by the CMC in April 2018. OMI augments the standard CMC device models supported in circuit simulators by providing users the flexibility to customize the standard models to fit their own applications, all without touching the native implementation of these models (Figure 1).

Figure 1: OMI model parameter update.

An example of where this flexibility is necessary is in Layout-Dependent Effects (LDE) modeling, such as Well Proximity Effect (WPE) and Shallow Trench Isolation (STI), where the LDE model is technology dependent. OMI hosts LDE equations, interfaces with standard models in circuit simulators, and secures technology IP through the OMI shared library. For reliability simulation, OMI also acts as an aging platform that enables modeling of all degradation mechanisms (Figure 2). This includes Hot Carrier Injection (HCI), Bias Temperature Instability (BTI) and Time Dependent Dielectric Breakdown (TDDB). It also allows for running reliability simulation and analysis in circuit simulators supporting OMI. The simulator-agnostic, OMI shared library can be provided to a foundry’s customers or to Integrated Device Manufacturers (IDM) internal design groups as part of their technology design kit. For reliability simulation, foundries don’t need to change the model libraries they provide to their customers today. They only need to add the OMI Aging portion to their package.

Figure 2: OMI Aging flow.

OMI is now available to the entire industry through Si2. By using the standard interface, foundries, IDMs, and EDA vendors can now focus their resources on supporting a single, common interface. Fabless companies and internal IDM design groups can also take advantage of whichever simulator they determine best for their technology mix. Likewise, EDA suppliers have an efficient path for broad customer support. The CMC OMI Standard, v1.0.0, supports the following models: BSIM4, BSIMBulk, BSIM-CMG, BSIMSOI, and HiSIM2. More models will be supported in the future.

Summary
Until now, in order to support multi-vendor simulation flows and aging models, foundries had to support a variety of model interfaces. With Si2’s release of OMI, the fabless-foundry-supply chain will benefit from the efficiencies gained by implementing and using OMI-compatible solutions. Since its official release, OMI has been steadily gaining traction within the industry. Mentor Graphics, A Siemens Business, implements this effort within the Analog FastSPICE (AFS) and Eldo tools, which both support the Open Model Interface.

References
1] https://247wallst.com/autos/2017/12/30/as-average-age-of-american-cars-nears-12-years-a-challenge-and-opportunities-for-manufacturers/

2] https://www.usatoday.com/story/money/2015/07/29/new-car-sales-soaring-but-cars-getting-older-too/30821191/

Ahmed Ramadan is a senior product engineering manager, analog and mixed-signal verification business unit, at Mentor, a Siemens Business.

Greg Curtis is a senior product manager, analog and mixed-signal verification business unit, at Mentor, a Siemens Business.