WP_Term Object
    [term_id] => 69
    [name] => Platform Design Automation
    [slug] => platform-design-automation
    [term_group] => 0
    [term_taxonomy_id] => 69
    [taxonomy] => category
    [description] => 
    [parent] => 14433
    [count] => 11
    [filter] => raw
    [cat_ID] => 69
    [category_count] => 11
    [category_description] => 
    [cat_name] => Platform Design Automation
    [category_nicename] => platform-design-automation
    [category_parent] => 14433

CEO Interview: Albert Li of Platform DA

CEO Interview: Albert Li of Platform DA
by Daniel Nenni on 11-27-2016 at 7:00 am

Platform Design Automation, Inc (PDA). recently closed a US$6 million pre Series A investment round, and the company has shifted its focus from providing SPICE modeling related software and services to forming a complete AI-driven ecosystem from probing to simulation. Albert Li was the GM of Accelicon, a leading EDA tool and service vendor on device modeling, and it was acquired by Agilent in 2012. Albert is now founder and CEO of PDA, a 4-yr old high tech company that provides a comprehensive set of products and services on semiconductor device characterization instruments, device modeling and PDK validation software.

SPICE Modeling market is saturated and shrinking, what new products or technologies that PDA developed to attract a major investment?
Yes, my team was known for providing SPICE modeling software and services for leading semiconductor companies, and it is also true that the marketing is shrinking due to the consolidations of semiconductor manufacturing and design companies. But with process continues to scale down, we did discover strong needs of having faster device characterization solutions to obtain large samples of measurements to account for the increasing process variations, and having sufficient silicon data is really the key to enable accurate simulations. So we expanded our scopes to device characterization solutions, where we have a lot of hands-on experiences and know-how, also we have been working on optimization technologies (for model extraction purpose) for more than a decade, so we were one of the pioneers to apply machine-learning algorithms to achieve faster simulations such as statistical simulations and we now apply these algorithms to achieve faster measurements.

Our first device characterization instrument is NC300, the world’s fastest 1/f noise characterization solution, providing 10X speed improvements over other products on the market and for production tests, it can achieve 100X speed improvements thanks to our AI driven technologies. NC300 was quickly adopted by the leading semiconductor companies, which gave us the confidence to develop new semiconductor parametric test solutions with broader scopes, we combined machine learning algorithms, user know-how and huge amount of previous data and experiences for the training and these have generated very promising results and we will soon release a new product line that can achieve much faster parametric testing speed for semiconductor applications.

What are the products that PDA is currently offering?

There are three main EDA products and the first one is for device modeling and QA called MeQLaband it can be used for applications like:

  • Device modeling for FinFET and planar devices
  • Statistical modeling and mismatch
  • High voltage device modeling, sub-circuit modeling
  • Built-in modeling library and model card QA
  • SRAM modeling
  • Noise modeling and circuit analysis
  • Design or process optimization

PQLab is a tool for automating the QA of PDK libraries, saving engineering time and can be applied to:

  • Foundry PDK developers needing to QA a PDK
  • IC designers verify that a foundry PDK meets their requirements
  • IC designers compare two or more PDKs

For 1/f noise measurements and characterization they have the NC300 system to apply at the wafer level, device, circuit or even with sensors.

You have been working on device modeling throughout your career, what are the challenges in device modeling posed by the latest process technologies such as FinFET?
Again, the increasing process variations are the key challenges, a lot of layout dependencies such as stress related layout dependencies, lithography introduced dependencies, need to be taken into account during device modeling. For FinFET, even though the new process introduced more dependencies, for example sometimes even the shape of metal2 over a device plays significant impact on device performance, but thanks to the rigorous design rule, there are only a few fixed layout combinations that are allowed by the design rule, so the required modeling efforts are actually less. To model the process variations accurately, the key is still to obtain sufficient silicon data to produce accurate statistics, and again faster measurements are in great needs. And for designers, the foundry corner models or statistical models are often conservative, in my experience, having some silicon data on the design side to adjust model or corner is the quickest way to achieve design margin, as the model/design kit itself is a source of margin loss.

What are the future technology trend for SPICE Modeling?
Having sufficient data is really the key to the problem, if data is sufficient, model can be automatically generated or synthesized. The concept has already been applied to the case of passive device modeling, such as modeling inductors. EM solvers play the role of proving more “data” or the synthesizers to generate models automatically. We’ve been working with the same concept for the active devices for quite a while, one way is to enable faster measurements, so that a lot more data can be collected and the other way is to achieve huge amount of data based on limited silicon through machine learning, which requires deep understanding of device behaviors, device modeling knowledge, data for the training and years of training experiences, we have already successfully applied the methodologies to our service projects, and tedious tasks such as model re-targeting is now purely done by machines.

Machine Learning enabled model targeting from tweaking model parameters to just defining the targets and let the machine finish the job automatically

What are other areas in semiconductor you see that Machine Learning can help?
We’ve published 3 papers in the past few years related to machine learning, and we used machine-learning algorithms to help on speeding up soft error simulation of logic circuits, automatic statistical modeling, and automatic RF front-end design,so the areas of machine-learning applications are massive. Algorithms, expertise, data and risk are the four key components to access Machine-Learning applications, take device characterization and modeling as examples, we have been working on the machine learning algorithms for over a decade, and we are definitely the experts in device characterization and modeling, we also have huge amount of data and models from previous projects, and these enabled us to train our software or instrument to achieve faster measurements and automatic model generations.

How do you position PDA?

Our products address device characterization, modeling and PDK validation, and all of our products will be driven by machine learning to achieve faster measurement and faster simulation. We have 3 dedicated Ph.ds from Tsinghua working on algorithm development for years, and we continue to increase our investment on experimenting different algorithms, and training. So I would position PDA as a solution company with AI as the core competence and our ultimate goal is to continuously improve speed and effeciency for our customers from probing to simulations.

Also Read:

CEO Interview: Mike Wishart of efabless

CEO Interview: Chouki Aktouf of Defacto Technologies

Executive Interview: Vic Kulkarni of ANSYS

Share this post via:


There are no comments yet.

You must register or log in to view/post comments.