Intento Design

French startup believes it has found a way to speed up analog analysis by breaking the rules.


While the United States is where most EDA developments have come from, there have been pockets of success at various places around the world and one that has produced more than most in recent years in France. Semiconductor Engineering spoke with Ramy Iskander, founder and CEO and Eric Laurent, worldwide sales and business development for Intento Design. What follows are excerpts of that conversation.

The company is based on 25 man years of research aimed at developing a new analog design methodology at Laboratoire d’Informatique de Paris 6 (LIP6) in Université Pierre et Marie Curie (UPMC) in Paris, France. The research resulted in a portfolio of intellectual property, including an EDA tool for the design cycle that automates the sizing and migration of analog and mixed-signal circuitry used in complex SoCs for a variety of connected applications.

Intento has recently raised 900k Euros in funding from Seventure Partners and the FOREIS Endowment Fund.

Ramy Iskander started his EDA career working in the digital arena doing High Level Synthesis. Then, at Mentor Graphics he worked on the front-end of AccuSim. There he found a lot of complexity and wanted to find the same kinds of solutions for analog that existed for digital. This became the seed for his PhD work. After looking around at the approaches being taken, he decided to go in a different direction and develop some new ideas.

As a result of his PhD, some technology was developed and then participated in a technology transfer which resulted in a start-up to turn the technology into a commercial product.

SE: Are there any reason why there have been so many startups from France?

Iskander: Universities in France are starting to focus on having autonomy and not to rely on financing from the government. This means that they have to use the technology that is developed inside the universities and to perform technology transfers. In 2011, the government signed a deal for 72M Euros to be spent over 10 years for technology transfer with a goal of creating 300 startups. That provided me with the opportunity to mature my technology.

SE: How difficult has it been to go from research to commercial product?

Iskander: Because I came from industry, my research never lost focus from the real application. First, I did the research in the lab for 10 years and then commercialized it in two phases. The first phase was to mature the prototype that proved the concepts and then in the second phase we went into the commercial deployment.

SE: How easy is it to get funding in France?

Iskander: It is very easy to get funding in France at the moment so long as you have a good idea. The funding is widespread from bio-technology, chemistry and to EDA.

Laurent: In the United States, venture folks are concentrating on the Internet of Things (IoT) and related stuff, but in France, EDA is still quite fashionable.

Iskander: In Paris, the people are very efficient and open-minded and a lot of good stuff is coming from the universities, so it is easy for young, talented PhDs or people from business school to set up a startup. We also have a very strong alumnus that are communicating with each other and try to help each other. Getting funds for EDA companies today is difficult, but what makes it easy for us is the technology.

SE: Let’s talk about the technology. What makes it different?

Iskander: Since the 1980s people focused on providing the fastest simulator with thousands of elements using matrixes and numerical approaches. They take a set of equations and solve them simultaneously. I bring is the physics to it. Instead of having a modified matrix where you have to solve a thousand by a thousand calculations, I explode the matrix into primitives. We have identified a few analog components in the same way that digital does. With these primitives we can build any kind of application, such as physical sizing and simulation.

We work on subsets of the netlist rather than trying to solve the whole circuit each time. This is much faster. The idea is to find a good operating point which gives us stable conditions on which we can calculate the analog behavior. We focus on the DC operating point and we can do transient analysis as a succession of DC analyses by doing it in a structured way. We have the same accuracy as SPICE because we use SPICE in the background and this is the source for our models.

SE: How do you migrate between technologies and nodes?

Iskander: The models are independent of the technology and we have been able to create this level of abstraction which hides all of the technology phase. There is no technology information, which means the same model can be run on different technologies and then you can just switch the technology file. The models we construct are a bridge between hand crafting and actual simulation results. It has some input parameters so when the designer fixes the values of these by hand he will get the exact dimensions he wants. We do not optimize the design sizing, we accelerate the design process. If the designer chooses a topology and a set of parameters to control, we accelerate the whole process. We do only the first phase of the design and do not do layout. All of the design setup is done before then.

SE: So design productivity is the primary goal of the tools.

Iskander: The tools used by the industry today create a design productivity gap. They cause many design iterations. To overcome this, you have to accelerate the tools, and Intento does that by bringing technologies such as sizing, tuning and migration. Many of the latest nodes are just too difficult for analog but we now have the technology to allow people to see this for themselves. They will see better results using our technology and they will be able to see exactly why it does not work. Which assumptions are not correct.
We also need new device types. CMOS cannot work for much longer so we need to look at finFETs and other kinds of devices.
SE: It would appear that the Internet of Things will create a huge opportunity for analog.
Iskander: IoT is much more challenging than just hardware or software. It will push all of the semiconductor companies to change their business model. They have been focusing on technology on silicon but with the IoT, there is much more focus on the system.

Laurent: Our focus will be on Mil/Aero and automotive and that is where most of the interest in our products is coming from. With the IoT we can expect to be getting interest as well.

Iskander: Our first released will be in March 2016 and they will be attending DAC in June as a first time exhibitor. Today, Intento has a partnership with Cadence through their Connections Program. They are using Spectre as their underlying simulator. They made this choice because they believe that a large percentage of the market uses this.

  • Kev

    Is this much different to top-down digital design with behavioral block modeling, but using Verilog-A (I presume)?

    • Brian Bailey

      In part yes. The problem that analog has had in the past is that everything affects everything else and thus the whole circuit has to be solved for every small change. Intento believes they have found a way to stop that and can solve things more locally.

      • Kev

        That’s more-or-less what “fast” SPICE simulators do, but most of those don’t do behavioral modeling that well – is in they can do it, but mixing Verilog-A and SPICE descriptions is usually manual and messy.

        A key piece in this kind of flow is being able to back-annotate from P&R into a behavioral block description, and I have not seen anybody do that.

        A secondary problem is that analog guys hate changing tools, so from a startup perspective analog design is a very bad market to tackle. My own strategy is to tackle the analog problems in the digital flow (power, DVFS, CDC, etc.), and pull in the analog design/verification as an extra.