Issue link: https://iconnect007.uberflip.com/i/1315894
38 DESIGN007 MAGAZINE I DECEMBER 2020 integrity, thermal, fabrication, assembly, and test. We are all sharing fundamentally the same information and coming up with ideas and solutions that all can agree to, or one of us says, "I have a problem with that because of this," but the next person says, "We can solve that problem with this." You've quickly col- laborated, and you don't make the mistake of jumping down the deep well that later requires a re-spin. Ford: The nice thing about this is that it's not just product-based. You can break down any product into modular components. You could have a small module of circuitry, which could be, for example, a power supply, perhaps an interface. The whole design can be broken into these modular components. You can even break down the modular components further into specific features, whether based on, for example, the shape of a land pattern, a type of component, or even specific reference designa- tors. You can get into as much detail you like. You can talk about the success or otherwise of those features in different processes. When you come to making a new design, you have all of the associated information about layout and component decisions that you're making at your fingertips ahead of any prototyping stage. Holden: It's much easier for AI software to look at those solutions and come up with general- ized models based on a lot of different things that the human wouldn't have time or incli- nation to figure out what's common with this and what's not common. Ford: Human intelligence (HI) is the best form of intelligence, but it's also the laziest (laughs). Artificial intelligence (AI) isn't yet quite so good creatively, as it's just a sequence of algorithms, but it's dogged. AI will quickly spend billions of computing cycles, finding the very best solution to a problem or task. The vast majority of work in the digital world is a repetitive application of rules associated with any problem. AI can, therefore, be very effec- tive at providing that first-off context of things and handing it over to the HI to do the final stages. We will see an increasing trend in that direction almost everywhere. Shaughnessy: Does all of this fit under the umbrella of predictive engineering? Holden: What I was saying was predictive. Other people were calling DFM and design rule checking, which is after the fact. I'm back to Dewhurst and Boothroyd: As we do the design, let's predict the problem and solve it so that it does not become a problem. Ford: It is all about prediction, though that is a difficult word because people often associate it with long-term simulation. Taking a prod- uct, a production line, and certain tasks to be done, let's simulate and optimize how things are going to happen. There is some established software to do this that will come up with an excellent solution. Two days later, however, the production situation changes, the simulation criteria are now obsolete, and you have to start the simulation again. Any form of prediction and optimization has to happen much faster now, in near real-time, using information that can be derived about every facet of the situ- ation. It's a new digital world, and this feeds very nicely into smart factories and Industry 4.0 because this rapid response flexibility is exactly what those concepts are driving. Shaughnessy: What has to happen to make this a reality? Is this going to be led by the EDA vendors or the designers? Ford: There are a few different stages. At first, there needs to be an understanding of the value of interoperability. We need to create the The nice thing about this is that it's not just product-based. You can break down any product into modular components.