Electronic Systems Design and EDA see big shifts in several key areas.

Cloud and Machine Learning Tools Released for Electronics System Design

Cadence recently announced several new services and products for engineers working in electronics system design. The San Jose, Calif. company has more than 30 years of experience as a software supplier. The new products and services revolve around artificial intelligence, machine learning and the cloud in addition to the simulation software already offered.

Cadence Optimality mines the full design space for productivity gains. (Image courtesy of Cadence.)

Artificial intelligence (AI) and machine learning are having a significant impact on electronics system design. Cadence Cerebrus is an AI-based intelligent chip explorer and Optimality Explorer is the companion product for system design. Beyond these new tools comes news of OnCloud, a new cloud project. These new options should make big improvements in the lives of electronics system design engineers, whether they’re using Cadence’s software on premises or in the cloud.

Using Machine Learning to Save Nanometers

Cadence Cerebrus is an intelligent chip explorer built on a machine-learning approach to electronics system design. When engineers are working on power, performance and area (PPA) concerns, several variables are in play for the positioning of dozens of components in a system. Machine learning gives engineers and designers the ability to work with the variables and find an ideal solution faster. When a user sets goals for a few variables, the algorithm can work through different combinations of components and run simulations to see if a design can meet these goals.

Cadence shifts the work from the engineers to Cerebrus. (Image courtesy of Cadence.)

Several technologies are driving the need for more interconnectedness. Autonomous vehicles, 5G, the Internet of Things, and hyperscale computing all require a higher volume of signal transmission, as well as a larger volume of connected products. System-on-chip (SoC) builds are becoming denser with the competing constraints of more functions needed and less space available. Saving space even at the nanometer level provides the opportunity for manufacturers to save development time, take advantage of a smaller footprint, and save material costs.

Cadence says that the current state is one where engineers are making a first best guess at flow and running a simulation, and then making incremental changes based on the results. This iterative process relies on the strength of the engineer’s skills and continues until PPA goals are met or the end date for the design phase is reached. Cerebrus uses machine learning to move from register-travel level (RTL) design abstraction toward an optimized graphic data stream (GDS) output. The reinforcement learning engine monitors the simulation runs as they’re happening and terminates runs that aren’t moving toward the PPA goal. The CPU can then move on to another variable configuration, instead of the current method of waiting for results before assessing the next iteration and starting another run. This RTL to GDS process results in shorter times to realize PPA goals with a chunk of the effort shifted from engineers to Cerebrus.

The company says that a typical 5-nanometer, 3.5-Gigahertz high-performance CPU design using Cerebrus can take a process that once required many engineers to work many months to one that can accomplish the same goals with one engineer working for 10 days.

Multidisciplinary Analysis and Optimization

The Cadence Optimality Intelligent Systems Explorer is built for speed. The company talks about the current state of systems design, where chip design, PCB design, and systems design are operating in silos. Using AI and machine learning to bring these three functions together can result in big time savings. The company estimates that cycle time for system design and development can improve by 10 times or more using the Intelligent Systems Explorer.

Cadence Optimality Explorer merges AI, chip design and system design. (Image courtesy of Cadence.)

Using the multi-disciplinary analysis and optimization (MDAO) approach, chips and PCBs can be designed at the same time with an eye on their eventual inputs into the system design. Taking this further, the different variables on a single chip can be optimized, varying the inputs from the transmit side through the PCB to the receive side. Machine learning allows the solvers to check several initial states for these variables and quickly zero in on optimizations by moving all the dials at once. The time savings comes from requiring a smaller number of runs to reach the final design.

Cadence uses the example of truck driving to illustrate the current to future vision of AI-powered design. Present-day truck driving requires a driver and a truck. Society is moving toward a driver sitting in the cab while the truck drives itself, with the driver in control of the system but not doing most of the work. This transitional period could one day result in a group of autonomous trucks being controlled remotely by one operator. Machine-learning design for electronics could progress in the same fashion. In the current state, engineers are building their own designs and analyzing each one. Machine learning can be used in the transitional phase to design systems much faster with the engineer monitoring and verifying the work. In the future, one engineer could shift to working on several machine-learning projects at once.

Using Cadence Tools in the Cloud

Mahesh Turaga, VP of Cloud Business Development at Cadence, told engineering.com that the cloud has never been more attractive. Companies of all sizes are no longer wondering why electronic systems design should be moved into the cloud but are asking how the shift should happen. Acceptance of the cloud as a necessary part of engineering has been shifting over the last five years but accelerated during the pandemic. Offices were closed, most engineers worked remotely, and collaboration had to take place online.

The Cadence OnCloud software-as-a-service (SaaS) and e-commerce platform was recently announced with several of the company’s services now available through the cloud. Advanced/Mainstream CFD, PCB design, and Multiphysics Analysis are the first tools that will be available. The system is already supporting around 4,000 customers and this announcement opens the gate for more customers to come on board. Turaga said that the cloud is attractive for small- and medium-sized customers that want to use the functions of the Cadence software tools without making heavy hardware investments. Using the cloud versions of the software can minimize or eliminate the need for high-performance computing resources and free up time for engineers to focus on design instead of working through IT concerns. Prospective users can register for the service and be up and running quickly using the current software offerings.

Security can seem a concern when dealing with the cloud, but Turaga mentioned that Cadence OnCloud takes a multilayered approach to security. Security is handled at the application, software, network and infrastructure levels and data is encrypted on both ends of transmission. The company also requires multi-factor authentication of its users, and OnCloud was developed in compliance with ISO27001, ISO27017 and SOC2 standards.

The Rug That Ties the Whole Room Together

These announcements each describe a new software or service, but the goal is the same. The company is working to enable its customers to run simulations faster and more efficiently with more confidence that the software is generating the best possible designs. AI and machine learning are great partners for the cloud, as the computing resources of the host are leveraged by the customers on demand.

Electronics system design is undergoing huge changes as constraints get tighter and the available tools for engineers are rapidly evolving. Artificial intelligence is gaining acceptance as a tool that’s widespread in industry to a point where companies almost need to use it to be competitive. Cloud computing is also widely accepted now, whereas five years ago companies may have been hesitant to transmit data multiple times per day. The efficiency gains in both software use time and engineer focus time show that using cloud platforms, artificial intelligence and machine learning are all well worth investigating for companies of all sizes. Cadence’s recent announcements show that the company is shifting to the cloud and AI, and is ready to capitalize as the industry catches up.