The word 'system' is extremely scalable; it can describe anything from a small SoC designed for wearable fitness trackers, to an entire cellular network, and even the infrastructure that enables data to be processed by cloud-based analytic algorithms. Each 'system' becomes part of a bigger 'system', giving rise to the term 'system of systems'. A lot of the complexity in those systems resides in software, but they depend on differentiated hardware to run those systems. This dependence is increasing as we enter a period of “domain specific compute”.
Take 5G as an example. It is expected to connect more things, in new ways, operating in more diverse scenarios than any other mobile network that has come before it. This means, as a 'system', 5G networks will extend far beyond the traditional concept of a cellular service. Even the representation of a cell in this new network will be redefined and must be far more flexible.
Similarly, the move to electric and autonomous vehicles will enable new types of service vehicles, able to operate without a human in the loop for long periods of time. Again, the concept of a vehicle, when considered as a 'system', will need to be re-evaluated.
We can also see there will be greater convergence of these new systems, as they overlap to enhance and enable each other – with 5G, for instance, being used for transportation, industrial automation and consumer cases alike. As this happens it will accelerate the rate of change and the underlying technology will need to keep pace with this acceleration, as will the technologies that enable their design.
Identifying complexity
Complexity comes in many forms; the evolution of cellular technology could be defined by its migration through the digital modulation techniques used, from xDMA, WCDMA, QAM with OFDM and MIMO and now, with 5G, a combination of these operating in the millimetre wave radio frequency bands. This will demand an entirely new radio, but it is also clear that the use of mmW technology will have far-reaching implications throughout the entire network, extending to every aspect of the RF design and at every point there will be a need for new integrated solutions that simply did not exist before the demand for 5G emerged.
This highlights the double-layered nature of rising system complexity. As the systems are becoming more complex, the devices needed to implement these systems are often themselves being developed in parallel with the applications. Managing this level of complexity, from a design point of view, requires a system-level view and a new, intelligent approach to design. Underlying it all lies Cadence's deep knowledge of computational software, which provides the fabric to enable algorithms for design, implementation and verification.
This scenario isn't unique to 5G. The automotive industry is going through its own, very similar transition as it moves closer to full autonomy. Cars have, for some time, been considered systems; now, the boundaries of those systems are extending far beyond the physical periphery of the vehicle. All cars will be 'aware' of all other road users, both those that are autonomous and those that are not. The vehicle's system will now include everything that it could potentially interact with, because every interaction will need a reaction, and that reaction needs to come from the vehicle.
This accurately illustrates the way that system boundaries are being redefined, to include more and exclude less. The same approach is happening within the EDA industry, where the tools must now work more systemically and consider much more than they were originally designed to handle.
The system level challenge
It is apparent that the systems being developed now will not exist in isolation. Autonomous vehicles will need high-speed, always-on, always reliable wireless connectivity to function. The 5G network will be important here, but so too will other forms of wireless communication, such as Wi-Fi. This complex landscape will define the specification of new solutions. And even within a car – the system of many systems inter-operating within an even bigger network with other cars and future smart city infrastructure – the architecture of the car is undergoing fundamental changes towards networked zones and several areas of centralized compute.
Meeting these requirements will require these new solutions to be developed with both safety and security at front of mind. Nothing can be put into service today without including some level of protection against cyber threats. Maintaining that level of security across system boundaries introduces a new requirement. Safety, too, becomes a multi-domain issue, as autonomous vehicles will rely so heavily on the 5G network, a safety-based approach to implementation will be a high priority.
These multi-modal, systemic design requirements are best addressed by subject matter experts. For this reason, many OEMs are now taking the approach of developing their own integrated solutions, often determining the usage of the underlying semiconductor IP or even developing their own silicon, causing a re-shuffling of the design chain. Manufacturers in vertical markets are (re)turning to in-house chip design purely because they are best placed to define the requirements and ensure they are met. The challenge they now face is engaging with EDA vendors that understand their needs and have the tools to meet them.
Today, integrated circuits are incredibly complex, with potentially billions of nodes that need to work flawlessly. Because of this, new designs often start with proven IP, but now that the applications have such domain-specific requirements, this IP must be more flexible, allowing for configurability.
Every aspect of the system has an impact on the overall functionality and performance, so the design of the IC must be approached from a system level initially. This means using a design environment that has been developed to support a more intelligent approach to system level design.
EDA tool flows must now comprise multiple tools, starting at the block (IP) and subsystem level with the right processor and design building blocks, to the “system-on-chip” level for defining the smallest physical details. That chip will need to be housed in a package and that package mounted on a PCB. The PCB will also need a housing and that enclosure may be placed within a larger cabinet or even a vehicle. The environmental conditions of that cabinet or vehicle will have an impact on how the chip performs, perhaps at the level of that smallest physical detail.
Observing how the operating environment will impact the chip once it is in the field is too late; that happens long after any design changes can be made to compensate for those environmental conditions. Design changes must be made when the chip is still in development, at the start of the process. This illustrates the scale of the design challenge and how it requires a new approach to system design in order to address it. Designers need to model the entire system using multi-physics analyses, including the AWR Design Environment platform for RF and microwave design, the Celsius Thermal Solver for electrothermal analysis, and the Clarity 3D Solver for electromagnetic simulation. Bringing this together in an integrated environment, including safety and security, enables faster time to sign-off. Cadence is enabling this integrated design process through intelligence, to save manufacturers huge costs and valuable design time.
This is why Cadence has used its expertise in computational software to develop its Intelligent System Design strategy. It starts with optimised EDA tools that work together in a common digital flow, covering the IP, packaging and PCB design across multiple domains, including thermal and electromagnetic. This multi-modal approach is extended through system innovation, to add new capabilities to the design flow. Bringing this together demands the adoption of leading-edge technologies such as AI and machine learning, which Cadence harnesses through its expertise in computational software.
System boundaries are changing, and trends including 5G and autonomous driving are two noteworthy examples. The use of AI from the core to the edge is also forcing manufacturers to reconsider everything. Tackling the design challenges this presents can be daunting, but through its Intelligent System Design strategy Cadence continues to enable manufacturers to meet these challenges and manage escalating system complexity.
Author details: Rebecca Dobson, Corporate VP EMEA, Cadence
