By Björn Zeugmann and Olaf Enge-Rosenblatt
The digitalization of industry is progressing in leaps and bounds, albeit not at the same speed everywhere. In many industries, processes can be digitalized well to very well — for example, because electronic control systems can be retrofitted from analog to digital relatively easily. In some cases, new industries emerge only because processes have become digital. However, some established and well-performing industries aren’t yet benefiting from digitalization to the same extent.
In traditional mechanical engineering, for instance, it’s often difficult to equip existing plants with electronic sensors and actuators. Digital systems currently available on the market are mostly off-the-shelf solutions that are not sufficiently geared to the needs of this industry. This makes them difficult to integrate into the high-precision production of plant and equipment, which is often based on years of experience. The most important other factors for equipping plants, besides price, are the space requirements of the electronics, their robustness, their interfaces for communication and — occasionally — their power consumption. It’s also not an uncommon requirement for the additional electronics to have very little effect if any on the mechanical parts of the system.
A technical solution to this dilemma might be to use a circuit design that has been customized for precisely these requirements of the specific application. Such an adapted system can be implemented in the form of an ASIC, or — in the case of more complex setups — a chiplet, in combination with necessary, but reduced, discrete external wiring. The hardware is optimized for the actual task and can be operated in conjunction with established control systems using standardized interfaces.
One disadvantage of the ASIC-based solution is the higher unit cost. A standard solution can be used in a variety of situations and thus produced at lower cost. With the customized solution, this broad range of applications falls away. However, there are several ways to reduce various different costs. For example, the design process can already be highly modularized, so that different circuit designs share similar bases and circuit components can be taken from the corresponding libraries. This is common practice. The choice of technology based on performance requirements also has a major impact on manufacturing costs. Splitting the design across multiple technologies and using chiplets for modularization similar to the design itself can also reduce costs when viewed across multiple systems/applications. The cost-benefit ratio must be carefully weighed on the basis of these and other factors, depending on the application.
Today, it is often the case that large amounts of data are first collected from machines or in processes, then transferred to a cloud (either on-premises or public) and analyzed there. This requires high-performance transmission media and computing capacity. Due to a lack of real-time capability, cloud-based computing doesn’t lend itself to controlling machinery and equipment in most cases. This calls for edge solutions, which are close to the point of use but offer limited computing power. In order to evaluate data directly at the edge, the algorithm must know as precisely as possible what it is looking for. If this algorithm is known — or it can be found through appropriate upstream measures such as machine learning — then customized hardware tailored precisely to the application purpose and the corresponding software can mean an immense advantage over standard solutions. Calculations can generally be conducted faster and more energy-efficiently directly in special hardware than through a software solution on general-purpose hardware. What’s more, omitting components that aren’t needed can also save energy as well as space and thus costs.
Despite their higher cost, customized, integrated solutions really come into their own in applications with special requirements. This is particularly true for scenarios where standard solutions come up short. In these cases, customized circuit design has a significant role to play in the transformation toward Industry 4.0.
Olaf Enge-Rosenblatt is group manager for computational analytics in Fraunhofer IIS’ Engineering of Adaptive Systems Division.
Björn Zeugmann is group manager for integrated sensor electronics in Fraunhofer IIS’ Engineering of Adaptive Systems Division.