The power of data standardisation in driving AI trends for electronics engineering

However, just like in other industries, existing legacy barriers have been slowing down the speed at which these changes can have a deeper impact and transform companies and processes. On electronics, one of the major barriers is the quality and quantity of data available in a comprehensive digital format, which stands as a critical factor to achieve the needed performance from models. To overcome these challenges, it is necessary to rethink how information is handled, exchanged, and processed between all actors in the industry, especially component manufacturers and engineers.

Inconsistent data formats create overheads

Several processes within electronics engineering depend on gathering and comparing data and running calculations. This information is predominantly sourced by engineers from the internet in the form of component datasheets, reference designs, application notes, or architectural drawings. The abundance of information, however, presents a challenge. It is not easy to find the most relevant numbers and facts, and to compare them, as usually those are embedded within PDF files, which aren’t truly ‘digital’.

Furthermore, the global diversity of hundreds of manufacturers and distributors, each presenting their data and definitions in their own terms, without any standardisation, increase the complexity of the engineering work, relying on the interpretation and analysis of each specialist to compare solutions and devise the best one for their application.

While component manufacturers and distributors try to close this gap by keeping a close relationship and providing design support for engineers, they tend to focus most of their time on large accounts as this approach is not particularly scalable. As a result, the majority of engineers engage with manufacturers primarily through their technical documentation and datasheets, without enjoying the benefits of direct interaction, and experiencing all the issues related to the lack of a data standard.

Standardisation examples to date

Data standards are not new in electronics and other engineering fields. They were created and grown in strength and adoption as digital tools have made their benefits clear. Gerber files, used now for decades, improved substantially how information is exchanged between engineering companies and electronics manufacturers. Emerging standards, like IPC-2581, offer greater capabilities, facilitating the split between design and manufacturing, the creation of a healthy competitive environment and allowing companies to focus on their core businesses.

Within the component realm, JEDEC recently introduced the JEP30 standard, targeting a replacement of the traditional datasheets in PDF through this new digital format, to streamline the information exchange between manufacturers and engineers. Similarly, CELUS has developed an internal digital standard to describe circuits, components, and their applications in a format called CUBO, enabling automatic processing by artificial intelligence algorithms and deep learning models, resulting in circuit and component recommendations, and automated generation of project design files.

The success of a digital standard in the industry depends on the balance between ease of adoption and attention to detail. These evolving standards transform how companies interact and collaborate, driving efficiency and innovation in the industry.

Data availability and interpretability are key

The transformation unlocked by data standardisation in the electronics industry goes beyond simply improving how engineers compare PDF files. Over the past year, there has been a massive increase in the number of applications of machine learning, particularly the ones driven by large language models. This is highlighted by the fact that ChatGPT and other similar models achieved a remarkable level of performance and enabled a direct application of this technology in many fronts where data was readily available and could be used for training and refinement. The key to this advancement was the widespread availability of the required information in digital form, allowing automated processing and preparation for integration on learning routines of machine learning models. To latch onto the trend of smart, AI-based engineering tools, unleashing the potential of these technologies to support electronic engineering processes, it is necessary to prepare the data to be used digitally and unambiguously.

Complex models require vast amounts of data to learn patterns and rules effectively. As a matter of fact, the volume of content related to electronics design available is only a small fraction of the total amount of written content across all fields. Implementing data standards helps to reduce the necessary amount of data to reach good performance by simplifying the learning process and removing a big portion of ambiguity. This gives momentum to the trend and accelerates the transformation of engineering software. These advanced tools will help to bridge the gap between component manufacturers and engineers in a much more scalable way than traditional support through application engineering. They enable a deeper understanding of applications and project requirements, simplifying design-in and both increasing design quality and chances of design-win.

In conclusion, the adoption of a common data standard in the industry brings numerous advantages. Firstly, it significantly reduces the effort required to identify the best component and its application circuit, eliminating the need to manually transfer data from PDFs into other software, as they can directly import parameters from digital assets. This also removes the need to spend hours recreating basic circuits in EDA tools since they are readily available to be imported, diminishing the likelihood of typical human errors during the process. Finally, it allows for the trend on smarter, AI-based tools to bring extensive benefits to the industry, where engineers can dedicate more time to innovative architectural solutions and design validation and automate repetitive work.

This trajectory follows the path taken by the semiconductor industry, where the focus on creativity and validation has led to notable advancements. Adopting such a standard is a step towards greater efficiency and innovation in electronics engineering.