Reading time ( words)
In Norway, a recent survey showed that one out of seven engineers thought that their job could be performed by artificial intelligence (AI) within 10 years. If we look at the engineers in our industry, I would say it’s fair to assume that the numbers are even higher, and that it will happen faster. But that will be nothing compared to what will happen to the rest of the players in the supply chain. To explain why, we need to look at the underlying needs that drive the supply chain itself.
As automation works its way onto the shop floors, it still struggles to replace humans in the supporting roles, such as designers, purchasers, brokers, and back-office staff. Where automation on the shop floor replaces humans in doing repetitive manual tasks, the supporting roles (at least some of them) require intelligence to understand and utilise information.
Even as tech-savvy as our industry is, it still builds upon old technology (just a quick look at the expressions still used, like screens and stencils, confirms this), and the human brain is great at making sense out of unstructured information. Some tasks are slowly being taken over by machines, but not particularly intelligent ones. One example is the ERP systems that are automating procurement and doing push and pull requests on deliveries or predicts needs and place orders.
The 50-Year-Old Legacy
The PCB industry carries the legacy of more than 50 years of development and language. To be manufactured, PCBs depend on people who understand them, compare them with their generic requirements, and find partners for assembly or manufacturing. Add to that the secrecy needed and the different formats used, and humans are pretty much your only option, for now.
This is where AI comes into play. This universe of formats, languages, dialects and expressions is complex, but still operates within a set of rules. As AI slowly learns these rules, it helps humans to make sense of it. Gradually, as the intelligence expands, it replaces humans altogether. Integr8tor by Ucamco is an example of this. Although it could be argued that this is not actual intelligence—just a set of algorithms and code—it replaces humans in building understanding. There are a number of other examples of software tools that do the same thing or performs related tasks; there will be more and they will become better. They won’t take your job, but instead, they will simply allow you to be more efficient. For them to replace humans, they need to step it up and integrate with other systems, and the confidence in them needs to grow. This will happen; hundreds of developers are already working on it.
In this chaos of formats, languages and dialects, AI is not the only path to choose, and it is not an “all or nothing” choice. CircuitData is an open source project that uses standardisation to replace brain power.
A Substantial Challenge
My colleague, Jan Pedersen, senior technical advisor, has more than 40 years in printed circuits, receiving each week numerous files with manufacturing data. The pattern is quite repetitive, and not in a good way. There is one substantial challenge that keeps coming back: insufficient—or even totally wrong—article specifications.
When he receives manufacturing data from 10 customers, he often gets 10 different ways to present and explain what’s asked for. Tolerances are expressed differently. Fabrication drawings are explained differently. Solder mask is called green mask or even green oil. Component notation is called legend, silk-screen, silk layer or something else. We can live with a different word, but not a missing requirement!
We have decided to challenge the world and improve this issue. We have started to create a standard PCB specification, a new language to share information. We are offering this new standard—or language—as an open source to the PCB industry. It is free of charge and will be a valuable source of information for any PCB designer, user or not.
A Language in Addition to, not Instead of So, what does CircuitData do? To begin with, it aims to be a supplement to the files that are currently used, such as Gerbers, ODB++ and IPC-2581, so that all aspects of manufacturing not provided are resolved. As the standard evolves, we want it to replace the need for transmitting the files throughout the supply chain.
This can be done by providing the different parties with only the information they need, and leave the exchanging of files to the first and last part of a set supply chain—the OEM and the manufacturer.
But, as with all free and community-driven projects, the value of the language increases simultaneously as the number of contributors and users rises. It needs to be developed, maintained and fueled like an old and beloved sports car. Because communicating PCB specifications is a dialog, not a monolog. It requires several pieces of information to come together so that all the facts are presented correctly. But the specifications are just part of what is needed; they need to fit with profiles (requirements) and capabilities.
To use the format, you start out by reading the documentation. Then you prepare your internal systems to be able to send and receive this kind of information. Over time, we believe that there will be several systems available that will help you utilise the format.
To read the full version of this article which appeared in the November 2017 issue of The PCB Magazine, click here.