“Our Main Goal Is To Develop Systems With High Inspection Performance, Using Entirely AI Based Technology”

Roberto Gatti, CEO, Delvitech

Roberto Gatti, CEO, Delvitech, speaks with Mukul Yudhveer Singh on how the company is implementing AI into PCB manufacturing machinery. Set up in 2018, Delvitech designs, produces and implements automatic optical inspection (AOI) systems for assembled printed circuit boards (PCB).

EB: What was the inspiration behind creating an AI machine for PCBs?
Delvitech is a new company. Starting out, it was clear to us that the market already had important and respectable competitors with an extraordinarily strong market share. It was also clear that the different types of AOI systems in use, too, were competing on the basis of microns of precision and resolution, even as they met diverse customer needs.

So Delvitech set out running in the opposite direction. We wanted to offer one system that was flexible, scalable, and easy to use, with an extremely high error classification capability.

T.V. Thomas, Head-Asia operations, Delvitech, SA

The question we faced was, “How to achieve all of this?” 
There was only one way and that was to use as much AI technology as possible at the heart of image processing. Customers need to have immense freedom in using an AOI system. This was the inception point of the machine that Delvitech is proud of today.

EB: Can you give us an overview of your D.ONE machine?
We can describe the D.ONE machine by Delvitech as a cost-effective post-solder AOI solution. It has been designed to deliver excellent results for improving the product quality and boosting productivity. In fact, our aim is to ensure superior accuracy and precision with minimised false calls.

Our system’s strength is the advanced optical head, which gives the highest performance in resolution, detection and inspection speed, by utilising five digital cameras. Let me say that Delvitech’s technology allows us to obtain different benefits.

First, the D.ONE has a relevant storage capacity. The integrated library system allows us to manage a significant amount of data, avoiding all the problems generated by a large number of images.

Second, the library is completely customisable— the machine works using the information from a complete data set, but every client can add different elements to the library. In addition, thanks to D.ONE Analysis and D.ONE Report, it is possible to use the repair and programming data stored to generate reports. D.ONE Core offers two different inspection methods—TBA (threshold based analysis) and MLBA (machine learning based analysis).
The advantage of having both is reflected in the high detection levels with low false calls. But our ‘diamond tip’ is without any doubt, artificial intelligence. We use the most advanced neural network techniques to give the best outcome when it comes to error classification. All the upcoming developments will progressively enhance the use of deep learning techniques to allow the system to be self-programmed and give nearly absolute results. In the next few months, we will share some important news. For this reason, I invite readers to stay informed through our social media platforms.

EB: Is Delvitech’s machine any different from traditional 3D AOI machines?
Compared to the classical 3D AOI machine, which currently dominates the market, our 3D machine is different, with unique and innovative features. For example, let us just consider that the acquired images come from five cameras (one top and four on the sides) placed at different angles.

Through these acquired pictures (without any rendering), the algorithms are capable of working on 3D for error classification. Obviously, these algorithms must have specific features, while the cameras and the machine must have an extremely high image acquisition speed. In addition, we cannot forget the use of artificial intelligence to improve all our performance parameters, which in turn enables our unique inspection capabilities.

We use artificial intelligence to obtain the best performance possible in terms of detectability, programming time and false calls. Delvitech entered the industry by innovatively converting typical AOI systems into an inline black box. This setup is implemented onsite. The machine learning algorithms used enable the collection of a lot of information desired by the client.

EB: Is there a price difference too?
The D.ONE system is in line with the cost of its high-end competitors. The difference is not in the cost of the system, but the benefits it offers.

The differences will increase dramatically once the new technology connected to the D.ONE system is introduced in the market. We are talking about a two-digit percentage difference, which translates to saving a huge amount of money

EB: How can the Indian PCB  industry benefit from using D.ONE?
The Indian PCB industry can benefit from the use of this technology in several different ways.
There is no longer a need to make very complex investments on very different production systems for different processes.

There will also no longer be any need for advanced AOI experts to do AOI programming, or worry about wasting long hours in programming, or in completely tuning programs to get better performances.

There will be a huge positive impact on costs due to the improved yield of any production line that incorporates a D.ONE system, as well as the new technology embedded in the latest model that Delvitech will shortly introduce to the market.

EB: How are you managing the Indian market right now? Have you built up a network of agents, distributors and salesmen? 
As of now, the Asian market is looked after by our head of Asia operations, T.V. Thomas. With him at the helm, we have signed a distributor contract with Sanson Technologies based in Bengaluru, thanks to which we can offer both commercial and technical support to our customers.

Our clients can benefit from the presence of local distributors, or from the direct support of the Swiss headquarters. This is because we want to guarantee the best international support possible to our international customers. Delvitech is also organising training for Sanson engineers to ensure they are up to date with its technologies.

As for marketing, it is managed directly from the headquarters in Switzerland. Marketing drives our sales strategies across the world and, for this reason, we have a high consideration for this activity. That is why we have decided to go to the market with a unique tone of voice, which has been defined at the headquarters in Switzerland, based on the strategic importance.

EB: Including India, what are Delvitech’s plans for the Asian market?
We consider the Indian market a strategic one for Delvitech but, at the same time, only our starting point. Delvitech wants to establish itself as a multinational company investing in the Asian market.

In recent years, for the ASEAN countries, the manufacturing sector has been (and will increasingly be) one of the main drivers for the economic growth in the region and Delvitech must be ready. We are planning to establish a new branch office in Singapore and Hong Kong, before expanding into China.

EB: What is your role within the company?
I have a technical and commercial background in the electronics field, where I have more than 30 years of experience in various activities. I started out as a support engineer for CAD, CAM and CIM systems. I then became a product manager, a sales manager, and finally I started managing my own companies—in Italy, Germany, Tunisia, the US and in Asia. Since 2018, I have been CEO of Delvitech and the chairman of its board of directors.

EB: Tell us more about how Delvitech was born.
Delvitech was born in 2018 based on a personal business idea, which was to create excellence in the inspection industry by founding a company in the field of AOI systems for PCBs. But the electronic sector is just a starting point since we are planning to expand into the food and pharma markets too.

At first, we planned to set up Delvitech in Virginia (US), but then together with the technological investors, we decided to open Delvitech SA in Tessin, Switzerland.
We started in October 2018 with TiVenture (Banca Stato Foundation) and a family office based in Lugano, Switzerland as main investors. Delvitech is a young company focused on cutting-edge technologies, with the support of important Swiss partners and universities specialising in AI, new technologies, and applied science. Our main goal as of today is to develop systems with high inspection performance, using entirely AI based technology.

EB: What is your core business then?
Our core business is to produce machines that can be used in the field of optical inspection for the assembly of electronic boards, in the stages of post-paste, post-placement, post-reflow, post-wave and post-selective soldering.

In the future, our business will also involve the development of optical inspection systems to be used in the pharma and food markets. Delvitech will be an increasingly technology-oriented company with only final assembly inhouse operations in our headquarters here in Switzerland.

EB: Tell us more about the growth plans charted out.
First, we will go to the market with the Head+, which consists of an especially important and enhanced upgrade to the existing systems. But the breaking news, especially for the Indian market, will be the self-programming software totally based on AI and entirely developed by Delvitech and its partner, IDSIA.

The self-programming software will allow a completely automated optical inspection with a significant reduction in programming time —by 80 per cent. In addition to reducing the optical inspection time, what is crucial is that the common errors that occur with manual programming will no longer happen.

We have a US$ 28.6 million development plan, which not only includes the release of the Head+ and self-programming software, but also the development of the D.ONE and a brand new technology for post-placement processes, which will be announced in November this year (to be officially installed by some customers starting from the second half of next year).


Please enter your comment!
Please enter your name here

Are you human? *