There are more connected devices today than ever before, and greater computing power than the world has ever known. But within each device lies a complex chip, powered by a semiconductor industry that has largely remained invisible amidst discussions of digital transformation.
From Integrated Circuit Design (IC design), to wafer production and back-end assembly and test, a multitude of hands and entities are involved in making chips. These companies, meanwhile, are supported by a host of others who supply the equipment and tools needed across the entire process.
With the advent of 5G, the semiconductor industry has had to design and manufacture an entirely new type of chip suitable for 5G-enabled devices. And these companies are turning to AI/ML, ERP software and other such technologies and applications to do this, and are utilising the very same digital transformation tools they are helping facilitate.
One such company is Singapore-headquartered AEM, which supports back end assembly and testing – the last mile in the chip-manufacturing process. We speak to Chandran Nair, CEO of AEM to find out more about how testing has gotten more sophisticated and complex over the years and how it’s ushering in the new technological age of 5G and IoT.
The semiconductor industry can be divided into phases, with companies doing Integrated Circuit design, companies involved in front-end wafer production, and companies that do back-end assembly and test. A few integrated device manufacturers (IDMs) like Intel and Samsung are involved in IC design through front-end production, back-end assembly and testing. AMD, Qualcomm, HiSilicon and others specialise in IC design. Front-end wafer production is done by companies like Taiwan Semiconductor Manufacturing Company (TSMC), GlobalFoundries or Semiconductor Manufacturing International Corporation (SMIC). Companies like UTAC, JCET, ASE Engineering engage in back end assembly. AEM provides testers and handlers to them. AEM equipment is mainly used to test and handle chips coming out of the manufacturing flow.
Within the semiconductor industry, what are some of the most important innovations in progress? And how are they going to impact enterprises?
Semiconductors play an important role in almost every aspect of our life today. Almost every new product has a high level of semiconductor content — it could be an automotive, mobility devices, aviation, of course, computers. Almost anything a consumer touches, almost anything new a factory uses has semiconductor content in it.
If we start at the factory level, almost all the production machines are dependent on control systems that use semiconductor devices. Modern factories and retrofitted factories require continuous monitoring of data to make sure that their factories and machines are working well. The impact of the workflow on things like yield and utilisation are monitored, again, through semiconductor devices. The ability to gain efficiency, increased productivity, are deeply influenced by the semiconductor industry.
Then let’s go one level up — what does one do with all these data? Let’s assume that it’s a factory which is well instrumented with sensors, with lots of automation on the floor. The data goes to the cloud — either private or an external cloud, depending on the kind of factory and the agreements they may have.
Then there is data analytics that helps the factory do statistical process control. If it’s a large enterprise, that helps the factory connect with the ERP, connects the ERP software to the factory level software on how many goods need to be produced. How much is going into inventory? What is being shipped out? All those are computer-based systems are heavily dependent on the semiconductor industry.
There is a lot of instrumentation and control equipment that goes into helping you become a better researcher or developer of products. The different kinds of instrumentation, computing processes and power available today are directly impacted by the semiconductor industry.
Finally, you sell to consumers, who could be mass consumers or B2B companies. If it’s the consumers, a large part of their spending cycle is based on innovative features — faster, better, cheaper. That is directly impacted by the semiconductor industry.
How does machine learning help reduce the cost of tests?
Machine learning is a widely used and sometimes abused word. When you look at the basics, data analytics and machine learning are integral parts of our future test and handling methodology. It spans the use of data for intelligent predictive maintenance on our machines, for example, or it can be also used to better slice and dice test data, to understand trends, patterns, quantify risks, make decisions. This provides the customer insights into manufacturing processes that will help them decide how much testing is needed, what kind of tests are needed, how certain processes may be speeded up, etc. The ability to collect valid data, to stream that data in a way that machine learning algorithms can use, and then to give valuable insights is a continuous effort that we have started making for our customers.
Now, the more general context of machine learning is widely used for prognostics and diagnostics of predictive maintenance – before something breaks, to be able to predict what is wrong and what you should be doing so that the machine doesn’t end up being broken. That is the broader context of one of the uses of AI amongst the millions of different uses.
5G is being deployed in certain parts of the world already. How’s that going to disrupt the industry?
5G is yet another word that is massively used and abused.
There are different levels of what people call 5G. One is the telecoms company being 5G-ready right, having the infrastructure to be able to support 5G communications — both at the cell level, at the mobile device level — and the ability to have large networks or micro networks on 5G.
But basically, the importance of what I call the 5G economy is the ability to transfer large amounts of data between devices, which is supported using 5G. In addition to having tremendous developments on edge computing, it is the ability to put high power processes everywhere. This communication, with sensors, allows a wide variety of applications to be taken to market that can have direct impact, positive impact on human lives.
From the ability for a doctor in a city hospital to be able to use the computing and connectivity provided by 5G to look at and diagnose a patient in a remote area — all the way to applications where terabytes of data can be sent to the edge or to the cloud that can then compute at a high speed and manage mission-critical control systems. There’s a wide variety of applications that can have a direct impact on human life. When you go one step further and talk about autonomous vehicles’ ability to talk to each other, know where each is with regards to the other, talk to infrastructure, communicate, restructure and then navigate autonomously — all these capabilities get powered, because of the ability of 5G networks and high speed computing. And fantastic sensors, of course, at a very reasonable cost.
How does AEM fit into the 5G economy or 5G innovations that are coming?
Many of the devices are getting more and more complex. You have system in package (SiP), system on chip (SoC) — multiple IP nodes being put into a single package. In addition, you have what we call heterogeneous packaging and chiplets. With all these things being put together, the complexity of the device makes traditional tests almost impossible. The only way to test it is to do system-level tests. And that is where we play a big role.
You can think about system-level tests in their simplest form: how do you know that a CPU works? You put the CPU in its environment — in a motherboard — download software and run it to know if the CPU is working. That’s a simplified system-level test. For complex devices, only system-level tests can assure you a zero-failure world.
Think about the Boeing 737 Max incident. It’s an aerospace product — every single part of it has been very well tested. So just because every part is tested does not mean that when it is put together and tested again, it works well. In the corner pieces where it does not work, it can be catastrophic.
You have cars and aeroplanes working on a complex chip. Can you afford to not have full test coverage? You cannot.
Do you have to come up with a qualitatively new way of testing, considering how complex these devices are? And what sort of specific innovations led you to be able to do these tests?
System-level tests have been around for about 20 years; but typically, they have been expensive. So people would do more of sample testing — one out of a batch of many. But now, as the consumer demands assurance, safety, no thought of failure, people need to do population testing — every product has to be tested.
Now, how do you take the system-level test concept to then allow millions to be tested in parallel at high speeds so that the throughput of high volume manufacturing can be maintained? The innovation comes in during the automation process, in the test methodology, in the way you use test data to make intelligent decisions.
What should enterprises look forward to, within this sort of 5G-enabled IoT world? And how can they leverage the power of these chips to conduct their own business?
The 5G-driven economy will enable organisations to be run efficiently, enable the benefits of technology to go down to the very far reaches of society, and enable better use of resources globally.
That can have direct impact on healthcare for the masses, on employment, on agriculture, and through technologies like blockchain that are also powered by the semiconductor industry.
COVID-19 has done terrible things to the world. But the one silver lining is that it has taught us how to use technology to be effective, without necessarily flying to all the corners of the world. This has a direct positive impact on climate change and on our planet. Now, it has also taught companies to hire worldwide. The silver lining in it is that we can continue to use the good practices even when the world returns to normalcy. The 5G economy accelerates that usage of technology for good.