AI could be the key to understanding the data collected by the IoT.

Big data is useless and all the sensors in the world are not enough. Contentious? Maybe. I've talked in the past about the prospects for digitizing the world and it's true that we have many of the ingredients to make this happen: tiny, low-power sensors including optical and MEMS inertial sensors that provide contextual awareness; connectivity technologies for almost every practical and budgetary constraint; low-cost processing power and mass storage.

We're well on the way to seeing almost 30 billion devices connected to the IoT in the next couple of years, and there is no practical limit to this. We have enough IPv6 addresses to cover the earth's surface many times over with smart "things." We can easily collect the data we need to digitize the world.

The bigger challenge is to understand what that data are telling us and, from there, determine suitable responses. The sheer volume, velocity and variety of data we can now capture through IoT devices easily exceed the capacity of humans to analyze and extract meaningful insights manually. AI is the perfect companion to the IoT, capable of providing the assistance we need. Bringing them together as the AIoT is the key to tackling complex challenges such as sustainability. Studying the climate and humans' impact, the effects of using natural resources such as energy, and the prospects for controlling and managing these are subject to huge numbers of variables that are impossible for us to analyze effectively.

ECTC revealed the latest developments in 3-D hybrid bonding.

More than 50 presentations on hybrid bonding filled rooms to capacity at the IEEE Electronics Components and Technology Conference (ECTC) in Orlando, as did the Tuesday morning panel session. That panel discussion, on "Copper Hybrid Bond Interconnections for Chip-to-Wafer Applications," organized by Infinera and Qualcomm and moderated by TechSearch International, included perspectives from design and EDA (Synopsys) and research institute IMEC, users of hybrid bonding (AMD with production at TSMC and Intel with its internal development program), equipment makers Besi and EVG, and yield and reliability specialist PDF Solutions.

Synopsys pointed out the importance of design tools and IMEC described key drivers for 2.5-D and 3-D integration technologies such as increasing system complexity, increasing need for heterogeneous integration, increasing die-to-die interconnect data bandwidth (more interconnect channels and higher interconnect speeds per interconnect), reducing die-to-die interconnect energy with shorter distance interconnect, scaled, lower capacitance interconnects, and lower voltage. IMEC noted the issue is not the number of interconnects but rather the available (local) interconnect density enabled by interconnect pitch scaling. Many of today's hybrid bonding applications using die-to-wafer structures are focused on the high-performance space where the cost can be justified, including stacking SRAM cache and logic-on-logic, as introduced by AMD's products in desktop, servers, and AI/machine learning.

Clear communication with manufacturers can ease that queasy feeling.

You are finally finished with that very complex monster of a printed circuit board (PCB) design while enduring a very tight project schedule. Now it's time to get the design data to manufacturing for fabrication and assembly, but there is a lingering doubt in the back of your mind, an uneasy feeling in the pit of your stomach as you hand off data to manufacturing. Why? I'll explain, based on my perspective gleaned from decades of designing PCBs.

Let me start with a question: Did you collaborate with your manufacturing suppliers up front? The answer for many is typically "No!" You may have simply designed a PCB without any manufacturer involvement whatsoever, generated output files and threw the data over the proverbial wall to the manufacturer in hopes that what you designed is buildable and will work as intended. Then, when a technical query (TQ) hits your inbox, you might get upset and wonder why you're being notified that the job is on hold because the design data contain issues or are missing information that needs to be addressed before fabrication can begin. In some cases, it's not just answering with a reply of "approved to modify as suggested," but rather it requires going back and redesigning! Sadly, this happens in our industry far too often, and in more instances than you would expect.

Manufacturing costs include more than just material, labor and overhead.

Possibly it's because inflation has ticked up over the past year or so, or possibly it's because despite a long career in manufacturing, and I am still not sure all stakeholders from production, sales, customers, suppliers – and especially accounting – really understand or agree, but I find myself now more than ever trying to identify and come to grips with the "real cost" of what I produce.

I am hardly the first to ponder this question. Truth be told, I spent too many years early in my career performing standards engineering and being responsible for product costing. All that experience, I fear, has left me more a skeptic than an expert on product costing. Too many companies in our industry have ended up foundering, in no small part because of their leadership not understanding the real cost of their product.