This country’s semiconductor chip shortage is likely to continue well into 2022, and a Georgia Tech expert predicts that the U.S. will need to make major changes to the manufacturing and supply chain of these all-important chips in the coming year to stave off further effects.That includes making more of these chips here at home. &nbsp;Madhavan Swaminathan is the John Pippin Chair in Electromagnetics in the School of Electrical and Computer Engineering. He also &nbsp;serves as director of the 3D Systems Packaging Research Center. &nbsp;As an author of more than 450 technical publications who holds 29 patents, Swaminathan is one of the world’s leading experts on semiconductors and the semiconductor chips necessary for many of the devices we use every day to function.&nbsp;“Almost any consumer device that is electronic tends to have at least one semiconductor chip in it,” Swaminathan explains. “The more complicated the functions any device performs, the more chips it is likely to have.”&nbsp;Some of these semiconductor chips process information, some store data, and others provide sensing or communication functions.&nbsp;In short, they are crucial in devices from video games and smart thermostats to cars and computers.&nbsp;Our current shortage of these chips began with the Covid-19 pandemic. When consumers started staying at home and car purchases took a downward turn, chip manufacturers tried to shift to make more chips for other goods like smartphones and computers.&nbsp;But Swaminathan explains that making that kind of switch is not simple. Entire production operations have to be changed. The chips are highly sensitive and can be damaged by static electricity, temperature variations, and even tiny specks of dust. The manufacturing environments must be highly regulated, and changes in the process can add months.&nbsp;The pandemic highlighted another challenge with the semiconductor chip industry, according to Swaminathan.&nbsp;“There’s a major shortage of companies making chips,” he says. “If you look worldwide, there are maybe four or five manufacturers making 80-90% of these chips and they are located outside of the United States.”&nbsp;This creates supply chain hiccups with the raw supplies needed to make these chips as well. Add in the fact that many of these companies only design their chips – they don’t manufacture them directly.&nbsp;“American consumers use 50% of the world’s chips,” Swaminathan says, which creates a serious challenge when the overwhelming majority of those chips are manufactured in other nations.&nbsp;In the short term, the costs of the chip shortage is being passed on to the consumer. We see this directly with products like PlayStations and Xboxes that are more and more expensive and harder to purchase when the chips necessary for the consoles to function are in short supply.&nbsp;Beyond 2022, Swaminathan says we need to work to revitalize the industry domestically.&nbsp;“We need to bring more manufacturing back to the United States,” he says. “The U.S. government has recognized the importance of this semiconductor chip shortage and is trying to address the issue directly.”&nbsp;That means investing in new plants to manufacture the chips, but America's journey toward &nbsp;chip self-sufficiency will continue to be a work in progress.“This is a cycle,” Swaminathan explains. “But this is probably the first time where it has had such a major effect in so many different industries.”&nbsp;But consumers can take direct action on their own in the coming year. “Reduce the number of times you purchase or upgrade electronic devices like phones and cars,” he says. “Then it becomes just a supply problem, not a demand and supply problem.”

Georgia Tech expert predicts that America will need to make major changes to the manufacturing and supply chain

Addressing the Microchip Shortage

For most of our nRF52 Series, we have a chip-scale package and a Quad-Flat No-leads (QFN) option. With our nRF52840, we introduced the aQFN package to our product portfolio. We also decided to offer the same device in a QFN package, so Nordic customers now have the option of using nRF52840 in QFN48. This has just been released to the market.There are two ways of looking at this new option. You can see it as a less complex nRF52840 with fewer pins and 1.7-3.6V supply voltage for our customers that don’t need USB and don’t need to run directly off Li-Ion batteries. It is also similar to an nRF52832 with additional security features, more flash and RAM as well as Zigbee and Thread support.I used this opportunity to read up on different package options, and reeducate myself. I will share with you what I learned.<h2>Then and now</h2>The first microcontroller I got in contact with was in a Dual in-line package (DIP). You don’t see them much nowadays, but DIP packages are still beloved by hobbyists. They are easy to solder, and you can do a software update with a screwdriver. Just pop out the microcontroller with a screwdriver and put a new one with updated software into the socket. This is not a modern or resource-friendly approach, but it works if you sell &lt;1000 devices a year that do not need much intelligence. One company I know of still uses this method in some of their older machines.On today’s microcontrollers, you mostly see Quad Flat Packages (QFP), QFN, and Wafer Level Chip Scale Packages (WLCSP). When it comes to QFP, you will see a lot of Low profile Quad Flat Packages (LQFP) and Thin Quad Flat Packages (TQFP). These are surface-mounted packages with the leads extended out to the four sides of the package.They have been around for decades and are still widely used. They are still solderable by hand (which I do not recommend), and the extended leads allow for visual inspection after assembly. They also make the solder joints more resilient to thermal and mechanical stress than leadless packages.The extended leads are also what cause trouble with these packages. If one of the leads gets bent during transport or assembly, that integrated circuit (IC) becomes unusable for automated assembly. Having all the pins on the side of the package also leads to bigger packages in general.The length of the pins is also why you do not see Nordic using any QFP package variants. The length impacts the inductance. Moreover, the capacitance between the pins is heavily impacted as the pins are parallel to each other. This increases the noise levels and coupling between the pins, particularly for devices using GHz radio.<h2>QFN</h2>QFN packages, as the name suggests, do not have extended leads. This reduces the coupling and noise between the pins. The contacts to the PCB are on the bottom of the package. Most packages also have an exposed thermal pad on the bottom, attached to the IC. It is usually grounded, which reduces the noise level further.QFN packages are a lot smaller than QFP-variants with the same number of pins. Some benefits are clear, like the smaller footprint and lower weight. No external leads and shorter bond wires mean reduced lead inductance. Besides, less metal is used since most bond wires use gold or copper.In short: Smaller package and better thermal and electrical performance, but visual inspection gets harder.<div class="fr-img-space-wrap"><img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1635922759572-QFN48+6+x+6+mm+.png" style="width: 765px;" class="fr-fic fr-dib">Figure 1 QFN48 6 x 6 mm package&nbsp;</div><h2 id="isPasted">WLSCP</h2>Array-type packages is the way to go for devices requiring high pin density. BGA is mostly used for big processors with hundreds of pins. For that reason, I will not write too much about BGA itself but focus on WLCSP here.When you read Chip-scale, it usually means the package is not more than 20% larger than the die itself. I mentioned before that QFN was a small package compared to other options. Let me put that into perspective. For our nRF52832, the QFN48 is 6x6 mm, and the WLCSP is 3.0x3.2 mm. So, the WLCSP is less than 30% of the area of the QFN. The pins are on the bottom of the chip and not only on the sides, which means we can drastically reduce the size and don’t have to use bond wires.For WLSCP, the die will be painted with UV protective paint, or a thin plastic layer is added on the top of the chip. To make the pins accessible, we add wiring to the bottom and place the lead balls like in a BGA.WLCSP is mostly chosen because of cost or size. It’s generally the cheapest option for the manufacturer, although varying with die size, pin count, etc. A 73-pin device will be more expensive than the same die with only 48 pins.WLCSP denseness usually requires more aggressive PCB designs. For a smaller design, you will also need via in pad or buried vias. This is great for reducing size but drives assembly costs up. However, since these are usually needed for size-constrained devices, this is not an added cost of the WLCSP.The main drawbacks are less resilience to mechanical stress and a more complex assembly process with additional layers on the PCB.<div class="fr-img-space-wrap"><img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1635922780959-WLCSP+3.544+x+3.607+mm+.png" style="width: 765px;" class="fr-fic fr-dib">Figure 2 WLCSP 3.544 x 3.607 mm package&nbsp;</div><h2 id="isPasted">QFN or aQFN</h2>Advanced Quad Flat No-lead (aQFN) is an improvement on the QFN for higher pin counts. It has multiple rows of pins on the bottom of the package as well as the exposed thermal pad. A 73 pin package would then be 9x9 mm or 10x10 mm as a QFN72 and about half as an aQFN73 (7x7 mm).With this package, we have even shorter bond wires than with the QFN and, therefore, less inductance. Again, a step up in terms of thermal and electrical performance.The main drawback with an aQFN package compared with a QFN package is that it requires more advanced PCBs. We typically see the need to go to 4 layers as well as more aggressive via and track width rules to be able to use all the pins, especially the pins on the inner ring.<div class="fr-img-space-wrap"><img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1635922801882-aQFN%E2%84%A273+7+x+7+mm+package.png" style="width: 765px;" class="fr-fic fr-dib">Figure 3 aQFN™73 7 x 7 mm package&nbsp;</div>To give you an idea of the different sizes, here is a comparison of a QFN, an aQFN, and a WLCSP with similar pin count.&nbsp;<div class="fr-img-space-wrap"><img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1635922826228-QFN%2C+aQFN+WLCSP+size+comparison.png" style="width: 765px;" class="fr-fic fr-dib">Figure 4 QFN, aQFN WLCSP size comparison&nbsp;</div><h2 id="isPasted">What does the PCB look like?</h2>To finish off the package overview, let’s look at some PCB designs for different packages. I have selected one of our nRF52833 reference layouts for each package option. We have a QFN, an aQFN, and a WLCSP package for the nRF52833.If you want to look at them for yourself, we have reference layouts for all our products <a href="https://www.nordicsemi.com/Products/Low-power-short-range-wireless/nRF52833/Downloads#infotabs" rel="noopener" target="_blank">in the downloads part of the product page</a>. This is where you’ll find the layouts I show here.<div class="fr-img-space-wrap"><img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1635922847663-QFN+pcb.png" style="width: 766px;" class="fr-fic fr-dib"></div>This picture shows that we can break out the device and route all signals on one layer when using a QFN. You then use vias to ground the thermal pad. Those vias help reduce the inductance and improve thermal coupling as well as improve the mechanical attachment of the SoC.<div class="fr-img-space-wrap"><img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1635922870010-aQFN+pcb.png" style="width: 766px;" class="fr-fic fr-dib"></div>If you are using an aQFN, you have the pins in two rows underneath the SoC. You will need at least two separated layers to route all signals. However, the area we use on the PCB is smaller, and we can place additional components closer to the Soc because we do the routing underneath.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1635922938371-WLCSP+pcb.png" style="width: 766px;" class="fr-fic fr-dib">If we now look at one of the reference designs for the WLCSP, we can see that we need another additional layer to access all the device’s pins. We also use even less area on the PCB for this package.<h2>Why nRF52840 in a QFN48 package?</h2>The short answer is more options for our customers. The nRF52840 in QFN48 (nRF52840-QF) does not support USB. Also, VDD and VDDH are connected inside the QFN48 package. Therefore, the device is only usable in normal voltage supply mode (1.7-3.6V). The customers using our nRF52832 could see this as an SoC in the same package with double the Flash, four times the RAM, Cryptocell 310, Zigbee, and Thread.It is also a cheaper and smaller version of the nRF52840 in aQFN73 (nRF52840-QI) if you don’t need the USB, 5,5V, or the additional pins. You might also be able to reduce the layers on the PCB. This is a great option for Smart home applications using Matter or Apple's HomeKit that don't require more than 30 GPIOs. There is ample memory for the different protocols and device firmware updated.<table id="isPasted"><tbody><tr><td>Package</td><td>LQFP</td><td>QFN</td><td>aQFN</td><td>WLCSP</td></tr><tr><td>Package cost</td><td>High-&nbsp;much material needed-&nbsp;longest bond wires</td><td>Less-&nbsp;smaller package-&nbsp;shorter bond wires</td><td>Effective-&nbsp;even less material used</td><td>Less-&nbsp;with least material used</td></tr><tr><td>PCB</td><td>Large area-&nbsp;big packageFew layers-&nbsp;no additional layers to access the pins</td><td>Medium area-&nbsp;Smaller package than LQFPFew layers-&nbsp;no additional layers to access the pins</td><td>Small area-&nbsp;Even smaller package-&nbsp;Components can be closer to the PCBAdditional layers-&nbsp;Needs additional layers to access pins</td><td>Minimal area-&nbsp;Chip scale package-&nbsp;More space for other componentsMore additional layers-&nbsp;Needs more additional layers to access pins</td></tr><tr><td>Design</td><td>Simple design</td><td>Simple design</td><td>More complex design</td><td>Most complex design</td></tr><tr><td>Assembly</td><td>Easy assemblybut leads can be bent during transport or assembly</td><td>Easy assembly</td><td>Complex assembly</td><td>Complex assembly</td></tr></tbody></table><hr>This article was first published on Nordic's&nbsp;<a href="https://blog.nordicsemi.com/getconnected" target="_blank" rel="nofollow" id="isPasted"></a><a href="https://blog.nordicsemi.com/getconnected?utm_campaign=2021%20wevolver" rel="nofollow" target="_blank">Get Connected Blog</a>.

For most of our nRF52 Series, we have a chip-scale package and a Quad-Flat No-leads (QFN) option. With our nRF52840, we introduced the aQFN package to our product portfolio.

QFN and aQFN options for the nRF52840

Taming the AI Demon - Arnaud Hubaux (ASML), Senior Technical Program Manager AINot AI but domain knowledge is what sets you apart from the restWhy should you invest in AI now? Both the expected ROI and risks are still highly insecure. Nevertheless, ASML invests heavily in applying AI. The company is not only one of world’s most successful high-tech companies and world-leader in making lithography machines but also a true leader in AI. Arnaud Hubaux (ASML), senior technical program manager AI, shares his lessons learned. His key take-away during a lecture at the AI Leadership Forum: AI on its own is not a Unique Selling Point but a commodity. Defining your use cases and applying AI in a smart way is what sets you apart from the rest. Your domain knowledge is what distinguishes you from your competitors.“Much is still unknown when it comes to AI, “says Arnaud Hubaux during a session organized by AI Innovation Center. “A lot of leaders in companies around me wonder if they should invest in AI. Reports promise the moon, but a lot is still unknown and the expected Return On Investment is still insecure. So why should you get on this train now?”Keep the end-customer in mind “But how do you decide? And how do you get an overview of the risks? I always keep my end customers in my mind. They all have dreams of their own. They want to explore space or make nanobots that cure cancer, or they want to solve mobility issues. They have one thing in common, they all need chips that are produced by using our machines. The linking pin is the silicon wafer. Our machines turn the chip designs into silicon chips that are produced on the silicon wafers.”Making chips faster and cheaper “Chip manufacturers want to produce as many high-quality chips as is possible per day. Manufacturers of devices want to produce as many high-quality devices as possible per day. It is about good chips and affordable chips, affordable devices. That is where Moore’s law kicks in. It is not just about having continuously more power but also about making it cheaper.”Demand for better litho performance “This requires new chip designs, which requires chip manufacturers to keep on producing new types of machines. This is how the industry works. Innovation is key. The demand for new machines generates R&amp;D budgets. ASML uses these budgets to continuously improve litho performance, our core business.”Using AI to innovate “This is where ASML’s thinking about AI starts. You look at what you can use AI for. This boils down to three things; 1. reducing costs, cost of machines, the cost of ownership for customers 2. reducing time to market and 3. improving customer value. Reducing costs for instance can be done by improving your chip defect rate detection by using deep learning technology or by detecting defective coating, by segmenting and classifying images to zoom in on damaged areas, qualification of parts goes faster.”Risks of applying AI “So AI can be useful and brings value, but AI also brings along certain risks. Let us call this the AI demon. These are a few examples:1. AI makes you build solutions that are unexplainable, this is due to models. Models learn from the input you give them and try to predict some form of output. A model tries to figure out some form of correlation but as we all know correlation is not causation. So, then you know something does not function well, but you still do not know why. If you do not understand and cannot explain why things go wrong that is not satisfactory to your customer, your team members cannot fix it and you neither improve the machine designs nor innovate. &nbsp;2. Context dependency is a second issue. Once you have discovered a defect you tweak everything in house in a lab situation, accordingly you achieve the accuracy level that you were aiming for but the model is sensitive to small variations. So, once you deploy it, you notice that it does not work.3. Machine Learning and AI are a commodity. You have to be aware and accept the fact that models and algorithms can be copied easily. Six months after you have built a model, your customers and competitors can do the same. AI is not what gives you a competitive advantage.”Delivering sustainable innovation by using AI &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; “Then why should you invest in AI? By countering the risks, so by ensuring your output is explainable, context-specific and by accepting that AI is a commodity you can make AI work for you. There are four basic steps your system can go through. They are displayed in the image below. The data comes from the system, from there you can calculate features (they are in fact physical models that give more meaning to these raw data). Next you put this input in a machine learning model, the output of that model can then give you several indications why the system is not ok or present the root cause with a certain likelihood. When it is a recurring issue, you want to schedule some maintenance on the system.” <img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1622451400990-asml.png" style="width: 748px;" class="fr-fic fr-dib">Invest in the blue boxes “The blue boxes are your domain knowledge, the unique knowledge that you are bringing to market. These are the subjects you should invest your time and money in: the physical features and the symptom analyses. This means that when you have data that derives from physics or from historic equipment corrections for instance you should use them instead of trying to predict based on data by using an AI-model.”Build in autonomous retraining and include customer in feedback loop “When it comes to the context specific element, beware of the fact that variations occur, especially when you are in the manufacturing field, that is completely normal behavior. That is why besides your real-time prediction, you should build in an autonomous retraining system, to make sure that the AI model monitors itself to stay accurate over time. Finally, include the end customer in your feedback loop. The customer can tell you if the result is acceptable. You must include that in your model.”Define the use-case in which AI can make a difference “Finally in most cases it is not necessary to build AI models yourself. As I’ve said before it is a commodity. By using existing models, you can go a very long way. What makes the difference is the quality of your data and the quality of your physical models. If you know your industry, then focus on those use-cases in which AI can make a difference. Make sure that is explainable and context specific. Your unique differentiator is what makes you successful and that is not Artificial Intelligence in itself. AI helps ASML to optimize for a very specific context, by adding a layer on top of what we do we can really squeeze even more performance of our system.”

Arnaud Hubaux, Senior Technical Program Manager at ASML

AI on its own is not a Unique Selling Point but a commodity. Defining your use cases and applying AI in a smart way is what sets you apart from the rest. Your domain knowledge is what distinguishes you from your competitors.

microchip

ORGANIZATIONS.

SHAPING THE INDUSTRY.

Particle

efabless

Latest Posts

Addressing the Microchip Shortage

QFN and aQFN options for the nRF52840

Not AI but domain knowledge is what sets you apart from the rest