Design and Manufacturing Perspectives from DISH Technology’s Les Beller
I recently interviewed Les Beller, a long-time PCB designer who is now a manufacturing engineer for DISH Technology. We discussed his company’s business shift towards 5G and streaming, and the stresses that puts on a design team. He also explains the greatest challenges that he’s facing with HDI and higher frequencies, and the added importance for designers to understand the manufacturing process and DFM tools.
Barry Matties: For our readers, give us a little bit of your background.
Beller: I'm a manufacturing engineer for DISH Technology. We used to be known as EchoStar Technology, until we were absorbed back in with DISH as DISH Technologies. Prior, I was a quality engineer, and before that I was a PCB designer for about 30 years, back to tape and Mylar. So I've done the nose of the dog, the tail of the dog, and now I'm in manufacturing engineering, which is sometimes on both ends of the dog!
Matties: And all the parts in between. Let's talk about design for a minute. I would think that's a foundation for everything that you're doing. When you look at designs today, what's your impression?
Beller: The first thing I look at with designs today is how they're utilizing the board space, taking into consideration whether it's a consumer product or a high-mix/low-volume product. I look at some of these traits from a designer's view as well as an assembly view, and how they’re interconnecting/packaging the PCBs. Are they designed for a resilient life? Are they going to be long-life products out in the field? Did the designers create a board that is going to be easily fabricated and inexpensive to build? I perform tear-downs occasionally and am amazed how well, and how poor, a product can be designed with respect to serviceability.
Matties: You mentioned the HDI; is that a new area for DISH?
Beller: Absolutely. We've traditionally been involved with designing set-top satellite TV products with as few PCB layers as possible, down to the cost per square-inch pricing. We have only had two or three HDI designs in the past. We now are performing a refocus based on the drop-in pay-TV services—people are doing more streaming. We've had to shift our service from purely satellite-based product to an OTT type streaming product, and have also had to learn to contract out to other OEMs that need this type of product. One of our newer niche products is in the multi-camera streaming and production market, SlingStudio, which is not at all related to pay-TV. What’s new in our future is going to be 5G IoT type products. Based on DISH's award of the frequency bands we obtained from the FCC, we're going to be setting up our 5G type network and designing everything under the sun that will be narrow-band IoT for the customers. Some of this IoT is already determined and in process, but the layers are just being peeled back in this new market. Everyone is rushing into it.
Matties: Is the push for HDI just purely a space functionality issue?
Beller: Primarily, yes. We'll still have different technologies of boards. HDI will be necessary depending on the chip architectures and device packaging. If we get down to very low-pitch devices, we're going to have to use HDI. This is based on assumptions that a good portion of the products are going to be more of a compact type of product. We will still steer our products to the technology they require to save cost, whenever possible. It’s our mantra!
Matties: There's a lot of conversation about HDI lowering your costs. Or does it actually cost more?
Beller: I would say the engineers tend to think that immediately it's going to lower the costs because it can be much smaller, but that is rarely the case. This is where a PCB designer needs to be intelligent, to push back and validate if the industrial design (ID) requires the end product to be that small. Let’s use a bit more space for the vias; let's not force the fabricator to put more labor into the board or force them into a low-yield process just to make my board easier to design. How about a look at panel usage, alternate materials or stack-ups? These habits will always save you cost and open up negotiation windows with the suppliers. I believe what primarily drives HDI for us is going to be product size (like a mobile phone or portable device) and then functionality (processing speed) that will drive the HDI requirement and cost. Then of course, there's the middle market, in which I think size doesn’t matter quite so much for a product. Some companies will tend to overdesign to get to market fast, then wont take the re-design path for cost-savings like we would all hope. This is a critical decision for the management: time to market or cost? There are many designers, ourselves included, who are learning more about HDI as we move forward.
Matties: Designers who I'm talking to are saying, “Well, we would never use landless vias because it's not spec’ed in.”
Beller: Customers are concerned with the reliability. The fabs like it, as it solves a little bit of a registration issue for them as the tech drives smaller. Somebody told us about using landless vias in our future products with them; as long as there is good information on the service life, it might help in some cases.
Matties: Because if you can get rid of the rings, you bring in a lot more real estate, right?
Matties: I don't know that there's an appetite to pursue it. For me, I don't understand why.
Beller: You should look at the big picture when you're talking about a newer technology. It's new to some people who never would have considered it before. My first concern with a landless via on a portable is going to be resiliency. Then during the manufacturing, do I have the right laminate that's not going to create a lot of Z-axis expansion and put the interconnect with the barrel at risk? There’s no pad to have pad-rotation on! There still needs to be consideration for when it's appropriate and when it's not appropriate.
Matties: It seems like if we're really looking at functionality, you must pursue this path. If you're looking to add more functionality in the space that you already have.
Beller: Like a power application. Today's switchers and our processors are running so hot that the thermal cycle in a product is going to just wreak havoc on any of the board in that area. That's why the data is going to be real critical to that kind of technology.
Matties: In your role, what are the greatest challenges that you face dealing with your suppliers?
Beller: Boy, that's a good question. There are indeed several. I would say that my biggest challenge has been the suppliers using and maintaining statistical process control (SPC) to control our more critical signals on the boards. The suppliers we continue to use have adapted, and through a) data analysis, b) process/quality improvement and c) development of their processes through our initial development and proto cycles. They have come to realize that using SPC is a requirement in some areas. Many profess that they utilize the prior steps and love to point out charts on the wall during tours, but only some actually benefit from long-term data collection that can show them where to drill down, and where to save labor. Don’t get me wrong; many suppliers use SPC to track a process window, but they do not understand completely how it can improve yields and allow them to accept higher spec products.
A good example would be USB lines or HDMI lines. If your impedances are off on some of those lines, it's not that noticeable to a user. As an example, many devices are designed with those variances considered (considering the cable loss as well). If a supplier is “drifting” out of process control on those types of signals, it may not be as big of an issue as, for example, WiFi or Bluetooth (RF). Now you affect how far Bluetooth or WiFi can reach or the power level over time (battery drain). As a supplier, you cannot just use one global rule for high speed, and this will be the challenge going into 5G. I have spent a good part of my technical meetings explaining why this is so important to the Fab customer. My acknowledgement is that the processes are still in effect upon a return trip. By far my greatest challenges when utilizing a Consumer Product supplier base.
Matties: For designers, having to be responsible for signal integrity is fairly new to them, right?
Beller: Yes and no. It is just more important than ever before. Integration has resolved the volume of the issues, but now there are more sub-systems integrated within one board, making for more critical items to track. In years past, you might have one or two boards to focus on at one time, before moving to the next critical pair of layouts in the rack. The engineer took on all of the SI functions during the layout and after. Now, multiple subsystems are included on one mainboard so, in effect, we actually complicated the overall integrity of the system. Noise and lower signal levels have just killed what overhead I thought my system had. Now the designer experience just climbed two notches at least. At least the tools have improved parallel of the design complexity, so we can track it a bit easier.
One challenge we have from time to time is that we may have three or four different 100-ohm pairs on the board. All need to be differential 100 ohms, but the different circuit engineers will insist on not standardizing their structures with other engineers in the interest of time. The average engineer, if there is such a beast, may think this is a “don’t-care” issue, but we have just tripled the amount of IPQC (in-process quality control) impedance inspections for the supplier—don’t forget all of the impedance modeling and tuning error opportunities. First thing they will ditch if they get a chance are these extra checks! This is what I believe is the responsibility of the designer today: realizing opportunities to improve the manufacturability and therefor the quality and yield of the product, and just simply raising your hand.
Matties: I keep hearing that there's not a lot of communication between designers and fabricators. How important is it for designers to really understand the manufacturing process?
Beller: Let me state it this way. I would always recommend that a PCB designer spend some time in a board shop on the front end and some time on the floor. It's important that the person who’s purchasing the boards and the people that checking the quality of the boards know the capability of the supplier, but also know the struggles that the suppliers have meeting the requirements (i.e., using more standardized materials, using the laminate capability data back into modeling, and knowing the difference between first, second and third-order challenges). I’ve personally learned that it's amazing to go in and see the challenges the board shop has trying to meet your specs. Understanding that is always going to make somebody's quality-of-job much better. It’s equally as critical with the engineering side as it is with the supplier side, back to the person who is purchasing the board!
Matties: When somebody is purchasing a board or hiring a designer, what sort of questions should they be asking?
Beller: If I was hiring designers right now, that would be one of my first things I'd be picking at them about. Explain to me some of these terms. What do you know about this component? What do you know about that laminate? What kind of problems am I going to have if I don't specify a low-loss material when I'm trying to control my traces in a factory and what are the variabilities? What's going to happen if the supplier goes with high resin and changes the laminate or changes the prepreg on me in the middle of the game to save cost? What are the challenges if I specify too many impedances on too many different layers? I would really want my designer to have that experience because the way our industry is going, we’re going to have such higher frequencies. Back in the day, a stubbing via was no problem. Now people are back-drilling to reduce the number of stubs. Five gigahertz? Not a problem, but in the next couple years, people are commonly going to be getting up to 25 gigahertz and even higher for some applications. That requires designers at the top of their game.
As far as purchasing a board: Audit the factory and utilize surprise audits when convenient. Use references or some experience basis as well as cost to select the supplier. Get references from current customers and ask about the quality. Be careful about just buying boards from whom your contract manufacturer selects. The CM is not closely tied to the performance of the supplier as much as cost benefits. Make sure your CM is re-auditing suppliers at some reasonable frequency. Stay involved with quality reports and don’t be afraid to throw in some of you own requirements for monitoring ongoing PCB quality data. Keep two suppliers if you can, so they can bid against each other.
Matties: One of the things that we see in North America is an aging design community. My understanding is that electrical engineers will be doing circuit design as well. It's not just going to be a guy who buys a tool and starts designing, for the very reasons you've just described.
Beller: I think that's the aging population we talked about this a few years ago too. We’re part of the aging population now. It's going to be an issue if the new younger designers coming in don't have good direction, or don’t accept the training by the experienced designers. The high-speed experts, the impedance guys, the RF gals, the materials and process people as well as many other tremendously talented individuals who have a lot of experience...we need to find a better way to get those experiences down to the younger designers. Through shows, online training, or in the classroom? These days, there are some specific training rules that work with the younger professionals. We need to focus there as an industry.
Matties: If we go with that tribal knowledge, are we just passing on the old when what we really need to be thinking about is the new?
Beller: I think you need to know how to grow a rose before you can start plucking the petals to make it look beautiful, right? I think that there needs to be an availability of information. I don't know if “joint venture” is the right word between old and new, but I think it's a merging that will do both. To support your concern, there does tend to still be the mature part of the workforce, and I've caught myself saying this as well. It's like, “Well, we've never done it that way before.” Well, that's not always the right answer; in fact, rarely is it the right answer. I believe it's a mixture of old and new because there is a lot of information that we've learned over the last 20 years that's critical. How do you get that information downloaded to a millennial generation of quick-to-move, quick-to-learn, download-me-something, e-taught designers these days? I really think that products like the designer's guides that are coming out, short reads, are very good on-the-move training info and just good information. I personally print out or forward articles to my phone for when I have free time or for when I feel my attention is better to read and learn.
Matties: What about the tools themselves? Are they evolving to the point where you think that there's lessons learned just from the tools? For example, all the automation features coming in and design rule checks.
Beller: I don't have a lot of access to a lot of different design tools these days. We use Mentor and share databases when needed. Now that takes patience! Probably the more management-pleasing feature we use is multiple designers within one database. We still find a high value to that, cordoning off the board and allowing the designers to work in their respective areas. That gets our designs out into the market faster. However, it seems like these days we still don't have a good internal inclusive DFM type of tool without buying the third-party solution or the extremely expensive solution. Valor is a good tool for a high-volume design department, but give me a tool where you don't have to pay an excessive amount of money for DFM, and that's going to be the tool for me. In our organization, we decided not to purchase Valor because having a person trained just to run Valor and the maintenance costs of Valor are high; it’s a high cost of entry for low design volume. We still do manual checks using checklists. We have experienced people who go over the designs during the design cycle and when they’re completed, but occasionally we can still miss something. It’s manual, but it mostly works.
Matties: Is that decision based on the volume of work that you're putting through, or on some other factor?
Beller: It's based on two factors: not spending money where we feel like we don't need it, and the fact that we don't have an excessive quantity of boards that we're designing like a service bureau would. A service bureau or a very large company that's doing 20-30 designs a month can benefit from the Valor tool and could justify the cost.
Matties: You have to weigh your specific need there. But at the end, you must do some sort of DFM.
Beller: Oh, absolutely. We've learned through the manufacturing world of hard knocks with a lot of different suppliers that the CM’s DFM rule list is not always perfect either. We've given them our designs, have them run it through their Valor checks, and we've gone through and looked at it. We found that a lot of them are canned responses, and through building the boards after that, we found many more issues that should've been caught that were not. Your DFM checker is only as good as the person who inputs the variables that it looks for.
Matties: The other topic that comes up is best practices. As Happy Holden says, “There should be a red light, a green light, and an orange light right on the tool.” Best practice, green. You're doing great. And the red is “What the hell are you doing?” But there aren’t a lot of tools like that.
Beller: Yes, like an online check for everything during the process.
Matties: I guess best practice is still subjective.
Beller: Yes, it is. It does depend on your market target as well. If you're doing a lot of low-speed consumer-grade type products, high-volume low-mix type products, you can shove them out the door and just put another person on the end of the production line to deal with your design issues. If the design issue is significant however, this can add labor cost. If you're doing high-spec work, then you need to hit the target right off as rework may not be an option. Right now, with most design schedules, you can't afford six or eight revisions of your PCB to get to market. You need to have it right in two versions. We typically had 6-8 versions of boards getting to MP volumes, but now have this at 2-3 versions.
Matties: That's the other case that Happy makes is we need predictive engineering. We need to be able to predict before we even design. Even with the DFM check, it's a spin and re-spin process in some cases.
Beller: For example, one of our international CMs has a design guideline, and it's a matrix—resistor to resistor, resistor to capacitor, IC to resistor—basic component spacing requirements. They'll have the minimum, the nominal, and the preferred. Well, of course their preferred is that they want a tenth of an inch from every component to every other component for “repair.” That's not reasonable except around some larger parts that might need rework (CPU/Flash, etc.), and it’s not even realistic. Some of those numbers were practical 20 years ago, but not now. We push the bottom end and every time, the CM that runs Valor gives us back a DFM listing that's a mile long. We try to get them to adjust, but it’s hard to do that with a canned/mature rules file.
Matties: What about the thermal issue in what you guys are doing?
Beller: There are two predominant issues in most of our designs. One is thermal and one is signal integrity. Thermal has caused delays in the middle of some compact products and requires that we think even more outside the box than we thought we had. We've had to completely change some very compact products and how they spread the heat away from the processors; some people call them heat spreaders or heat sinks, but most are custom fabrications. As the processors get smaller and more complex, they run hotter. As we ask switchers to do more, they're running hotter. In a small, compact product where you don't have a lot of air movement through the product, you're just going to keep building up heat until the product just gets too hot. Like a lot of cellphones now. Use them for 10 minutes and you feel it. Thermal is one of our big issues within our products.
Matties: How do you deal with that? Are you using predictive engineering?
Beller: We've gotten much better over the last five years on the predictive side. We've installed thermal modeling software. We plug in the envelope sizes and the temperature ranges of all the chips in the product. They've got all the thermal characteristics of the laminates, etc. They plug all that in and we try to see where all the heat problems are going to be. We'll use that to change the case ventilation and add a fan (worst-case scenario) or go with a larger heat spreader to try to dissipate that heat out.
Matties: Along those lines, what about materials? There is a lot of talk about the copper foil market where a lot of the foil was being consumed by the battery market. Standards are lower, so it’s easier for the suppliers to deal with that.
Beller: Let's talk about where we're going with copper. That's an easy one. That's a low-hanging piece of fruit. We have had significant price increases across the board over the last year, month-over-month in some cases, where we've heard the prices are going up again! I'm concerned with the longer term, not just this year or next year. I've understood that there are some changes in the mining industry that are going to cause a shortage of copper, if what I read was correct. It's based on how they're exhausting some mines and bringing other mines online. There could be a lag in the supply chain. That's only going to become more of a crazy issue, especially with China moving more to electric cars. They're already talking about it in California now with new electric car requirements. The procurement people need to stay on top of this as batteries become larger and more common. We're going to hurt ourselves in other areas if we can't come up with a different battery technology. So far, we haven't had any phone calls saying, “We can't get the laminate,” or that the laminate is shipping late. It's just literally on every quarterly business review, the prices will go up maybe 5-7%, quarter over quarter. Right now, it's just a cost issue. Which for us, if you've got a set product price, it will be affected. Luckily, some suppliers we use have better agreements with their laminate fab than others, so we may not see quite the increase.
Matties: Bottom-line dollars.
Beller: We can't keep increasing our product price to our customers.
Matties: It's interesting to think about the satellite market switching, because I don't know how many people are putting satellites or dishes on their houses anymore.
Beller: Yep, that's true. We've seen a trend for our business somewhat declining because the availability of content online these days is just amazing. I myself stream a lot of content, and I've got a satellite system. I utilize both. We know that it's on the decline, so therefore we've been putting a lot of our technology into Sling TV. We've got one of the top three streaming products out there. We've done well with it and now we're going to move into 5G using the FCC-awarded frequencies. We have a plan together to work on infrastructure using some partnerships with some very well-known companies to help get us to a network that will support our IOT products that we're designing. We have a couple already in the process. People want to get their content faster, easier and cheaper over the internet. We're going to move forward with that technology.
Matties: No doubt you will. Just some final thoughts. What advice would you give a young designer?
Beller: I would try to position myself with larger corporations or design bureaus where you can have a lot of different design experiences. That was a benefit in my past. Working with a lot of different companies, I had to become a chameleon and learn something new every day, and work with many different engineers and with varying technologies. Position yourself in a company that has a wide variety of power products, high-speed products, etc. Design bureaus are always great for that. If you don't have that luxury and you tend to be experiencing one type of product design that you're working with within a captive design environment, spend the time to get out and get the education. Surf the Internet; there's a lot of information about layout and design. Educate yourself. Get your CID rating. Use your tools. Get familiar with other tools if you can. Get demo copies. Understand the strengths and weaknesses of different tools, because, let's face it, if you have to leave a company and you're an expert on tool A and that other company has tool B, you're not going to be as good of a bargain for them as the person that's experienced with tool B.
Matties: Is there anything that we haven't talked about that you feel like you should share with the industry?
Beller: I spend probably 25% of my time working with the supplier base and understanding their capabilities with respect to the circuit boards themselves. We've had to disqualify some of the suppliers because they won't move along with our recommendations to improve different characteristics of their factory. I would say don't just treat your supplier as though they're behind a curtain and that they're going to know how to do everything the best. Go behind the curtain and audit the suppliers using the basic info you can find on the Web. It’s a fun project to develop a PCB audit program that focuses on basic quality points and ensures the supplier is utilizing SPC. Look over their shoulder and spend time in their factories looking at what they internally audit.
If your company has a supply chain person who's not technically based, who's going out and doing all the work with the suppliers, go with him. Insist that it's required so that we can assist that supply chain person and go in as a partner should. Be very aware of what your supplier is doing relative to laminate use, modeling, and what they're doing with their critical nets. Basically, it’s very important for a designer to be involved in the whole front-end process in the fab shop these days.
Matties: Great advice. Les, thank you so much. Always great to catch up with you.
Beller: Thank you, Barry.