Big Compute - Innovation in Antenna Design

Episode Date: April 11, 2019

In this Big Compute Podcast episode, Gabriel Broner hosts Mike Hollenbeck, founder and CTO at Optisys. Optisys is a startup that is changing the antenna industry. Using HPC in the cloud and 3D printing they are able to design customized antennas which are much smaller, lighter and higher performing than traditional antennas.

Hello, I am Gabriel Bronner, and this is the Big Compute podcast. Today's topic is innovation in antenna design. Traditionally, RF antennas have been built out of multiple parts, approximating a design that would work for the customer. These antennas would be very costly and heavy. More recently, Optisys has embarked in using high performance computing in the cloud to design and 3D print antennas,

especially built for each customer. These antennas are one piece and fit on the palm of your hand. To discuss this new approach to 3D print customized antennas, our guest today is Mike Hollenbeck. Mike is founder and CTO at Optisys. Welcome, Mike, to the Big Compute Podcast. Thank you for having me.

It's a pleasure to have you here and talk about what you're doing in terms of this new style of designing antennas. Mike, let me start by the beginning. How have antennas been built in the past? Yeah, so antennas, as with any piece of physical hardware, have historically been built using the manufacturing tools that were available to engineers at the time. And in particular, the specific kind of antennas that we work with are the ones that you make from metal and air. So they're basically like hollow metal tubes with intricate internal features. And these have a number of advantages in being lower loss than other methods and being able to

support higher power levels and getting a really complex RF function without paying a big price in losing energy in the process. So the traditional tools available to make these structures were subtractive manufacturing processes, something like a CNC manufacturing process or similar, or some of the even early additive manufacturing type processes like electroforming, casting, extruding. But there are these manufacturing methods that all require subtractive manufacturing of either the block of metal itself or the mandrel or the mold that will make the actual final part. And so a subtractive manufacturing process is one where effectively you start from a giant block of material and you slowly remove material away until you have the structure you

want. And that means that when you're an engineer designing to that process, you need to be able to design a geometry where you can access every single feature with the machining tool you're using to cut material away. And for an antenna structure, what really matters to the structure is what's on the inside of the part. We like enclosed volumes that can trap an electromagnetic wave as it propagates through the structure and performs functions on the wave through those individual features inside the part, and then assembling them together, whether you're bolting it together or brazing

or using a number of other additional fabrication techniques, you're still trying to effectively make dozens or hundreds of pieces to achieve a specific antenna function, to achieve a specific antenna performance. So if a customer showed up in the past, not with your technologies, and asked for an antenna, they'll ask for an antenna for a particular frequency, then you go build pieces or you go find pieces that exist and build an antenna out of those multiple parts? Yeah, so both approaches have been taken by companies and it really depends on the specific application

and how stringent the requirements are. I come from the world of aerospace design for antennas where the requirements are very stringent. And so some parts you can get off the shelf, but many times you're designing a custom part. And especially with antennas, you'll get a specific frequency allocation to work within because frequency spectrum is a limited resource. And you can't just make one antenna to do everything because it's like going into a loud restaurant with music playing.

You can't really have a conversation because everyone's talking at the same frequencies. So with antennas, you want to separate that out to multiple different frequency bands. And so that requires a lot of custom design. Okay. So in the old days, I needed an antenna for my particular use case. I'd tell you what the frequency is. You go design several custom parts, and then you build an antenna out of multiple parts. Would that be a slow process? Would it be costly? What's your view of that process? That's exactly it. Yeah. So you have to actually go and break the problem into individual chunks that you can design as individual parts. And so each one of those is a design step and each one of those is a fabrication step. And then there's assembly that follows that. There's tuning at each

step because these are very sensitive components. And you also have tolerance stackups and individual cost stackups, and it's expensive to assemble things. So each of those adds to the cost, the weight, the size, the complexity. And so as you get a more complex antenna system, it makes the total antenna assembly much more expensive. Yeah, and when you say the way you build this out of multiple parts, how many parts are we talking about? What are we talking about typically? Yeah. So for a simple antenna structure, I mean, it can be as simple as a few parts, but that's not the stuff that we're primarily

interested in, right? Well, for a lot of complex systems, the complex systems, they can be hundreds of parts, especially when you include hardware and fasteners and O-rings and all the other little ancillary pieces you need to close up all the joints between the parts. In the old days, you'd have to build an antenna out of multiple parts, may have been a hundred parts, and you have to put them together and build the antenna. That takes time. That was expensive. expensive and the typical example you've mentioned aerospace who'd be the users of this type of antennas just to so we all follow the the conversation more clearly yeah so so uh in terms of uh use cases and users so so it would be things like uh you know government uh defense airplanes Governments, defense, airplanes, ships, remote terminals that have to be carried by, say, a soldier to some remote location. And they have to be high performance but lightweight.

Things like that, right? Where it's a lot of use cases. In fact, I think antennas is one of those things where most people don't realize how ubiquitous they are around us. Everyone has their smartphone. And that has a fairly simple antenna on it. people don't realize how ubiquitous they are around us. Everyone has their smartphone, and that has a fairly simple antenna on it. But on the other side of that is the tower that has a much more complex, expensive antenna. There's satellite antennas all around to pump the data down that's being generated in space by all these cameras and other sensors up there. Antennas are really

something that keep the world connected and are just all around us. Yeah, that's good. We may see, I'm seeing on the rooftops that I'm looking at in San Francisco, lots of antennas. That may be just an example, right? Exactly. Yeah. So, you know, that was the way antennas have been designed in the past. And you guys have taken a different approach. What has changed in the world to go to a different approach that has enabled Optisys to do things differently today? Yeah, so it's really been the confluence of three key technologies. So there's big compute in the cloud, the simulation capabilities,

and then the advent of the fourth industrial revolution in manufacturing, which is additive manufacturing. where before you would have to actually write your own interface to a high-performance computing server and negotiate with the individual companies that have it. It was a much more complex way of actually getting your individual software to run on an HPC cluster. And now there's companies like Rescale that are out there that are just these wonderful interfaces that have just a nice turnkey solution

where everything's integrated in, it's super easy. The software, so we use ANSYS software that is a full end-to-end of the electromagnetics, mechanical and thermal simulation of these antenna structures. And so we're able to create really complex structures in the virtual environment and optimize them very accurately. And that's something that, you know, the accuracy and the complexity you can do in a model

has really just been around in the last five to 10 years to where we can start to get to larger assemblies and actually get to real world structures. And then in addition to that, metal additive manufacturing, it's truly a disruptive manufacturing method where it's been around for 30 or 40 years. And even 10 years ago, the structures that it produced were nowhere near good enough for any kind of electromagnetic structure or even most applications. Their density wasn't there, the material properties weren't there. But then in the last five years, that's all changed. And And that's really where we've succeeded is capturing all three of those and realizing when the confluence of those three would hit to allow us to actually make functional parts at interesting frequency bands in the millimeter

range. Yeah, no, that sounds very good. Just to clarify, is additive manufacturing 3D printing or is it not? Yes. So additive manufacturing is the term that I use when I go to technical conferences and present papers. And 3D printing is the term that I use when I'm at trade shows because it actually is much more understandable. Yeah. Okay.

Those two terms are interchangeable. In particular, there's a couple of different kinds. We're focused more on printing in metals and predominantly in aluminum, just because it has this excellent strength to weight ratio, excellent material properties in terms of conductivity. So it's just a spectacular material for antennas, especially that has been used for probably a hundred years now in antenna structures, but now we can print with it. That sounds very interesting. So I think you were one day realizing the world has changed and the way antennas were built in the past,

which you described early on, could be different because of all the changes that have happened in terms of 3D printing, in terms of how performance can be in the cloud, in terms of the applications, et cetera. So how did this happen? What happened at that moment? Can you describe, okay, here is a possibility.

We've got to go and start a company to do that. Was that the thinking? Yeah, it was kind of like that, yeah. So the founders, we all had been tracking additive manufacturing just because it was really cool. And the way that most people were approaching additive manufacturing was saying, I have designed this thing. How can I take additive manufacturing to produce what I've already designed? And we realized pretty early on that that's just the wrong question to ask. And the question should be, how do I learn this process, figure out what it's good at, and then design structures that have never been created before that can create the RF

performance we need that take advantage of the strengths of the process. And that was something that no one else was doing. And so we realized that it was time to jump and to do it. And so that's where Optusys came from. That sounds great. And you were a startup, a relatively small company coming up with this idea at the time. Is that fair? Yeah, that's fair. So we were a little different than some of the more traditional Hollywood view of a startup where it's one heroic founder that is the technical expert and the business expert and saves the day on everything. We were actually a team of four deeply technical people that come from expertise in additive

manufacturing, expertise in mechanical engineering, expertise in RF design, expertise in systems design. So we had a view of the big picture, and we'd all been in the expertise industry for quite a while. And so we came together with a lot of experience, and together we're able to tackle this problem. Because it's really, it's too big for any one individual person to solve all aspects of it. Yeah.

And I imagine at that moment, you came up with the idea. Did you face any particular challenges you remember at the time? Yeah. And I imagine at that moment, you came up with the idea. Did you face any particular challenges you remember at the time? Yeah. So the biggest challenge really was convincing people that the little antenna structures that we were making worked and were actually made of metal. So we were coming in with these designs and we jumped straight to the most complex thing we could possibly think to make to prove that it could be done. And it was something that took a really complex antenna and reduce it from over 100 parts down to a single part. It took what would typically take 10 to 15 inches in length of a part and reduce it down to less than two inches. And the weight associated with that, if you have all these, you know, over a hundred parts that have to be, you know, machined and put together, you know, would have been traditionally 10 to 15 pounds easily.

And we were less than two ounces. And so dropping that into someone's hand and telling them, here are all the components inside and here's the performance it achieves. You know, it really just had this level of disbelief that we had to overcome. And we did that by actually providing measured data at the same time. So we went out and took measurements. This wasn't just PowerPoint design. It wasn't just, you know, pretty little parts. We wanted to prove that, no, this really can be done. And so we actually, we had a number of papers that we went out and published to show, you know, here's the performance of the process. Yes, it follows the laws of physics. Here's some of the intricate things you can do. You know, it was this education process

that we had to go through for the first probably year and a half of existence. So you went to a world that was used to having antennas that were 10 to 15 pounds in weight and made out of 100 pieces. And you showed up with an antenna that fits in the palm of your hand, that weighs a few ounces. Maybe that was a bit radical for people to accept it immediately until you proved it. And that's kind of what you face on the early stages. It was. Yeah. So we actually, we had the, actually quite the range of experiences. So some people obviously had, you know, sheer disbelief that it could be done. Others, when they saw it,

we actually had a couple people that, you know, had the, you know, just total shock because they did believe, because they knew our reputation, where we came from. And it was just completely shocking to them. And, you know, they're really, I almost, you know, look at some of our early pieces as works of art, even. They're just, it's amazing what can be done when you really push the limits. That's great. So in the early days,

people showed up and say, I need an antenna. And you go build a bunch of parts and you build a system, et cetera. What happens today at Opti-C? Somebody shows up and says, I need an antenna with these characteristics. What do you guys do after that? Yeah. So, so we actually also had to change the way that a lot of these companies look at engaging with antenna companies, because they're used to saying, I need this part. I need the specific antenna part of the simplified component. What can you do? And we, we, you? And we were like, no, no, no, that's not the best way to use us. We want to talk to you at a system level. What do you want your antenna to do at the system level? And we'll tell you which parts make sense for us to combine into a single printed part and what makes sense

to break off and use traditional processes. So it's really, we like to interact at the specification level for the top level assembly and be more almost a partner as opposed to just a vendor that you go buy apart from the catalog. Yeah. So when people say, I need this antenna and then what do you do? You start running simulations. Can you tell us about what goes next? What happens next? Yeah. So, you know, from the start of the

engagement, typically it's, it's, they, they say I have this, you know, top level antenna performance that I need. And so we'll go through and we'll do some basic, you know, quick calculations that, that are just your basic physics of what any antenna has to do to meet those performance characteristics and you know once we come up with an approach and they agree to that approach that's when we jump into simulation because simulation is still you know fairly intensive and it's typically once we've agreed on you know this is the approach we're gonna take that's when we jump into it yeah and you simulating actually this customer's antenna, right, with the specific requirements the customer has, right?

Well, actually, so the way that we approach it, we're bringing to market what we refer to as mass customization of antennas, where if you look at an antenna as a black box system, you can write a pretty generic set of specifications for it. And those specifications can be achieved by a set of Lego building blocks that we have developed internally that we can mix and match in simulation

and then hit print and provide a custom solution. So those building blocks are already there, and they're scalable and functional. And so it's not that we're printing someone's specific design they give to us. It's that they're saying, I have this black box specification, and then we can create an antenna very rapidly in simulation

and then print it that meets those specifications. Yeah. Great. So you basically, using the building blocks, you simulate the antenna that you're going to provide this particular customer and then you're able to print a customized antenna. That's right? Exactly.

Exactly. This is what you call in your article in Signal this month, mass customization instead of mass production. Is that right? Exactly. And that's a big thing of the 21st century. So in the 20th century, there was this big move towards mass production where you could reduce the cost of something by making a million of them. With additive manufacturing,

we have mass customization where we can significantly shorten the development and fabrication cycles to the point of getting them to almost commercial levels. In fact, better in some cases, the development and fabrication cycles to the point of getting them to almost commercial levels, in fact, better in some cases, and even approach the costs of commercial mass production for custom components. Yeah, that's very good. A very good example and very good to hear. So, Mike, early on in the process, you decided to use HPC in the cloud. Can you tell us about that choice? Of course. So we have a core belief at Optisys not to in the cloud. Can you tell us about that choice? Of course.

So we have a core belief at Optisys not to reinvent the wheel. We have a special set of talent regarding additive manufacturing and antenna design. We don't necessarily have talent in IT or high-performance computing. And so it was really kind of an obvious decision to move towards using HPC as our computing resource from a very early stage. So as a simulation-intensive company,

especially in the antenna space and other traditional engineering spaces, you need to have these really expensive, really powerful software packages that can make very accurate models of physical objects and simulate the physics around them. And when you are designing something and you want to go buy hardware to actually run the simulations, you need to anticipate the largest problem type

that you will ever have to run on that hardware when you are going out to buy it. But you're only going to be using that 10% of the time of the year. And so it's really wasting money to have to have all that expensive hardware in-house just to run the simulations to be able to compete in this industry. So what HPC allows us to do is to actually dynamically scale our hardware resources to meet the needs of the day so that we can really focus internally on design and have a relatively cheap system in-house and then use Res you know, Rescale

to do any of the heavy lifting for our simulations or to free up our local machines so we can work on other things. So it really was just a spectacular solution that happened to be available right at the time we started. Yeah, no, it's good to hear. So you're using HPC in the cloud to run your simulations, you go 3D printing. What kind of results are you achieving? What kind of antennas are you producing? What kind of results are you seeing with your customers today? produce structures that are orders of magnitude smaller and lighter than any traditional process. And one of the really fascinating things is, you know, with the traditional process,

like I mentioned earlier, adding complexity adds to costs, lead time, weights, you know, all those issues with the traditional antenna system. But for us, adding complexity actually effectively reduces the cost in a sense. Because the price of the part doesn't really change much as you change the internal complexity. Because it's all done in simulation. And we're able to really just minimize the structure down to the absolute minimum volume. And so as we add more complexity, we become much cheaper and much higher performance compared to any traditional process. Very interesting. we become much cheaper and much higher performance compared to any traditional process.

Very interesting. So in the old days, you tried to minimize complexity to save on cost and weight in manufacturing. But now you can increase complexity. It's going to cost the same. It's going to take the same amount of time to build. It's going to be the same weight, but you get much better performance. Is that fair? That's fair. That's exactly it.

Interesting. Once you take a step back and look at the problem differently, you come up with different innovative solutions where you go to here. Mike, what kind of challenges do you guys have in front of yourselves right now? So the biggest one right now really is time and talent. I guess that's two challenges really. So time, there's only so many hours in the day, and this is such a wide open area of discovery for the different ways that we can use AM to improve these systems. And it takes just experts, that special talent of finding people to add on,

and we're just growing at an explosive pace, and choosing which avenues to explore next that are the best to improve our intended performance and systems. We can let you advertise your open positions here if you want. People have done that in the past. There you go. We'll send them to your website and they can look at your open positions. That seems to be a common theme.

Startups are pushing the boundaries like yourself. Finding talent is a bit of a challenge. Exactly, exactly. That's good. By the way, you're in different locations, right? I understand you're on the East Coast. Some of the people are in different places. Is that fair?

Yeah, that's fair. So we're actually headquartered in Salt Lake City, Utah. And I personally live out in New York. And so, yeah, the, the cloud is spectacular for that. You know, technology nowadays, you know, we're, we're connected on a daily basis. And the, the fact that you know, the, the rescale platform allows me to throw my simulations from anywhere, absolutely anywhere. It's spectacular. Very good.

Hey, Mike, you're ahead, right? So you're pushing the envelope using the combination of HPC in the cloud, 3D printing, mass customization. Are there any thoughts you have for people that are not necessarily in antenna design, but are thinking about similar problems in your area or other areas? Any, any thoughts you want to share about your experience in having, you know,

rethought the problem and, and, and push the boundaries a little here. Yeah. I think you appropriately stated it when you said you need to take a, when you take a step back and you look at the big picture, you can really accomplish some great things.

And that's really, you know, the thing that I would say is a lot of times people have learned how to do something and they keep doing it that way because it's worked. And a technology or a fabrication process like additive manufacturing

come along and allow you to do something you've never imagined doing before. I kind of liken it to the Peter principle where people are promoted into obsolescence. So they're promoted as long as they're excellent at the job until they hit the point where they're no longer capable of performing the job. And there's something very similar in design, where people have learned how to design something or to apply some technique, and it worked. And so they keep applying it, even when it no longer makes sense. And so there's a lot of value in asking, you know, am I taking the right approach? Is there something novel that I could be doing here? Is there some new tool at my disposal that allows me to use a different set of rules for how I engage a problem? That sounds very good. It sounds inspirational because what

you're doing applies to all of us and what we do every day. I mean, it's been really good to hear your stories about how you really took a step back, look at what you had in front of you and how things like high performance computing in the cloud, how 3D printing had evolved, how you could really rethink antennas and make antennas even lighter, more complex, more cost-effective, higher performance than before just by rethinking the problem. And I think that's been great to hear your story. Before you close, anything else you'd like to add?

Yeah, so I'm really excited for what the future holds for HPC as well and other non-traditional uses like what we're doing. So I view us, the way that we use HPC, as kind of an off-label use for HPC where traditionally you have a gigantic data set you need to process and there's no other way to do it within your lifetime besides just scaling your resources. But for us, we're looking at it from a small business perspective saying we don't want to drop $100,000 a year on hardware. We'd like to dynamically scale our hardware capability on a day-by-day basis. And HPC allows us to do that, even though it's not

required for every problem we use it for. It's such a huge improvement over our ability to do that in-house with in-house hardware. So I'm really excited for what HPC has in other uses that have not traditionally been explored yet. Very good. And maybe cloud will enable this democratization of HPC. You don't have to buy the big system.

Other people can be thinking about new problems, new solutions. Exactly. Great. Very good, Mike. So I would like to close. I would like to thank our guest, Mike Hollenbeck, his founder and CTO at Optisys,

for being in front in the area of antenna design and for sharing his experience that we may apply in other areas. Till next time, I am Gabriel Bronner, and this was the Big Compute Podcast. Thank you.