- Interviews >  Interview with an Expert: Spencer Wright of pencerw.com and nTopology
22 May 2017 9:37
Interview with an Expert: Spencer Wright of pencerw.com and nTopology
Almost everyone involved in the world of manufacturing will have browsed Spencer Wright’s blog or manufacturing newsletter at some point. Spencer is a longstanding expert in workflows for metal additive manufacturing and head of research & partnerships at nTopology. Spencer was kind enough to sit down with RP Platform to discuss his background in both AM and conventional manufacturing, his current work designing optimised workflows, and how he sees the field evolving in the near future.
How did you originally get interested in 3D printing?
I became interested in metal printing just because it sounded cool!
My background is in traditional manufacturing. I am a project manager, product manager, a pretty decent mechanical designer… I built custom bicycle frames for a number of years. After that, I worked in a prototyping shop where we were developing electromechanical assemblies, so I got some experience in embedded systems design.
I moved to New York in 2012 and was taking a step back and thinking about what I wanted to focus on, while continuing to work on conventionally manufactured stuff. But I was looking for something a little bit different. At the time, Makerbot and Shapeways were really big in New York’s hardware scene. Both of them are really interesting companies, but having a background in mechanical stuff – structural, functional things — I became curious and heard about a couple of different metal printing technologies. This was around the same time as GE acquired Morris, which made a big splash here in the US. I became friendly with a bunch of people who worked in a strategy company called Undercurrent, which was working a lot with GE at the time, and ended up joining them full-time.
It was through that work that I became aware of metal powder bed fusion. There was lot of marketing out there about how 3D printing was going to change X, Y and Z, but I thought “Let’s get some context here”. We have extremely stable manufacturing processes. We have extruding, we have casting, we have forging, we have machining… These are multi-billion dollar industries. How does metal printing compare with them?
So I started looking into it and found these printers cost around $1 million, which seemed like a lot of money compared to a CNC mill. So the question was, “What can they print?” The build platforms were about the size of a breadbox, so what’s a high-value part that fits in a breadbox and might benefit from being lightweight, or having some interesting design elements?
Most of the metal printing industry today is parts for aerospace, or medical implants, or oil & gas, but the answer that I came up with was bike parts. It’s actually a great industry, as people will pay a lot of money for something that’s just a little bit lighter or fits a little bit better. If you can print a part that looks really unique, that is clearly futuristic, people like that.
I didn’t know anything about the technology, but I knew how to design parts, I knew how to source parts, and I knew how to ask questions. So I used those three skills (and the fact that I had a blog and a newsletter, which was a big hole in the market) and kept working at it. I was designing and printing parts, then writing about it, and it just became a thing. As I was working on stuff, I started getting more and more calls out of the blue from people in companies like Siemens and Philips, who are way better at this stuff than I am. It got my voice out there and it gave me access to people at these companies, so I could ask questions and find what they were going through.
Around that time, I ended up looking for something a little different in my career. I decided to focus full-time on manufacturing and design workflows, so now I work at nTopology, which makes design software for industrial 3D printing. Most of what I do is figure out how our software fits into the rest of the tool chain. How do we trace our customers’ workflows across this entire cycle and make that experience better?
It seems like your background in traditional manufacturing means you’re approaching 3D printing in a much more organic way than many people do: starting with a problem and identifying 3D printing as the right tool to fix it. Would you agree with that?
Yeah, definitely. I will note that at the same time, my interest is in using the right technology for the job. I’m constantly tinkering with design projects. I print some of them, but I CNC some of them too. I get people emailing me every week saying they have a business idea and they want to 3D print it. My question is always “Why do you want to print it?”
This is really a unique thing about printing today, and I think that the tech press maybe hasn’t done a great job at establishing why someone would ever want a printed part. There are lots stories out there saying “We’re going to print X and it will be awesome”, but I think the press has not been good at questioning how that would be a benefit to anybody. When I’m looking for something, I don’t care about whether it was printed. Consumers don’t care if things are printed and businesses certainly don’t care. What they care about is whether this product performs better, costs less, or is more readily available.
The bottom line is that for the vast majority of parts, printing is not a good solution. For something that needs to be lighter, it is, as you can incorporate internal passageways and simplify assembly. Those are actual benefits. For me, it’s been figuring out those applications, and the heuristics we can use to look at an industry and figure out whether it will be a good bet for printing.
My advice to people is, don’t try and shoehorn your application into printing. There are really good things for printing and we should focus on those.
What would be some good examples of those recently?
We work in a few primary industries. Across all these different industries, we’re designing lattice [lightweight] structures that have the exact mechanical properties that our clients want.
The biggest one is aerospace — rockets and spaceships — where the regulatory environment is very complicated. There are significant questions about how you inspect parts and ensure they are still good after X number of flight hours, for example. Aerospace has the most stringent requirements, so if you can meet those, you can usually meet the needs of other industries!
We also have a big presence in medical implants, where the design needs are very different, but are still very highly regulated. We also work with a lot of consumer tech companies, who are making parts that absorb energy in a particular way. Footwear companies are very public about it, but there are other sportswear companies looking into printing things like padding. When a person’s body collides with something, you want to absorb that energy so that’s it’s not transferred to their bones. Foam is fine at that, but you can use different structures to dissipate energy outwards, rather than through the pad.
Do you find that there’s much of a learning curve in terms of companies’ workflows, especially when multiple software platforms are involved?
These are pretty complicated. There are times I find myself running $50,000 worth of engineering software on my computer at the same time! The good news is that no matter what you’re working on, whether it’s rocket parts or shoes, the calibre of software we see is typically pretty similar. There are half-a-dozen CAD programs our customers typically use, then there are a handful of different analysis packages and manufacturing software platforms.
While there are differences, the overall requirements are very similar. These are serious engineering companies with serious needs, so you have to maintain traceability, you have to meet the regulatory environment, you have to have documentation, and you have to be able to trace analysis results, which can be tricky.
The reality is that while the engineering software world is relatively mature, manufacturing software is not. The packages companies use (mainly Autodesk Netfabb and Materialise Magics) have not been around that long, in the grand scheme of things. Every year, they update with pretty significant differences. In addition, you have file formats that are constantly changing, so it can get complicated.
What do you see as the answer in terms of streamlining all this?
Right now, our focus is making this process as smooth as possible. Ultimately, the area where we can affect that most is within our software, Element. We can make that experience great. Within this industry, that’s difficult. What are the file formats you’re going to use to go from CAD, to Element, to your build processor? Right now it’s STL, which is a mess of a file format.
Lattice structures might have millions of beams. When you use STL, you’re describing the surface of that structure with triangles. For every beam, you’re going to have a minimum of maybe ten triangles, although in most cases we find it’s better to use at least 50 triangles. With one million beams, this means file sizes that are just insane.
It’s just not practical to describe geometry in that way, and so inside of Element, we don’t use tessellated geometry at all. We use a graph structure, so every beam goes from one node to another node. Each node has an X, Y and Z location, and a radius assigned to it. The beams then connect those nodes. With a relatively small amount of information, we can show designs that have many, many beams. We create designs within our software that way, then we use an open-source file spec that allows you to export them and use your own slicers and orientation tools.
We’re working at integrating this into the 3MF file standard. The goal is to have a much simpler file presentation to communicate lattices. At the very least, this means that if you have to email a file, you’re not waiting 20 minutes for it to upload. The file transfer is easier and the rendering is a lot easier, as the simplified geometry means your CPU isn’t working as hard. We can slice that representation just fine as well. We’re also able to import this right into your FEA software, so we can run beam analysis – which is way easier than using solid elements.
How do you see the uptake of this new file format?
Honestly, I’m not that worried about it. We’ve seen fits and starts in file formats before, but for us, there is such a clear advantage to this format that I suspect it will move very quickly. Most printing file formats are still using triangles, which means the geometrical representation all exists within the same paradigm. 3MF also has triangulated geometry (it’s a better format for a number of other reasons) but we’re adding a completely different representation on top of that, so we can represent the same part with a much smaller file. I suspect the benefits of that are going to far outweigh any difficulties.
Do you see customisation for clients’ specific needs as a key factor in successful software for AM?
Yes and no. We are a product company. We sell a product that we’ve built for the most demanding users and seek to make the best possible version of this. Having a consistent approach to problems makes the experience considerably better. For the more fringe CAD applications, what often happens is a customer says “We love the features, but we really wish that there were context menus everywhere”. The CAD company really wants the sale, so they give them the context menus. In a couple of months, you get context menus everywhere, they work well, and they get the big sale.
In a year, someone else comes along and says “We love your software, but we really want a better command folder system, as the context menus don’t really work for our engineers”. The CAD company spends six months on that, and then you have two ways of doing the same exact thing. What you end up with is a piece of software that’s schizophrenic.
We tend to prefer having a single interaction. Any new feature we introduce will be addressed the same way and will be consistent across the entire workflow. We listen to what our customers really need and come up with something that fits into our framework for managing interactions but also fits those needs.
But adding to that, the need for flexibility in workflows is definitely a point. With the new file format, we have written it in such a way that it is very easy for anyone to grab it, manipulate it and understand what’s going on.
What about streamlining AM workflows? What ways have you found of making the overall processes tighter and reducing any disconnects?
That’s hard. Ultimately, when you look at our aerospace users, you’re constantly bouncing between different things. Even with a relatively simple part, you end up with a part number, then version 2, version 3 etc. and it kind of explodes. One thing that we’ve done is to allow multiple versions of the same design in a single file. You can fork out in different directions, but keep all that within the same actual file, which helps somewhat.
Ultimately, however, interoperability is a bigger deal than streamlining. We’ve thought a lot about making a simpler version of Element as a plug-in to other CAD software, but then we’re working with someone else’s geometry kernel and user interface. Tailoring those things is more difficult than doing it in your own piece of software. In addition, do you choose the most popular CAD software and build for that? Do you pick the one that your customers use?
So for now we’ve chosen to maintain our flexibility. You import into our software, but we’ll make that process as easy as we possibly can. We know that if we keep most of the work in our software, we can guarantee a good user experience.
In wider terms, how do you see all this evolving?
I’m excited for metal printing becoming more mature. Humans have a century of cutting metal under our belts. We’ve done this a lot and we’re good at it. It’s repeatable and it’s reliable, so I don’t need to know what machine a part is made on. With metal printing, that’s completely different. It’s not as repeatable, so if I go to several different manufacturers with the same file, they’ll build it up in different ways and we’ll get a different result. I hear people talking about how we need more materials or speed, and those things are important, but what I really want is maturity. That’s what I’m really excited about.
We have this big effort to build design constraints into our software. If you design something in Photoshop, it has a colour palette that you apply to your design. They don’t let you choose colours like ultraviolet, because you can’t see them, so there’s no point designing something that way! With CAD software, it’s easy to design things that are not manufacturable. Our main focus is on helping our users understand what is and isn’t manufacturable and building that intelligence into the software, so they get real-time feedback on how their design is going to print.
It’s really difficult for us to do that in a world where not only do machines print parts differently, but the manufacturers don’t publish specs on what’s printable. What I really want is for machine manufacturers to publish API’s that give you feedback on printability and point out where the problem areas are.
In closing, what do you see as the ‘next big thing’ in the coming five years?
It’s not a sexy answer, but I’m looking for a more reliable, more repeatable, more stable process. Or at the very least, I want a flow where people can do crazy things, but that offers reliable build times to provide a solid foundation for that.
I think that it’s good to see more of the machine manufacturers developing more integration with parts manufacturers and opening up their own service bureaus. They’re going to learn more about printing parts, and will hopefully feed that intelligence into their machines. At the same time, we’re seeing machine manufacturers get more tightly coupled with software.
There’s a thing called Conway’s Law that says companies tend to build systems that are a facsimile of their own organisational structure. For example, if you’ve got a part with a printed circuit board that’s attached to injection moulding and the mechanical engineering team are in a different building to the embedded systems engineering team, then the PCB is going to be falling off the injection moulded part.
The reasonable corollary is that if you want an end-to-end workflow that is seamless, you need to integrate these teams more closely. Have them working in the same building and bumping into each other at the watercooler! Having more companies bring these different roles together is something I’m very keen to see.