0:00 Okay, great. Hey, good morning. My name is Andrew Sink. I’m an applications engineer at Carbon and I help our customers to develop 3D printed parts for the Carbon platform. Today, we’re going to be talking about mass customization in additive manufacturing, specifically the Physique 1:1 custom bike saddle. But before we dive in, I do want to take a quick step back. This is a very forwardlooking kind of conference, so I want to take a moment to acknowledge how far we’ve come.

0:23 The last time that I was metaphorically here was on stage at CDFAM Berlin May 2024. And the name of that presentation was the road to 1 million custom parts per month. So the intent was really we were trying to build an automated pipeline to enable mass customization of 3D printed parts with the goal of making 1 million custom parts processed every single month. So at the time it was a little bit of a stretch.

0:46 And really when we define custom parts we mean you mail us something. You give us a string of text, a pressure map, a piece of geometry, and somewhere else on a 3D printer, a part starts printing. Very hard to do as it turns out. But it’s exciting because as of Q3 of this year, we’re currently averaging 920,000 custom parts per month. So, we’re making real progress towards that goal.

1:06 And we’re going to be talking about a small subset of that 920,000 parts today. First, a little bit of background for anybody who’s not familiar with Carbon. Carbon is a 3D printing company founded in 2013. Focuses on hardware, software, and material science. So, we make our own 3D printers, we make our own materials, and we make carbon design engine, which is a lattising software designed for making pretty much everything.

1:29 So, really what we’re focused on is enabling our partners to make consumer products. So, this is a VAT photopolymerization process called digital light synthesis. We have UV light. We project it into a liquid and we pull a solid part out of it. It’s going to come in handy later. So, quick introduction to physique. Physique is an Italian cycling company. They’re focused on high performance aftermarket accessories for the cycling industry.

1:53 So, think saddles, helmets, shoes, bar tape. Basically, if you know somebody whose bike costs about as half as much as their car, they’re probably riding some physique parts on it. They’re very well known in the industry. So, this was kind of a natural fit because they know what makes a good aftermarket accessory and we really know what makes a good 3D print. So, in 2019, we launched the Ontar Adaptive.

2:13 It was the first 3D printed saddle with carbon technology on the market. Since then, we’ve launched about 10 more of these. So, this is really a strong focus for us and it’s a really good fit. If you’re asking yourself, why are we 3D printing saddles? What’s the benefit? Typical saddle can be made from anywhere from five to seven different materials in a variety of thicknesses. So, it’s a really complex manufacturing process.

2:34 It is not as easy as it looks when you’re looking at a saddle and you’re thinking it’s just a block of foam. Turns out there’s a lot going on under the hood. So, we’re able to digitally create things like zones, relief cutouts, pressure channels, things that are really hard to do conventionally. Turns out are a really good fit for 3D printing. So, back in 2019, there was a cycling news article that was talking about the first 3D printed saddle with carbon.

3:00 And really what they were focused on was the reviewer who must have had a time machine or crystal ball or something said, “Hey, this is really cool, but you know, really the promise of 3D printing is that things can be made custom. So, wouldn’t it be great if there was a way for someone to go to a bike shop, sit down on a pressure sensor, and have a custom part made for them and mailed to their house or mailed to a bike shop to have it installed?

3:21 So, flash forward a couple years later, and that’s exactly what we’re doing. So, what do we need to create that that workflow? The materials, the computational tools, and manufacturing at scale. So, we already have all of those. We can 3D print many parts. That part is is fairly known. The tough part is that onetoone rider pipeline. And so this is where the German cycling experts at Gabiumize come in.

3:46 We’re not going to talk about them too much today, but they are really critical to this process. So they know what makes a comfortable saddle. They’ve spent decades defining comfort in the context of three of saddle of cycling. So they’re really a you know important part of the of the process. Given that we’re at CDFAM, I think most of the talks today and tomorrow are going to be focused on the CD, so the computational design side.

4:11 We’re going to be talking about the FAM. We’re going to be focusing almost exclusively on the four additive manufacturing and the 3D printing of these parts. So, this is the typical workflow for a custom saddle. So, you have somebody who goes to a cycling store and they go and get a bike fit. They go through the process of making sure that they fit on their bike, that they’re comfortable, and then their data is collected.

4:30 And so, this is done through dynamic real-time testing. So, in a variety of positions, you have cyclists who are kind of moving around on the bike. So, you want to make sure you’re mapping this pressure accurately. And the intent here is really to reduce peak pressure during cycling, especially if you’re riding for 30, 50, 100 miles at a time. Comfort becomes pretty critical. Once that data is collected, it’s then sent to our design engine software where it modulates a lattice and that lattice is now designed custom for that user.

4:57 It’s then printed and then delivered. So, today we’re going to focus basically on steps two and three. So, we’re going to talk about design engine. We’re going to talk about printing. But the question I want you to ask yourself and just something to kind of think about in general for this type of process is how do I guarantee printability of a part that’s different every time? How do I make sure that I know this part is going to be sent to a printer and print successfully without operator intervention, without somebody having to check the support material, check the parts, things like that.

5:23 So, here’s what the customer sees. You can go to a bike shop. There’s about 50 shops or so right now that have this setup. You bring your bike, you put it on a roller, and you’re asked some questions. So, this is part of the algorithm that determines your pressure map. How tall are you? How much do you weigh? What is your typical ride style? Do you have an asymmetrical gate?

5:41 And more importantly, they’re checking for comfort. So, the most comfortable saddle in the world won’t be helpful. If your seat post is up so high that your feet can’t reach the pedals. So, we’re really trying to solve for does this bike fit you right now. Once it does and we know that, then we can start to focus on the the customization of the saddle. What’s really interesting is it’s possible to go through this entire process and not know that the saddle is 3D printed.

6:06 The 3D printing is sort of information that’s there if you want it, but for most consumers, they’re really competing on performance. They’re they are buying this because of of performance, not the novelty of being 3D printed. So, it means we have to actually make this part work pretty well, too. Okay. So that pressure data gets sent to us. But where is it going and what is it doing?

6:25 So this is an example of the base saddle geometry that’s inside one of these 3D printed saddles. We use carbon design engine to create a tetrahedral lattice. So through tetrahedralization, we’re basically filling a design space with a bunch of little pyramids. Turns out tetrahedral lises are great for making parts with known mechanical performance, but they don’t look really great on the surface. They can look a little chaotic.

6:45 So, one of the challenges we had to overcome was to create a uniform surface on the top of the lattice. So, what you’re looking at here is a cage lattice. This goes over the surface of the design space. And this was inspired by UV texturing. So, we’re really unwrapping this part, laying it flat, applying a hexagonal surface to it, and then creating that interior lattice as a second step.

7:03 So, there’s basically three types of lattice here. There’s the cage, the interior, and then a stitch layer that connects these all together. So the intent here is to make something that we know has a a known mechanical performance. We need to have a specific there’s a specific set of requirements. And when you’re generating strut lattice like this, there’s typically three levers you can pull. You can change the configuration of the struts.

7:26 That’s your unit cell. You can change the size of that unit cell. So that’s how big the blocks are in space. And then you can also change the diameter of the struts. So for strut base lattice or beam lattice, those are your three levers. For a custom project like this to even be possible, those first two levers have to be locked. We have to know the configuration of the lattice.

7:44 We have to know where the struts are every time. And we have to make sure that that repeating block is the same size throughout. So really what we need to make sure is we can modulate those struts to minimum and maximum diameter without having to affect the performance of the the printability of the part. So here’s an example of the stitching. So what we’re looking for here is finding all of the nodes on the exterior of the surface and then finding all the nodes on the top surface facing layer of the interior lattice and stitching them together while avoiding underpopulation, overpopulation or unprintable struts.

8:15 This is one of those things that looks really really easy on paper and then when you go to do it, it turns out it is just a little bit harder. And so we spent a lot of time trying to make this lattice printable and trying to understand how can we create a mechanically repeatable part. So, the printability aspect is really never in question. But we’re not quite done.

8:35 As it turns out, most people who are interested in riding a saddle or or having any kind of part 3D printed, they don’t just want a block of lattice. There’s also a skin. There’s support material. There is all these extra sort of accutra that make it go from a a strut based lattice into a 3D printed product. And so, that also has to be baked in. So, before we’ve even assigned diameters to these struts, everything you’re looking at here is a center line.

8:59 We’ve also added this external geometry so we know that we can send this to the printer. And what we’re trying to do before we start customizing it is we’re checking for the for defects. We need to make sure that we don’t have any defects that pop up during printing. So the easiest way to do that is to assign bounds to these strut diameters. So we have a minimum and a maximum strut diameter.

9:18 So this is the softest the saddle can be and then the firmst the saddle can be. So with the minimum start diameter, this would be the equivalent of putting a feather on the pressure sensor and then hitting go and you wind up with a lattice that you can almost crumple in your hand. The maximum diameter would be if you were to park a truck on that same sensor mat and it fires full blast that this is what comes out.

9:36 And so we want to make sure that these are both technically valid outputs from the pressure mapping. So we need to make sure that they’re both printable. So the easy way to do that is to print the entire lattice at that specific strut diameter. So we have a minimum and a maximum. We go through and we’re checking to to make sure that these parts are printable. And we’re really focused on trying to define defects.

9:58 So the two types of defects we’re looking for are print related defects and design related defects. Print related defects are sort of one-off problems. This is the printer turns off during printing, somebody bumps into it, the resin isn’t mixed properly. You know, these are problems that can be solved by turning the printer off, turning the printer back on again. Typically fairly straightforward. Design related problems are a lot more challenging.

10:19 These are typically problems that are between the keyboard and the chair. And so this really has to, you know, has to do with the model is made. And these are these will absolutely plague you if you don’t track them down initially because they can occur intermittently. They can occur consistently. And if you’re going to be making an a potentially infinite number of custom parts, you need to get rid of these quickly.

10:39 So here’s an example of a design related defect. So in the first picture in the sequence on the skin of the saddle, you’ll notice there’s a little tear. Turns out I can’t actually see into the part to really understand what went wrong. And so I have to figure out, okay, how do I fix this part without being able to understand where it started. So I can’t see into the lattice, but I have a machine that does.

11:01 And so in the second picture in the sequence, this is a CT scanner. So this is an example of a CT scan part. So, we take the part, we bring it into a CT scanner, we run a full scan, and now we have a watertight mesh that we can scrub through sort of in like a 5 to 10 micron increment to understand where did something first go wrong, where did this part start to delaminate.

11:18 And through that process, we’re able to see the third picture in the sequence. This is an example of the slice video. So, this is what the projector is projecting into the lattice. And you’ll notice a sort of like fang shape. And this is caused by in that initial tetrahedralization process, there were two struts that were near horizontal. So on the inside of this lattice, we basically have a strut that’s not quite flat, but not really at a self-supporting angle.

11:42 And then we have another strut that builds off of that. So when the strut diameter drops below a certain amount, this strut will sort of flip over and cause all kinds of problems. Difficult to solve. So the easy way to to to fix that is to create a vertical support column which runs through the center of that node. So now we know that these struts have a column that attaches them to to zero.

12:03 So we’re printing at the minimum diameter, we’re printing at the maximum diameter, we’re trying to understand the the problems that could occur from at a minimum strut diameter, is there enough surface area on the bottom of the saddle to stay adhered to the platform during printing? And at the maximum strut diameter are the voids is the the holes in the lattice are they large enough to allow us to evacuate the resin.

12:26 So back to the beginning we use a photopolymerization process. You can think of our resin is sort of the viscosity of honey. And a 3D printed lattice is unfortunately kind of like a sponge. So imagine you’re dipping the sponge into honey. After printing we have to spin it. We use a centerfuge to pull that resin out but we also need space for it to evacuate. So we’re really trying to we have all these problems we’re trying to solve for initially because once we have that print ready design, we can repeat it as much as we want.

12:54 And that brings us to the optimization step. So now we have minimum, maximum, we have some rider data that we’ve printed and now we’re also using a checkerboard process. So here we’re going through and for each of these individual sensors, we’re actually changing them to min and then max. So, we’re trying to understand, will nodes have problems if half the struts that are attached to them are a very, very small diameter and the other half are much larger.

13:15 And we’re basically just trying to throw everything at this part trying to make it fail, including even intentionally lowering the dose of UV light to cure the part. So, the part is actually curing weaker in this initial stage. We know once we hit the ground running that we’re going to be making hundreds, thousands, tens of thousands of these parts. So, we want to get ahead of these defects as quickly as we can.

13:36 So, we go through all this process. We have the geometry locked down. We have the strut diameters assigned. And then we have the actual strut lattice itself is sort of locked. And there’s only a couple more steps. And one of those steps is doing the whole process nine more times. So this is the physique 1:1 saddle range. So these are all of the saddles available for this project.

13:55 Some of the geometries are fairly straightforward. You can think of the geometry of a saddle like a frisbee. It’s kind of consistent in the center and then it has a sharp taper on the edge. This can be difficult to fill with a lattice. Some of the triathlon and the time trial saddles, they’re much thicker. They have a much more forgiving amount of lattice in them. So, we can make those a bit more conformal.

14:14 But we go through, we do the same process for all these saddles, for all the sizes, and now we have these parts that are ready to print. So, the question then becomes, okay, so we’ve done all this work. We have this custom pipeline. We know that we can produce these parts in an automated fashion. We know the geometry is printable, but the real core question is, does this actually work?

14:35 So here’s our report card. Six of the top 10 finishers in the 2025 to France road saddles that were powered by carbon. So these are saddles that were printed on our printers using our materials, the majority of which were made using our design software. And the overall segment winner, Tad Bogachar, was actually riding a physique 1:1 saddle. So he went through this same pressure mapping process that we just talked about.

14:54 And it’s not just for elite athletes. I also have one of these saddles myself. One of the cool benefits of working on a project like this is you get to be involved in the testing. So, I have a Physique 1:1 saddle. I went and did the same process that everybody else would. And since I have had it made back in March, I’ve put 1,200 m on it or about 2,000 km.

15:13 So, it’s my go-to saddle for longer rides. So, this is a really interesting one because it’s not just designed for proathletes. This isn’t for NFL players. This isn’t for tour cyclists. This is really a true consumer product, which is what we’re really focused on. So on that note, that kind of brings us towards the end here. We’re really focused on growing that number of 920,000 parts. We want to hit that 1 million parts per month.

15:38 And this is where you come in. If you are a category leader, if you make a product that is best-in-class and you know that you can define performance and you can define what makes a good product, we want to get involved with you because we know what makes a good custom product. We have that pipeline worked out. So, we do this with CCM, for hockey helmets, for Adele, for football helmets, with physique, for bike saddles.

15:58 These are all things that we’ve these are all products that sort of exist out in the market today, and we’re looking to grow this segment. Personally, mass customization for me, I’ve been an additive for a little over a decade. This is really the sweet spot, being able to make something that’s a little bit different every time. But you can have that reusable workflow, those building blocks. Okay, so we have two QR codes on the screen as Duann mentioned.

16:21 One of them is follow Carbon on social. If you scan that and then follow us on LinkedIn, Instagram, all that good stuff, we can give you a t-shirt. The other one is a really fun one. This is powered by Carbon. This is a new page that we just launched. This is a basically a shop landing page for all the products on the market today that are made with carbon technology.

16:42 So, these are consumer products. Everything from insoles to midsoles to saddles to helmets to medical devices. It’s basically if you can go to a store and buy it, it exists on this page. So, we’re really trying to highlight our work in the consumer space. On that note, thank you for having me. I’m excited to come back and say that we’ve hit that 1 million mark at some point in the future. Thank you.