I’ve always been fascinated by precision bearings, but I’ve never been a fan of their price tag. Most of us are familiar with standard ball bearings—cheap, ubiquitous, and great for high-speed radial loads. But what if you need something with a large inner diameter, a thin profile, and the rigidity to handle both axial and radial loads? This is where the cross roller bearing shines. These precision workhorses are found in heavy machinery, robotics, and other high-precision applications. They're also notoriously expensive, often costing hundreds, if not thousands, of dollars.

That's the problem I decided to tackle: can I build a functional, large-diameter cross roller bearing for a fraction of the cost using a 3D printer?

My Roller Bearing Rethink

A traditional cross roller bearing uses cylindrical rollers, oriented at 90-degree angles to each other, to manage loads in all directions. The problem I faced was that printing a bearing 100% out of plastic often leads to disaster. Plastic-on-plastic friction is bad, and plastic itself is too soft, deforming under the high loads a bearing is expected to handle. It might look good off the print bed, but it quickly becomes a wobbly, useless mess.

My solution? A hybrid design. Instead of fully 3D printing the entire bearing, I decided to use the 3D printer for what it does best: creating complex, custom geometries. I'd integrate hardened steel roller bearings into a 3D printed inner and outer raceway. This eliminates the plastic-on-plastic friction and significantly improves the bearing's rigidity and durability. While the loose rollers aren't free, costing about 10 cents to a dollar each, it's a small price to pay for a dramatic increase in performance.

From PLA to Polycarbonate

For my first prototype, I used PLA with two perimeter layers and 25% infill. A simple micrometer test showed about 0.2-0.3 mm of static deflection—not terrible, but not my target of 0.1 mm. The play was clearly coming from the plastic flexing. This led me to my second version, a significant upgrade.

For V2, I kept the overall design but made a few critical changes. First, I switched to a stiffer, tougher engineering plastic: polycarbonate. Second, I doubled the perimeter layers to four. The most innovative change, however, was incorporating steel wire into the raceways. I designed grooves into the internal faces of the bearing to inlay piano wire, a material known for its extreme hardness. To solve the issue of the wire being too springy to stay in the grooves, I built a custom 3D printed wire-bending jig, which allowed me to pre-bend the wire to the perfect diameter.

This improved design yielded impressive results. A new test showed a static deflection of around 0.1 mm—hitting my target and feeling significantly more rigid.

Costs and the Future

My V1 bearing came in at about $50 in materials. The V2, with the polycarbonate and steel wire, was a little more at around $80. Compare that to a comparable off-the-shelf, all-metal precision cross roller bearing which can run upwards of $1,500. I'm on track to save over 90% of the cost.

This isn't a replacement for a high-precision, off-the-shelf bearing for every application, but for slow-moving, high-load projects where a little friction is acceptable, it's a game-changer.

The Problem With My Hands and a Dial Gauge

Up till now my “testing” was a bit unscientific, involving a dial gauge and a lot of guesswork from my own two hands. I knew I needed to do better to get meaningful data. So, I designed a test rig to reliably measure two key parameters: axial twisting moments and axial runout (how flat the bearing rotates). These are the unique strengths of a cross roller bearing and crucial for my target application.

My test rig design mounted the bearing securely in a base and used a stepper motor to precisely rotate it. I added a rigid collet to one face of the bearing to attach weights, using dumbbells for a variable load. My first attempt with a tiny Nema 17 stepper motor was, to put it mildly, a bit of a failure. It simply didn't have the torque to rotate the bearing under a 4 kg load, which nicely illustrated the downside of open-loop stepper motors: they just stall without telling you.

This hiccup meant I became a human stepper motor for a while, manually rotating the bearing and taking measurements. While this was far from ideal, it did give me my baseline data for the Version 2 bearing, which was printed in Prusament PC Blend. I measured its static deflection at 0.13 mm with a 4 kg load and 0.27 mm with an 8 kg load.

Part 2: The Carbon fibre Challenge

While I was waiting for a bigger, better Nema 23 motor to arrive, I decided to print a new version of the bearing using a carbon fibre-filled filament. My hope was that the added carbon fibre would make the plastic stiffer without losing too much toughness, thus reducing the flexing I saw in my initial tests.

Why does adding carbon fibre work? Carbon fibre particles act like rigid boulders within the polymer chain “spaghetti.” The chains have to move around these particles, making the material more resistant to deformation. This increases the stiffness but can also make the material more brittle, so it's a trade-off. For my tests, I used a Prusament PC Carbon fibre filament with milled carbon fibres, which is designed to increase stiffness and reduce warping.

With the new carbon fibre bearing and a powerful Nema 23 stepper motor (equipped with an encoder for closed-loop control to prevent lost steps), I was ready to test. I ran the new setup, but hit another snag: the timing belt kept slipping because the motor mount was bending under the torque. After reinforcing the mount, I was finally able to conduct some proper tests, including running the unlubricated bearing for over 2,000 revolutions with a 4 kg weight. I even managed to measure the static load deflection with an 8 kg load, though once again, I had to be a human stepper motor.

The Results

The data gave me some interesting insights.

Axial Runout

Across the board, the mean axial runout was very similar for both the polycarbonate and the carbon fibre bearings, at around 0.1 mm for a 4 kg load and 0.15 mm for an 8 kg load. This suggests that the addition of carbon fibre didn't significantly improve the bearing's overall flatness during rotation.

Static Deflection

This is where the carbon fibre really shone. The static deflection of the carbon fibre bearing was significantly lower than the pure polycarbonate version.

• 4 kg load: The polycarbonate deflected by 0.13 mm, while the carbon fibre version only deflected by 0.07 mm. That's a 45% decrease in deflection!

• 8 kg load: The polycarbonate deflected by 0.27 mm, and the carbon fibre version only deflected by 0.23 mm. This was a smaller but still notable 15% difference.

Final Thoughts

What did I learn? My 3D printed hybrid wire raceway bearings aren't ready to replace a high-precision, off-the-shelf metal bearing, especially under very heavy loads. An 8 kg load on these bearings creates a significant amount of friction and static deflection.

However, for a 4 kg load, the carbon fibre version's static deflection was a mere 0.07 mm, and its mean flatness was around 0.1 mm. This makes my design a serious contender for real-world applications where cost is a primary factor and the load is light. And I also learned that if a motor looks too small, it probably is!