Epoxy rack

I have been working on a “rack and pinion” style system using a rack cast in epoxy. Here is the result so far:

I glued some GT2 belt to the bottom inside of an aluminum C section tube, closed the ends, and poured epoxy in it. It makes a perfectly matching rack for the belt. I clamped the rack to the underside of the frame tube so dust or chips cannot settle on it.

I also designed a new roller that moves over this epoxy rack using a closed-loop GT2 belt, which looks like a tank tread. It’s pretty much a drop-in replacement for the standard roller.

I did some rudimentary analysis of the stiffness. When the steppers are on, the roller moves about 0.10mm per kg of load (on the roller itself). The exact number partially depends on the tension of the closed-loop belt. This compares quite favorably with 0.35mm per meter for plain GT2 belt.

Note that the deflection does not vary with the position of the gantry, in contrast with GT2 belts. Also, you can essentially cast a rack of any length you want. Several meters should not be an issue at all.

There are some things I would do differently in a future iteration though. I need to think of a better way to cast the epoxy. I got huge air pockets in it. Doesn’t seem to be an issue so far, but it could be better. Just plain epoxy is more flexible than I thought, so I will try to embed stranded steel wire in it while casting the rack, to see if that makes any difference. (I put it in aluminum C section tube now, hopefully that won’t be necessary anymore). The roller design could also be improved in a few ways.


Here are some more details of how I cast the rack. I tried two approaches.

One attempt used a single PVC C section tube like this:
I glued a 9mm GT belt to the inside bottom, glued some printed plastic parts on the insides of the ends to close it off, and just poured in the epoxy. I did not use any release agent, and it was quite easy to de-mold. Epoxy does not stick well to PVC, nor to GT2 belt. This was very easy to do, but has two downsides. For larger distances, you might need to make something to hold the PVC straight while it cures, it has a tendency to warp a bit. Also, the top surface came out very rounded (high at the walls, low in the center). It would require some post-processing to get the same thickness for the rack everywhere. Should be fairly easy though.

The other attempt used two different sizes of aluminum C section tube. The smaller one had an outer width of 15mm, and an inner with of 12mm. The total height is 8.5mm. I glued the 9mm GT2 belt to the inside bottom, like for the PVC tube. However, this time I got a larger C section tube and put that upside down over the smaller tube. After curing, I removed the large tube, but left the rack inside the smaller tube. This is the one you see in the picture. I did use PVA release agent, and it was still quite hard to de-mold. The height of the rack is the same everywhere though, and it is stiffer because it’s still encased in aluminum. It does have huge air bubbles where the epoxy just sank too far and seeped out.

I feel like the PVC approach could work better if I embed steel cables and/or other filler material to make the final rack stiffer, so the outer aluminum is not needed anymore.

If I use two aluminum tubes again, I would try to hold it vertical for a little while to try and get the air bubbles to rise to the top. It’s not exactly water tight though, so I think epoxy would leak out.


Awesome! I love the use of the rail as rack support. No extra beam (which could deflect) needed.

If you are brave you could cast it in place directly adhered to the rail. EMT is cheap if you screw up a few times. :slightly_smiling_face:

I wonder, how much stiffness does the motor have? It would depend on current and it might be nonlinear as it gets closer to skipping, but for a given gear ratio (pulley diameter) it would be an upper bound on roller stiffness. If you reach that number then your mechanism is perfect and you are done.

1 Like

I thought of that, and I was hoping that it would explain the deflection I saw. I can actually see the pulley rotating ever so slightly when I pull on the roller.

However, I think the stepper only moves 0.080mm before it skips (do you agree?). I can pull 4kg on this stepper before it skips. So 1 kg would only be 0.020mm, and I measured (at most) 0.100mm… no idea about non-linearities though. With the current design, the stepper axle and pulley are kind of enclosed, and I cannot easily mount a laser pointer or something to measure the angle of the stepper.

The closed-loop belt needs to be quite tight. If not, there is slack in between the bearings because the belt wants to make a slightly larger radius. This slack is then taken up when there is load on the roller. I’m pretty sure that a very tight belt gave a deflection of around 0.060mm or so. But at that point it takes quite a bit more torque for the stepper to move the belt. I was a bit surprised at how tight it has to be.

I briefly considered a 2:1 gearing. That would deincrease the stiffness of the stepper itself, but also limit its top speed. 24V should be fine, but I only have 12V. And it would take more parts.

I think the roller can be designed to be stiffer than it currently is. I would like to take more advantage of the stiffness of the gantry tube and the stepper housing, and have less plastic over longish distances. The current design definitely served its purpose as a proof of concept though.

That might not be a bad idea! However, I would coat the rail in PVA to make it release after curing. Epoxy does not stick well to hard and smooth surfaces that well anyway. But you’d have the perfect shape, presumably without any air bubbles. Definitely worth a try.

I also read up on “epoxy granite” this evening. Some people use it to cast a base for a very rigid CNC machine. Granite sounds expensive, but it’s actually 20% epoxy and 80% plain sand! So even cheaper than 100% epoxy. It gets very hard and rigid. No idea how brittle it becomes for thin and long parts like the rack though.

1 Like

How much was the epoxy you used?

Could you just epoxy a belt to a bar and call it a rack?

1 Like

It can work. Brauns CNC works that way. The gt2 belt profile has a convex rounded tooth profile and might not grip another belt as well as a matching concave shape. Or there may be some other shape that is even better.

Thanks or the explanation.

Regarding the bubbles you probably already thought of these things but I thought some might be of help.

I have had luck getting ride of bubbles in two par systems that have a temperature dependent cure rate. Just put it in a chest freezer.

There are competing rates in the system. As the temperature drops the viscosity increases and the bubbles move slower. As the temperature drops the rate of increase in viscosity due to the curing. Add to that the epoxy releases heat.

If I learned anything in differential equations (Or maybe that was infinite sequences and series?)
it was that competing rates hard to predict.

It also helps to pour the epoxy from a good height. The thin stream makes the bubbles break on the way down. Also pour it in one place and let it flow so you don’t trap any new bubbles.

Hope that helps.

You could do a small scale test.

1 Like

I bought 800g of this epoxy, which was 24.99 euro ($27.50). It should be enough to cast 4 ranks of 2 meters each or so, even if it’s just epoxy without any filler. You just need two mixing cups and something to mix it with, that’s all.

1 Like

There are two issues with that. First, epoxy does not stick well to the rubber used in GT2 belts. (Also the contact cement I used does not bond well, but for my purposes is was good enough). The belt will come loose after a while I think. If I recall correctly, BraunsCNC also had this problem.

The other issue, as Jamie said, is that GT2 belts do not mesh very well with each other. I did read about people using T2.5 belts, which mesh better with themselves. Haven’t tried it though.

1 Like

There are two kinds of bubbles in my cast. One kind are huge bubbles that are similar to the concave shape I got with the PVC attempt. It’s more like a void than a bubble actually. The PVC attempt did not have these. I think this issue is due to the way I tried to cast and should be avoidable.

The other kind are tiny bubbles all throughout the epoxy. Almost impossible to see with the naked eye, but easy to see under a microscope. It makes the cured epoxy look a bit hazy instead of fully transparent. I’m not too worried about these actually. I don’t care about the appearance. The strength may be a little less than optimal but I expect a load of at most a few kilos should be fine for the teeth.

In an attempt to get rid of the small bubbles anyway I went over it with a butane torch before it was cured. Maybe I should have tried to pour it layer by layer and go over it every time with the torch. The PVC has the disadvantage here, because it is more sensitive to the heat of course. But you can go quick enough so the PVC does not get too hot. The internet is full of tips and tricks to avoid these tiny bubbles. The long thin stream is one of them. A more professional approach is to put the resin in a vacuum (but I don’t have the equipment for that). I had not read about the freezer yet though.

1 Like

Just an idea, back to the cast it to the tube idea. Drill a series of holes down the centerline of the tube, then when you fill your mold, put the tube holes down on top of the mold. Should push epoxy up the holes locking it in place. Depending on the depth, you could even put some self tapping screws into the holes and just leave them sticking out a touch.


At first I was thinking that the vacuum would not be practical because it would have to be so big but you could just put it in a PVC pipe and connect a vacuum pump to the end. If you added a Tee you could put a piece of polycarbonate over the hole so that you could watch the bubbles under vacuum.

1 Like

At one time, I was also playing around with rack and pinion drive for our CNC machines… though I used a 3d-printed “herringbone” r&p setup, similar to one I’d seen for a 3d printer. I’d thought about mounting a rack under the MPCNC conduit rails similar to what you show but, at the time, was working on a different machine altogether (which is a flipped version of what you show) so never got so far as to adapt it to MPCNC. Here’s what the printed r&p looked like

It seems the discussion so far is more about fabricating a cast rack itself however… and, sadly, I have nothing to offer so I will shut up. But I’m really interested in the r&p drive setup you’ve shown in your photos and would love to see more… the r&p drive idea in particular.

1 Like

Nice machine!

I plan to release STL files, but not just yet. There are still a few minor flaws in the design. I think I will release the second (future) iteration somewhere.

1 Like

I like the idea. But I hated drilling holes in my stainless steel tubes for the Z axis. I had no experience, and even with cutting oil and starting with 2 mm drill bit I found it quite difficult… maybe EMT is lot easier, I don’t know.

1 Like

Oh fun!

I would add putting the teeth on the outside would be better, you have it so no stuff can fall on the track but it can still fall on the stepper side. Rotate it 90 and you are away from the dust and gravity helps.

Drilling into stainless is a NIGHTMARE.

The next time I plan to do it I’m going to invest in some quality drill bits.

EMT will be a piece of cake in comparison. Use a centre punch or a jig of some kind to keep the drill bit from walking.