Ultimaker to Hbot conversion

[Hrvoje Čop]

Hi all!

Great to find a group of fellow Hbot enthusiasts :)

I'm trying to convert my Ultimaker to an Hbot just for the fun of it. Now, since I haven't properly researched Hbots and never seen one live, I ran into the gantry twisting problem, which was so severe that my Y axis motion often lost steps.

It seems to me that the most problematic scenario was when one motor was moving, and one was still, ie. diagonal moves.

The whole thing was done in one day, so it's not really refined or well thought through, but I wanted to see it moving as fast as possible. :)

I wanted it do be more like an add-on of sorts, so I don't have to do any mods on the Ultimaker itself. What I wanted to ask is if anybody here has any tips as to what affects gantry twisting the most? How to reduce the problem?

And does the position of the pulleys on the carriage vs. the X and Y bearings or bushings have any impact on that? 

Thanks, 

chopmeister

[iquizzle]

Cool project. In my experience, the gantry twisting is due to the slight angle that the linear rods can move through the bearings/bushings on the sides. One thing you can do to minimize that is to space out your bearings as far as possible... but this has the disadvantage of reducing the useable area of your printer. I have only encountered a slight distortion -- especially visible on interior circles. I would still like to find 'the perfect solution' though, however it might involve more expensive parts.

The other thing you can do is switch to the corexy geometry. If you look at the way the forces are applied, I think it has a slight advantage. Over the past few days, I've considered making a mod for the UConduit to implement corexy-style geometry. It might take a little more re-thinking though.

[Hrvoje Čop]

Thanks for the reply. Like you mentioned somewhere here, I also found out that a single bearing/bush works much better than two per axis due to that slight misalignment. I think those extra long LM8LUU's could do the trick here, they're 45mm long. I did want to reuse most of the parts from the UM though, but those bronze bushings seem even more susceptible to this problem than bearings.

I'm not sure where to cram in the coreXY in the ultimaker frame though but it could be interesting. 

[iquizzle]

Yes, I agree. The bushings I've used have a rather large clearance from the linear rod. This is unacceptable for the h-bot geometry. I know that you can buy "specialty" bushings from Thomson or PBCLinear that have a much closer fit. You have to buy them matched to a specific grade of linear rod as well. I haven't tried these, but they're expensive. I'm not sure they justify the cost vs linear rail.

[Hrvoje Čop]

I was looking into IGUS bearings and their aluminum rods, but haven't had time to go see their local distributor yet. That would be a better solution maybe, because they are much lighter than the sliding blocks of classic linear rails. Having an Ultimaker for a few years now, I've learned to fully appreciate what low moving weight does for the speed and quality of prints. 

But I admit it's tough to do a system which is both very light and at the same time very rigid, without it being really expensive. But what can I say, I love challenges :)

This UM addon is just a test for a bigger Hbot printer I intend to do from scratch, so I'm using it to learn about the ins and outs of Hbots. And well, I sure am learning, hahaha. 

[Hrvoje Čop]

So, after some changes in the setup, I've got some movement. :) You can take a look at it here.

It still twists a bit in the video. I've since tweaked it a bit to there is even less twisting than that, and only at certain moves. I have to hook up an extruder now to see how it performs while in use.

[iquizzle]

Nice progress! Are you using a single LML8UU on the ends? You might be able to reduce the twisting if you use a pair of regular LM8UUs on the ends and space them farther apart.

[Hrvoje Čop]

That's exactly what I did, I used two. I found that vertically aligned X rods twist much less than horizontal ones like I had on the first version. Too bad my extruder fits much better on horizontal ones, so I need to do some 3d modelling magic to fit it all on this head. But I'll try to whip something up today. I'm very eager to see how it actually prints, and how much twisting happens during printing. It's hard to see, but it seems to me most twisting occurs when homing, not so much while the head moves around the center of the platform.

[Chris (crispy1)]

The twisting occurs whenever the X direction reverses.

I tested an h-bot setup and it's basically impossible to get good prints because the crossbar shifts/twists.  At that point you have basically 2 alternatives: make the frame and bearing assemblies as stiff as possible to counter-act the twisting force, or switch to a CoreXY belt setup which completely elminates the twisting forces.  I decided the latter option was easier.

[Hrvoje Čop]

Well, I've seen some Hbots on the web print pretty good stuff, so I definitely think it's possible to achieve the required rigidity. I can't fit a coreXY gantry in my current add-on frame, so that experiment will have to wait until my new laser cutter arrives. In the meantime I will still try to do some prints with this, if nothing else then to explore how exactly gantry twisting affects prints. I'm currently printing the last parts I need so there will be some test prints today. :)

[iquizzle]

You can reduce the gantry twisting, but at some level I don't think you can't ever completely get rid of it. At this point in my h-bot, I can't visually see the gantry twisting in my h-bot. For me, it's rather small and doesn't affect most prints. Where I notice it the most is on small to mid-size interior holes. The change in direction causes a small distortion of the circle along a 45 degree angle (depending on if it is printed CW or CCW).

The outside rods of the h-bot can move through the bearings at a slight angle. When you propagate that over the "lever arm" of the gantry, it becomes more significant. I think one solution to keep the h-bot geometry is to use more precise bearings on the rails. 

[iquizzle]

I made a useful graphic that should illustrate the physics behind this:

[Cameron MacLachlan]

For this reason I think I will attempt a UConduit mod for CoreXY, and see how that works out.  Should be relatively straight forward.

[iquizzle]

apologies folks -- the correct drawing should look like this:

same result (with opposite direction of gantry torque). I was a little hasty in drawing the forces.