Triperon Motion

oliof

@fma joint design is one of the main challenges. Wider joints than on @Apsu s 2016 design will help with the backlash I encountered in https://forum.duet3d.com/topic/16588/colinear-tripteron-progress ... but I am currently doing other stuff.

Best,
Harald

Apsu

@droftarts said in Triperon Motion:

I don’t think it’s a delta movement. If the angle of the Tripteron effector is 45 degrees, a horizontal movement is translated into a 1:1 movement in the corresponding axis. However, I don’t know for sure; @apsu or @oliof would be the ones to ask!

You are indeed correct that it is not delta movement. It is completely linear, and described by a linear matrix/system of linear equations. All Tripterons are (including my 2016 Delteron/Colinear design), and as such are uniquely different from SCARAs or Deltas.

And lastly, just to say it, the orthogonal Tripteron layout is also Cartesian.

Apsu

@o_lampe said in Triperon Motion:

@fma Right, designing these joints and then print them are two different shoes...

Sure, but these machines do work. It's not like they haven't been built and we're just guessing. Questions about specific rigidity/resonance targets are a separate thing, but one of the benefits of parallel kinematic systems is that they are redundantly constrained (not overconstrained). Meaning the rigidity and dynamic behavior isn't just a question of individual joints but the system as a whole.

Or to think of it another way, the whole is stiffer than the sum of its parts

droftarts

@Apsu thanks for the info. Is there a link to how the joints are currently designed? I’d think something like a motorcycle/bicycle head race (tapered roller bearing) would be ideal.

Ian

Apsu

@droftarts https://i.imgur.com/o2IsTWH.png is the original joint I settled on in my 2016 build. Double-shear, dual-back-to-back preloaded radial bearings (608), with steel washers used for compressing plastic between metal, and separating inner from outer bearing races.

Tapered rollers or angular contact sets have been suggested before, and I think they're a fine idea, but I think larger bearings (like the 6806s in Seward's latest videos) and using square tube (printed or otherwise) is a much bigger improvement over my old 2020 arm idea. Extrusion does not resist torsion force very well due to the small core diameter, and there's a large number of elbow angles where the force on one or more of the arm segments is largely torsional.

My most recent Tripteron prototype (https://www.youtube.com/watch?v=7LAmA4Q6-o4) build is using 2040 segments with single-shear joints, radial bearings on each end 40mm apart, and a thrust bearing inbetween. It's decently stiff, but I still think single-shear needs much larger bearings to improve the surface area over which the joint halves interact (stiffening the lever arm of the interface).

Seward's current joints in the video mockups are pairs of 6806s in single-shear with bolted in caps to secure them together, and it's a very reasonable approach. I also believe you could use a 6806 and 608 in combination with a similar setup and it'd be fine. The biggest question overall here is honestly joint 'wrist' geometry/cross-section, and the material/cross-section of the arm segments given their torsional and bending forces. Everyone focuses in on the joint itself but these other two points are much bigger in my experience. And I built dozens of these things to test

Anyhow, that's my $0.02

fcwilt

Interesting stuff.

What is the raison d'etre for such machines?

Thanks.

Frederick

Apsu

@fcwilt said in Triperon Motion:

What is the raison d'etre for such machines?

There are very few machine designs -- for printers or otherwise -- that are linked in parallel, in 3 or more axes, and also linear kinematic mechanisms. Of those, the multipteron family are the simplest.

The general benefit of parallel machines is high dynamic stiffness without needing to move one actuator with another (serial linkage), which can compound the error of one axis with the other due to the coupling. The side effect is you can build a 3D printer with a fixed bed and all fixed motors.

fcwilt

@Apsu said in Triperon Motion:

The general benefit of parallel machines is high dynamic stiffness without needing to move one actuator with another (serial linkage), which can compound the error of one axis with the other due to the coupling. The side effect is you can build a 3D printer with a fixed bed and all fixed motors.

Thanks.

Is it hard to get the "slop" out of the joints?

I understand your comment about serial linkage but it would seem, from an "outsiders" point of view, you've traded one engineering challenge for another.

Frederick

Apsu

@fcwilt Joint slop is quite easy to take out with one of these types of arrangements, as preloading radial bearings is exactly the method used for doing so in machinery. There are many articles and data sheets discussing taking slop out of radial bearings by preloading, as well as different preload approaches (and when you might want to use angular contact or tapered rollers, etc) out there in the wild. Like https://www.nationalprecision.com/info-library/technical-data/bearing-preload/ or https://www.skf.com/binaries/pub12/Images/0901d1968065f1f4-Bearing-preload_tcm_12-299896.pdf for example.

Back to back and face to face are both very common setups for wheels and shafts, from bicycles to engines to skateboards. Getting the joints stiff is pretty easy. Designing the housings and connections and arms to handle the large lever and twisting forces is the real challenge.

fcwilt

@Apsu

Thanks.

They do look like they could be a fun and challenging project.

Frederick

o_lampe

If you need another excuse to build a monorail Tripteron, imagine you have four of them surrounding the printbed.

dc42

@Apsu said in Triperon Motion:

You are indeed correct that it is not delta movement. It is completely linear, and described by a linear matrix/system of linear equations. All Tripterons are...

If that is the case, then standard RRF supports Tripterons, and there is no longer any need for custom kinematics. Use the M669 command to specify Cartesian and use the XYZ parameters to modify the matrix.

speedpete

Hello,
I'm just trying out the delteron from Apsu, does anyone have a working config.g for duet wifi board? ...could you please show this here?
My hardware is printed, smart effector is here, so I can start building this into my Kossel XL.

Best Regards
Peter

oliof

@speedpete you'll need to build your own firmware from my fork https://github.com/oliof/RepRapFirmware/tree/colinear-tripteron

oliof

@speedpete also you need to brace for disappointment, there is a lot of inherent backlash in the Delteron

https://photos.app.goo.gl/Kk1dJpv59M4AB1mi8

speedpete

@oliof
that looks more complicated that I can handle ...What is with "If that is the case, then standard RRF supports Tripterons, and there is no longer any need for custom kinematics. Use the M669 command to specify Cartesian and use the XYZ parameters to modify the matrix." from DC42?
Doesn´t that work? I´ve found no one showing such a config.g

Best Regards
Peter

oliof

@speedpete there are two different kinds of Tripterons:

Orthogonal Tripterons in a box frame -- These use Cartesian kinematics and can be run without any changes to firmware. You could use the RepRapFirmware configurator to create a config.g for these.
Colinear Tripterons in a delta frame like @Apsu s delteron -- these use their own kinematics as implemented in my fork and need firmware changes. I'll need to check whether it can be done as @dc42 says, but back in the day I couldn't make it work (but that's probably my lack of understanding and not a limitation of the firmware).

oliof

@dc42 these are the inverse kinematics for colinear tripteron

/*
Inverse Kinematics
     *
     * 3x3 reduction coefficient matrix * 1x3 coordinate matrix
     * [a_x -a_y 1]   [x]
     * [b_x -b_y 1] * [y]
     * [g_x -g_y 1]   [z]
     */

How'd that be expressed in M669?

But besides that I needed arm angle and the height at which the machine homed plus some correction factor (which thinking about it today might be dropped in favor of adjusting steps/mm...) for proper movement, as well as delta style homing at the top (because the motors move the arms colinearly and not in a Cartesian fashion). I may pull out my prototype and do some testing; I assume I can drop a lot of my code if M669 matrix definition works.

speedpete

@oliof
That would be very interesting ...nice of you Truly I really depend on you in this case ...so please give it a try

Best Regards
Peter

dc42

@oliof said in Triperon Motion:

Inverse Kinematics
     *
     * 3x3 reduction coefficient matrix * 1x3 coordinate matrix
     * [a_x -a_y 1]   [x]
     * [b_x -b_y 1] * [y]
     * [g_x -g_y 1]   [z]
     */

How'd that be expressed in M669?

If a_x, a_y etc. are constants, then this:

M669 K0 Xa_x:b_x:c_x Y-a_y:-b_y:-g_y Z1:1:1

with appropriate substitutions for a_x etc.