How would you create the functional joints in this spherical contact linkage?

[pmesten18]

I gave my best crack at modeling this amazing linkage I found on this youtube video.

https://youtu.be/aLaqMreVj9o

and for some reason I cannot get fusion to play along with the joints, to simulate range of motion. 

I can sort of get the model to move by dragging the top plate around, and if I throw the mouse around a lot, I can get it to snap/fold over to a different orientation. Maybe I'm missing something, and modeled in something in interference, but i bet someone smart here can get it to work, because I'd think this is within the capability of Fusion.

https://a360.co/2ol1lrh

Thanks for checking it out!

[TrippyLighting]

I'll take a look at it. I saw that vid just yesterday.

[hpekristiansen]

At first glance, one can be lead to believe, that the two revolve joints in the end of an arm works as a ball joint, but that is not true. -the arm can not twist. I can not tell, if it is mathematical impossible to make it work -and what we see in the video is just loose revolve joint close to each other. -or if there are some geometry where it works.

Here is a workaround, where I cut the links in two and enable them to twist:

If you make the center cross section circular, the joint will be invisible.

In your design, you have a move that depends on a joint, to combine the two symmetrical halfs of the arm. -like that it is very difficult to make changes later. I suggest to put all your joints in a design like this at the very end.

[TrippyLighting]

While you've made a few choices that could potentially back you into a corner, I don't think there's anything wrong with your design. It sees to simply be a hard nut to crack for the joint solver in Fusion 360.

@jeff_strater this is an interesting "thing".

I'll have to model this myself and see if it will exhibit the same behavior.

[TrippyLighting]

---

@hpekristiansen wrote:

In your design, you have a move that depends on a joint, to combine the two symmetrical halfs of the arm. -like that it is very difficult to make changes later. I suggest to put all your links in design like this at the very end.

---

Yep, that is exactly what Im mean by the OP having backed himself into a corner.

[TheCADWhisperer]

I got it to work (grudgingly) in Autodesk Inventor Professional - but your design is very tight and not possible to manufacture.  There is not realistic clearance between components at the joints (in the real world a 4mm pin will not fit into a 4mm hole without a big hammer or an arbor press - and then it isn't going to turn in the hole).

If you were to fix this issue I might be inclined to experiment a bit more in Inventor.

(And as noted - it is not spherical DOF.)

[pmesten18]

Would a move with an align of the new position, and not a joint be better form? or would there be a better way to design that flip for a more robust design to changes?

[pmesten18]

I appreciate your insight, but am a little surprised that you'd want my model to be manufactureable before you look at it, why?

As I've been a machinist for the better part of the last decade, I have a lot of experience with tolerances and how components ought to fit together, and what I have found is that a nominally designed model is far superior for the intended manufacturing process of this design. (in this case; machining, you can see the flats in the base which may allow for indexing to drill the pivot holes, as well as the linkage arms only being a 2 setup machined component)

99% of the times where we have to send a model back to the designer, or we have issues hitting our tolerances on the first try, it is because a designer tried to model in their design tolerances. I will cede that modeling intended clearances into a design is smart, but in this case it would leave it to be sloppy and unsatisfying as it's intended design as a desk toy, when a simple offset in the CAM, and drawing definition would suffice. 

out of curiosity for my main question though, would modeled clearances or slop make the joints work better?

[TheCADWhisperer]

---

@pmesten18 wrote:

As I've been a machinist for the better part of the last decade...

---

I spent 8 years out on the shop floor as a certified Journeyman Machinist.

I taught manual and CNC machining for 5 yrs.

Used various CAD programs for 32 yrs.

Good luck with your design.

[kb9ydn]

@pmesten18 wrote:

I appreciate your insight, but am a little surprised that you'd want my model to be manufactureable before you look at it, why?

---

I'm baffled by this as well, but anyway...

---

@pmesten18 wrote:

out of curiosity for my main question though, would modeled clearances or slop make the joints work better?

---

I was about to say no, but then it occurred to me that you could actually do this (or sort of fake it at least).  So I decided to try it in Solidworks (because I'm faster at it).  There isn't a way to add slop to concentric mates but there is a way to do it with coincident mates; that is to set them as distances with a narrow max and min..  Interestingly it did kind of make a difference, but ultimately the movement was jerky at best and most of the time it just locked itself up.

This confirms what I was originally thinking in that the system of equations that the solver has to deal with to resolve the position of the moving parts is just too complex to do in real time.  So the solver just basically gives up and allows little if any movement.  In Fusion you would have 12 joints that all have to be solved to determine the position of the moving base.  And those joints are all interrelated.

It is a fascinating problem for sure.

C|

[TrippyLighting]

I tried something similar in Fusion 360 by replacing some of the revolute joints with cylindrical joints and giving them small min//max joint limits.That had the same effect you described.

If all joint limits are removed from the revolute joint  it moves some but prefers to snap to certain positions after a little bit of jiggling. I did the same thing in. ZW3D and there it also likes to snap tp certain position after some initial jiggle.

@kgrunawalt previously has provided very knowledgable replies on joint problem. 

[kgrunawalt]

Very interesting. This is a challenging model to understand intuitively!

I tried my own simple version and may have similar behavior (driving joint far enough caused jump and then rotations were allowed but not sure if motion is correct). For some reason, I've not received the download email yet. Note that I did not model the full link geometry but did join things up properly (I think).

I tried removing some redundant constraints by using cylindrical joints. I will try some more combinations by converting all revolutes to cylindricals and then selective locking some joint occurrence sliding DOF. Like a simple fourbar, this mechanism does not need all the constraints on translation that the revolutes provide (I think). It is possible that removing some redundancy would unstick the solve.

The model with the added revolutes above allowing twist certainly looks like it gives the right motion. I'll have to try that too. The real mechanism does not look like it relies on flex -- the arms look pretty rigid. I wonder if a small amount of flex is needed?

Sorry, but no solution at this time!

[iampete]

Hi all, 

after a hour or two of searching for papers the one describing this joint is:

'Quaternion Joint: Dexterous 3-DOF Joint Representing Quaternion Motion for High-Speed Safe Interaction'

[TrippyLighting]

@iampete got a link ?

[iampete]

not that isn't behind a paywall, sorry

[iampete]

I have had some time to read over the paper. this is the TL;DR. 

The joint is essentially a 2D anti-parallelogram mechanism, it does not move as a true rolling sphere but rather as a a rolling ellipse. If you choose the dimensions correctly this approximation is very close with a max error of 0.0046 mm in there example. However they must also have some slop in the mechanism, they quote 0.109 mm that 'can be absorbed in the mechanical tolerance of the pivots and links'.  They also give the 'Omni-Wrist III' as a example of a true rolling sphere linkage, that looks like quite interesting in its own right.

[kgrunawalt]

Oh interesting! Thanks for the info. Joints in Fusion don’t have any slop so that explains the problem. I guess the added joints allowing some twist effectively allowed for the slop.

[TrippyLighting]

---

@iampete wrote:

I have had some time to read over the paper. this is the TL;DR. 

The joint is essentially a 2D anti-parallelogram mechanism, it does not move as a true rolling sphere but rather as a a rolling ellipse. If you choose the dimensions correctly this approximation is very close with a max error of 0.0046 mm in there example. However they must also have some slop in the mechanism, they quote 0.109 mm that 'can be absorbed in the mechanical tolerance of the pivots and links'.  They also give the 'Omni-Wrist III' as a example of a true rolling sphere linkage, that looks like quite interesting in its own right.

---

I'd still would like to have a link to the pare regardless whether it is behind a paywall. Given that the purpose of this forum is to share knowledge I find your reluctance to share it rather disturbing!

[iampete]

I didn't provide a link because a google search of the title I provided gives you a page of links to the paper behind a paywall, I find your lack of familiarity with the internet disturbing.

[TrippyLighting]

---

@iampete wrote:

I didn't provide a link because a google search of the title I provided gives you a page of links to the paper behind a paywall.

---

I had forgotten about your earlier post in this thread.

Thanks!