OK, guys...this one is new to me. It's a ROTARY hexabot platform.

http://www.hexel.com/rotobot.htm

It's addressed a lot of the problems for home builders, like variable length legs. Its legs are a fixed length, and the base length is adjusted. Brilliant...it's inspirational.

BUT, being the cheapskate that I am, I WANT to build one of these, but on the cheap! Any ideas? It appears that each of the legs has freedom to move, only until it hits its neighbor...there *is* interference between them. Could it be as simple as a set of nested "lazy susans?" In their graphic, the center looks to be nothing more than stacked ball bearings with attached "arms."

Any thoughts as to the ball joints? I wouldn't mind one of these units, but I'd want something small...maybe a socket-wrench u-joint could work, instead?

And, is it just my imagination, or are the strut connections to the ball joint at the top of the platform, co-planar? i.e. just a hinge? That would simplify the construction tremendously.

So, what do y'all think? Worth it to build one of these, or not?

-- Chuck Knight

P.S. I know EMC has support for hexabots...would it control something like this?

very ineresting design, i think your right about the nested lazy susan idea, i wonder about the pivot points and the overall rigidity, and about how it is locked into position.
are those set screws by the lower pivots, at the end of the arms.

Well, it's a commercially available design, so overall rigidity can't be too bad. And, since errors are averaged in hexabot design, rather than additive like in X-Y-Z designs, the system should be able to hold to higher tolerances, without a vastly increased tolerance for each individual part. A homebuilder's dream. The downside is that there are no fewer than 6 steppers included in this design, and it's overkill for most folks.

My primary concern is not in building it, but in controlling it. Since posting my message, I've run through several dozen variations of this machine, in my mind. (Doesn't everybody do this?) Making it won't be a major problem...it'll be complicated, but not impossible, even with just a home woodworking shop.

My concern is that, as the struts move apart, they move in an arc...that means non-linear equations. They're not that hard to program, but does any easily available software support them? (And, does anyone have the math skills to figure them out? I've got a background that includes differential and integral calculus, but I'm rusty.)

It also means that resolution will be variable...increasingly coarse resolution as the base length shrinks. I think that's right...but that aspect is common to all hexapod designs, I think.

The genius of this unit is that it takes an already very flexible 5 DOF machine, and integrates a 6th DOF into its fundamental design. By rotating all the struts around the path, simultaneously, you get a rotational axis with minimal increase in complexity, and one which still averages the errors.

So, is anyone interested in helping me figure out the dynamics of this gizmo? ;-)

-- Chuck Knight

it almost seems as though 4 axises(<--- is that right) would cover it in a straight line (well 3 and a rotory, bear with me here it will make it more complicted but probably easier to program)
because really its a 3 pivot idea that goes to 6(for the up and down) so if there was a way to move it up and down(maybe a screw) that pushed directy at the upper pivot point (the 3 )
and was nested in the middle of the corrosponding lower pair of legs, i don't know i explained that right but it seems like a good idea, makes a 4 axis a 6 axis (other axis on the rotary)

and your right about the arc,.5" will give you a foot at the end of travel, so being able adjust it thru a screw mechanism will give you greater control

I think I see what you're saying, but I'm not sure it will work. This platform not only tilts, but *moves* across its working envelope in terms of height, tilt, pitch, yaw, etc...

Take a close look...the hexapod design is a variable geometry octahedron. All sides are triangles, which makes it stable. Adjusting the geometry of those triangles makes it move. Go to the videos section of the Hexel web site (under gallery, I think) and you'll see how these things can move all oer the place.

Now, if you removed the triangles, and just put in a single vertical strut, then it would no longer be stable. Same principle as geodesic domes. It would also not move, except up and down, and tilt.

Check out my avatar...my picture. Notice the model I'm holding in my hand. I'm a dome nut..."natural geometry and geodesic structures", is one of my hobbies. Inherently stable structures are impossible to build, without triangulation...in the model I'm holding, the panelized approach creates "virtual triangles" since the corners are connected by the panels themselves.

Back to CNC...I think this design will only work if it is built pretty much as-is. We might be able to simplify or modify some of the details, but the overall geometry is going to have to remain the same. The control has got to be done in a layer of software...be it from within EMC, or in some custom software that would have to be written.

I wonder if Hexel would make their control routines available to us? The home workshop is definitely not their target market...

-- Chuck Knight

that design is ingenious and looks very doable to build but math is another story as you said. regardless, it would still be interesting to build.

with this design, the top half of the machine is trivial with fixed legs and table. I dont quite think a standard hinge will work for the joints but that is just a gut feeling. something like ball joints from suspension or steering i think would be better.

what if you used an automotive flywheel for the base with the six motors walking around it, granted the teeth are coarse but it is flat. also I am wondering if the arms at the base have to form a triangle at the center, if not the arms could be shortened slightly and mount to a ring that rides a single bearing in the center.

This site: http://freeandeasy.sourceforge.net/rotproto.php
may give some hints to the basic math required to get started on the software.

robotic regards,

Tom
= = = = =
Do not pray for tasks equal to your powers; pray for powers equal to your tasks.
- - Phillips Brooks, bishop and orator (1835-1893)

You might be able to get away with three motors. I see two potential ways:

1. Each corner of the table has two legs. If you were to fix one of these legs to the base, then you would only need to move the other leg. This however would involve more complex math for some simple movements. For example, if you wanted to simply tilt one edge of the platform down, you would have to take into account a slight twisting of the platform since only one of the corners supporting legs is moving, the corners position ins 3D space will travel proportionately in the same direction as the moving support leg. This would necessitate a counter rotation of the entire table to counteract.

2. Have the two corner legs attached to the same motor. This would mean that the two legs travel the same distance in opposite directions for any movement. A real upside for the do-it-yourself’er is that any single movement will be restricted to less complex set. That is to say, if you wish to move corner A, then the two supporting struts will travel an equal distance and corner A’s position in space will be easier to calculate.

--bb99

Originally posted by jimbo
with this design, the top half of the machine is trivial with fixed legs and table. I dont quite think a standard hinge will work for the joints but that is just a gut feeling. something like ball joints from suspension or steering i think would be better.


I don't know that I'd call it trivial, but it's definitely home buildable. With nice modular parts, like ball joints, it could almost be a "tinkertoy" approach. My thing is, I'm a cheapskate, and I'm not very familiar with automotive parts. But, I seem to remember that the smaller automotive ball joints have a very limited range of motion. That's my concern.

I seem to remember some toy, years ago, that utilized ball joints...I doubt it was very accurate or long lived, but as I remember, they just "snapped" together. Toobers and Zots, or something like that? Oh well...it's been a while since I played with such toys.

what if you used an automotive flywheel for the base with the six motors walking around it, granted the teeth are coarse but it is flat. also I am wondering if the arms at the base have to form a triangle at the center, if not the arms could be shortened slightly and mount to a ring that rides a single bearing in the center.

I wonder what kind of strength the motors would have to have? I have several ~50-60 oz-in steppers just lying around...

I love the idea of the automotive flywheel -- the teeth gave me a good idea. Essentially a rack and pinion system for motion...they're supposed to be quite good, very fast, and easy to adjust for backlash. The teeth are already precision ground on the flywheel, and the price would definitely be right. I *know* they're cheap, on the used market.

As for the center shaft, I wonder if it'd need to have bearings, or if a bronze bushing would be sufficient? We're talking about relatively little and slow motion, after all...lots of starting and stopping, though, so it'd probably be a good idea to use bearings.

Now I'm wondering if the central shaft is even necessary. Assuming a downward thrust, the model I posted above would hold together without the central shaft. I'm guessing that it would help hold it in alignment, but a reasonably "tight" construction around the outside edge would do the same thing. And, a pinion driven around the outside, wouldn't need any connection at the center.

Power distribution might be helped, though -- it might be possible to make a 360 degree electrical conection through the bearing? Otherwise, the rotation could cause all sorts of interesting twists and turns in the wiring.

I think this is a GOOD START!

-- Chuck Knight

Originally posted by bb99
You might be able to get away with three motors. I see two potential ways:

1. Each corner of the table has two legs. If you were to fix one of these legs to the base, then you would only need to move the other leg. This however would involve more complex math for some simple movements. For example, if you wanted to simply tilt one edge of the platform down, you would have to take into account a slight twisting of the platform since only one of the corners supporting legs is moving, the corners position ins 3D space will travel proportionately in the same direction as the moving support leg. This would necessitate a counter rotation of the entire table to counteract.


I don't know that this is true. You can get reverse threaded screws -- a turnbuckle is a perfect example of this. One rotation, and even motion out both sides. This would eliminate the twisting.

Down side is that it would also remove the rotational aspect of the construct -- we'd be back down to 5 DOF, but that's more than enough for most purposes.

Thinking about it, though, I'm not sure that would work. It'd raise and lower the corner of the platform, but I think it would eliminate any side to side movements that were available. I think that two legs *have* to be independently adjustable.

Check out the motion in these videos...
http://www.hexel.com/video.htm

-- Chuck Knight

You could always build the base as a square and not a circle. The legs would then move along the sides of the square. Two legs per side, with the max height when they are in the middle and min when they are at the corners. Eight legs total (an OctoPod :D ). Moving in a straight line might also simplify the math.

But it would eliminate the rotational capabilities, as well. You're right about simplifying the math, though...

The genius of this design is not only in its construction, but also in its *integration* of the rotational movement into the overall design. It's not an "add on" feature...it's designed, from the ground up, to do that. Brilliant, really...an elegant solution.

-- Chuck Knight

i didn't actually mean use single struts..../|\ the two struts will still be holding the shape and the screw either a turnbuckle design on its edge( --- ) or a screw on the verticle (|) between the /and the \ , and you can still get the sixth axis with a rotory device attached to the bottom, if you have a picture of it i could draw what i mean.

double post ...sorry
remove me plz

Chuck, as cool as it is who is going to program the gcodes? I am sure there is nothing out there that can do it.

Eric

I’m not sure that the g-code would need to be any different for this table. It all depends on the controller software that interprets the code. For example, Mach2 ingests the g-code and produces the signals that go to the CNC hardware. So, someone would need to write a Mach2-like package that would move the table correctly. If done right, none of the g-code emitters would need to be modified.

I’m a computer programmer and am willing to donate time to this if someone is willing to build the h/w.

--bb99

the rotary hexapod at this site might give some hints.

http://www.prsco.com/r1000.htm

Looks like a geared ring on outside perimenter. The ball joints I was considering were heim joints. They are on ebay, for $2-3 a piece, so 12 joints is still not too bad. I think they would have enough dof for the movement.

I believe you will need to keep the center post along with a couple bearings that ride the perimeter to counteract the push-pull of the legs depending on the orientation of the table.

[edit]removed cable wrap issue

Another thing, any idea what lateral motion available, say for a 2ft dia base with 1ft legs, and a 1 ft dia table as an example. Most of these seem to have a very small work envelope even those on a knee mill. Need to make a prototype because thinking about this in my head is too hard.

I was looking through emc docs, there is support for a cable based hexapod which I would think would require modifications to work with this.

I do some programming too so I could help.

http://www.i-way.co.uk/~storrs/lme/HexapodSoftwareModel.html

I found this site with some software information that might be helpful.

T

Chuck, as cool as it is who is going to program the gcodes? I am sure there is nothing out there that can do it.

Good question. In fact, that was part of my original inquiry...whether EMC could control something like this. It does support hexapod systems, but this one is a bit unusual.

How is the EMC code structured? Does each machine have a "definition file" that gets loaded as a plug in to the basic code, or is it all just hardcoded in the main program? If the former, then it might not be too bad to add the capability to EMC...otherwise, we'd be looking at custom software.

Alternately, it's obvious that Hexel has control routines and software -- maybe they'd share them with us, or sell the software separately?

-- Chuck Knight

I’m not sure how I miss this after spending 2 days at the tool show in Chicago. I’ve been thinking about it since the first post. I doubt if it’s solid enough for heavy machining of metal but it might be fine for wood. I would be very interested to assist in any way I could. I’m over 50 so I can say it’s cool.

Gary :D

Chuck,

I don't know if you've given up on the rotary hexapod project but I know that the EMC code is modular to a degree and there are different definition files. For example there is one for a bridgeport (which I think uses 2 parallel ports for extra features) a generic and a minimill.

Brian

Chuck - it seems like you have given up on this? but heres some hints:


OH - the secret to using EMC to control a hexapod is getting appropriate GCODE generated.. i mean, a machine with not control is.. not that great.. SO> rainea's cnc toolkit does hexapod post-processing!

the math to figure this thing is called kinematics. its trigonometry. basically, its algorithms that let you use the angles of input joints to find the position in x,y,z of an end effector. in the case of this device, there are 3 2-degree joints that need the Kinematics\inverse kinematics, then there is the plate where the axis join - this all has to be modeled mathematically, which is doable because there is software to do it! its free! its public domain, and university supported. ive used it for 5 axis robot arm simulation. it can do hexapod simulation also.. theres lots of other software also. heres some links:

Hexapods:
http://www.physikinstrumente.com/micropositioningsystems/8_4.html
http://www.rainnea.com/cnc_hexapod.htm <--- Some basic hexapod info MUST SEE LINKS!
http://www.parallemic.org/ <--- LOTS OF INFO


Robot Kinematics Software:
http://www.easy-rob.com/link_left_1.htm <-not free
http://www.newtonium.com/ <---has free demo version
http://darwin2k.sourceforge.net/ <--- Check this one out. Freeware


General Engineering Simulation Links:
http://www.engr.psu.edu/ae/steelstuff/andesoft.htm

I wouldn't say that I've given up on it...I'd still like to build one, very much!

Having said that, it's definitely on the back burner, and has been demoted to a "someday, maybe" project.

I've figured out that, despite its amazing capabilities, its working volume is fairly small relative to its footprint. This makes it more difficult to justify the space, in my garage...er, um, shop. :-)

Of course, most of the "tough to make" pieces are fairly small, like connectors or jewelry, so a large working volume might not be necessary.

-- Chuck Knight

If you want to build a hexapod just because its an interesting gadget then the only answer is to do it.

But if you want a way of poking a router cutter into a piece of wood anywhere there could be simpler ways to do it.
start with a normal xy table
put a rotary table on top of it
put a "C" shaped frame on top of that.
The C frame would need to be 2 piece to get the z axis.
fit a router to the top end of the C frame with a 2 way pivot.

You could then go over or under an object in the work area or revolve anywhere around the sides.

And it could probably be done as an add on accessory to an existing machine just to sort out the principles

The programming would be interesting
As soon as you go rotary you need to use sliprings in the circuits.

That is the best idea Ive seen in a long time. If you didnt see the video of one in action it is a very small file worth looking at. http://www.hexel.com/hexabotvideo2.htm

I didnt read all of the posts so 4give me if it was mentioned. If the plate was mounted on a swivel ball fixed in the center underneath, then you would only need 3 actualtor devices to make it do the same as what it does in the video. Surley the actualtors wouldnt be hard to make.

About the comment that the math would be hard, if the math was divited into two parts I think it could be done. Imagine if broke it down like this. The rotation of the plate was divided into 360 degrees. You would tell the plate to face lets say 10 degrees and then tell it at what angle from horizontal it shold be.
The computer would not have to calculate how to drive the actualtors but go to pre recorded points to achive the result. I dont know how you would go about creating paths to move from one position to another. Hey lets face it, its not that dissimilar to a radar dish. The maths is out there.

Chuck, Is this what you were thinking when you said " Nested lazy susans"?

I cant draw well, but here is an attemp at a side view.

I guess it cant be called a hexabot, and must be called a tribot or quadbot if you count the middle support. I wonder if it would be as stable though.

strange.

i posted a message with key pieces of info on a-figuring the math for this thing b- free software to help c- a free solution to generate hexapod gcode.

no comments...

thats ok... apparently its easier for some people to type than to read.

Sorry Vac, dont be upset, you are right, I didnt read very closley. Anyway I was talking crap. I went to bed and thought about what I wrote and was completely wrong with both my pictures and and math theory. I didnt take into regard X Y Z movment and only was thinking about the pan and tilt of the platform. Although if the mill did the X Y Z part and the table did pan/tilt movement it may work.

Hey Vac, your my favorite member in the forum. I like your enthusiasm, and LOVE your pictures. Dont be upset :) I didnt think Evil Robots had feelings anyway !

vacpress,
Your information was not lost on everyone. I am very much interested in building a hexapod. I am however trying to finish a standard 3 axis router to put to work to finance a hexapod. So unfortunatly right now I am absorbing info and planning my attack stratagy. I have found for me anyway, is I like to do a lot of lurking and absorbing, not much talking. I will however keep your info in mind and will be posting to get help later as I begin building. Thanks to you and all the others who post. Thanks Ron

yneeb- thats not true exactly - you at least attacked the problem - did some drawing... thats productive. i think that if you spent 2 or 3 hours reading those links from the rainea page, you, of all the people here, could prolly rig a working example.. without even learning trig, oh yes. and you will do it in 2 months, and on the 3rd month make your first grand off it. :)

im not mad.

i just spent 10 minutes digging up quality links, and was hoping for a discussion that had moved from general "i want to build a hexapod" to more pointed "based on the 6-axis interpolation algorithm, i hope to use the cnc toolkit to control my 6 stepper-driven linear actuators. i think i am going to model it first in roboworks, to get the inverse kinematics algorithms" or at least

"what are inverse kinematics?"

i dunno. its all good.. :)

cuz i love this topic, and would very much like to see someone actually build a DIY hexapod that works, here is some gathered info on how to do it. something most people dont know is that the 6-dof "parallel kinematics" platform was first used industrialy as far back as 1930 for carnival type rides.. just like modern simualtors.

http://www.rainnea.com/hexapod_labeling.gif

This diagram represents a typical hexapod system. while not exactly simple, the hexapod is a doable project, even for a beginer. the trick is to start with a breadboard. dont go for a real machine first!!! I will try and outline howto build a totally makeshift hexapod frame to test ideas on.

http://www.i-way.co.uk/%7Estorrs/lme/Drawings/2-1.gif

this too large image is a screw driven lienar actuator.

OK. heres my quick 2-cent description of how to make a hexapod.

1 - goto hardware store, get : some PVC pipe, and some jsut a bit smaller. you are going to also get some threaded rod to put instide this and make a nut-driven linear actuator that looks basically like a car suspension strut. hell, use car-suspensionwill be.struts. just not for this quick test version.

. anyways - the more identical these are, the more accurate your device .anyhow. the hard part is:
2 - take your 6! steppers and hook em to your 6 PVC struts.
3 - now you will have 6 identical struts that extend or contract via step\dir signals. great.
4 - hook these to a triangle of plywood or whatever, basically like this:
http://www.mel.nist.gov/photos/hex/hxpro1.gif

5- these joints in the picture are ball joints... use rubberbands or something to make flexible joints. remember.. this test unit is a proof of concept.

6- get rainea's cnc toolkit, and put your hexapod geometry into the software. if you cant get 3Dstudio to use with CNCtoolkit... i dunno what to say now.. get it. "borrow" it. cuz ****, this isnt a comemrcial endeavor.. your not gonna hurt discreet..

7- get EMC, or other 6-axis cnc controller software. load cnc toolkit output.. enjoy watching your machine follow paths... start imagining the next version.

check this out:


www.amm.mw.tum.de/Forschung/forsch/arbgeb/entkop/entk_hex_d.htm

its in german - but it shows an example of an easily DIY hex-design that could be made with your guys current mdf machines!

make it of alum. and mdf!!! DO IT! i dare you.

http://www.amm.mw.tum.de/Forschung/forsch/arbgeb/entkop/hexapod.jpg

Vac and others,
rainea's program works just fine with gmax from discreet and it is free. I am just now tinkering with it and personally think I am going to spend my time learning Rhino. Gmax has the edge for gaming and you have to use it for the 5 axis program of Rab's I think because you can animate the part within Gmax. It can be done in Rhino but you have to buy the plugin.

Mike

im sure you can model in whatever, and import to gmax, or 3ds, or 3dstudio VLZ, or plasma.

as for animation... does it use the animation features? i wouldnt think so. but then, ive never used cnc toolkit.

what does this plugin do? is it a cnc toolkit version? or does it let you use 3ds plugins with rhino?

Vac it won't work with Rhino, aleast not to my knowledge and Rab's web site. It will work with the two Discreet products, and one from another company. I am not sure about the animation but Ithink that is how he gets the spindle to always stay perpendicular to the surface.

Mike

Rab if you happen by let us know.

That was my hangup -- cheap linear actuators that actually worked. Thank you for the idea.

I may have to prototype one of these, now that I see how simple it is. I'm thinking CPVC pipe, because it's got a small diameter and is commonly available.

Why ballscrews? Since error is not accumulated in parallel designs, cheaper components should be able to be used...

-- Chuck Knight

And, for this I am eternally grateful. It was very helpful, and has reawakened my interest too.

Actually, I'm not sure whether I should thank you...or not! ;-)

[quote]"what are inverse kinematics?"

This is a bit of a guess, but...

Kinematic equations determine an X-Y-Z position from a set of positions for the 6 legs. Inverse kinematics probably is used to determine the movements necessary to result in the desired X-Y-Z placement of the table...the inverse of the equation?

Of course, I might have them backwards...

-- Chuck Knight

[I wonder if it would be as stable though.

The way you have it drawn, Ynneb, you have a triangular top and bottom, and 3 parallelogram sides.

A parallelogram is not a stable configuration. For stability, you need triangles...it's the only inherently stable configuration.

If you analyze the traditional 6 legged hexapod, it's nothing more than a simple octahedron. Triangular top and bottom 180 degrees out of phase, and with the corners connected.

Other stable configurations are found in the Platonic solids...the tetrahedron (less useful), the octahedron, the hexahedron (cube...not stable), and the icosahedron. They're the regular Platonic solics, and they'll work. I'm only including the regular, non stellated solids, for simplicity's sake.

The octahedron was chosen because it presents a flat face at the top, which can be moved in 3-space. The tetrahedron and icosahedron both have a single point at the top...the tet might work if you held a cutting tool at its vertex, and the icosahedron just shows a LOT more complexity for little gain, though it would be a workable configuration.

-- Chuck Knight

[quote] The way you have it drawn, Ynneb, you have a triangular top and bottom, and 3 parallelogram sides.

A parallelogram is not a stable configuration. For stability, you need triangles...it's the only inherently stable configuration.[quote]

I still think my idea could work. (Forget the lazy susan bit) Did you consider the support post in the middle? Doesnt that make it a tripod? The middle post is fixed and solid. I am trying to understand your thinking but cant. I still recon it would work.
I am happy to be wrong, but cant understand how I am in this case.

Ok,
Here's what is in question for me in the Hexapod scheme. Zeroing the machine. I can build one mechanically. I understand how I can do that. EMC will control it on a linux box no problem, maybe a few little challenges but solvable. What I want to know is How do we zero this thing and how do we measure the legs accurately enough to work out the kinematics so they are accurate? Remember most of us are working with basic tools and measuring devices. Please enlighten me as I am in the dark. Ron


edit: I think I even have the ball socket challenge figured out, need to test my theory.

I'll let you guys handle the math. :) I've been thinking about this a lot, maybe WAY too much, but it seems
to me that all the ball joints I've seen aren't gonna do the trick. There are probably ones out there but like I said I haven't seen them. Instead of ball joints how about 3 concentric rings? The inner most fixed to the arm and can pivot up and down. This is fixed to the next which can pivot side to side, which is fixed to the third that's fixed to the base or the table depending on which end we are talking about. If you don't understand my description, think about the old ships and how they held lanterns on the wall whilst counter acting the ships rocking and rolling.

There is a name for these things, but it escapes me at the moment. So if we are talking easy and cheap I think these might be a good idea. Just to see what the table travel will be I think I'm gonna make a wooden mockup with some dowels and some scrap ply I have laying around.

Brian

BrianS- What you are thinking about is a Gimbal Bearing, where all of the pivots hav one common center. I think that a gimbal bearing would work, but would not be nearly as stiff as a ball and socket. Speaking of the ball and socket, is there any reason that a spherical rod end (heim joint) could not be used? I believe that is what is being used on several of the commercial hexapods that I have seen. The other type of joint that I have seen are nested ball joints. Where there are two balls in one socket. There is the outer socket, and then one ball that is hollow, inside that is the third ball. The only problem is making the balls! I had seen one large hexapod machine during construction, and they were using this nested ball design. I could model it up if anyone is interested in seeing what I am taking about.
Keith

Thanks Neatman I knew they had a name, just couldn't think of it! I could be wrong but I don't think heim joints will allow the amount of travel needed. Also they all apear to be a 90 degree offset and we would want 180 degree, no? What we want is like the cheap swivel vise ends with a ball in a socket, like
the nested ball joints you mentioned. (someone else mentioned this before, I'm not trying to take credit from anyone)

I'm sure the ball joints would be stiffer and a nicer solution but, if I couldn't purchase them personally I don't know how I would make them accurately. I was just trying to think of the simplest solution although it may not be the best.

Brian

i would thing 1/2 - 3/4" or bigger threaded rod would work fine for the leadscrews in these actuators. the thicker, the less other support will be needed. the fun thing about this design is its modular. get 1 strut built - build 5 more. 6 old HP LJ 2 printers can provide the steppers, and a gearbox.. so no tooted belts to add $$.. that would prolly be 5-10 per printer. ive heard there is a way to use the driver also, which has a built in powersupply.. havent gotten it to work though...

i saw a $5 harbor freight part that would work for the "balljoints" its called a "u-joint". these would be similar to the ones in that last machine picture i posted.

2 per strut = $60... not TOO bad..

http://www.harborfreight.com/cpi/photos//36800-36899/36846-t.gif <--- 2 for 8

http://www.harborfreight.com/cpi/photos//41500-41599/41569-t.gif <--- 15

hmm. couldnt find the exact one i was looking for.. odd.

I'd thought about using a U joint, but there's usually a lot of play in them. What type of accuracy would be required of them? I suppose the lateral play wouldn't matter, so long as they "rotate" smoothly about the pivot point.

Hmmm...might be interesting!

-- Chuck Knight

I think a gimbal bearing could be quite workable, experimenting with a pen on my desk it looks like you only need two degrees of freedom ie the connecting rod doesn't need to be able to twist. You could use a block with some recesses for rollerskate bearings in the middle of the joint with the base and rod connected with U arms arround the sides.

Block sort of like this

tachus - beauty! that is very simple and elegant.. assuming my logic is right.... it seems doable..

after i get the paitner going, i may build a hexapod prototype to try cnctoolkit.. very curious how that software performs.. will help me decide if my planned 2-3k router table should be more than 4 axis... i know i want 4 from the start.