View Full Version : How would you build a Dynamixel torso joint?

Pi Robot
04-29-2011, 11:18 AM
Hello Mechanics Gurus,

I am trying to build a torso turntable joint for Pi Robot that uses a Dyamixel servo as the actuator. I have attached a picture of Pi Robot's base with a 6" lazy susan lying on top in roughly the location I want it to be mounted. The rest of the robot would then be mounted on top of the lazy susan. The lazy susan is sitting on a 1/4" sheet of acrylic and there is only about 1/2" space beneath that sheet and the laptop that you can see below it. So I am thinking the servo would have to be mounted in the robot torso and connected to the top half of the lazy susan. Alternatively, I could add another acrylic layer above the current top layer if it makes more sense to mount the servo to the base.

I built something like this before using a 4" lazy susan which was too small (allowed the torso to rock too much) and I created some ugly hack for attaching the servo's rotation platter to one half of the lazy susan.

Anyway, how would you go about building this kind of joint? Would you even use this kind of lazy susan or are there better turntable bearings out there for this purpose? Would you drive the turntable directly using the servo, or use a belt and/or gears?

BTW, the robot above the turntable weighs about 6-7 lbs.

Looking forward to some ideas!



04-29-2011, 11:28 AM
How much rotation are you looking for and which dynamixel?

Pi Robot
04-29-2011, 11:30 AM
Good questions--I would like at least 90 degrees rotation to each side, so 180 in total. I have spare AX-12s lying around but would be willing to upgrade to a higher torque model if required. (Speed is generally not an issue.)


04-29-2011, 02:37 PM
VEX has just released this turntable


Pi Robot
04-29-2011, 03:20 PM
Thanks for the link! I actually happened upon that turntable a few days ago when I was ordering some stand-offs from Vex but thought it might not be a large enough to prevent wobbling with 6-7 lbs mounted on top of it. But it might be worth a try at $19.95. I'd have to mount a Vex gear cog to the dynamixel servo somehow and then figure out a way to mount the servo such that the cog meshes with that outer 60-tooth gear...


04-29-2011, 11:24 PM
Hi Patrick
I use a similar thing for my arm base (200mm dia) I just use a servo in the center and a bracket i made a spacer to connect the servo to the arm base the height from the bottom of the servo to the bottom of the arm base is 70mm when i redo the top half i will mount the servo on the arm base to reduce to height reqiured.
I do have some issues with the rotary base its not very stable i used blocks as slides to try and reduce the wobble of the structure (ball rollers would be better) the other issue is the rotation is not smooth may be due to the friction of the blocks and the offset weight of the arm.

The press steel and bearings type lazy Susan will always have a wobble. A 150mm dia thrust bearing running in grooved washers with the balls held in a race and the higher tolerances should reduce the wobble'
the other drive method is a timing belt that way you can change the gearing to increase torque


Pi Robot
04-30-2011, 07:51 PM
Thanks Peter--I had never heard of a thrust bearing before so that is a good lead. I'm looking for something as rock steady as possible since the entire upper body of the robot including arms and head/camera (Kinect) will be riding on it. And when it comes to inverse kinematics of the arms and PCL stuff with the Kinect, I don't want any unnecessary wobble.


04-30-2011, 09:41 PM
Patrick, you may find this implementation by Upgrayd (http://forums.trossenrobotics.com/showpost.php?p=44060&postcount=82) to be a useful reference.

05-01-2011, 06:58 AM
I think even with a large thrust bearing fast and out of balance loads will still cause a problem. A while ago i tried using a taper bearing (wheel bearing ) with a bolt to provide pre load this was more stable than than my 200mm lazy Susan bearing. I was thinking to combine the 2 lazy susan on top to transfer the load out and the taper bearing on the bottom to provide pre load to stop the structure from trying to lift with out of balance loads the draw back is an increase in friction.
while dining with some friends i seen a new type of lazy susan bearing its a ring 30cm in dia but 1.5cm *1.5cm bearing enclosure (larger bearings?) what impressed me was its quite(no bearing rattle) and stable very little movement in the bearing assy but heaver that a normal lazy Susan bearing (for some reason you get strange looks when you dismantle someone table before dinner :happy:)


Pi Robot
05-01-2011, 08:22 AM
Patrick, you may find this implementation by Upgrayd (http://forums.trossenrobotics.com/showpost.php?p=44060&postcount=82) to be a useful reference.

Thanks Stobs. I wonder what Upgrayd is using as the bearing itself? Or is that washer attached directly to the servo plate?


05-01-2011, 11:36 AM
You're welcome Patrick. It looks to me like he's using an aluminum thrust plate (attached to the servo), and the white Delrin as the thrust bearing itself - which the thrust plate nestles into. gdubb2 also had the idea of using a piece of Formica laminate as the bearing (suitably supported), as a low cost alternative, since it too would provide a low friction surface. Might contact either of them to see what they did/meant.

05-01-2011, 11:41 AM
Not sure how big Pi's base... but what about one of the smaller "Large Diameter Solid Surface Turntables" shown at the bottom of this page with McMaster Carr (http://www.mcmaster.com/#turntables/=c47upm). You could bolt the outer ring to the base, drill some holes in the inner surface to attach a servo horn to from the underside (the servo body would be rigidly attached to your base), and then attach whatever on top of that surface to build up the torso.


Pi Robot
05-01-2011, 07:09 PM
@Stobs: Interesting--it never occurred to me that you could slide two materials over each other like this without bearings inbetween. I wonder how much friction there would be if the two plates were larger (about 6" diameter) and carrying 6-7 lbs?

@Fergs: Thanks for the MMC link. Pi's base is an 8"x13" rectangle so the "Large Diameter Solid Surface Turntables" wouldn't quite fit (smallest appears to be 10" diameter). But what you describe is pretty much what I did before using a pair of acrylic discs bolted onto the upper and lower halves of a 3" square lazy susan like the ones shown at the top of the MMC page. Then I mounted an AX-12 servo to the underside of the lower plate and attached the horn through a hole in that plate using long AX-12 bolts to the top plate. The bottom of the robot torso then mounted to the top plate. The small diameter of the lazy susan introduced a little rocking (but not horribly) and those skinny AX-12 bolts allowed a little twisting flex when rotating the plates under load. It was also a PITA to line everything up to be perfectly aligned axially.


Pi Robot
05-01-2011, 07:28 PM
Look at the prices on these "Plain-Bearing and Extreme-Capacity Turntables" (http://www.mcmaster.com/#slewing-ring-bearings/=c4dtw1) at the bottom of the page. Yikes!

05-01-2011, 10:34 PM
I don't think it would be a problem for the Delrin at all, and maybe not for a Formica-based solution either - although I've never used Formica in such a way I know it can be a pretty slick surface.

I hadn't heard of Delrin before coming to this forum, but I know that some plastics do have a very low Coefficient of Friction (http://www.machinist-materials.com/comparison_table_for_plastics.htm). ... After doing a bit of web surfing for CoF listings as well as information on Delrin, I'll also consider using nylon (performance oriented automotive suspensions often use nylon bushings instead of rubber).

Before seeing Upgrayd's post I was considering a similar installation, but using a type of Teflon tape I came across back in the early to mid-90's. Unlike the white (plumbing) or yellow (gas) tape you can find at almost any hardware store, this tape was used inside of Xerox Docutech printers on some of the surfaces of sliding assemblies. I've tried sourcing it on the web, but so far I haven't found the product I recall the maintenance/repair technicians using - which was a translucent, battleship gray/brownish color and several times thicker than standard electrical tape. On the up-side Teflon has a considerably lower coefficient of drag than Delrin but, on the down-side the tape (that I recall) is significantly less durable.

http://www.maximumfloorsafety.com/bot.pdf (Formica vs. Teflon; and no, I haven't, as yet, read the pdf.)

05-01-2011, 11:23 PM
Yeah, McMaster can be a bit expensive and they do not estimate shipping, but living not far from one of their warehouses I get same day delivery :veryhappy:.

Not quite as stiff/strong as Delrin, but with much higher abrasion resistance than either Teflon or Delrin (and dynamic coefficient of friction much closer to Teflon) is Ultra High Molecular Weight Polyethylene. And at least when bought from onlinemetals.com, UHMWPE virgin sheet is less expensive than Teflon by quite a bit (shipping is UPS and seems to be about $13 at minimum for me).

05-02-2011, 05:54 PM
Some slick tricks:
Low friction plastic that is machinable, look to plastic cutting boards. Also if you just need a low friction surface you can tape to bearing surfaces with industrial tape, try the thin flexible cutting boards.
http://lh4.googleusercontent.com/public/d51QunHh-y3N0i9h8yNN0j44urWAt8_GXMnm_ayH8JL8r8CmzWk66v2MgPZ oD1LtTjaxvzuhtSkGwEeRixoCdJvzfad782DynSeJc7V8RbEJG cBOWSl5V4dnbNo7cBrEjhjZcxT2ja4AQURMSTpkoA



05-03-2011, 07:48 PM
Look at the prices on these "Plain-Bearing and Extreme-Capacity Turntables" (http://www.mcmaster.com/#slewing-ring-bearings/=c4dtw1) at the bottom of the page. Yikes!

Sorry, somehow I missed replying to this earlier. At pretty much a guess, I think lnxfergy was referencing the bearing assemblies towards the bottom of the page since, being square and having two mounting brackets already attached, they'd be a fairly straightforward install. On the other hand, the lazy-susan style assemblies at the top of the page will be more than capable of handling the six to seven pounds of needed capacity you'd mentioned earlier.

Personally, I'd go for a Delrin/sheet plastic with nestled thrust plate design if compactness was a higher priority than budgetary concerns, and I'd go for either a Formica/Formica, Formica/sheet plastic or a lazy-susan based design if the budget for that particular project was a higher concern. The weight range you mentioned wouldn't impact my decision on which way to go; I'd think the nestled thrust plate would be quieter; wear-wize it'd all depend on the quality of the components, and aesthetically I'd go for the Delrin/sheet plastic with nestled thrust plate solution. Ok, my fingers hurt now, hope you're happy! lol :)

05-03-2011, 08:01 PM
@jhertzberg: I'll definitely keep that in mind! I'd probably avoid cutting boards with a texture if I'm using a metal thrust plate due to wear-induced wobbling (depending on alignment precision requirements), but ...for a light-weight application one might be able to use the same material for the thrust plate as well? Hmmmmm, food for thought jhertzberg! :)

Pi Robot
05-04-2011, 11:27 AM
Thanks for all the great suggestions. Unfortunately, most of the McMaster turntables are too large for Pi's base which is 8" wide. I'll already have one of the 6" square lazy susans and it actually seems to have very little play. Given my inability to cut a straight line or drill recessed mounting holes into plastic sheets, I'll probably try the lazy susan approach again (which is what I had before, but with a 3" square). I'm guessing a belt or chain drive would introduce unwanted slop into the rotational position so direct drive is probably what I'll do again. The only trouble I had before was that the AX-12 servo seemed to have some overshoot when rotating the torso to a new position. I think there are some parameters in the servo config one can set to "tighten" this up. Or maybe I'll need a bigger servo.


07-18-2011, 12:04 PM
Regarding rotational position "slop:" are there any position sensors you could use to actually measure the position of where the torso is at? For example, you may think you've rotated it 45-degrees, but you've really only rotated it 43-degrees. Would there be a "physical feedback" mechanism you could incorporate to know an accurate position of your torso.

Food for thought.

Pi Robot
07-18-2011, 12:33 PM
Good thought. However, as it turns out, I'm not so much worried about positional precision as I am just simple mechanical wobble. For example, if I rotate the torso to move an arm/hand toward an object to be grasped, I don't want the arm/hand to overshoot and whack the object.