Changing Dynamixel to RC Servos (mechanical)

One of our first objetives is use other kind of motors than Dynamixels. In our case we are going to use HS-7954SH and HS-7950TH. (The electronic modification will be publish in other post).

The idea is follow the philosophy of the original project and try to keep the silhouette. Our first change is to create a external case for the typical dimensions servos (40x20x37mmm) and adding a rear shaft strong enough.

As you can see, the upper and lower case need be attached with four M4 screws. The distance between the two faces will be 42.5mm (more than the 30mm, so the legs needs to be modified). For the rear shaft we use a rivet nut and a flanged bearing, offering a second axis support!

The colours use in this design is only for showing the different parts, and all the components be easily 3D printed.

3 Likes

Hi again!,
continuing from the previous post, the base idea is to create two types of ‘arms’, one for he normal output axis and the second for the rear shaft. To easily connect it, we are going to use a hub (https://www.sparkfun.com/products/12543)

The full leg kit will look like this, more or less (it’s still a beta), but you can figure :

But looking the concept of @Alexandre and the ideas of @Matthieu from the post (http://poppy.discourse.group/t/special-foot-with-a-double-rotation/71) some modifications will be made :smile:

As you can see our idea is to obtain a 6DOF leg. Any advise about the stability of the structure?

2 Likes

It is a very good idea to create an adaptor to make the shape of the servomotor looks like the one of Robotis Dynamixel.
Also modifying the structure to just change the width should be rather straightforward.

On a project organisation aspect, I would advise to create a Git repository with your motors (similar to the Robotis_library I created) and in addition, only fork the legs repository to include your modification (and use the motor repository as a git submodule).

For the 6 DoFs leg, @Alexandre has already made a good one. It is not yet included in the leg repository so maybe I will take a look soonly (maybe this WE) as it seems very useful for you.
Afterward, you will have only to change the width parameters to fit to your servomotors and it should be done !

I did not find the drawings of this hub but it looks a bit different from the Robotis one, you may have to modify the horn fixation parameters too (normally it should be done easily).

It is a good idea this project, because this will reduce the cost of Poppy. Moreover, this motor is stronger than Dynamixel MX28

What is the configuration for the knee you use? Because there are two configurations, MX28 and M64. I believe you use MX28 configuration, but with MX64 configuration you have less modification for applying the Hitec servos

@Matthieu I have the 6 DoFs leg, so if you want, it is possible for me to give this part @poppy_esaii

We start from the MX28 knee configuration design, We have not yet find the MX64 configuration. @Alexandre do you have any links?

Thanks!

@poppy_esaii Here you are leg and thigh with MX64 knee configuration : https://drive.google.com/folderview?id=0B6T9dkX7VsL-flNpaWR0MTJmdFU3M3JCYm5meUtzaG1MQmc1WXlweXBJb1BOWUFuTEVtUlk&usp=sharing

Must say awesome try. :smile:
I really really like the way you guys going. This will actually pull heads in without cutting your pocket :smile:

Please keep us informed on more progress.

Regards
NicoX

Hello everyone! Just wanted to say hello. I’m designing a new motor for biped knee joints that will be about $100 in parts (maybe less) and will provide high torque and positional control. I’ve got the mechanicals for a prototype built but I need to get some controllers made. I’ll open source everything once I finish so that anyone can work on it. But my hope is to make a biped like poppy that is much cheaper. I would love to make it work with poppy but I prefer not to use screws in my designs so I may do it differently. Anyway just saw this thread and wanted to add that. It will be some time before I finish though so it’s great that others are working on replacing the dynamixels too!

2 Likes

Hi @tlalexander
Super like from my side.
I am also stuck in my development (currently gathering money) due to the huge cost of MX28/MX64.

As you were developing I thought I could share my findings such that it may be useful to you and to the forum as well.

  1. Can you consider AX 12 outer case for your new design. Reason behind is MX 12W (http://www.robotis-shop-en.com/?act=shop_en.goods_view&GS=1610&GC=GD080100).
    MX12W is with absolute magnetic encoder and we can use poppy framework. MX 12W has the same 3D dimensions as AX12 or AX 18.
    So we can get electronics with PID control in small form factor as AX 12/18.

  2. Problem is MX12W can be used only in wheel mode and the gear ratio/motor is such it can not take joint mode torque as needed in poppy. May be @Thot can add some points as he might have seen this actuator by person.

  3. @tlalexander If we can keep the design same as MX 12W (ie the case dimensions) but if we can use metal gears / motor suitable to generate torque as MX 28AT or MX 64AT, that might help. For low cost motor Faulhaber (http://www.faulhaber.com/) may be good I guess. One man has already replaced MX28 motor with it and made a Darwin OP clone :smile: (https://www.youtube.com/watch?v=RrdbYt8fs1w)

Regards
NicoX

Take care about mx-12w, we can use it as a joint mode but consider the torque is below 5kg.cm. I don’t know if we can cross easily an mx12 with an ax12/18. Design all the shapes for ax is very risky.

Actually I’m literally designing a new motor and constructing it from scratch, and is a high torque brushless motor design that requires my custom controller, which will also be open sourced. The motor has sufficient torque that I hope a 1:10 gear reduction will be enough to drive a biped leg, and this low gear reduction allows for the use of (large, wide tooth) printed gears. The driver is current controlled and uses torque-vector control for moving the rotor, so integrated hall effect position sensors (analog) are used to accurately sense rotor position to a few degrees.

The new motor with high torque and low gear reduction means it safely backdrives, and the cost of the system is reduced by not needing an expensive gear train. It is largely theoretical at this point but I have extensive manufacturing experience and have built a handful of brushless linear motors in the past using this controller so I’m not totally dreaming here. The first prototype motor works in static torque tests but the coils were a little undersized to I have just sent files for quote on a new stator design. The previous stator was 3D printed stainless steel but for the next version I am looking at waterjet cut steel sheet that I will stack. This will have better magnetics properties and will likely be cheaper. Ultimately I will want to have these manufactured in china though with the construction method I use it is pretty cheap to make in the US. The rotor is printed plastic.

I am doing a lot of research into 3D printed mechanisms and hope to help designers reduce the cost of their robotics through the concepts I am developing (which will all be open sourced). For example here are some slightly out of date pictures of the 3D printed remote control car I am building which already works extremely well. With chinese tires (not shown) it has enough grip (and power) to do standing wheelies on asphalt/tarmac!

I suppose I should upload some pictures of my current motor design as well…

Anyway, the complete “servo” design is going to be almost entirely 3D printed aside from some skate bearings, and the motor consists of a printed or laser/waterjet cut stator with hand-wound coils. Basically everything else will be printable on a home printer. Unfortunately I have a “real job” which interferes with my ability to design robots, but it’s a self-run business so I try to split time when I can.

5 Likes

I’m currently working on an adapter housing myself which involves removing the factory mounting tabs and the bottom plate of the standard hobby servo and placing it inside the new casing retaining the stock servo case mounting screws. Still in beta testing phase. For more Info contact me at cliffordcannon101@yahoo.com. will be posting part file and link here as well as thingiverse.com.

So far testing goes well the fit is near perfect. Sat on my bicep so further testing is needed after new part prints are finished.

2 Likes

The motors (mechanical and electrical) part are ready!

Next step, print the modified legs, and finally… control :wink:

5 Likes

It went ahead with the project in question ? what’s new ?

Hi @Rubex,

Yes it did! In 2-3 weeks I have to deliver the thesis, I will share the progress afterwards.

Dimitris.

2 Likes

hi, any news about this promising solution?

Heres another similar solution that uses LX-16A serial servos.

https://rockytherobot.com/