I have a question about the actual speed used and the ranges they operate (angles) for the 4 MX-64 servos.
The reason I’m asking is the following:
After reading the @poppyrate experience I made a few calculations and indeed it might seem possible to drive a scaled down version of Poppy with XL-320. So I’ve started to make the changes (in general the scale rate is roughly 0.72) with some surprising results: everything is very light. To give an example a whole arm (with 4 servos and all the plastic) is roughly just a little more than 100g. Compare that with the weight of one MX-28: 72g.
What you see in the picture is the upper part. If you notice the abdomen MX-64 are now replaced with the AX-12A. But I’m not very happy with this:
they are heavy (I never though I would say that…)
I can run them at 7.4V on the same bus with the XL-320 but I had to hack the cables and I don’t think this is for everybody
I’ll have to mix 1.0 and 2.0 commands which is not that simple
So here is the background of my question: I’m thinking to replace these servos with XL-320 but fitted with an additional reduction gear. I’ve found some very nice ones and I have quite a good option to integrate them nicely in the design. I have 2 options: 4:1 ratio or 3:1 ratio. The result of adding this additional reduction (apart from multiplying the torque) would be to divide the allowed angle and maximum speed by the same factor. So at a 4:1 ratio I will only get 75 degrees (300/4) or range and 28.5 (114/4) rpm maximum speed. The numbers for the 3:1 ration would be 100 degrees and 38 rpm.
For the torso I think the 75 degrees (ex. -37.5 to 37.5 range) is more than enough because these servos are unlikely to need something outside of that range. For hip_y it might be another story - although also I think this range would be enough for good walking.
But the main question is about the speed. In a real Poppy do you ever go above 30 rpm for these 4 servos?
Once I’m done with more testing for the robot I will post the parts.
For the time being I call (her?) Poopy Seed. It’s because it’s one of the smallest…
Hi @sonel, great work! It is impressive how you made the shrinked parts fit so nicely with the XL320.
Unfortunatly, I do not have much answers for you but more questions
Do you directly scale by 0.72 the parts from the humanoid and then modify the surfaces to make them XL320 compatible? How do you do that in practice (which software, which starting format (stl, step, …))?
How/Where do you print the parts?
Do you have links to the reduction gear? How would you attach them reliably to the motors and parts?
You can find all the parts on Onshape here. I hope the link will work. For the time being the file is a little in disarray. But all the parts and the assemblies that are at the beginning with a “small” in their name are for the modified parts. In general the workflow I follow is the following:
I remove all the holes and fillets that I don’t need (using a combination of “delete face” - this works good for holes, “move face” works for removing counterbore holes and “modify fillet”)
sometimes it is simpler if I get simply add a cylinder and merge it with the part
sometimes it’s easier to extrude the bottom of a hole up to the upper face - this works particularly nice when the upper face is not a flat one
when all the cleanup is done I use the “uniform scale” from the transform tool to scale to the size I need; the strategy I use is to adapt the part so that the horn to contra-horn distance will fit the new servo. For instance in the case of the forearm the original distance is 41mm to accommodate the distance between the MX-28 horns. For XL-320 the distance is 30mm (if you include a 3mm bump that mirrors the horn) so scale will be 30/41 = 0.7317. After scaling check that the distance between the holders is 30mm.
then I start adding the features to support the new servos; mostly holes. Sometimes I need to move faces to make enough material so that I can place the holes where I need them
The printing is done on a very cheap Chinese-made 3D printer branded CTC like this. I used it for about a year now and understand how you need to treat it. I always print in ABS and learned the hard way not to get cheap on the filament. I use now one that is made in the Netherlands (FormFutura).
I do a lot of filing and sanding after print to make it to my likening. Sometimes I also use the acetone vapour trick (you put acetone on some paper towels and line the inside of an air tight box, then place the parts in the box without touching the walls and leave it for 10-15minutes, depending on the temperature. The parts will look smooth and shiny and, most importantly, the process will improve the binding between the layers and will make the parts even stronger).
The gears that I ordered are branded by RS and I assume are widely available in other countries. For instance the 48 teeth gear is this one. If you search on RS Pro POM you should get a whole range. I’m using the 0.5 pitch (metric) form and ordered the 48 and 60 gears and the 12, 15 and 16 pinions. An idea about how they come together is for the 12/48 combination in the assembly “check_geared_abdomen” in the Onshape document.
Thanks for the detailed explanation! The onshape link works, it is really interesting to see your work in detail.
The quality you achieve with this filament printer is impressive. The postprocessing makes it look like SLA technology, that is why I asked
About the gear system:
You plan to glue the pinion to the horn_adapter?
You plan to drill holes into the Gear_48T to put Ollo rivet?
Given the substantial torque they will have to hold. Aren’t you afraid that the gear_adapter part ‘bends’ and lead to misalignment of the gear and pinion? My concern is that it is attached to the XL320 only from one side. If everything is tightly packed by the abdomen_small it might all work fine though.
Looking forward to see a video of this creature moving!
To a certain extent the filament is good, but the rest of the secret is here:
It’s actually not that hard to give a very nice smooth finish to ABS and that sponge abrasive is a nice find.
Yes, but I also have made the hole in the horn adapter about 0.2 smaller than the pinion boss and I will press it in the horn adapter with this:
It’s a watch glass press (I have been dabbling with some watch repairs for a few years…).
You plan to drill holes into the Gear_48T to put Ollo rivet?
Yes, I will have to make holes in the 48T gear. There were a few other options of gears that came with holes but they used much larger pitch or were very expensive. So I decided it’s not really that big of a complicated thing to make 4 holes. But I will not use rivets in this case because there is not that much space for a 4mm hole so I will use 2mm Allan screws.
That is indeed a serious concern. I wanted to first try a simple modular approach. If that doesn’t work I might have to make a hole in the gear adapter and place a heat sert and then screw with a ball bearing - like the one in the idle horns of MX series - to the support in the abdomen (for the x axis) or the pelvis (for the y axis).
I hope the gears are coming tomorrow and I will be able to put it together over the weekend.
This is amazing, when I first saw it I though it was a normal Poppy. The switch to XL-320 is very well executed.
Concerning the AX-12 “problems”, depending on your goal, maybe it’s a good idea and easier to have legs with AX and upper body with XL-320, this way you just have two buses with DXL1.0/12V and DXL2.0/7V.
The use of gear looks tricky, it could work for the abs but I do not know if it will be efficient on the leg where you will need more torque and still large range of motion.
First of all, this is a very nice work !! and open-source… I think I will copy it since I need a smaller Humanoid robot.
Take care about the XL-320, they are impressive but can burn very easily. And when I see the lever arm of the elbow…
Concerning the “abs” articulation, the 75deg are really enough. I reach the stops only when Poppy “falls” on himself.
Concerning the speed, 30rpm is enough. If you go faster, you risk making shocks on the upper torso and it will be flexible moves.
Did you consider the AX-18 ? They are more powerful but more expensive… and make some “geiger” noise.
First of all, this is a very nice work !! and open-source… I think I will copy it since I need a smaller Humanoid robot.