Testing: Cycloidal vs Harmonic Drive 3D Printed Reducers

over the last few weeks videos which you can find in my channel i produced a prototype cycloidal drive reducer and also a strain weave or harmonic drive reducer i previously only tested these with a sensorless skateboard esc which made the motor turn but it only uses the back emf from the coils to work out what the motor position is this makes it hard to apply holding power to the motor and if the motor stalls then torque drops significantly my version 2 cycloidal drive reducer uses two cycloidal discs offset 180 degrees from each other this reduces vibrations and also helps convey torque more efficiently i did some tests with the skateboard esc just lifting some mass at 100 millimeters from the center of rotation which was relatively successful but it doesn't realize the motor's full potential due to the use of the skateboard esc the cyclodal drive is all printed in pla it is a 10 to 1 reduction which means that it can be back driven this is required to give dynamic robots such as my open dog version 2 some natural spring in their legs my harmonic drive reducer uses a flexible spline 3d printed in tpu but the rest is also standard pla i found that in my version at least this was less efficient than the cycloidal drive it's still a 10 to 1 reduction so it can compare with the cycloidal drive but i could stop the output of this reducer with my hands at least when using the skateboard esc it is also not back drivable which limits its use in dynamic robots this time we're going to do some better tests and compare the two reducers we really need to use a motor driver that's aware of the motor's position so we can apply power to it effectively so this time i'm using an o-drive 3.6 with a 6-cell lipo the o-drive uses an encoder on the back of the motor so it can be calibrated and knows what the motor's position is this means it can energize the motor's coils at the right time the o-drive will allow both position and velocity control of the motor and i've used these in many robots before including open dog version 2 and the really useful robot let's test the harmonic drive first so i've screwed that to a piece of wood and clamped it to the table we're using the o drive and the encoder to control it so it should have more torque than the skateboard esc but it appears i can still stop that with my hands if i'm very careful which isn't right at all even open dog version 2 with 5 to 1 belt reductions is more powerful than that i really couldn't hold back the dog's leg with one finger as i'm doing here with the screwdriver even though i've got a lever so that doesn't seem right at all so that's not right i shouldn't be to stop that with one finger when it's a ten to one reduction and we've got the same parameters as the open dog motors which are only a five to one belt reduction and the whole thing stands up and walks along so there's probably something wrong with the efficiency of my design there some way the teeth mesh or just flexing that flex spine or something's causing too much friction so i was much happier with the cycloidal drive version two when i made it so that's the one we're gonna test next the cycloidal drive was much smoother moving and also back drivable which means we can use it in dynamic robots again i've put the encoder on the back so we can use the o drive to drive the motor properly so we've got a very similar setup i've already attached a lever to this one because i know it was already more powerful with the skateboard dsc and yep there's no chance of me stopping that with one finger so that just seems much more powerful altogether the other one was really easy to stop for some reason but this has a lot more torque in it and it's still 10 to 1 and still of course all 3d printed in standard pla and if we try and hold position with the o drive in the encoder then we get a virtual spring back driving that whole gear box and back driving the motor with the motor driver trying to hold the position in place we can see that's actually back driving the whole drivetrain if you see where my finger is you can see the motor moving as i move the output shaft so that works really well for some proper torque tests we're going to hang some mass at 250 millimeters from the center of rotation which is more than double from the previous video to start with i'm using an 8 kilogram kettlebell which we previously did at 100 millimeters so not too many problems there but let's move up to 16. so it won't lift it but we want to find out what the stall talk is of course i can't lift 16 kilograms with one finger so i think we've found almost the sweet spot either here or slightly less an eight kilogram mass hanging on a string if you want to lift it up that requires 78 newtons of force in the earth's gravitational pull so a quarter of a meter that's going to be 19.5 newton meters of torque which is what we easily get from this gearbox obviously with 16 kilograms it's going to be near a 30 so we're probably somewhere in the middle there realistically for stall talk around 25 to 30 newton meters so 20 newton meters is 198.8 kilograms centimeters if you want to compare with hobby servos and obviously this is an 850 watt brushless motor though so we would get a lot more torque but we've only got a ten to one reduction so with the higher gear reduction we could get a lot more torque but we really need that ten to one so we can back drive it and we can make a springy dog leg which is the entire purpose of developing this so we now need to do some more comparisons with an actual dog leg and i did these tests with the five to one belt reduction when i started developing open dog version two and now we're gonna do the same test again with this with the previous test leg when i started open dog version 2 we used two 5 to 1 belt reductions i made a test leg set at 45 degrees the most i could push down on scales was around 9 kilograms and you can see the motor's back driving and the leg is bending quite a lot so this isn't the maximum that the dog could support in a normal standing position though we can apply more force to the scale with less torque on each of the joints doing this test i was able to push just about 12 kilograms down onto the scales and you can still see that the joints are back driving so we're going to do the tests with the cycloidal drive and so i've made a virtual leg which is 250 millimeters from the end of the wood and 250 millimeters to the foot and this is the same as the open dog version two dimensions and i'm setting that leg at 45 degrees to start with we're using the o drive to hold the position again and at 45 degrees i can push 17 and nearly 18 kilograms down on those scales with the leg much straighter it's much much more around 25 kilograms of force that i can push downwards on those scales so pretty much double the previous tests so that seems pretty good we could reduce mass from this gearbox as i said when i developed it by taking all of those bearings out and putting in bushings as well so we could reduce the masses so we wouldn't be doubling the mass of the robot dog we would be doubling the force and the torque that the joints could withstand so i don't think there's any issue with this really it'd probably make a more powerful dog that's less wobbly and that can walk much better but we do need to make sure that the dog can jump on the ground a lot without smashing the gearboxes to pieces and that's why belts are good because the worst is the belt skips and nothing bad happens and the pulleys can easily be 3d printed in pla so we do need to do some more resilience tests for this to check it's actually going to last before i make 12 of them and try and make another robot dog [Music] [Music] [Music] [Music] [Music] [Music] um [Music] [Music] i went quite some distance like that but the footage of one wheel going around isn't very interesting but it does seem still to be running okay so let's open it up and have a look at the drive so yes that wheel was a bit wobbly but that's down to the molding of the wheel everything else is running true on the hub that i put on to attach that to the reducer so let's open up the gearbox and everything still seems to be running absolutely fine and there's no sign of any damage or any melting and bear in mind everything is still made of pla in this model let's take out that top disc and have a look at the other one and everything seems to be perfectly intact so i'm pretty happy with the way this was designed and i think those bearings have helped a lot i'm pretty happy with the robustness and the talk that we get out of this design and is just 3d printed in pla so thanks to 3d fuel for the pla for this project the orange is 3d fuel industrial pla and the yellow and the other colors is just 3d fuel standard pla and obviously we can upgrade to pet g or some other material or 3d printed nylon or laser-cut something or cnc aluminium for those discs or anything else if we need to make it stronger in the future but 3d printing makes it really accessible and i'm going to publish this design as always as open source and you can find that link in the description to this video to github so there are several improvements we need to make to this design though several of you have pointed out the cyclodal disks aren't the right shape and that's because i just approximated them by putting lots of cylinders in a circular pattern kind of rounding off the edges check out this video by levi janssen who actually explains how to do it properly and i'll be referring back to that video in future versions when i make another more refined version obviously this is full of bearings that makes it quite heavy and also makes it quite costly so i plan to replace those bearings around the outside with just bushings probably on shoulder bolts and that will probably reduce lots of the weight and if we use nylon or some material like the spacers i put in here just get lots of those and use those as bushings that remove quite a lot of mass as well so as i say this is open source and if you'd like to support me through patreon or youtube channel membership then those links are in the description below and patrons and youtube channel members can get access to all the videos up to a week early and sneak peeks and pictures of what's coming up to be part of that discussion alright that's all for now [Music] you

Channel: James Bruton

Views: 326,387

Keywords: Cycloidal reducer, harmonic drive reducer, 3d printed gearbox, 3d printed cycloidal drive, 3d printed harmonic drive, 3d printed strain wave reducer, torque testing 3d printed gearboxes, how strong is a 3d printed gearbox, strength testing 3d printed gears, strength testing 3d printed cycloidal drive, how much torque from 3d printed cycloidal drive, testing cycloidal drive, testing harmonic drive

Id: KodT7lQyYkk

Length: 12min 41sec (761 seconds)

Published: Tue Apr 13 2021