Now I know about the Abacus Drive, thank you so much! As I understand it, it's a bit different: it uses a second cycloidal wheel on the wave generator instead of a round wheel.
@@MishinMachine Interesting! It would be cool to see a comparison between the two devices, including full characterizations of the drives. This solution seems a lot easier to build, so I might incorporate it into some of my projects!
You're the first guy Isee measuring backlash with that pointer on a wall, after me, many years ago :D (I used to do it at 10-12m distance, for greater acuracy) Careful: at 8:35, going one way, returning, then going that same way and returning you're not measuring backlash, but repeatability. To measure backlash you need to go one direction, return to zero, mark the position, then go the opposite direction and then back to zero, mark the new position and measure the difference. You might also want to measure the same under a relevant load. You will find backlash increasing with the load. That's called torsional rigidity and the smaller is the better. With plastic I imagine quite a load of it. On my Aluminium + Steel pancake system I was measuring 0.75 arcminutes under zero load (like you measured here) but 10 arcminutes extra per every Nm of torque added. Anyway, amazing work, especially explaining it and creating the parametric model!! That's awesome!
I think the backlash measurement was when he attempted to manipulate the actuator by hand. It would require a lot of care to automate a backlash test because any asymmetric load is likely to take up the backlash (e.g. the action of gravity on the mass of the laser pointer in this setup).
@@cooperised He's using a laser to measure, which is clear evidence he's perfectly aware that any external load can have in impact on the measurement, and that measuring at a distance is orders of magnitude more precise than the test people usually do (wiggle it like there't no tomorrow and concluding there's no backlash based on the perceived stiffness). So I don't think he was trying to Measure backlash when attempting to manipulate the actuator by hand, just to Assess if there's a gross amount thereof. Also, the laser pen is so light that the friction forces inside the unit are overcoming any imbalance the pen might induce, so that's negligible. Ask me how I know :D
This is great! It would be very interesting to see the actual load carrying capacity and the efficiency of these drives. Thank you for the great video.
This is a really really cool design! I love it ! The only thing I'm wondering is what about using mini ball bearings for the rollers? I can see that the roller cage is one point where it's sliding friction. It's probablly not bad, but seems less optimal. It is certainly not an option for small gearbox, but maybe an option for large gearbox?
@@MishinMachine Actually looking again at the video, I'm not sure how would you use the ball bearing as roller. I just realize the separator is not moving in the ecentric motion, which mean the bearing needs to go in and out on the seperator.
Nice video! It's great that you're including some testing too. ( I agree with the other commenter saying that for backlash you should approach from different directions) Have you seen @LeviJanssen 's video of a co-planar compound cycloidal gear? "The Compound Cycloidal Drive - Something Novel" would love to see you make a video exploring this idea.
I tried to test the load, but I haven't done it before and only had laggy Ikea scales. The results also really depend on the current of the motor driver. In my tests, the highest load was 1.4 kg on a 20 cm lever for a 20:1 reducer.
@@MishinMachine So around 28kg*cm or 2.8Nm? holding torque and active torque can be a bit different but if we say nema17's are around 0.2-0.3Nm So its somewhere between 50-70%efficient? Would love to see more tests ^^
wait what, you can use the outer surface of the bearing? to me that blows my mind, because up until that point my brain was still working in strain wave gearbox space!
What a great design, and an excellent vid explaining it! Thanks for the model too! The backlash seems quite good as it is, but I wonder if it could be even further improved by introducing a slight offset between two phases? That could make assembly difficult, but it might not be too bad if you assembled the two phases separately and then applied some torque as you were screwing the two parts together. This would increase friction and thus cost efficiency, but that might be fine in applications where zero backlash was important. With two sections 180 degrees out of phase with each other, is the rotation completely smooth or is there still some input/output nonlinearity?
Thank you! I don't quite understand how to offset the phases. But you can, for example, increase the size of the wave generator wheel a little and use precision hardened steel shafts as intermediate rollers. In my test I used nylon standoff spacers and a tiny offset in separator holes because my old 3d printer is not quite precise. Actually I haven't tested 180 degrees out of phase setup yet. Got carried away by another variant of this drive where this balancing issue should be resolved.
@@MishinMachine The slight offset I think he is referring to is; instead of 180degrees, make it 179degrees for example. On the second phase. This intentional angle offset is used for involute gears now for the same result. Because the phase will be "misaligned" by 1 degree in this case, the tooth will likely collide, hence "no backlash". But they only work well with relatively light loads and present other eccentric output speed problems. (unless you rotate the Cycloid profile as well by 1 degree, but that eliminates the offsets purpose). Which is usually a deal breaker. Realistically even if you try for perfect 180degrees offset between crank lobes, you wont get it. So the natural tolerances do this for you in a way.
What would you think about replacing the rollers with small bearings like MR62ZZ? I know it would be more expensive, but do you think that would have a significant performance gain?
Why not. But in this case you also need to use output pins setup like in cycloidal gears instead of a separator. The diameter of these pins should be less than the inner bearing diameter by the amount of eccentricity. So with MR62ZZ bearings and 1mm eccentricity, the pins should be 1mm in diameter.
Can you share the reducer suitable for the 42-step motor in the video? It seems that the onshape software cannot modify the reducer into a square one. Thank you very much!
Interesting design! But since the rollers have to move radially to follow the squiggle, I don't think there's any way for them to push tangentially on the output carrier without sliding friction. That will probably be your ultimate torque limiter. I wonder how it compares to a cycloidal of equal size.
Yes, I tested them. It was my second option after the brass round standoffs shown at 2:39. You can try different options as long as they are round 😄. The model allows changing the diameter and height of the rollers.
I found a similar gearbox in which the direction of rotation of the output shaft is discussed, depending on the number of rolling elements in the gearbox: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-hf8vh8ee9hc.htmlsi=duANiOU8uvaQRpdP I checked, and indeed, if the number of rollers is one more than the number of arc-shaped grooves on the working profile, such a kinematic dependency is created where the output shaft rotates in the same direction as the eccentric. The synchronization of the direction of rotation of the eccentric with the output shaft is interesting, as well as the possibility of increasing the number of rollers by one more than the number of arc-shaped grooves to ensure their displacement on the other side of the groove crest. Could you make such a version of the gearbox? Thank you!
