My New Linear Actuators are SO MUCH BETTER! | PDJ #21
ฝัง
- เผยแพร่เมื่อ 4 พ.ค. 2024
- In today's deep dive, we journey through the evolution of Jugglebot's actuators:
🔍 Recap of the Old Design: We take a nostalgic look back at our original actuator design, appreciating where we started and recognizing its constraints.
❌ Challenges with the Old Setup: While our old actuators did the job, they had their limitations. Discover what hindered us from reaching peak performance.
🔧 Unveiling the New Actuator Design: Change is afoot! I introduce our brand-new actuator design, optimized for speed, modularity, precision, stiffness, and strength.
📊 Putting Them to the Test: Witness firsthand how these new actuators outperform their predecessors in head-to-head comparisons.
📢 Quick Housekeeping Note: We've launched a Zulip site dedicated to Jugglebot's journey! More details in the video.
🛠 DIY Assembly Guide: For the makers out there, I'll guide you step-by-step on how to assemble these cutting-edge actuators for your own projects.
Your feedback has been a driving force behind these improvements. Thank you for being a part of Jugglebot's evolution! Let me know your thoughts on the new design and if you join our Zulip community. Happy juggling and building!
🔗 Links Mentioned in the Video:
▶️ Previous Linear Actuators Video:
• Designing High-Perform...
🔧 Brake Resistor / Regen. Braking Info:
Find a summary of my attempts and the current status here:
pdj.zulipchat.com/#narrow/str...
💬 Join our Zulip Community:
pdj.zulipchat.com/
🛠 Build the Actuator - Bill of Materials + Files:
www.printables.com/model/569606
🎪 Support Jugglebot's Evolution on Patreon! 🤖
Join the Jugglebot family today: / harrisonlow
00:00 - Intro
00:25 - Design Requirements
01:20 - Overview of Previous Design
01:49 - Pros and Cons of the Last Design
03:21 - How is the New Design Any Better?
04:13 - How are the Strings Routed?
05:01 - Performance Metrics
05:20 - Endurance Testing
06:24 - Speed Testing
07:08 - Precision Testing
08:46 - Strength Testing
09:27 - Performance Summary
09:45 - Possible Improvements
11:51 - Final Remarks and Zulip Update
13:16 - Assembly - วิทยาศาสตร์และเทคโนโลยี
🚀 Hey Jugglebot Enthusiasts! 🤖
I'm thrilled to finally share this update of the actuator journey with all of you! Your insights, curiosity, and passion have been pivotal in this progression.
If you were designing this actuator, what would YOU do differently? Or better yet, if you could see one feature or capability added to these actuators, what would it be? 🧐
Don't forget, if you want to learn more about the project, or to lend a helping hand, check out our Zulip site, here: pdj.zulipchat.com/
Please don't add music. I really wanted to see how this was made but just couldn't with the music.
@@MichaelMantion Hmm interesting. I'm in two minds about this. Do you think just making the music a bit quieter would be better? Or would you prefer just none at all?
Would love to make a design for you but i just cant do things for free anymore
Love the series and the focus on the process of basing the design in proper design requirements. Inspiring thank you, clearly you are on a great path with the project 👊👊👍
Cheers Martin!
For the bearings, I would suggest a delrin or Teflon roller. You can get a pack of these easily, marketed as 3d printer rollers. They would stay on the bearings better, resist abrasion, and have a self-lubricating effect with the tubes. For the synchromesh cable, you may be able to design a clamping plate similar to those used for bike cables.
Very good suggestions, cheers!
I think the bike clamping plate will probably be the final solution for this problem; super simple, small and light. Great thinking!
The wheels used by 3D printers to roll on 2020 could also be used. The wheels are really cheap. There are also different shapes. Maybe a concave shape is better for rolling on the carbon fiber. You can buy them with a groove. Or if you can control the spinning modify them with a static spinning drill bit and slowly turning the wheel.
Why not linear bearings? Recirculating ball type used by many machine tool makers for linear movement. Robust, accurate, hard wearing and cheapish.
@@seeker1015 Weight to stiffness ratio is worse (if you include the rails) than carbon fiber. More weight means: requires bigger motors.
Some of these have a Delrin "v" groove profile that would give you 2 points of contact per bearing on your carbon tube. This is the first thing that jumped to mind when I saw your printed sleeves....I would imagine that this would keep things located a little more positively (at least, short of a concave profile to match the tube radius) . You might be able to use 3 of the same on the captured end of the moving tube in place of the 2 guide rollers you currently have riding one of the outside tubes...basically the same thing you have holding that moving tube inside your printed end, but flipped inside out. That would both keep it centered AND prevent axial rotation given the 2 contact points inside the V. Anyway super interesting project, and very promising.
Is it just me.... or every time I watch video of this guy I say: "I really like this guy". Awesome project! very good idea and straight to the point!
Cheers for the feedback 😊
Every time I watch this guy I say, "I hate your perfect jawline and chin."
It's a great design. You should also measure the force you can apply in extension - the opposite of your weight lifting test.
Cheers! I did consider that but I couldn't think of a good way to mount the actuator to support a load when upright... I imagine it'd be fairly similarly strong in both directions, though?
@@harrisonlowyou could use string and pulley so that it pushes the string downwards and the string pulls some load. But youll have to have at least two strings so that side forces cancel out and dont deflect towards the string
@@TrueHolarctic That'd definitely work and I think I considered something like that but quickly abandoned it after thinking of the direct lifting test that I showed in the video. I'm not super concerned about strength and I would expect the two directions to be similarly strong so I didn't bother with testing the other direction 😛
@@harrisonlow an extension strength test would test buckling characteristics.
I love the cable routing. Flipping the motor 90 deg may be better and less complicated, but I love complicated mechanisms. It just looks awesome
Sort of an aesthetically pleasing Heath Robinson affair.
Indeed, if there's no strong reason to keep the motor in-line (or not use of a 90° gearbox), it could make things much simpler.
It certainly looks cooler like this though
Regarding the MBS of that string, you're going to get something closer to half of that once you tie a knot in it.
It’s also worth mentioning that some strings are stretchy. Using string with low elasticity may improve precision too.
FWIW, strength of ropes (atleast in climbing scenarios) are usually measured in newtons/kilonewtons. The fact that the string was given in Kg (assuming they didn’t do the conversion themselves) is a bit of a red flag.
I don’t think the material was mentioned in the video, (fishing line?) but dyneema, if they can find a suitably thin version, seems like it would work well here.
@@techheck3358 The color size and rating make me think kevlar. Spectra, dyneema, dacron and kevlar might all be good choices. You can put them in a woven dacron sleeve and tie a knot in that to get a strong loop on the end without loosing as much strength... Look up trick kite lines and bridals for that info.
@bkuker I agree, though this does change based on what knot you tie. For the string strength testing, I tied figure of eights, which (AFAIK) are fairly good with not severely diminishing rope/string strength. Interestingly the string actually broke in the middle, far away from both knots 🤷♂️
As for the string in the actuators themselves, this is a big reason why I wanted to avoid knots altogether. I believe the system is fairly good with having no "stress concentrations" on the string, and when it does brake (eg. from the motor not slowing down fast enough and the actuator ramming into one of its ends) it seems to be braking in different places each time; some of these breakages being in sections where the string is perfectly straight at all times!
@techhech3358 Interesting point re. stretchyness and precision! I've just been treating these strings as perfectly (axially) stiff, though you do raise a good point that this, of course, isn't true. Perhaps I should test how stretchy these strings are compared to synchromesh cable.
I should've mentioned in the video, but these strings are made of kevlar and are usually used for kites. The MBS/WLL is actually quoted in lbs and I see where you're coming from re. a "red flag" though I'm not sure I'd expect kite-hobbyists to care too much about the differences between weight and mass 🤷♂️
@@harrisonlow The strength loss of knots is a factor of the bend radius of the knot, knots like the figure-8 have a pretty big radius; but the way you're mounting the string also has a radius. I do think in this case it's best to avoid knots, so maximizing the radius of the mounting points will help. I can't tell where it failed in your lift test so this might not be the issue now but you could run into it later on. A good strategy is often to coil the rope around a perpendicular rod a few times before securing it. This coil forms a friction knot and will take up a lot of the force before it gets to the mounting point. Maybe add a large radius to the 3D print and then coil around axial to the carbon fiber rod?
Regarding terminating synchro mesh cables: You could probably create a part with an internal "thread" that matches the cable. Perhaps create a two/three piece part that can be pushed onto the cable and match its shape. As the pieces are forced together with the cable between, they'd pinch the cable and hold it tight. Then you pinch that part in a collared holder that prevents it from slipping out and provide pressure to keep the cable captured. Very hard to explain, but perhaps you get the idea. Not sure if it'd stand up to the tension, but it probably could. Let me know if you want me to elaborate and produce diagrams!
I was just about to post a comment like this. My Prusa 3D printer does a similar thing for terminating the axis belts. You push the belt sideways into a cutout that matches the profile of the belt, and it engages several of the teeth to keep it in place.
Interesting! I think I get what you mean and I wonder if I could use this for the spool end of the cable? Something like: a conical part with a cutout in the side that the cable slips into. The conical part then fits into a similar conical cutout in the spool. As the cable is pulled tight, the first part (the one the cable is actually touching) is pulled further into the conical cutout, compressing it and strengthening its grasp on the cable.
Is that along the lines of what you were thinking? I like the idea, and I imagine it'd be fairly straightforward to turn this into a mechanism that would allow for quite fine adjustments to be made to the tension.
Cheers for the idea!
@@harrisonlow I'm sure there are many ways - I'm thinking of a cylindrical "nut" where the thread of the nut fits the cable. A typical nut probably wouldn't hold the cable because the clearance needed to rotate the nut would make it sloppy. But two half nuts with the proper thread could be clamped onto the cable with a set screw.
Perhaps you could even tighten the cable by rotating the two half nuts in their clamping recess - though I believe you'd have to both pull the cable and rotate the nut to overcome friction while fine tuning. I'd design the half nuts to have some kind of flange to allow turning it.
@@mattiasfagerlund Hmm, I'm not sure I follow exactly what you're describing. Are you suggesting having the synchromesh cable be fed through a cylindrical tube that has a sort of internal helical cutout? The synchromesh cable would need to be twisted to feed through, and shouldn't pull back through if exposed to a purely axial load (no torsion)-assuming the friction between the cable and the tube is high enough.
If you want to share images etc. may I suggest posting to the Zulip Topic on the matter? I'll post a link as a reply to this comment (in case YT removes it). If that doesn't work, you can find it in "#General > How to terminate synchromesh cable" on Zulip.
Link to Zulip Topic for Terminating Synchromesh cable:
pdj.zulipchat.com/#narrow/stream/399279-General/topic/How.20to.20Terminate.20Synchromesh.20Cable.3F
What an excellent project update. I love the collaboration happening, thanks for sharing.
This is awesome! I'm so glad to see your progress!. This is awesome! I'm so glad to see your progress!.
*Old Design Issues*
- Bearings wore out quickly.
- String clamping led to wear.
- Assembly required hammering, causing parts to crack.
- Not modular; any issue with one actuator affects the entire system.
- Bowden tubes kinked and had slop.
*New Design Improvements*
- Strings internally routed, making it sleek and modular.
- Motor contained within the actuator for easy replacement.
- Clamping mechanisms simplify assembly.
- TPU sleeves minimize wear on bearings.
*Performance Metrics*
- Endurance: Over 5 hours of continuous runtime.
- Speed: Max 3.4 m/s.
- Precision: Less than 0.06 mm.
- Lifting Force: Almost 7 kg.
*Future Improvements*
- Use motor body as the spool for the string.
- Consider using synchromesh cable.
- Replace TPU sleeves with heat shrink.
- Modify bottom cap to reduce torsion.
*Community Engagement*
- Set up a Zulip site for better communication.
*Positive Points (Sorted by Importance)*
1. *Modularity:* The new design allows for easy replacement of individual actuators, reducing downtime.
2. *Endurance:* Over 5 hours of continuous runtime, passing 50,000 cycles with minimal wear.
3. *Speed:* Achieved a maximum speed of 3.4 m/s, meeting the design requirements.
4. *Precision:* Less than 0.06 mm, close to the theoretical limit of 8.4 micrometers.
5. *Lifting Force:* Capable of lifting almost 7 kg, exceeding expectations.
6. *Ease of Assembly:* New clamping mechanisms eliminate the need for hammering parts together.
7. *Reduced Wear:* TPU sleeves and internal routing of strings minimize wear and tear.
8. *Community Engagement:* Set up a Zulip site for better communication and sharing of ideas.
*Negative Points (Sorted by Importance)*
1. *String Wear:* The string broke during precision testing, indicating potential durability issues.
2. *Torsion Issue:* When fully compressed, the actuator has little resistance against torsion.
3. *TPU Sleeve Problems:* TPU sleeves can slip off and are not perfectly smooth or symmetric.
4. *Complex String Routing:* The design requires complex string routing, which could be simplified.
5. *Braking Limitations:* Difficulty in slowing down the actuators, indicating a need for better braking mechanisms.
6. *Assembly Cracks:* The old design required hammering, leading to cracked parts.
7. *Old Design Wear:* Bearings and strings in the old design wore out quickly, affecting longevity.
Wow! What a thorough summary!
definitely not openAI output.
Good boy. You paid attention. Have a chocolate fish.
@dine9093 tbh even if it were ChatGPT-made, I reckon it's still pretty good! I wouldn't think it'd be so effective just from (presumably) the transcript?
@@harrisonlow It would if it also used your previous videos transcripts. keep up the cool work though!
For the TPU on bearings, you can use a concave instead of a convex shape.
I'm glad you mentioned this! I couldn't think of a succinct way to include this in the video but that was actually my first approach to this problem! Unfortunately there was an ungodly amount of rolling resistance between the 6 TPU sleeves and the tube. For reference, the 6 TPU sleeves had a concavity that perfectly matched (within 0.3 mm tolerance) the profile of the CF tube
@@harrisonlow If you make a rigid sleeve around the bearing with a semicircular groove, I'm sure you can find an off-the-shelf rubber o-ring that will fit in the groove snugly and provide friction.
@@RobertLockhartMakesGames Fantastic idea! Very simple and would be easy to replace parts for. I'll get some O-rings next time I'm at Bunnings and give it a test. Cheers for the idea!
To ensure symmetry you could print your bearing sleeves in two halves.
Would it be better two have two o rings per bearing or just one centered o ring?
Neat concept and build. Look forward to seeing your updates. Cheers!
The TPU sleeve over the bearing will work way better if you design them to fully engage with the surface area of the carbon fiber tubes. Make them like a circular indent like doing a sphere revolve cut in your cad program where the sphere is the same size as your carbon fiber rods. It will also help hold them centered if you add a bit either side of the bearing on the flat side that helps lock them in, like a U shape that comes down about a millimeter from the outside diameter of the bearing towards the center of the diameter. Hope this helps. 👍👍
For your linear bearings, a lot of newer industrial bearings use metalized ceramics. Systems that used to be rows of ball bearings on hardened steel rails have become ceramic sleeves which fit snug to the rail. I don't know if that applies to carbon fiber tubing, or much for your project - but a good lool at linear bearings might give you some ideas for how to stabilize your tube.
I would even recommend going with some kind of graphite coating on the rods that can act as a dry lubricant. An oil or grease probably won't fit the application well, but can make a world of difference in longevity. Gears that run for decades with oil in them won't last a day without it. A little can go a very long way.
You could look at just trying a teflon sleeve or set of pads that form to the wall of the tube. Play around with contact area and pressure to get what you need. The resin binder for the carbon fiber is probably going to be easy to wear. You may want to consider adding or trying to identify tubing made with a harder material. A silicon dioxide coating may improve wear characteristics (or perhaps an alumina spinel - quartz/saphire coat).
If you want to go aerospace, you could try a sputtered titanium nitride coating. Not sure how well (or horrible) it would fuse - but if you're looking to go off the wall with it - that would be a direction.
For the string, you might actually look into amarid/kevlar. The nomenclature to use is "tow" when looking for strings of fabric material (same with carbon fiber) - I spent days losing my mind looking for a carbon fiber string because I didn't know that was called a "tow". Well - a tow is unbraided, so I suppose not technically a string/line.
You might look at fishing lines for off the shelf solutions and even talk to those communities about line routing.
Super glue probably doesn't stick well to TPU. Scotch Weld PR40 adheres very well to TPU. You could also print the TPU sleeves to be concave on the outside so they are self centering
This was a well thought out project.
One thing you can do rather than use brake resistors is have a common rail for your motor drivers, so when one motor is generating it can power the other motors on the rails, and this will tie in with regenerative breaking too
When adding heat inserts, as soon as you pull the iron away, flip the part over and press the face with the insert (and still hot plastic) directly against your workbench for 3-5 seconds. The flat surface of the workbench will keep the small raised ring from forming around the insert as it cools, so you won't need to shave it off manually. It will also help make sure the insert is aligned perpendicular to the face. Alternatively, you can press a small block of copper of aluminum (like a spare heatsink) against the fresh insert to accomplish the same task.
That's a *really* good idea! Thanks heaps for sharing it!
You should look at 1-1.5mm Dyneema rope for your cable, very light, very flexible, very strong, low friction, low elongation/elasticity and easy to splice to create loop or termination
Kevlar would be better as dyneema does exhibit some degree of creep when maintained under tension for long periods. Kevlar is absolutely stable. Also is it possible to arrange it as a continuous loop running back to the capstan (motor) where it would be anchored by a tensioning screw widget. Would you ease the termination clamp problem.
This is awesome! I'm so glad to see your progress!
Cheers! I'm very keen to finally get Jugglebot moving around and I finally feel that these actuators are good enough for the job at hand 😁
This is Friggin Awesome!
I once took apart an old photocopier and it had a steel cable moving the photocopier head around, so i think you might have some success with steel cables rather than nylon/other.
And it really did surprise me how supple the steel cable was, so there might be some specific type of cable for applications like this.
Interesting application! After the last video I actually picked up a bunch of steel cable and the sizes that are flexible enough weren't that much stronger than the kevlar line (and a lot more expensive) so I ultimately decided to stick with the kevlar.
Cheers for the suggestion 😊
There are v groove pulleys which will increase your contact surface area for the bearings. Doing the heat shrink around the surface will increase the grip while allowing some compliance. I would use spectra cord if you can since they are about as strong as you can find and very stiff. For torsional stiffness you would be far better off using a square profile tube rather than round. If you use v groove bearings you could run with 4 bearings, 2 on each corner which should give you plenty of stiffness.
hi dexterm, have you become a flat earther yet?
Great video by great creator!
Really happy to hear and follow your project
Cheers for the support 😊
Very good improvements! I am looking to make some actuators for a very different application, but endurance is important to me, keeping cost in mind. I'd very very keen on seeing you implement your planned improvements.
If you need to dump break power for a very short time with significantly longer time between breaks, you could use supercapacitors, as they will essentially be shorts when they're empty
Just what I thought too. As a bonus it reduces energy consumption.
Terminating the synchormesh cable can be done with a guitar string winder. (The part of the guitar that turns the string to adjust the tuning). This also gives a nice way to adjust the tension. I'm wondering why the motor isn't just turned 90 degrees instead of having the strings go through pulleys. I think regenerative breaking would be ideal and would allow your juggler to run on batteries for shows where power isn't available.
Making the motor wider would mean a more complex/flimsy swivel mechanism between the actuator and the base
Or make a long tin hole to thread the synchromesh cable through and then have a 2nd perpendicular hole through which a screw is tightened down to impinge upon and lock the cable in place in the hole.
@@Superkuh2 that makes it hard to adjust the cable
@chrisBruner Addressing your points in order:
1 - Good idea! I don't play guitar and this idea hadn't come to mind. I'll do some research into how these work and I'll see if I can make a similar mechanism at the small scale/weight that I'd need. Cheers!
2 - The main reason I didn't set the motor perpendicular to the extension axis is because I don't want the motor to be sticking out the side of the actuator. I have a gut sense that this would likely cause issues down the line once I put all 6 actuators together into a Stewart platform so I want to keep everything as "neat" and in-line as possible. I think using the motor body as the spool would still work, though, because I'd be able to keep the centre of mass of the motor approximately coincident with the extension axis.
3 - I agree. I love the idea of Jugglebot running off batteries and being fully portable, though I have precisely 0 experience with these and don't want to find myself in the position of having to pause work during troubleshooting in order to charge batteries (not to mention any physical risks/dangers of high-power batteries)
@Superkuh2 I agree with techheck here - the synchromesh cable needs a fairly clear path to be pulled through and I think a convoluted tensioning path would be tricky to make fine adjustments to. FWIW one of my patrons had a few great suggestions that I've written up on Zulip. I'll post the link as a reply to this comment in case TH-cam takes a disliking to the external link. (The thread can be found in General > How to Terminate Synchromesh Cable)
Cheers for the suggestion!
This is a cool project and the discussion gets me thinking. Great job. :)
For terminating the synchromesh cable, you could use a metal ring attached to the end of the the rod/the rod stop, and simply thread a small hole for a screw that then clamps on the cable. Im having trouble explaining it, but if you google crossfit jump rope, it's the little screw at the end that I'm talking about. It should be strong enough with simply the clamping pressure, but you could always allign the screw up against the black coiled wire. I hope this helps. Fantastic work btw!
Two things you didn't consider when you thought of using rotor as spool: it'll have way worse precision and way less torque, by a factor of 3 it would seem like.
I think the precision would be if anticogging were set up, but after doing that for one motor + ODrive I'm not too keen on to do it for all of the motors. It's a very finnicky process that occasionally just doesn't work 🤷♂
Good point re. torque! Cheers for the input 😊
This is such a nice design! I love cables in actuators, they feel like an underated technique!
It's been a treat watching this be developed! Some suggestions for future testing:
- Conduct endurance testing while the actuator is (reasonably) loaded, to better match end-use conditions
- Implement multiple load cases into the motion accuracy test (and record deviations in both absolute and relative positions with each move). Under a single constant load (or no load) the stiffness largely "disappears" since the parts' deflections under stress are not really changing. In the final application, the load will be constantly varying and this may have a large effect.
To improve strength, you could have 3 pairs of strings. You also have to be careful when considering cord strength. When you're pulling the bag up and down by hand, the jerk you're applying is definitely high enough that it could reasonably be producing tensile loading in excess of the rating. The literal weight/mass rating doesn't matter, you need to be thinking in terms of force, which will make it more obvious how jerk and acceleration can create very high peak forces. If you improve the motion controller to enforce jerk limits, it should improve the performance of any cord or cable, and also reduce shock forces on the machine overall. This kind of thing is being used in 3D printers to improve print quality, reduce ringing, and go faster with fewer consequences.
Haha I was wondering if anyone would call me out on my sketchy testing of the string strength 😂 I agree that my "jiggling" of the bag would increase the force in the line, but it felt like my jiggles were fairly small and (I would think) shouldn't increase the force on the line by a factor of 4 (MBS/bag weight).
As for the motion controller, I actually don't know if it's possible to set jerk limits with the ODrives. I'm currently planning to use trapezoidal trajectory planning (with velocity being the trapezoidally-controlled parameter) which does a pretty good job of smoothly moving between setpoints. It's yet to be seen how well this does when the setpoint is being updated at eg. 100 Hz...
Cheers for the input!
@@harrisonlowAt work we have a garage door, and you pull it shut with a cord. When I started this was generic 90lb nylon diamond weave. It's not hard to close but the way the guys pull on it, not even really a yank, was causing the rope to fail every month or two, it would just rip in half mostly at mid-span. I'm the knot person, boss asked me to fix it, I just did mostly the same thing but using 550lb rated paracord, and a knot that's better at distributing the load. It's been maybe 8 months, no problem even with spending hours a day in direct sun.
It's just that easy to generate high shock loads by accident. I assume your cord, due to being smaller and not diamond weave with a useless fluff core, has a shorter elastic range.
I'm not sure what cord you've got, but in the vein of paracord there's "1.18mm Micro Cord", a lot of paracord dealers have it. Supposedly it's "rated" for 45kg, but they're not clear about what kind of rating that is. It's some kind of weave, which may widen the elastic range and make it a little more forgiving, but whose to say how the overall stretch will differ.
Honestly the other commentors seem to have your synchromesh cord solved, I've never seen that stuff before, it's cool.
Edit: one more thing: all paracord melts, any cord you can fuse the end on can melt, they can melt from rubbing and from loading. Once they start melting the mechanical properties are out the window. You won't be able to feel this heating, the cord is just too insubstantial to heat your finger. The paracord family could easily just be useless to you.
@@Ziraya0 Haha I love that your work has a "knot person" 😂 What do you do for work?
Cheers for that writeup! Useful thought experiments re. woven vs braided thread and elastic range. I haven't really paid too much attention to the structure of the thread before! (FWIW, the kevlar string I'm using is made from two twisted threads)
As for melting: I mentioned in the video that I accidentally routed the prototype actuator incorrectly. This routing had the string going straight over a printed part at a rather aggressive angle. Cut straight through it! For better or worse, the string is far more robust than the printed parts so they tend to fail first.
@@harrisonlow You probably have enough space in this design to go up one or two sizes in kevlar line literally without changing anything else. The line I'm seeing isn't two threads, but it is 0.8mm, 45kg, safe working load limit 9kg. 90kg/18kg line is only 1.1mm, and 136kg/27kg is 1.5mm. If the synchromesh gets too complicated, this seller goes up to 476kg 2.6mm cord; at that point it's mainly a question of how big you may need to make the spools and how that'll impact the torque across the range of motion.
@@Ziraya0drop Kevlar for Dyneama thread. Size AA/30 thread is 0.28mm, holds 40+ pounds easily. It also stretches less than Kevlar. Thread is easier to track down, and that's all kite fancy kite string is.
Regarding the Pulliies..
you could turn the green "middle - direction" pullie such that its output guides the string on the tangent of the main pullie.. this would result in an really odd angle but would reduce the number of pullies. But I think three is much simpler, as you don't have to "think" about this odd rotation.. :)
Very good thinking! I wonder if it's true then that any two "skew" lines can be joined with just two pulleys?
I thought exactly the same.
@@harrisonlowyes any two lines in 3d space can be connected with two pulleys.
the way to think about this is choose any two random points, one on each line, and then connect a third line between those points. then you place pulleys at each of the points, tangent to one of the original lines and the newly constructed third line.
hope that makes sense
@@waxt0n Ah of course! Great way to think about it. Cheers for the explanation 😊
Awesome design, I am working on a project that needs a linear actuator and nothing I have found online will suit my needs. I cannot believe it never occurred to me that I could make my own, thank you for inspiring me.
very smart, thank you for sharing.
This is an amazing project.
you should be able to hold the synchromesh cable by having two small metal plates that clamp together with two bolts, when you tighten the bolts it will crush the cable in between the plates and hold onto it real tight, I am taking inspiration from how electric cables are held into some fixtures, its often molded into the plastic in consumer electronics so that when the chassis of whatever is screwed together the two mating surfaces pinch the cable to act as strain relief on where the cable is attached inside.
High tensile threads need real care in end termination; if you create a good load sync with a gentle bend radius you'll be able to get the MBS. Try using a boss at each end with a number of "dead" wraps which will distribute the load into the thread. If you use the thread as one continuous wrap you'll only need two ends as well. You can eliminate one bearing by aligning the bearing to the tangent of the drive drum as well. If you put a counter rotation feature in the fixed end you won't have to pay the weight in the moving end as well ;)
Cool project. Good job. 👍🏻
for the sincro-mesh cable put a helical cut into the 3dprint to match the synchromesh cable and a tab to hold the cable into the groves (tab could be grooved too but not required)
Interesting project. I'll look at your other videos.
Choice, and terminating of the cable/string: see if you can source nylon strand, or Kevlar thread.
As for terminating a metal cable, get a blade-style, swageless eye terminal and attach it with your bolt. Be careful with the weight of the terminal as it may affect the harmonics.
Great video thank you for sharing
I'm liking your design! Here's a couple ideas for ya. Change from the string to the cable as you mentioned. Set the cable tension with a adjustable pully. Put the first pully from your motor to be on a slot, and a screw on either side to adjust it to set the tension. Oh and the tension should be just tight enough, so when you pluck it you get a thump and not a strum like a guitar. Then the ends of the cable could be securely mounted, perhaps looped over a rod you have ran thru your carbon tube, then crimped, just like the common method for larger cables.
Use a collet block to terminate the synchromesh cable. The bonus is that a collet will draw it tighter as your cinch down on it.
For the cables, perhaps a wire rope end stop with set screws for easy adjustment, or a threaded cable stud that you could adjust the tension of with a nut.
To terminate the synchromesh cable the first thing that came to my mind is a two-piece metal clamp style endpiece with interlocking ridges...like a bike brake/gear cable clamp.
I think you could take the first bearing out of the blue string set if you align the central rotational plane of the second bearing tangential to the spool
Lock the synchromesh cable through three screws running in a line (more if needed), where each screw is tightened down in sequence with the cable at an angle. The cable is mostly straight, running through the screws. Spacing of the screws can accomodate the bend radius of the cable.
Great work! Thanks for keep us posted. I can think of many uses for this kind of actuator in robotics etc.
Cheers! If you ever use them in anything, I'd love to know 😊
I would absolutely let you know. They look so cool, it is tempting to create a project just to make use of them. I am thinking light weight quadruped robot. Or bouncing bipedal balancing bot. If they can handling gravity fast enough for juggling, they should be able to handle gravity fast enough for bouncing?@@harrisonlow :)
@@whitneydesignlabs8738 Both ideas sound awesome! Some things to keep in mind when drafting ideas:
1. These actuators have quite a bit of rotational inertia. If you pull the push-rod out (quickly) and suddenly stop it, the whole actuator twists around fairly aggressively (thanks to the rotational inertia of the motor). Thankfully I've found a very cool solution to this issue for Jugglebot (will be in a future video) but it's worth keeping in mind
2. I don't need to worry about the weight of the actuators, since I've got the motor (by far the heaviest part) right at the base, so it moves around as little as possible. I feel very confident that the actuators would be able to "jump" themselves off the ground, but you might not have a huge amount of leeway for the mass of whatever other bits are needed for the robot.
I love the idea of a quadruped, and it'd be handy to be able to scale this design down to fit a smaller robot. If you do go down this path, keep me posted on Zulip (link in description) and I'll happily lend a hand where possible 😊
Harrison as a Chartered Engineer I am blown away by you work, well done!
For the synchromesh cable termination you could consider the design in principle for locking nut tuners for guitars
On the TPU sleeves, you should make them concave instead of convex. This will allow the bearings to have more positive contact with the carbon fiber sleeves. This will make the system more rigid.
I actually originally had the TPU sleeves being concave instead of convex and they had an insane amount of friction (though they were a very good fit for the CF tube). Someone also recently had a great thought about this on Zulip:
"The problem with sleeves that follow the CF tube is that you will have contact points with the tube that are at different radii from the axis of the bearing."
Very clever way to think about it!
Regarding all the bearings, but more specifically the end cap alignment thing, I once took apart an HP inkjet and found that they use felt bearings. That's right! It was a piece of felt with a hole a bit smaller than the rod so it held firmly. there were metal rings on both side of the felt centering it around the polished linear rod. This helped keep the mass down, cost down and noise down. What I'm suggesting is to create triangular end cap with all 3 holes to prevent twist. It would be 3d printed and hold 3 pieces of felt-washers.
You would need thick felt (1mm or more) and could mass produce them if you created 2 hand made punches. Take a metal tube that is slightly smaller than the carbon fiber tube and sharpen the inside edge of it. Then hammer down on the tube against the felt to create an inside punch. Do the same for the outside punch. Make tons of them and redesign the end cap to encapsulate 3 felt washers and all 3 rods. You can get rid of the wheels and bearings this way and solve for that odd twist. Make the end cap organically shaped to reduce weight (like a semi-collapsed triangle, don't know how to describe it).
I'm unsure if this suggestion would also work for the primary alignment bearing block but try it if the end cap proves useful. You would essentially create a new block with many felt washers to center and hold firmly the center rod. It would be cool if you could drop most of the bearings to reduce cost and rotational inertia and wear on the shaft! Good luck.
BTW, I couldn't create an account nor use the Google signup... Anyway... On that previous video I was the one to recommend the synchromesh wire. Thank you for covering that in your video I feel honored and excited to continue to help!
Here's an idea for terminating the syncromesh cables: route the cable around an eyelet and crimp the end onto the cable. That is a cable terminal that will not break (the cable itself will break before the terminal does). Then you can use a hook screw and a hex but to tighten the cable as needed.
Totally Cool Project and Approach to the Solution. love these mechanical solutions.
5h running time sounds very little
Beading cable might be a good option between the synchromesh and the thread (depending on the diameter you use). It's cheap as chips too and I've used it running a toy cable car across a 10 metre run, but there was some slop in that length.
It might also improve the accuracy a bit as well as it should have less elasticity compared to the thread.
Randomly suggested video and I'm def subscribed.
Cheer for the support 😊
In terms of the slowing down, you may want to look into powered steering, a technique called drive by wire. It's essentually adds pressure from one motor that is generated from another to create a force. One motor controls the other. You also have to remember moving mass is heaver mass in terms of the string. Terminating the line you can drill a hole in a screw and feed it though and tighten, or use a couple of washers.
This drilling-a-hole-through-a-screw idea keeps coming up! I'm very surprised as I'd never heard of it before. Have you used this termination method before? What for?
So cool! Great job!
Cheers!
I suggest you to put the bearings in different orientations, 3 to 3 with some angle (I sopose 60 degrees in between) to get more points of contact with the bars and to prevent them of moving around its axes.
Great design, thank you!
To use the Synchomesh cable a set screw that holds and a tension screw that pulls the cable could be nice.
Not sure if you tested bearing sleve on the rod?
Love the idea with using the motor housing
You should be able to get the pulley count down to 2 per string:
* for the extension string, just rotate the existing bottom-of-tube pulley around the tube's axis until the input string is tangent to the drive pulley, and remove the other motor-side pulley.
* for the retraction string, chose a random point "A" in the plane of the drive pulley ~3cm away, draw a line from "A" tangent to the drive pulley in "B"; then set a point "C" on the central axis ~1cm above the pulley; the string should follow the path "BA" "AC" then up the central axis from "C". The 2 pulleys are semicircles at "A" and "C", tangent to "BA" and "AC", and "AC" "central axis" (in CAD, create a plane from each pair of lines and just add the circle) [for clearance, you might need to swap the extension and retraction drums]
@110:30. For the bearing surfaces, can you use an o-ring of something like teflon or viton? The viton would even give some resistance to tube rotation.
Nice evolution of your system and I really enjoyed seeing it. If you want help making the motor driver brake resistor not overheat, share your schematic with the motor driver. Cheers Mate!
A couple ideas come to mind. To terminate the twisted cable try a double plate clamp, 2 screws with a bump in between should do. The pulley "shoes" as I would call them, having them rounded may decrease their friction coefficient however making them concaved instead of rounded would reduced their odds of slipping off and also serve as an self-aligning guide sort of speak. I wouldn't make it exactly the same as your center shaft, maybe 1/2 a mm wider then the shaft on each side. effectively giving you a maximum of 1mm tolerance and ever so slightly reducing the friction vs a matched curve to shaft. really like what you did so far. hope my advice helps. feel free to message me if I wasn't clear and you want clarification. I can fab something up easily if needed
Definately use the dual wheel design on all 3 outter shafts to stop torsional forces. And for braking you can use another mini actuator to press a cylindrical bearing down against the wheels in the bearing block and the harder it presses the more it brakes the bearing and an electrical actuator is very percise and fast so with the right algorithms it can precisely brake the movement.
Just a quick idea on terminating that funky syncro cable, how about a copper and/or brass compression lug used in wiring electrical service, but with a plastic sleeve such that you're not directly compressing on the syncro-cable
Realy nice project would like to see more of this.
One safety approved could be add some end stopper for that middle pipe so it dosent fly out of when string cut off.
The subscribe button is just one click away 😜
I do like that idea, but it's also kinda fun not knowing if the central tube will shoot out or not 😁 (I probably will add safety features like this to the completed platform though. Depends how often/severely it happens in testing)
Cheers for the feedback 😊
A dirt cheap and possibly very small termination method for synchromesh or other steel cable: use the brass core from a good quality "domino" screw terminal (common in europe, eg. reference "034211" from Legrand, or "1563996" from Conrad ) , held in a 3d printed part with a screw -or with a thread all around the holder and a nut- for tensioning. (tensioning should only be needed at one end of the pair of cables)
have you tried to use fishing line for your string. fishing line or the string they use for bows (as in bow and arrow) is fairly stretch resistant and can handle a lot of weight.
You may wanna look at the Threadless ballscrew design that was popular a few years ago. Very simple design but needs a linear feedback mechanism
Have you tried polymer bearings like sleeve bearings or bushings (e.g igus drylin) instead of roller bearings?
To terminate the synchromesh, consider a banjo bolt then solder a plug or even a cable stop clamp
if the outer spiral is tight on the synchro mesh cable, you could make a plastic piece on the end of the actuator that it slots into and just turn the piece to tighten the cable
That's a cool idea! I have no idea what it's made from, but the outer spiral is bonded extremely securely to the central cable. I worry that this idea might be able to slip if the friction between the plastic part and the cable is low enough? The cable might be able to pull out (while twisting) if pulled hard enough?
ahh, darn you're right it would need some part to stand proud of the spiral so that it wasnt just the friction holding it, unless you manage to have a very long length of spiral in friction.@@harrisonlow
Someone has detailed (on Zulip) an idea that I think is quite similar to this and I think it would work pretty well, though it's a bit bulky. Check it out if you're interested 😊
pdj.zulipchat.com/#narrow/stream/399279-General/topic/Terminating.20Synchromesh.20Cable/near/388988893
@@harrisonlow That mechanism is what I was thinking of! ( but better lol ) It would be too bulky unless you could hide it inside one of the pre existing parts.
These actuators are awesome! I think they'd probably be useful for a really wide range of applications
Cheers! I'm keen to see if anyone ever uses them for anything 😁
If you get the angle correct on the spool pulleys, you only need 1 coming off the pulley instead of two. The strings will wrap more than 180* around the pulley though
To terminate the syncromesh cable, use a set-screw inside a metal screw insert. They're used a lot in woodworking. Drill a hole into the insert near the bottom, then feed the wire through, screw I to the insert.
For the wrapping around the motor housing, that would decrease the speed, but would massively increase the strength I think.
This is amazing progress. Not sure what you're planning on using this for, but it's really cool!
Interesting! I imagine you'd have to use fairly large inserts to do what you're describing? To check that I'm understanding you correctly, the wire would be "pressed into" by the bolt, but would it have anything touching it on the other side? Or does the strength of this idea come from the wire being bent by the bolt, disallowing it from being pulled back through?
Cheers!
Great work, especially with iterations. One thing that can't go un-noticed is the cyclic testing. With no load, almost every actuator outperforms their intended application. It's the testing of cycles under different loads and disturbances that can really speak volumes for a design. Loved the video. I wonder if you shouldn't consider extruded aluminum hexagonal cross section, as the part in motion. Would make for better load distributions. You could also add steel wear surfaces if you figured a good way to bond thin shim stock to aluminum.
I agree, cycle testing under load is going to define how well the design will hold up. The forces are going increase dramatically depending on how fast you accelerate and decelerate.
To hold the synchromesh, take a bolt and drill hole in it large enough for the synchromesh to fit through, thread on one jam-nut so it goes below the hole, insert synchromesh, thread on 2nd jam-nut and then use the 2 to clamp the synchromesh in place. Could be changed in seconds and wouldn't require any design changes and parts could be made in minutes.
I don't know if you are having differential tension problems between the two stings, but if you are there is a way to use a single loop to accomplish the same thing. I don't know what it is, but airplane ailerons are set up this way. Should be easy to find.
I really liked the way you make decisions for design. 🎉
Cheers 😊
Hi, I am a maintenance guy & working on industrial automation systems. In one of my plants there are multiple 3 phase motors sharing single load. All motors are driven by different VFD drives. Each drive is equipped with breaking resistors which tend to burn frequently.
I had removed all breaking resistors and connected DC bus of all drives in parallel using type C fuses for protection and monitor all drives DC bus in plc for system shutdown in case any deviation observed..
After this change we didn't face any tripping in system.
While some motors tend to overdrive the load, remaining low rpm drives generate reverse Power which resulted in dc voltage increase. This reverse Power is actually consumed by the first lot of drive which tries to increase speed of load.
@@TechnoAutomation Interesting! So to check my understanding: you have multiple motors that are all hooked up together so that if one starts braking, it sends power to the rest of the motors (with fuses in place to stop overloading)?
I'm not sure that such a setup is possible with the ODrive motor drivers I'm currently using, though I'd be very happy to be wrong! I think the closest thing would be a battery powered Jugglebot, which I'm all for, but I need to do a fair bit more reading about how to (safely) build such a system
Amazing design and great video, thank you for sharing all of this knowledge with us!
Did you think about using square CF tubing instead of round for the middle shaft? It could use 2x4 bearing, giving much better stability sideways and it would be resistant to rotating.
I honestly didn't even know that square CF tubing was a thing! (despite my supplier selling it 😅). I'm hesitant to go for square tubing though because I don't want to add even more bearings to the bearing block, though I do like the resistance to rotation. Others have suggested hexagonal CF tubing and I quite like that idea, but I can't find a supplier in Aus...
@@harrisonlow I bought some Hexagonal tubes from China for an old project, I think that should work with 2x3 bearings,, but honestly I love redundancy so I would add more bearings (but also this increases complexity, resistance and weight). I think I've seen some triangular tubing too, using that might be the ultimate solution: you can keep the current 2x3 bearing setup, all faces would be covered and it should be resistant to rotation.
For the termination of the cable you can put it through a tube, with a hole for a bolt, fixed to the piece at the end. Then you just have to press it with the bolt.
I think it's the easiest way you can do it.
For string termination, look to the guitar world. Be that the mechanism within locking tuners (the clamp action, not the tuning / rotation), or the design of a floyd rose locking nut.
the cheapest way to terminate the sychromesh cable would probably be a simple ring-tongue terminal that you crimp on the end. this way you already got a nice point to screw them into place. only problem i see with this is maybe some backlash you get, if you don't get the length exactly right on both cables, creating some dead space until the slack is picked up. Some kind of idler/tension pulley that you can adjust somewhere near the motor can solve this. I'm thinking some kind of set screw that changes the position of one of the pulleys, so you snug it up once and the whole system is ready to go until something else gives, stretches etc
For the synchromesh cables I would suggest terminating them with crimped electrical connector eyelets (that take a screw terminal like you see on high power capacitors)
Speaking of capacitors, you should be able to "charge" a couple of super capacitors when braking and then letting them discharge through a bleed resistor. This way the instantaneous current sink will be much larger while the discharge would happen over a longer period of time.
5:53 Those bearing slicks 👌
I think you can have fewer than 3 pulleys per side if you offset the spool insted of constraining it to be coaxial with the moving rod. The minimum might be with the spool tilted to be perpendicular to the rod.
Rock west prefabs hex section carbon rod which might be useful for improving the torsion issue.
If mounting flex leading to the string generating derailing forces can reliably be ignored, then absolutely he can use just 2 pulleys per string with the design shown. If derailing forces can't be ignored, then using pulleys with a "taller lip" (to retain the string better) may be enough to compensate. It's probably not worth focussing on right now, but definitely something worth considering on the maybe-last version.
I can't give any formulas for how to do the calculations off the top of my head, but the concept is fairly straightforward:
1) Each pulley corresponds to a plane, which it's embedded in (and parallel with); one of the sections of string (e.g. "from the spool" or "to the attachment point") must also be embedded and parallel with this plane, while the other section won't be;
2) The task is to intersect these planes with each other in such a way that:
2a) the intersection is tangent to both pulleys, AND
2b) the section of the intersection between the two pulleys doesn't pass through anything else (e.g. the spool), AND
2c) the plane with the string section that passes to the spool should be as close to perfectly tangential to the spool as possible (to minimize the risk of derailings).
Look into the PID loop of your actuators, all the other math seemed spot on to me. Cool video btw!
Yeah that PID will need some updating once the Stewart Platform is working. My current settings are "just-good-enough" values for preliminary testing 😊
Cheers!
I have had luck using capacitors in place of brake resistors on linear servos in factory automation. Bonus you can supplement acceleration current and reduce load on your supply.
Regenerative breaking can be done with a 3 phase rectifier and some way to either store the energy or dissipate the energy. A good place to start is looking up "delta alternator diode layout". Delta alternators have nearly identical wiring to most brush-less motors and you can use the same type of rectifier layout because of this. The trick however is using PNP transistors instead of diodes. By using PNP transistors, you can monitor the motor driver's voltages at the transistor's base and limit the current of the rectifier when the motor driver is engaged. Keep going on the project, very interested to see where this goes!
Thankfully the ODrive is capable of handling the electrical stuff for regen braking, but the problem is that my electrical knowledge is subpar right now so I need to learn/figure out what batteries or brake resistors to use. I've tried the naive "increase brake resistor power, keeping nominal resistance the same" but that didn't work for some reason 🤷♂
I am new to python coding and that endurance test counter would be interesting to see.. But this whole project is amazing. Thank you for documenting and sharing it!
The way that counter worked was to simply add 1 each time the actuator reached the end of its movement 😊
FWIW, I have a bunch of the code up on Zulip so check out the link in the description if you're interested to see it. And if you want any specific code, just let me know and I'll happily share it with you 😁
Add a large spring on the motor/strings such that it would pull the actuator to the middle position from the edges.
Crimp a cable lug to the end of the cable then attach a small adapter block to that (basically a right angle adapter with two holes) then use a bolt along the axis of the string to change tension
Wow, what an elegant design. Kind of reminds me of something you'd find in an IBM Selectric typewriter.
What an amazing mechanism! Cheers for bringing that to my attention!
I've had to sand flexible filament before it will bond with other materials. Also, you could try a concave profile for the bearing sleeves, that might give it more surface contact with the rods vs the current round profile. Def see the rolling brace being more stable when against all 3 rods. Look at how roller coasters work, you should be locked in place with just 2 bearings per rod if all are angled for opposing forces. Congrats on the progress! Looking great
Very interesting point re. sanding. I'll try that out!
I actually originally had the TPU sleeves being concave instead of convex and they had an insane amount of friction (though they were a very good fit for the CF tube)
I agree that four bearings are all that should be necessary, but I can't think of a good way to position them so that they're robust against forces in all directions!
Cheers 😊
@@harrisonlow you'll also want to sand the metal surface of the bearings where the TPU bonds - I use rubbing alcohol to clean both sides before gluing. I'd start w/ equal angles on all rods - 2 bearings per rod, on opposite sides of the rod (angled more than 90 degrees to each other) so the the tension is pushing out, away from the center equally. Provided the sleeves don't make contact w/ each other, they could be all aligned - so a ring of 6 bearings in a sort of star pattern. If you test the TPU sleeves w/ dif infills, you might find the right density to keep tension w/ out too much friction
For the ODrive and braking, you can use a lithium battery pack as power source to absorb the braking power instead of resistors. Then add a power supply to battery in parallel with ODrive (with diode so current doesn't go to power supply) to keep the battery at nominal (not full) voltage. A large pack won't heat much from the brake current. There's a setting to not use resistor in ODrive settings.
Does such a simple system work well? I very quickly ended up in a Battery Management System + Active Balancer Board rabbit hole when I started looking into LiPo batteries (coming from a place with 0 experience with them) and (temporarily) gave up 😅
Or, just use a very large super capacitor... you can get one that can handle way more current than all your motors combined will generate, so you don't have to worry about heating up a lipo into old age.
You could consider using a very large super capacitor as a "breaking load" for the ESC. If you dump into the capacitor and then pull power back out, you should end up with a net loss in power usage, thus the batteries will be at net discharge without worrying about chemistry issues.
Attach synchromesh cable with an i bolt. Twist to attach. Continue twisting until you reach the desired tension, like a guitar string.
For the brakes, use magnetic damper less work on the brake, plus it will give it a shove back the other way.