[peteeng]

Hi Kis - Looking at the Core XY system I think its overly complex. The rope/belt length is twice or 4x as long as it needs to be (and length means more stretch) for no kinematic benefit that I can see, except that the gantry drive does not need to have a motor on the gantry. I do like belts and have used them in several projects including launching a piano at a theme park and 3 routers. Was a very fast piano.. 60mm wide belt. I do not like U drives as they reverse bend the belt and to be correct, they require very large idlers, so the belts do not fatigue. Since you want a long life, fatigue is important. I think your answer lies in using timing belts not rope and not Core XY. I estimate your gantry will be 6-10kg and the if you place the gantry motor on it, it will be an extra 4kg? So if your clever I think 10kg all up will be a good figure. Since the gantry motor only drives the saddle & payload its got lots of torque available as the saddle and payload will be light (except it will have gravity to deal with in one direction). Since the gantry is driven by two motors again it has twice the torque to do its job. The XY system seems to load share oddly and that may be a pitfall.

I've drawn what I suggest. The benefit of using belts is that they are fully specified so you know their stiffness and their design guidelines. Going rope will be an adventure with various issues to solve. The sawdust image is Brevis one of my small machines. It uses 10mm AT belt. It moves very fast If I program it that way. Around 300mm/sec (18m/min) I also use it for laser work and the speed is useful there. You can get endless belts called ATL which is for precision applications. The teeth are oversize and squeeze into the pulley better... Simple is best someone once said... Peter

edit - Brevis uses one std car on its Y axis each side. This saves weight. You will need medium preload if you go this way. I'd suggest 20mm long cars medium preload one each side.... Brevis is a good light duty router so even a std car with med preload works fine for P&P... Brevis has one bearing for the Z axis and this lets it down. But it was not intended as a router it was for demonstration purposes as a plotter. But every one wanted a router so I moded it to a small router...

[kis-mat]

Hi Pete,

Regarding the rope system, I acknowledge the potential issues but believe they can be mitigated. My understanding of the rope concerns is as follows:
a) Rope stretch over time implies that a tightly stretched rope today might be looser tomorrow. You're referring to a gradual stretch under tension, not an immediate stretch caused by motor movement that could lead to vibrations from gantry motion, provided there's sufficient pre-tension in the rope (right?). To address this, I plan to install a screw adjustment mechanism on an idle pulley for each rope, allowing for manual maintenance and tension adjustments. A guitar tuner could then be used to accurately gauge rope tension.
b) Rope slippage: I've experimentally determined the friction coefficient between the rope and PLA (for a 3D printed pulley) and found that five wraps are more than sufficient to prevent any slippage.

The system you have outlined could be a possible solution. I just find it unintuitive to move the motors along for a system that is supposed to generate maximum acceleration (especially since the motor for the z-movement is not yet taken into account); I still hold on to the idea that ropes are the lesser evil. But I'll do the math this weekend...

Don't you think that you don't need two drives for the Y-movement to prevent the grantie from jamming? That would reduce the power of the motors even further with limited electrical power available.

Best,
kis-mat

Ps: How many systems have you actually built...impressive. You do that full-time, don't you?

[peteeng]

Hi Kis -
1) All things can be mitigated but is this a software project or are you wedded to the CoreXY enough to solve the rope issue? There are two elements to rope stretch 1) construction stretch - this is the initial internal looseness in the rope that needs to be worked out or bedded in 2) then there is creep and relaxation. This is material , tension and force dependent. Dyneema suffers in this area. Your application of high loads. high velocities and high accuracy all go against using this sort of material
2) being a sailor I could have told you 5 wraps will hold anything, but 5 wraps at speed will be difficult to manage. The wraps climb along the drum due to the wind angle (and internal twist of the rope) and how will that be managed? I have been involved with cranes and keeping winds organised on a drum is a big deal and they move slow.... I think the rope idea is a can of worms. Which if you want to solve that problem thats fair enough. But it seems to me your interest is in another area and that's where you should put the effort.
3) commercial robots move a lot of motors and move incredibly fast. You do the maths and size and dimension everything correctly. Understand the kinematics before you buy anything. Design the system correctly before you commit to the build. Meet the specifications on paper that's how to do it. Sure moving mass is an issue but if you see large mills doing 2G+ stuff they are moving themselves and 1000's of kgs of steel as well... You do what it takes to solve the mission not create hurdles within hurdles for yourself
4) My proposed system only moves one motor plus the Z system. It also moves things in a direct manner. Ultimately its up to you. Intuition does not replace the maths. Intuition leads you down interesting and frustrating paths. You need to walk many paths to find an answer and often its back to a basic bog std solution
5) I'm unclear by "don't think you don't need two motors for Y movement? " I would use two motors for Y that's in the diagram.
6) I'm a mature mechanical engineer that has been involved in robotics and bespoke machine design for 45 years plus. In the commercial area I analyse and design machines for companies (medical, mining, yachting, automotive, movie stunts, sports equipment over time) I'm particularly interested in advanced materials and build carbon fibre stuff for various people. The Milli thread goes through many development ideas
Over the last few years I have been developing small kit routers for sale. I have built belt drive, screw drive and now a rack and pinion large router so I can build small machines. I have several threads in the forum on the development builds. Brevis, Lanky, Frankie, Milli and others. I've sorted suppliers and basic cost ratios and stuff to go commercial with the kits. This has been part time. Designed 100's of cnc machines and filled a couple of HDs to filter out the chaff. Build about 2 machines a year once they have a purpose and work on paper... My commercial work is now part time, want to move into kits part time. Projects find me it seems...

In the front end of a development project all ideas are put on the table. This is called the Fuzzy Front End. Once you sort out the real requirements then you filter out the good ideas and decide on the basic approach. Then refine from there. Your still in the FFE. Keep at it. Peter

with high speed systems, often you have to start at slow speed, work out the bugs then move into high speed. Thats what prototypes are for.... Be clear in your objective don't solve things that you don't have to solve. That will add years to your project.

addition - cables that are used for drives have internal contra twisted lays to minimise torque effects created by rope tension. Timing belts use this and they use linear lays, steel cables etc that create stiff straight, true drives. Parasailors and hang gliders etc that use small ropes do not have these issues so the rope maker does not account for those effects. The devil is in the detail....

[peteeng]

Good Morning Kis-mat - Thanks for picking up the dia vs rad error. Your calcs are not attached for me to check Peter

[kis-mat]

Hi Peter,

Thank you very much for your detailed explanations! I learn a lot from your texts. Please don't get me wrong if I continue to present arguments in favor of a cable drive system. I can only imagine the experience you have, and I certainly don't want to argue against it. Instead, I aim to learn from your presentations.

Regarding the issue of the cable "creeping up," I thought of eliminating it as shown in the following image, where the cable is passed back and forth between two pulleys, preventing the cable from creeping up.

cable


In the industry, there are often three-dimensionally operating cable robots, similar to the one in this video:

Video

Aren't the requirements for the cables similar to those in a coreXY system?

However, at the end of the day, you're right. I really need to consider which solutions are worth pursuing. Since this is a side project and I do it for fun, not for a living, I must be careful not to get lost in too many minor issues and overlook the main project.

Apologies, the calculation was lost; here it is:



Best regards and thanks a million again!

Kis-mat

[peteeng]

Hi Kis - I haven't said don't do it. I see it as simply a can of worms which is a hurdle to your main mission. If your happy to solve the rope issues, then go for it. In the Milli thread I tried to develop a cold castable material for machine parts. Sounded simple at the time turned into a multi year project. That put me 2 years behind making a Mill. If I had aborted the effort early and made a mill from aluminium I'd have one by now. But I'm stubborn in some areas... But now I have a fuller understanding of the materials and machine geometry so all good! Amongst all the other things to get done... Tensegrity robots are an interest as well... so many things to do. Peter

explain to me why there is a gear ratio in the coreXY system... the tension is the same in all the rope. is the attached diagram your proposed circuit?

[kis-mat]

Does this help? The first equations are from your attachment.



greetings
kis-mat

[peteeng]

Hi Kis - Been a while since I've done matrices , but good on you for doing it. BUT here's how I look at it, lets say I'm playing devils advocate (I think the math is misleading)

1) If I make a FBD at the pulleys the only body force is the inertial force so the reaction has to equal the body force. This means the pulley sees 300N each side. or 300x0.0159=4.8Nm per motor for a 6kg gantry at 10g. Now I understand that the circuit also moves in X&Y and that when going diagonally the forces maybe greater or smaller. Need to think about that but in this condition you need 4.8Nm....

Now the question of preload arises. There has to be enough preload in the strings to prevent the "compression" side from going into compression. So I made a computer model of the system. Made the strings 1mm dia steel and upped the pretension until that occurred. That was 300N which makes sense so I ran the model at 350N preload to ensure the system stayed preloaded. It says that the reaction at the pulley is 365N and the offside is 71N ie ~300N to provide torque. There is a slight difference between the machine sides. I also did this to check what stretch is in the system. This model is 1mx1m and there is 3mm stretch in the steel at this condition. So a 2m machine with 1mm dyneema will be even stretchier.. maybe 4x the stretch

2) What I observed is that this geometry is symmetric, when the saddle is at a side the strings are asymmetric and stretch differently on each side thereby creating displacement differences on each side. This will need to be resisted by the bearings and I suppose it will average out. But it will come back on track when the gantry slows down... The stiffer the rope/belt the better...

FBD - free body diagram

I'm checking through my model now to be certain a couple of the pulleys aren't quite right... I think

[peteeng]

Hi all - I sorted the model so it now has symmetric outcome. One of the elements was spec'ed as a beam vs a string so it behaves slightly different on that side. Just makes the answer more precise and 366N pulling and 66N pulling which makes it an even 300N nett force pulling on the pulley. Peter

[peteeng]

Hi Kis-mat - I have been thinking about the combined motion case and have concluded that as the X dirn strings are symmetric there is no extra loading created by the X motion ie at the drive pulley this load creates equal push and pull..... I added a load into the model and this confirmed that thought. So the inertial load in the Y dirn is the dominant load to design to. So now that's clear we can add in gravity, friction, aero loads, rotational inertia and acts of God... and then we're close to reality.... Peter

So now you need to detail the gantry and the z axis to determine accurate masses to design the drive system...

[kis-mat]

Hi Peter,

thank you for the detailed analysis...I'm not familiar with FEM analysis myself; maybe I should learn that too.

You are talking about a 4mm stretch...that means that when the saddle is pulled, the cable stretches 4mm on the other side? This ultimately leads to a "wobbling" of the saddle, right?

And with Timingbelts I wouldn't have this problem / have it to a lesser extent?

Greetings
Kis-Mat

Edit:
I have the feeling that I need to go into a practical analysis.

I have a 1 meter rail here from igus...I think I will soon make a simple setup where a servo accelerates the carriage on the rail via cable drive and in a second setup with a timing belt. I simulate the planned cable length of the planned 1.2 x 2 meter with several windings parralel to the rail. With an accelerometer on the carriage or a high speed camera I examine the behavior of the carriage.
I think this is a quick to build setup that will give some insights.

[peteeng]

Hi Kis-mat-

Dyneema comes in various grades the best is SK-87 this has a fibre modulus of 110Gpa. Steel is 200Gpa. But a rope is about 50% by volume and it has twist for stability, both reduce the practical modulus. So 110x0.5*0.9= 50Gpa for the rope. You can get steel wire with a plastic coating for driving cables by the way. Being steel at 50% volume means its about 100Gpa as a rule. This is what timing belts use. Many steel wires, each neighbour is twisted in the opposite direction so its internal torque is countered so it stays flat. Heres some belt cattledogs...

So a 1mm diameter dyneema rope has an AE of 50xPIx1/4=39GPamm2 and a steel one would be twice that. Belt stiffness's are in tables in the cats... enjoy during morning tea time.. Peter

prototypes teach a lot - The stretch in my model does not matter. Need a real machine size and real rope stiffness numbers... Your rope supplier maybe able to provide the AE of the rope...

I get my rope from Hampidjan they have excellent technical staff...

https://hampidjan.com/

re timing belts vs rope. Belts will be stiffer due to their steel construction (also depends on what sizes you are comparing). But find out the EA of the rope and you can compare....

stretch = ( Force x length)/(modulus x area)

Home - Kuraray Vectran (vectranfiber.com)

[kis-mat]

Hi Peter,

thanks!

I'm going to read up a bit on cable robots in the literature ( for example https://arxiv.org/pdf/1806.10632.pdf ) and try to get a bit smarter from a practical test setup.

You have helped me a lot! As soon as I'm a bit smarter, I'll get back to you here.

Best regards and best wishes
Kis-Mat

[peteeng]

Hi Kis - Scanned the article and I think they do not understand what creep really is, plus they did not find vectran which would solve their issues. Even better is carbon fibre ligaments. Stiffer and definitely no creep or relaxation... Peter

[kis-mat]

Hi Peter,

why do you think they did not understand what creep is?

Anyway...since I planned to 3d print my pulleys I think creep is another topic on its own.

Best
kis-mat

[peteeng]

Hi Kis - Creep is a big subject. Can't find how they established the creep value. Ref {7} is about underwater stuff...Dyneema is non crystalline so they probably measured construction stretch and relaxation not creep. But "creep" is used as a general term about things changing shape over time... Re reading they did find vectran and PBO. I expect they used dyneema because its easier to get.... Peter PBO (zylon) is a great plastic, stiff as steel try to get some.... PBO has a modulus of 270GPa and its crystalline

[peteeng]

Hi Kis - I looked at some rope and belt specs. A D2.5mm vectran rope is the same stiffness as a D0.6mm steel wire and an AT5 10mm wide belt. AT10 belts are twice as stiff as AT5 belts,,, Hmmmm Peter

10mm belt AT5 teeth are rated at 500N so 12mm would be a good start for the 10g case... SF=4.0 you want endurance... easy get 10mm or 16mm here in OZ in AT5 or AT10....

Hi Kis-mat - I keep coming back to R&P. R&P gives you the same very high stiffness everywhere, helical teeth gives you zero backlash and its all easy to sort and available. If the gantry, z axis and motors came to 20kg (but I think 16 is doable) at 10g you need 2000N to move it. With a pitch pinion of 30mm r=15mm you get 200N from a 3Nm motor and you have two. Weve calculated you need 1Nm to move the motor so you have 4Nm to move the mass. Thats more then needed to get the 10g using suitable gearing. Use R&P across the gantry and everything will fall into place. The rope idea will take forever to sort and won't be stiff. The belts can be sorted but again won't be stiff. These accels are really high and vibration mitigation is going to be paramount. Stiffness kills vibration so I come back to R&P.... R+P is infinitely stiff compared to rope and belts.

Then even though the motors can provide the torques and get to 10g you have to have extremely smooth motion to stop start at these velocities and accels and small moves. So thats the real hurdle for you. Jerk control, motion planning etc will be paramount. So what is the question your trying to answer? or what answer are you looking for? Most of these things these days can be simulated no need to build something... Peter

[peteeng]

Hi Kis- recently came across this controller. Would suit what your trying to do, has excellent motion control capability. Peter

https://youtu.be/tOFb8tOTMlM