Is a sub 1 min 3dBenchy possible? Ultra Fast Motion System Prototyping! (Episode 4) 

Some years ago I built a fast cable driven 3D-Printer. There I used two parallel cables on a single shaft for each drive axis to constraint the rotation. I think this would be applicable to your 2D-motion platform as well making it more precise. I also have a video about the design on my channel if you are interested.

You don't need the springs, you need reverse kinematics. The only reason you need the huge springs is because at high displacement, one axes will start to affect another. This can be compensated for in software, it just takes some trigonometry. More/better springs will not keep you from skipping steps, as you are still commanding the steppers to positions they don't want to go to, you're just compensating for the inaccuracy by adding play into the system.

a lot of this is over my head but its fun to watch this project develop!

If you attach pairs of rotationally adjacent cables at the same point, so the puck is attached at two points 180degrees, you will get much higher axial stiffness and thus decrease the amplitude while increasing the frequency of the rotational oscillation. Placing the motors so you get 90 degrees angle from the motors connected at the same point will give you similar acceleration in both x and y. If you are wrapping the dyneema cables several times around the motor pulleys, the springs will have little to no impact on the dynamic tension of the system-you will still have to control the motors so they let the appropriate slack out for cross-wise motion.

So excited to see how this series progresses! 😁

I say as long as you open source the project, go full on with using the best motor solution for the project :)

I think that you need to tweak the design to remove springs from the equation. They are always going to oscillate as that's what they do. Having the air bearings to reduce friction is potentially going to cause more of an issue as there will be less natural damping due to friction.

Subscribed! You're a mad scientist for sure, this stuff is mind-bending!

Wow that last slow mo is wild. Crazy how stable it actually looks. FWIW I'm sure you're aware, but the further away from the build platform you put the motors, the less angle change the ropes will experience. Only thing I can think of to increase performance is reducing mass quite a bit. Using some kind of lighter high strength line combined with a lightened platform maybe? Really awesome work!

As you demonstrated, when one axis moves really far, it causes the other axis rope to expand. For it to expand, the rope must slip on at least one of the stepper pulleys. If you keep the motors on the same side or just use 2 motors then the rope doesn’t have to slip.

I think the springs to hold the tension on the slack side are trouble. So how about this: rather than the one spring on an axis loop, have a third motor with a pulley of varying position (like a pulley attached to an arm off-center to the motor). Then as the rope length of the axis changes, you can adjust the length of the rope path to match in software, maintaining constant tension on both inside and outside sections. Maybe a geared motor for this? Not sure. (With a geared motor, rope tension would take up backlash so it seems fine?)

Very cool! A suggestion: do input shaping, it helps a lot to increase accelerations because you don't have to fight against your own resonances. You can expect even a 50% performance boost

I'm not crazy about the loops of string you are using to attach to the sled. It feels like they are going to have more slop than a dedicated connection. I think you are better off printing a rigid eyelet or ball and socket, or even crimp on an electrical ring connector.

Thanks for welcoming the girlz - because we are here too :)

You could achieve higher real acceleration and travel speeds by moving the print head AND the bed in opposite directions. Each set of drivers gets 1/2 the travel distance and one is negative.

I would also suggest making the stepper motors closed loop by adding an encoder such as the AS5600

The problem with the cable lentgths can be ommited by having a inverse rod system beneath the table. Look at the Marionette printer by dyzekat. Might add a little moving mass to accellerate but you could get rid of two of the motors. Or just stick with 4 motors and use the motor winches themself for tensioning the cables. But then it would be wise to have a servo feedback loop to control the tension

Awesome to see the progress!

If the spring is attached to an anchor and a pully, then the spring will not add inertia mass to the rope. The pully could be at one corner or at a mid point, making a strait into an obtuse angle.

Have you checked whether your motors have the power for you wanted acceleration? At some point the magnetic fields move faster than the rotor mass can move.

Hey mate, if you want to make the steppers work harder you need to slow them down and load them up. A gear multiplier will reduce your stepper accuracy but make them work harder for the same travel distance, which will improve your travel speed for the same step rate... That or pulley diameter increase or both... You could also gravity tension your system the same way, a 10:1 pulley ratio on a tensioning weight will give you the capacity to support 10g of accel, but the spring system you've got now works a treat, so don't need to bother with that. With regard to the flexible linkages you're using, you're creating a virtual intersection when you're tying off which can result in multiple valid solutions for position when you're tying off on 4 points, not 2.

There is always going to be a torque on the sled when it is off centre unless you pull it from a common point. But then you'll need another mechanism to prevent the sled rotating.

seems like the faster it goes, the more stable it gets

With proper puck design the air bearing can be powered through holes in the table like an air hockey table. Using 2 motors, attaching each motor to 2 opposing corners and running the cable under the base instead of routing it around the outside. This avoids 2 motors fighting each other, causing step loss. Another option is 4 motors with drums to spool up the cable, mounted on a pivot this allows your platform to act as it's own Z axis by tightening all 4 cables at once, and avoiding the need for an air bearing.1 issue is, for multilayer winding needing to compensate for cable thickness or using a large drum for single layer winding and sacrificing step accuracy.

1. Check out the wire routing and the math of the of the Hangprinter project. They've made a few videos about it. 2. Print a cube or anything with a normal extruder and hotend at normal speed to verify the motion system is correct before going crazy on accel and flowrate.

Hey there, I think I randomly came up with a partial solution to your data issues :) If you hvae access to a camera that can either do highspeed video or can achieve short exposure timings you can have it take a photo at a specific point in time and as you know the theoretical position from gcode/software you can compare it to the position obtained optically. If possible you can either take a series of photos during similar movements you should get robust data. You can add high contrast pattern/point to the center of the air bearing and then have a python script or similar piece of code obtain that position and it's deviation :) Rotation would need a pattern on the air bearing. Hopefully this helps, keep up the nice work :)

Admire your goal of keeping in reach of the home CNC comunity. Increasing motor size (length to keep the inertia down) I do believe would help but first I would get drivers like Geko or Trinamic that are capable of driving the possible larger motors and trying with the current motors. But if you are loosing steps , then your underpowered. The vertical vibration of the stage looks alarming. magnets or a lubricant layer or rising the stage pulley points ? Air hockey ! Love the project. edit, watched an earlier vid and of course you had air floated .

this is just a thought and me throwing random numbers. But what if you made the hotend/extruder assembly into, basically a pendulum. and then added a motor and pulley set to it just like the bed, then in theory you could move the hotend a little in one direction at the same time as the bed moves in the other direction that way you can acheve higher total acceleration. For example you have 200K acc on the bed and then the hotend moves in the opposite direction with 50K, then you have a Delta of 250K acceleration. i know the hotend is heavy but if it moves 10-20% of the distance at 10-20 acc, then the bed only has to move 80-90% of what it normally does, meaning a increase in print speed of 10-20% for example. Benchy is ~6cm so if the hotend could move just 10mm at 33.3K acc for a 1:5 movement ratio resulting in a 16.6% acc increase from the bed's 200K acc. to 233K acc. in total. As for the pendulum part, if the hinge is 300 away from the nozzle and that only moves 10 mm (5mm off center each direction) then you only get a Z diflection of 0,0417mm at full swing. now is it worth the hassle to effectivly double the motion system for what i estimate can be a 10% increase in acceleraton. Probably not, but it might be the next step after you reach the limit with the current system.

For an approximation on a constant force, you could use a much longer spring. It's not constant, but the relative length change is much lower.

A Y bridle on the hookup points might reduce the rotation too. You would still have 4 points, but always two arms for a adjacent Y bridles would be connected.

you could add a perforated metal plate under the motion stage, add some magnets to the motion stage plus use the air pillow to reduce friction. With that you could eliminate the jitter while still maintaining low friction.

I would recommend using longer Springs. The longer the spring, the less the force will change for small amounts of displacement. You could also experiment with constant force springs.

Given that that rotational inertia is the proportional to the radius squared, using nema 23 would probably give diminishing returns as most of the inertia (rotational or linear) is probably already inside the motors. If more torque is desired, longer motors are probably your best bet. Great content!

Why do you have those 4 squares at the side of the bed? Does it have any mechanical function, and if not could it be removed to further increase feed rate?

Tbh for crazier accelerations I'd suggest you to use a bit more beefier motors along with some belt drive system with multiplication, they're readily available in case you'd build 1:3 or 1:5 multiplication. Direct drive is easy to implement, but the main issue is that any stepper could (and would do) 2 things: resonate and lose steps at higher speeds, so by introducing beefier motors you'll let them use their torque and keep their speeds in safe zone

one direct modification, just use two motors, with one wire, that loops around in a bowden cable, to the other position on the other side. that way only two motors and the position control is accurate, same cable both sides. or through pulleys, to route the belt. essentially. yes 2-axis belt drive. only two motors, one per axis. just uncouple the wire axis from each other. just like the 2d-yoke screw solution. yes the wires only ever need to go straight, not be bent. bro you are now just fixing your own problems. resulting from the initial choices. speed is your issue. yes both the machine and the video. yes, you want to get to the goal as fast as possible. it will make poor decision. slow down. well Roetz 5.0 fixes up what 4.0 did not do properly rushing in doing stuff. well 1.0 could have started in close proximity of the actual solution already.

Sweet Norov BedXY machine.

excellent work sir!

Steppers are fine I think, metal motor mounts and heating would do a lot for motor cooling. There are high rpm fans and server fans for the board.

Each cable needs to be a Y. 8 points of contact to the center wheel. Triangles are stable. Also, softer springs or more amps will help with the skipped steps. Steel cables would be best and could allow you to remove the springs all together. Add cross bars across the top of the pulley pairs to stop any movement in the pulley shafts. I also wonder why the head isn't mounted to the cables and the part stationary. Gonna be hard to make the part stick at those accelerations

Instead of having each axis have a single line on each side, you should have two parallel lines for each axis. That would further help with yaw rotation, and position instability; and would allow for 8 lower inertia motors and higher power. You must already be compensating for line stretch in the stepper motor command, thats a lot of stretch and its all happening behind the motors. If the compensation is not accurate I could totally see that causing skipping.

you could create a 5th axis for free by doing some variation of the following: take one pair of opposing motors, run their cables 1 turn around the main platform then attach them to their mount points. make sure they both go around clockwise. do the same thing with the other two motors and cables, but do it counter clockwise. i think this would let you tune out / offset the induced rotation in the platform. but it might just be a huge pain in the ass haha

I think it will be better to add XY guide rail underneath the bed. The bed right now seems easy to bounce.

The highspeed movement is kinda insane! How much slicer tuning is allowed in the speed boat challenge? It should be possible to reduce the sharp turns and build benchy entirely out of curves. I'm just wondering, I'm not the guy to build a custom slicer for it.

Really enjoying this build! Would it be possible to replace the pulleys at corners with some cams that compensate for the additional length in the rope for the opposing axis as the other axis rotates the cam? Same thing could likely be accomplished in software as well, but might be cool to do it mechanically.

Edit: Posted this before Watching :) The oscillation is likely caused by uneven pull forces from the motors. I think it can be fixed by crossing over the cables pairwise of the X and Y Axis between the motors and the sled. This will restrain the rotation axis by giving the pull ropes better angle of attack (two pairs of cables will attach tangentially on two points to the circumference of the sled instead of 4 individual bearings) With the existing kinematics it could also be solved with a fast software feedback loop using a gyro and a compass sensor in composite.

You got some very good concepts united in your Machine. This might revolutionize fast fdm printing. it is time to throw away heavy linear rails and slow steppers.

You can also do this with 2 motors. You only need 2 pulleys on the opposite side. The drive seems to have the same kinematics as a Core XY.

Is the vertical spacing provided by the air bearing consistent as it moves around? Is there enough tension on the platform from the cables to keep it from wobbling as it moves around?I know the "wobble" would be tiny if there was any. If thats already been covered just say "yes" lol

Nema 23 would be sick, think you should follow the rules of the speed boat benchy. But also push some crazy amount of power through the nema 17 and add some Active cooling

Wondering if it’s beneficial to embed small ball bearings in the platform to reduce surface contact. I had an old computer mouse that did similar. Four 1mm ball bearings. It felt like moving an air hockey puck when on a waxed surface (mouse pad). The mouse is the Rotokiller for those interested Rtr-720.

If you watch the neighborhood of the print bed: it is moving so fast, that the granite desk beneath is wobbling/rotating like pudding, too. Maybe some weird quantum or relativistic effect due to the movement near the speed of light.😁

The most efficient & accurate system would be to calculate the 'off axis angular offsets' and adjust the steppers in code. Getting rid of the springs & extra wire would reduce inertia - the formulas are, in theory, simple (pythagorus?) & something that is likely waaay above my pay grade but there has to be someone in the community can do the maths? This may be where a 3 axis system is your friend as I think it has been done for other systems.

Try delta kinematics. Like, if you’d take a delta printer and squish it onto a table, the delta motion columns (linear rail + belt + motor + rod) would be placed in star formation, and the rods would connect to your center platform instead of connecting to hotend mount. Would need to modify g code somehow to ignore Z for the 3 motors.

My Idea might be a bit difficult to make, but how about a sorta reversed delta setup? instead of the printhead beeing swung around, the printbed will be?

Hmm did you think about using a roller with a gas sprint to tension the things? Car Radiator + Liquid Cooling. A other option is Cryo Cooling but i doubt you are allowed to handle FCKW stuff. I do wonder if these tiny motors can be cooled efficently.

I am out of my depth here, but what about doubling up the motors? Could you halve the strain on the steppers by putting another in-line behind each one? Seems like it would keep the spirit of what you're going for with not using industrial components and things that anyone has access to. But I don't know if it would cause other issues that I'm not thinking of.

Are you seeing the natural frequency of the springs and the related phase shift causing the rotation? If so, could you add dampers?

Didn't expect the tony shirt

Posting before finishing watching. Would a printed/lasercut flexure possibly solve the rotation issue? That could allow x/y while isolating rotation. In order to have whatever 6cm of travel it might need to be quite large.

well, got my answer about the gas springs lol

A metal plate with a magnet in the print bed might be a solution to the noise problem.

Do stadium cable cameras have any useful tricks to share?

Does it make sense to have the build plate freely move on linear rails? Adds weight and friction but can reduce rotation / vertical rattle. Sorry if previously discussed.

maybe add some UHMW or Teflon tape under the platform i feel stiction might be effecting things a little or maybe see if Igus will send a roll of filament to print the platform out of. they make TF and UHMW pads for gaming mice if you can only put 3 pads. you can add the MKS servo backpacks or the ones btt makes feed it step and direction BUT don't plug the stepper motors in keep using your setup this is not closed loop. but you can use the logging to track grouse pos error id say go to nema 23 with a larger pulleys to take advantage of the torque but id also try making the pulleys 10-15% bigger and smaller at this point you reducing the RPM of the motors to reduce the torque the need to overcome the angular momentum. in the world steppers precision is tied to speed and for this size stepper they are rarely used past 180-200 rpm where a/c servos in the nema 23/24 size are used well into the 1500-2000rpm range

What about using synchromesh cables for less backlash? It kinda reminds me of The Positron.

Stepper Topic: As you said it might be worth an investigation to go with bigger a smaller steppers. That was a thought i got along that video as well. From what i know stepper compared to servos working better in lower rpm's to achive there best working torque. Your steppers don't do a lot of rotation (maybe not a whole rotation at all?). Would be interestet how much rpm the steppers would theroaticly have and if its to much. Maybe it would be best to overthink the gearratio with you pully thing on the shaft and maybe it is not wise to have the pully driven directly from the stepper? I'm not so much in to this but in a thought i would test a NEMA 23 decoupled (maybe geared?) from the pully that runs the cable. I dont think higher inertia effects the speed when you have them spinning slower and running a little more toasty. (I don't have proofs or enoght knowledge, these are just some thoughs i want to leave here)

Do closed loop stepper motors count as stepper motors? (Ten seconds after I wrote this, you mentioned them haha)

If you have 2 motors pulling on the same cable (in opposite directions), you could theoretically remove one of them, because it will do the pulling in both directions all by itself... oh well, and then you pretty much have a CoreXY printer! 😁

Umm, could the skipping be because you're moving all 4 motors at the same time and the strings are stretching because of the forces and that's letting the cable pull back on the steppers a bit hard? as you demonstrated 26:24

Name idea: Inselbegabung because it can only print benchys xD

do you need constant force springs or linear springs? if you need constant force you csn buy gas springs

the joint between the cable and the middle thing looks suspect to me.

so are you gonna have to drill a new hole in the granite to supply air to the air bearing? cause it got moved out of the existing grid with the new cable setup...

Maybe use external 5160 drivers like the ones from Mellow, they can provide much more power and dont get that hot/are easier to cool as they are larger. Also consider running them at 48v if you arent doing that already.

You are loosing steps because the spring is behind the motors, when they expand/contract the dyneema must slip on the pulley, if the friction is too high instead of slipping the dyneema will make the stepper turn, hence loosing steps. If you really want to use springs they need to be between the motor and the sled, and probably one for each motor so that two springs will contract/expand at the same time so that the sleds stays centered

Keep going, cheers!

you can have some rails under platform for xy so it will resist z rotation

okay so yea maybe only use 3 motors instead of 4 and try to use it like a delta, from what i see, klipper doesnt understand the movement system as when say u move x , then y also is moving and cables are pulled inwards rather than in one direction. so this caused strain on motors and stuttering , coz its like sudden jerk(brakes) is applied.maybe they skip steps coz of inertia. i feel its less about loosing steps and more about cable slip

Are you aware of hydraulic piston encoders? Basically a string attached to an encoder. could make the center platform closed loop to deal with missed steps

Do the benchy rules allow twin differential motion systems? So move part and head.

Turn stealthchop off if you haven't already, will help massively with step loss

what speaks against having the "bed" anchor point directly in the middle, connecting all motors to one point? (something ive been wondering since last video)

Even if you solve all the problems, how will you deal with bed adhesion at those acceleration levels? Won't you create a supersonic benchy missile with a corresponding benchy-shaped hole in your shop wall?

I think the stepper question if you should go higher is a mathematical question ... Does the torque increase outweigh the inertia gained ??!

Not speaking from any knowledge of such things but if instead of having the cables parallel to the table, if they were mounted higher on the floating part and angled down to the motors they might stop platter lifting so easily.

The best thing we can have is a 3D printer that no longer makes me wait 2 days until it is finished and a 3D printer that uses cheap granulate for 2€/kg instead of filament for 20€ per kg, and if it also says "High Speed ​​Filament" on the label then it's 30€.

Can't you make this with just 2 servo motors, one for each axis? maybe under the table. You could have a closed loop string for each axis, tensioned like an auxiliary belt of a car... ?

great video!

out of curiosity, what material did you use for the motor mounts? 10:29

Maybe you should try using the TMC5160 with a higher voltage, such as 48-60V.

"Technology everyone can afford" builds a 3D printer what looks like a Johansson Cordimatic.

Try a vortex tube for part cooling

yooooo 1080p60!!!

bigger stronger steppers

Why don’t you just get like a normal corexy setup with like mgn7 or 9, and attach a Little bed to it? Would solve a lot of problems ( no springs, no bouncing bed)

Wow

👍🏻

There are constant pressure gas cylinders.

Why not 8 steppers