Designers of marble fountains who don't use computing to design the paths run into reliability issues: sometimes balls derailing out of their track. They have to observe the contraption, identify problems (balls getting jammed up or jumping out) and then guess at the root causes and make manual adjustments.
That's the thing here: he has it running for hours presumably without any ball jumping out.
Most of the tracks consist of two rails, so the ball has two contact points. I'm no physicist but it seems like the goal would be to have ideally nearly equal forces at the two contact points at all times during the ball's descent. In other words, the track has to be perfectly banked so that the gravity and centripetal acceleration vector are balanced by a normal vector perpendicular to the rails. During a derailment, the ball has to lift away from one of the two contact points, so the normal force must have dropped to zero.
It's actually much weirder than that: banking changes the axis of rotation and thus kills the rotational inertia. The tracks bank super aggressively in order to prevent the ball from accelerating too much and hopping the track. This is part of why the descent is so smooth and all the balls move at more or less the same speed.
Also to be fair the final system does lose a ball every 30ish minutes. The tuning was largely me staring at the run or taking a video trying to catch where they get lost. Instead of hand tuning I would just update the generator and print another one. I'm considering closing the loop with a camera but that would be a whole new project.
First, I thought about Ansys or CATIA software but I couldn’t find any module specialized for simulation of balls.
But I think that people from those companies could help as well and participate in simulation as an interesting usecase. (These software are expensive for personal projects.)
My point was that these software could help to find weak parts in trajectory - so instead of trying to figure it out by looking where balls are too quick to fall from the ride - you can simulate it. I saw real tramway simulation done in Ansys.
I think the physics are different, a ball is basically a car without a differential, so it's going to behave differently on the tracks. I'd imagine the ball is harder to simulate because of that.
One of the results for hilbert curve marble tracks, mentioned elsewhere in the thread, was a video showing how to make one in blender, which has a physics engine so it can simulate it pretty well.
But for hobby purposes I would suggest to contact some university, they have such software, and they could find simulation of balls motion at marble fountain interesting for research (and educational) purposes.
I haven't actually measured it but that's a good thought, I may borrow a thermal camera and do some testing! It's not noticeably warm to the touch but this functionally a system that converts potential energy into heat and sound so there's probably a measurable change.
The state of each ball can be described by 9 parameters: the current location of the center of mass (x,y,z), the current linear velocity (vx, vy, xz) and the angular velocity on 3 axes.
I don't think the forces acting on the rails need to be similar -- they just need to be such that the acceleration of the ball is always parallel to the track. Unfortunately the equation of motion will look pretty ugly and optimizing the system will be quite a challenge.
And finally, the system has to be stable, ie. small perturbations should be cancelled rather than grow - if a ball gets a little too fast there should be something like a bend that slows it down, but that bend should at the same time not slow down a ball that is already too slow...
Another parameter - as a track designer you can manipulate the width of the track to change the ball speed. It raises and lowers the ball on the track, changing both the rolling diameter and the center of gravity. This can be used to make subtle changes to the ball speed before a turn.
This is beautiful. It would be amazing to have the tracks encode/decode audio, you know? Like, the track of the marble can be used to generate different frequencies...
I actually attempted this, the idea of a python script that converts a midi track into a marble run is just too good to not try. I printed a large drum with different track structures inside so I could test various "slopes" by changing the speed and it just doesn't work, the balls bounce around too much to get an audible pitch. A less rigid material or a larger bearing would likely work better but I decided to focus on getting the normal version working well.
Maybe at the bottom your marbles could land on surfaces with different accustic properties. Track selection would determine the surface and release time would determine the timing.
The particle simulation approach to generating an organic "tree-like" support structure is super creative! If I'm understanding correctly, you defined some laws of physics and then ran a simulation with the "time" dimension mapped to the z-axis? Is this a well-known approach, or something you came up with?
Either way, it produces a beautiful aesthetic. I'd love to play around with this idea.
That's pretty much it! It's the simplest method of supports I could come up with that allows for robust keepout zones. I did have a bunch of issues at first with supports blocking the path but with a little tuning it became surprisingly consistent. I doubt I'm the first to come up with it but I have not seen any similar systems.
Thank you! The emergent forms are much more interesting than they have any right to be for such a simple system.
This is a great example of a good use case for 3D printers. The smooth marble run action combined with the interwoven organic forms would be a huge PITA to fabricate with any other method I can think of, even if your just making one.
A good use case for 3D printers is random, small, custom household items that greatly increase my quality of life, much more than it is a unique sculpture.
I'm not the person you asked, but depending on what kinds of quality of life improvements you're looking for, your budget, your 3D design abilities, and your tolerance for working on the printer versus printing with it, the answer will vary in terms of what works for you.
If you're casually interested, the Bambu Lab A1 combo will do most things you'd want it to do, fairly reliably, but with a closed ecosystem.
If you want something more robust, go for a Prusa, but be ready for a more hands on experience.
If you want an entirely customized bespoke with a high learning curve, go for a Voron.
I'll give you an example from my life. I got a set of cheap LED lights for a closet. They come with a little remote control to turn them on, but I didn't have a good place to put the remote control, so I made a little wall holster that's sized exactly to hold the little remote.
I also got one of those SimpliSafe home security systems. It came with a door sensor, but the sensor didn't quite fit our door frame. So I printed a tiny piece whose dimensions exactly matched the SimpliSafe and my door frame, so it allows the parts to meet up but doesn't look weird.
Of course, 99% of what I print is useless stuff that looked neat on Printables, but sometimes I make stuff that actually serves a purpose!
I've actually done clear prints with LEDs installed. The bottom is much brighter than the top and it just look kinda tacky. I briefly hollowed out the supports and tried running fiber optics but it didn't help much.
I'm realizing now that I tried a lot of weird shit during this project that just did not work at all or make it into the final product, I should do another video just of all my failed abomination marble runs.
Minor suggestion/request: would be great if you added a final STL file to the github repo of a working example. Might be easier for people to try if they can't get the python code running on Linux.
(I haven't tried yet. But I'd love to just send an STL to my printer to see how well it prints.)
I appreciate the work. It's really beautiful and checks so many of my "oddly satisfying" boxes as a builder. It seems it hit those for you too, obviously.
Separately, the timing of seeing this is uncanny. I've been using marble runs to explain probability to my kids and was filming a marble run conversion lesson. Seeing this at the top of HN felt like someone was reading my minds.
Mesmerizing is the right word. "I can watch them for hours" was the key bit for me - I have always been fascinated how humans can stare at a random visual generator for forever, if it's the right one.
I think there is an instinct built deep in our lizard brain somewhere for this. Humans will happily stare at a fire, or an ocean, or a wave in a river, or (sometimes, especially children) a TV screen - and all I have worked out why is because it is constantly changing in an unpredictable way.
This marble run shouldn't even be unpredictable - clearly the paths are fixed and the cadence of balls is regular - but somehow it is still mesmerizing.
Not the person you asked, but look up "Hilbert Curve" on Thingiverse, Printables, or Thangs and I bet you'll find somebody who uploaded a marble run with more information for you.
I wish there was a write up of how some of the code works. There's a lot of Python in the repo. Looks fascinating, seems to use Python to generate OpenSCAD code, I think.
It's specifically using SolidPython2 to generate the models. I originally wanted to do a quick code cleanup and have the specific math be much clearer but by the time it was working that was an absolutely massive undertaking. If I touch the code again I'm probably going to refactor the entire codebase and use a different 3D engine.
Mostly speed, I'm mostly doing large boolean unions of primitives or chain hulls and OpenSCAD chugs pretty good at large numbers of operations. Don't get me wrong, they're great tools for what they're good at. I need to do more research before I start a port, SDFs seem like the best option but I'm not 100% confident. I am considering using your SDF library though (github.com/fogleman/sdf) but need do do some experimenting/benchmarking first.
Not sure if this matters for you or not, but my understanding (with some experiments) is that the "slicers" implicitly do a union. As in: you could have an STL with a bunch of overlapping blobs and the 3d printer slicing code just checks isInside -- which is effectively a union.
At least that's what I found when I was generating STLs in code.
Will this is incredible! I’ve been learning openscad and making 3d models with it using python and this is such an inspiration. Check out the latest nightly version of openscad it’s significantly faster, like seconds vs multiple minutes on my exports.
This is really quite interesting and similar to a project I'm working on. I've been using procedural generation to generate a marble dexterity track similar to a Perplexus. My tools are mainly Python, the Build123D library and a 3D printer.
Beautiful, I wonder what kind of craziness would be possible with this, at scale. Whole buildings being printed and assembled block by block. Real life Minecraft, if you will
Blame! is a manga where in the future humans have robots that build, and are controlled by people with Net Terminal Genes. Something happens and those humans die leaving the robots building non-stop procedurally for eons. By the time our protagonist moves about in the world, its said the Megastructure reaches from Earth all the way to Jupiter.
There are serious efforts and working prototypes of printing houses. This works surprisingly well, allows construction in days instead of months, and shows a lot of promise. It’s a great rabbit hole to fall into!
That guy who makes marble music recently worked out a ball funnel that uses inserts in a different material to eat most of the noise. But in that case it’s also eating momentum as well so I’m not sure how that would work for this design. Maybe some bushings in the supports to reduce harmonics.
Now my brain is doing it's usual "over complicate things to the extent a project cannot possibly get started" thing"...
Active noise cancelling. Vibration detectors on oscillating parts of the track with LRAs or similar actively driving opposing vibrations. Might be able to use whatever the cheap active noise cancelling electronics headphones have? Might be able to use a high speed camera and video motion amplification to work out the best places to deploy it?
How you mount things matters a lot, and adjusting shapes to prevent harmonics might be something this guy could add to his algorithm.
I saved a couple friends in college from getting into fights with their downstairs neighbors by finding them milk crates to set their speakers on so the bass doesn’t all end up in the floor. Isolating from the base or making the base of TPU could likely help.
I bet this would look baller with green rails and brown supports. Also might help with losing the balls visually as they get to the bottom. The visual noise makes them harder to track.
Designers of marble fountains who don't use computing to design the paths run into reliability issues: sometimes balls derailing out of their track. They have to observe the contraption, identify problems (balls getting jammed up or jumping out) and then guess at the root causes and make manual adjustments.
That's the thing here: he has it running for hours presumably without any ball jumping out.
Most of the tracks consist of two rails, so the ball has two contact points. I'm no physicist but it seems like the goal would be to have ideally nearly equal forces at the two contact points at all times during the ball's descent. In other words, the track has to be perfectly banked so that the gravity and centripetal acceleration vector are balanced by a normal vector perpendicular to the rails. During a derailment, the ball has to lift away from one of the two contact points, so the normal force must have dropped to zero.
It's actually much weirder than that: banking changes the axis of rotation and thus kills the rotational inertia. The tracks bank super aggressively in order to prevent the ball from accelerating too much and hopping the track. This is part of why the descent is so smooth and all the balls move at more or less the same speed.
Also to be fair the final system does lose a ball every 30ish minutes. The tuning was largely me staring at the run or taking a video trying to catch where they get lost. Instead of hand tuning I would just update the generator and print another one. I'm considering closing the loop with a camera but that would be a whole new project.
For roller coasters there is a software for simulation. It is imho similar situation compared with balls in your Marble Fountain
https://www.nolimitscoaster.com/
First, I thought about Ansys or CATIA software but I couldn’t find any module specialized for simulation of balls.
But I think that people from those companies could help as well and participate in simulation as an interesting usecase. (These software are expensive for personal projects.)
Well except for this is SIM only whereas the OP (WillMor) is making them for real with a 3D printer!
My point was that these software could help to find weak parts in trajectory - so instead of trying to figure it out by looking where balls are too quick to fall from the ride - you can simulate it. I saw real tramway simulation done in Ansys.
I think the physics are different, a ball is basically a car without a differential, so it's going to behave differently on the tracks. I'd imagine the ball is harder to simulate because of that.
One of the results for hilbert curve marble tracks, mentioned elsewhere in the thread, was a video showing how to make one in blender, which has a physics engine so it can simulate it pretty well.
https://www.youtube.com/watch?v=8YeXyUNCnhM
I'd imagine that the 3d-printable models could be imported into blender, so it's 'just' adding balls and motion to the lift.
You can simulate everything in these professional (and expensive) software.
https://ansyshelp.ansys.com/public/account/secured?returnurl...
But for hobby purposes I would suggest to contact some university, they have such software, and they could find simulation of balls motion at marble fountain interesting for research (and educational) purposes.
Does the temperature of the track change much after thirty minutes?
I haven't actually measured it but that's a good thought, I may borrow a thermal camera and do some testing! It's not noticeably warm to the touch but this functionally a system that converts potential energy into heat and sound so there's probably a measurable change.
Good thinking! Although I think that would result in a change of the failure rate, whereas in this case it appears to be constant.
Could (or do) you include a catch basin at the bottom to automatically return the odd errant lost ball to the queue?
Not to dimish the achievement, but TFA is pretty clear about the limitations of the piece:
> I was able to get it working consistently, although it did lose 2-3 balls an hour and could only run for a few hours without the motor overheating.
IMO that's more impressive to hear than if he hadn't mentioned it at all. (I would have assumed more marbles getting lost.)
> That's the thing here: he has it running for hours presumably without any ball jumping out.
You can see a ball on the ground at the end of the video :-)
You are missing inertia!
The state of each ball can be described by 9 parameters: the current location of the center of mass (x,y,z), the current linear velocity (vx, vy, xz) and the angular velocity on 3 axes.
I don't think the forces acting on the rails need to be similar -- they just need to be such that the acceleration of the ball is always parallel to the track. Unfortunately the equation of motion will look pretty ugly and optimizing the system will be quite a challenge.
And finally, the system has to be stable, ie. small perturbations should be cancelled rather than grow - if a ball gets a little too fast there should be something like a bend that slows it down, but that bend should at the same time not slow down a ball that is already too slow...
Another parameter - as a track designer you can manipulate the width of the track to change the ball speed. It raises and lowers the ball on the track, changing both the rolling diameter and the center of gravity. This can be used to make subtle changes to the ball speed before a turn.
Just nitpicking, but there is at least one ball next to his contraption in his video :-)
Doesn't make the whole thing less remarkable.
My naïve guess would be that you can't change the route of the ball without asymmetrical track pressure.
This is beautiful. It would be amazing to have the tracks encode/decode audio, you know? Like, the track of the marble can be used to generate different frequencies...
I actually attempted this, the idea of a python script that converts a midi track into a marble run is just too good to not try. I printed a large drum with different track structures inside so I could test various "slopes" by changing the speed and it just doesn't work, the balls bounce around too much to get an audible pitch. A less rigid material or a larger bearing would likely work better but I decided to focus on getting the normal version working well.
I am not surprised you tried given your earlier project :) https://hackaday.com/2022/09/25/this-found-sound-organ-was-m...
Maybe at the bottom your marbles could land on surfaces with different accustic properties. Track selection would determine the surface and release time would determine the timing.
Did you try changing thickness of rails instead of bumps to produce different sounds?
There's a Tom Scott music about a musical road in California: https://www.youtube.com/watch?v=Ef93WmlEho0
The particle simulation approach to generating an organic "tree-like" support structure is super creative! If I'm understanding correctly, you defined some laws of physics and then ran a simulation with the "time" dimension mapped to the z-axis? Is this a well-known approach, or something you came up with?
Either way, it produces a beautiful aesthetic. I'd love to play around with this idea.
That's pretty much it! It's the simplest method of supports I could come up with that allows for robust keepout zones. I did have a bunch of issues at first with supports blocking the path but with a little tuning it became surprisingly consistent. I doubt I'm the first to come up with it but I have not seen any similar systems.
Thank you! The emergent forms are much more interesting than they have any right to be for such a simple system.
> The emergent forms are much more interesting than they have any right to be for such a simple system.
hehe I wonder if this is how evolution in nature "comes up with" beauty :-D
This is a great example of a good use case for 3D printers. The smooth marble run action combined with the interwoven organic forms would be a huge PITA to fabricate with any other method I can think of, even if your just making one.
A good use case for 3D printers is random, small, custom household items that greatly increase my quality of life, much more than it is a unique sculpture.
What printer do you use for that and are you happy about it?
I'm not the person you asked, but depending on what kinds of quality of life improvements you're looking for, your budget, your 3D design abilities, and your tolerance for working on the printer versus printing with it, the answer will vary in terms of what works for you.
If you're casually interested, the Bambu Lab A1 combo will do most things you'd want it to do, fairly reliably, but with a closed ecosystem.
If you want something more robust, go for a Prusa, but be ready for a more hands on experience.
If you want an entirely customized bespoke with a high learning curve, go for a Voron.
If you want something more open source go for a Prusa. I don't buy that a Prusa is more robust than a Bambu.
I used to have a Prusa Mini and now have a Bambu X1C and it is a world of difference. I would never go back.
Really curious to hear a few of these!
I'll give you an example from my life. I got a set of cheap LED lights for a closet. They come with a little remote control to turn them on, but I didn't have a good place to put the remote control, so I made a little wall holster that's sized exactly to hold the little remote.
I also got one of those SimpliSafe home security systems. It came with a door sensor, but the sensor didn't quite fit our door frame. So I printed a tiny piece whose dimensions exactly matched the SimpliSafe and my door frame, so it allows the parts to meet up but doesn't look weird.
Of course, 99% of what I print is useless stuff that looked neat on Printables, but sometimes I make stuff that actually serves a purpose!
This is cool! Great job on the video - Simple voiceover, synced music, and the fountain speaks for itself. Bravo.
Super cool! I would love to see a white / clear one with LEDs. Rainbow road :)
I've actually done clear prints with LEDs installed. The bottom is much brighter than the top and it just look kinda tacky. I briefly hollowed out the supports and tried running fiber optics but it didn't help much.
I'm realizing now that I tried a lot of weird shit during this project that just did not work at all or make it into the final product, I should do another video just of all my failed abomination marble runs.
Minor suggestion/request: would be great if you added a final STL file to the github repo of a working example. Might be easier for people to try if they can't get the python code running on Linux.
(I haven't tried yet. But I'd love to just send an STL to my printer to see how well it prints.)
Please do! Also just some footage of the machine running from specific angles for a few minutes would be lovely! Nice work! :)
White/clear is just a matter of picking filament, I'm not sure LEDs would be easy to incorporate into the build
I appreciate the work. It's really beautiful and checks so many of my "oddly satisfying" boxes as a builder. It seems it hit those for you too, obviously.
Separately, the timing of seeing this is uncanny. I've been using marble runs to explain probability to my kids and was filming a marble run conversion lesson. Seeing this at the top of HN felt like someone was reading my minds.
Mesmerizing and beautiful in a simple way, I really like this type of thing.
Mesmerizing is the right word. "I can watch them for hours" was the key bit for me - I have always been fascinated how humans can stare at a random visual generator for forever, if it's the right one.
I think there is an instinct built deep in our lizard brain somewhere for this. Humans will happily stare at a fire, or an ocean, or a wave in a river, or (sometimes, especially children) a TV screen - and all I have worked out why is because it is constantly changing in an unpredictable way.
This marble run shouldn't even be unpredictable - clearly the paths are fixed and the cadence of balls is regular - but somehow it is still mesmerizing.
Be sure to turn on sound for maximum mesmerization.
Very cool! I've designed a lot of Hilbert Curve marble tracks, using OpenSCAD and python
Small world, I opted for a Gosper curve myself, but when it got too big, cut out specific pieces and connected those.
It gives everything a uniform look while allowing to fill the space in a different way.
That sounds super interesting, do you have a link?
Not the person you asked, but look up "Hilbert Curve" on Thingiverse, Printables, or Thangs and I bet you'll find somebody who uploaded a marble run with more information for you.
I wish there was a write up of how some of the code works. There's a lot of Python in the repo. Looks fascinating, seems to use Python to generate OpenSCAD code, I think.
It's specifically using SolidPython2 to generate the models. I originally wanted to do a quick code cleanup and have the specific math be much clearer but by the time it was working that was an absolutely massive undertaking. If I touch the code again I'm probably going to refactor the entire codebase and use a different 3D engine.
Have you considered Build123D for CAD code?
I am also procedurally generating marble tracks and 3D printing them for about a year now and found that library very useful.
The community is very active and its very similar to features we know from Fusion360/SolidWorks but all in code.
What did you not like about the SolidPython2 / OpenSCAD approach? What would you want from a different "3D engine" for this?
Mostly speed, I'm mostly doing large boolean unions of primitives or chain hulls and OpenSCAD chugs pretty good at large numbers of operations. Don't get me wrong, they're great tools for what they're good at. I need to do more research before I start a port, SDFs seem like the best option but I'm not 100% confident. I am considering using your SDF library though (github.com/fogleman/sdf) but need do do some experimenting/benchmarking first.
Not sure if this matters for you or not, but my understanding (with some experiments) is that the "slicers" implicitly do a union. As in: you could have an STL with a bunch of overlapping blobs and the 3d printer slicing code just checks isInside -- which is effectively a union.
At least that's what I found when I was generating STLs in code.
Will this is incredible! I’ve been learning openscad and making 3d models with it using python and this is such an inspiration. Check out the latest nightly version of openscad it’s significantly faster, like seconds vs multiple minutes on my exports.
This is really quite interesting and similar to a project I'm working on. I've been using procedural generation to generate a marble dexterity track similar to a Perplexus. My tools are mainly Python, the Build123D library and a 3D printer.
Beautiful, I wonder what kind of craziness would be possible with this, at scale. Whole buildings being printed and assembled block by block. Real life Minecraft, if you will
Are procedurally generated rollercoasters a thing?
Blame! is a manga where in the future humans have robots that build, and are controlled by people with Net Terminal Genes. Something happens and those humans die leaving the robots building non-stop procedurally for eons. By the time our protagonist moves about in the world, its said the Megastructure reaches from Earth all the way to Jupiter.
Also, the movie Fracture features these cool marble machines. https://www.youtube.com/watch?v=Y-v6E9H6nh0
Back when movies were made with unique good scripts and not marvel slop.
Where did they get the material for it?
They never really specify beyond the planets themselves being consumed wholesale. Some of the structures are just hollow spheres: https://preview.redd.it/7tvkbj5bp2hb1.jpg?width=1951&format=...
Also conceivable that meteors etc crash into the megastructure providing it with endless resources.
There are serious efforts and working prototypes of printing houses. This works surprisingly well, allows construction in days instead of months, and shows a lot of promise. It’s a great rabbit hole to fall into!
Does it though? I have yet to see a 3D printed house that would be cheaper than SIP panels.
This is awesome! I'd love to print one for my office. Any chance there is an stl around I can print without having to get the script running?
Stl?
Awesome project, I love to read about all kinds of different procedural generation approaches!
Stupendous project and video as well! The music is very complementary.
Amazing on so many levels!! Thank you also for including the source.
"ramble about marbles"
nice one!
A pun? I don't get it, could you explain?
Anagram, perhaps.
I meant same letters, ramble -> marble
Ah, nice, thanks!
I wonder if you could get it to run silently, or close to it.
That guy who makes marble music recently worked out a ball funnel that uses inserts in a different material to eat most of the noise. But in that case it’s also eating momentum as well so I’m not sure how that would work for this design. Maybe some bushings in the supports to reduce harmonics.
Now my brain is doing it's usual "over complicate things to the extent a project cannot possibly get started" thing"...
Active noise cancelling. Vibration detectors on oscillating parts of the track with LRAs or similar actively driving opposing vibrations. Might be able to use whatever the cheap active noise cancelling electronics headphones have? Might be able to use a high speed camera and video motion amplification to work out the best places to deploy it?
How you mount things matters a lot, and adjusting shapes to prevent harmonics might be something this guy could add to his algorithm.
I saved a couple friends in college from getting into fights with their downstairs neighbors by finding them milk crates to set their speakers on so the bass doesn’t all end up in the floor. Isolating from the base or making the base of TPU could likely help.
It looks like Bones from Hades 2. Beautiful, and super cool.
Is the banking of the curve for a specific velocity?
It says that the banking is intentionally excessive.
Why isn't the top-down footprint a square?
Up next: https://www.youtube.com/watch?v=IvUU8joBb1Q
Is it printed / sintered in metal?
Looks like fiber infused filament.
This is magical. Thanks for sharing!
This is absolutely brilliant
Maybe it’s the color and this would look better in a brighter shade, but I hate it. It looks wrong. Malignant.
I bet this would look baller with green rails and brown supports. Also might help with losing the balls visually as they get to the bottom. The visual noise makes them harder to track.
!play
Just wow.
Gorgeous!
so cool!
[dead]
[dead]
[dead]
I would pay silly money for one of those things on my desk