It's been almost SEVEN years of constantly switching between "IT'S SO OVER" and "WE'RE SO BACK" I'm no longer surprised when the pendulum swings over anymore.
@@XIIchiron78 He will reject every design anyone sends him cause it wont be to his level of perfection. He wants to make random perfect and it will never happen and he can't get out of his own way long enough to realize it. But its what happens when a musician tries to be an engineer with no willingness to learn how to be one, if he did he would realize and learn it will never be perfect there will always be bugs to work out he is just unwilling to compromise. He is making him self into Sisyphus and he may just be enjoying it so he wont stop.
@@NGabunchanumbers god these comments are unfortunately so true, yet so depressing. I still sometimes tune into this channel but I've been immensely frustrated - I've supported the "I Believe" for years with fandom and bought merch and stuff and then Martin's never ending circle of pursuit of perfection > failure > realization > attempt to create a good enough solution > falling into pursuit of perfection again really drove me off.
The thing I like best about the "each instrument has its own path back to the top of the machine" is that instruments can be tested in isolation, and the path will behave the same way that it would if all instruments were playing. Part of the problem with MMX was that the behaviour of the machine changed when more instruments were playing: more marbles using the return lanes, dividers, lifts, etc., and when that happened, it tended to break because those features hadn't been tested at that flow rate. If an instrument has its own flow path, then you can get it working at its own maximum flow rate, and then more or less forget about it when you move on to the next instrument, because, unlike MMX, adding another instrument shouldn't affect it.
Okay, but the source of that problem is also a core part of its fundamental design that cannot be "fixed": It's one machine that plays multiple instruments at the same time. Once again he's forgetting what he's working on, making choices that fight against things he can't change. If he wants to go in this direction, he'll end up making a bunch of separate single-instrument marble machines, which is extremely inefficient and sinks what's actually impressive about the project.
@@Mark_LaCroix I think one machine that handles multiple instruments is still one machine. it is true that each instrument wont be "connected" through the marble channels, but its still one machine handling the movement of those marbles
I agree with this sentiment, though I think there is an extreme form of this where he doesn’t end up with a single marble machine so much as he does several marble instruments. I’m sure it would be fun to watch (reminiscent of Pipe Dream from Animusic), but it feels less like a single wondrous instrument. Maybe that’s for the better though. He clearly thought the MMX wasn’t worth the insanity.
Yeah, i had kinda lost interest in the proje t from the original sketches saying hes building several seperate instrument modules. Not as cool anymore.
@@antiseth3964oh for sure. Like if ever instrument is completely separate, then he's just controlling a bunch of machines. But it seems that for right now, the power system and programming is all still central, so I would consider it a single machine
Silicon has a different coefficient of friction and could change the jam behavior. But I wonder if adding some rubber glue on the bumps or screwing to the plywood with felt underneath would improve the sound.
This definitely needs a few more likes and comments to move up. This is an important test to conduct before moving to a larger-scale production of dividers
Sound would still transmit through the screws unless the same rubber shock absorbers used in some CD drives and other shock-sensitive devices can be implemented here.
@@custos3249 You could use a spring retained screw to maintain position on the assembly but provide some dampening between the screw and the track in combination with a felt or rubber backing layer. Felt is a pretty cool building material for stuff like this, so that's what I would go with. but either should be effective.
This marble divider discussion reminds me of simpler times in my life ❤ These videos played a bigger role in my life and career choices than i give it credit for.
14:42 blink and you'll miss it, but the type 10 divider already had a potential permanent jam issue on the left most lane. You'll see it again around 14:48. I suspect the 'wiggle' method is introducing a much smaller but stacking chance of error on each extra bend it uses. Sadly, I have no good ideas beyond an agitator of some sort, but that introduces complexity no matter how it is implemented that will likely add an even more annoying source of error. Tighter tolerances perhaps? But I suspect that is just going to achieve a reduction, not elimination, before it simply jams from friction. If there are never any marbles pushing down on a splitter when it is not in action, it would allow for much more reliable designs to be put into effect. So perhaps double gating, one at the note playing end and another synced with it at the feed-in for that splitter? I'm still not fond of that as it introduces more points of failure and complexity. Sorry, I'm not particularly helpful here.
I also noticed it but realised that it was due to his makeshift marble gates. He had stopped them with half a marble through, which would never happen with a true marble gate. This upset the height tuning, causing the jam.
Well spotted! I actually think that any design without moving parts where you can get unlucky enough to balance the ball on the middle (pin or ridge) may potentially jam. And I think making it out of silicon will increase friction which would make balancing a single ball ever so slightly more probable. Perhaps the gate must be intentionally asymmetric and the weight of the incoming balls pushes balls to another channel over a small ridge on the outside curve? That way the divider would always fill one side first but as long as you feed enough balls, it will fill out all channels.
A row of marbles waiting in any kind of track will never have a fixed length. This because there needs to be some room around the marbles to reduce friction, and this room will allow marbles to configure themselves in different ways with different lengths. The only solution is a "height insensitive marble divider" indeed.
@@jurjenbos228 It seems to me that the "fixed length" requirement was about "difference must not exceed half the radius" instead of "length must be identical". I totally agree that the latter cannot be accomplished, but the former can be accomplished in most cases with adequate design.
Knights of the Divider i Call upon thee, ride forth and search the Holy Grail Marble Divider! Follow the trail of the Design Requirements into the deepest design spaces. Find me the Holy Grail Divider and your name shall be written into marble machine history forever! See details in the end of the video, do not email suggestions only send 10 second videos, including your credits on the video screen to the link below. Submit your 10 second video submission here: www.dropbox.com/request/pZO96wQ4I35eQkqZvIDz Regardless if we can solve the height dependency or not, i do think the 2D Y-split is the best option i am aware of. So much simpler and so much better than the solution on MMX, my current thoughts from all the info i have available today. Check out the videos from Mr Gonzonator: ruclips.net/video/AfHOlcR3Vso/видео.html ruclips.net/video/zmhG4W6Oavo/видео.html Check Out the video from Jesse Roelfs: ruclips.net/video/YW-nDEG2PBA/видео.html
Why not just go with the prototype by Rossero from the video "Engineering Basics: The Best Part Is No Part"? It was height-independent. Just cut it down to 4 lanes.
Could you explain the 2d requirement? I feel like that's not the actual requirement, that's an assumption based on another underlying requirement, but it's not clear exactly what.
I think having separate loops for banks of channels is the wisest design decision you could have made for MM3. The story of MMX essentially revolves around the troubles that came out of the sheer number of marbles some parts had to be able to handle in a single thread. MM3 is going multi-threaded. As a software developer, I approve! 👍👍
Now he just needs to learn one of the most important tools we software developers have: "Most problems can be solved by adding another layer of indirection." 😉
@@GeneromanHe could use different marble color for each channel of the machine. If a marble drop on the floor, he just have to look at the color to find the source of the problem.
@@DaVoKanfrOooh that'd be really handy! It'd also keep him from having to somehow count how many are in each one to get loose marbles back to where they need to go to keep the number of marbles in each loop constant, just look at the colors and they're all back where they belong!
A fill gauge would be another option. If he has enough marbles in each lane it doesn't make much of a difference which lane he puts it in unless the particular lane is running below his low level fill mark.
I noticed you didn't address an old enemy of the MMX: marble pressure. While it was primarily an issue for the gates, these divider designs look like they have the potential to operate differently when under load.
@@karlvestklev unless he builds the machine with a large gently sloping feeder they will all have a load on them and will operate differently causeing him to yet again scrap it and start over.
@@karlvestklevthe instruments that are used the most (e.g. Hi hat) would have a higher load so yes I guess this is a valid point that should be addressed
I love it when you incorporate music into your videos. Haven't seen that in a while and missed it sorely. You're a talented musician and it would be great to see more of that side again.
How many more years is it going to take before he finally designs some kind of catch-pan below the tables he works on to catch falling marbles on a daily basis?
Design? Buy $20 worth of gutters, a drill and 15 minutes and he will never have a marble roll off the table ever again edit: or even just do a snooker-style table where the edges are raised
Ok we have seen these marble divider tests before right? But we can finally see at 14:20 that Martin screwed a ridge to the bottom of the wood board, so that the marbles don't fall to the floor or need to be caught in a container! That is already a step up in my book!
I personally think the height dependency issue is a major reliability concern. If, for example, due to stacking a longer marble path is a bit inconsistent, it could cause issues. If marbles have minor size differences it could cause issues and if the parts during their life just stretch / wear a little bit it could cause issues.
If he uses silicon or any other sound dampening material for marble paths then marble pressure can have major effect on marble position. Also wear and tear after 10 000 or 100 000 marbles will mean constant jams since different paths wear to different lengths and height is no longer exactly same.
Assuming he's using mass-produced industrial ball bearings, the uniformity should be far in excess of what he needs. The tolerances are measured in thousandths of a millimeter.
I took a break from the wintergatan videos like a year or more ago and I cannot express the hillariousness of feeling like I came back to the exact same problem I left on just an x number of iterations down the line. Loove the work spirit! Keep it going!
You can tell that Martin understands the engineering mindset when he knows that telling engineering types that a solution doesn't exist and that "you can't do it" is almost a guarantee that someone will do it and come up with a solution.
@@beeble2003 , this only demonstrates that you don't understand the engineering mindset. When you are pushing the boundaries and doing something new, failure is always an option. Taking those failures and learning from them is at the core of the necessary mindset. The easiest thing in the world to do is just play safe and not stretch your abilities. If you are not failing, you are not trying hard enough. Also, your definition of success and failure is skewed. The success criteria for the original marble machine was its ability to play a tune. As his famous video showed, that low bar was met, regardless of what broke or how many marbles it threw on the floor. Martin was very clear that MMX was to be taken on tour and had to be used in live concerts. It had to be infinitely more reliable than the original and as his testing showed, not throwing any marbles on the floor was a deal breaker. MMX did not meet those far higher requirements even though it was far superior to its predecessor. Taking what he has learned and injecting it into the third version is admirable. To me, as an engineer myself, I was most concerned about his emphasis on the physical appearance of MMX over basic functionality. This was one of many lessons he has learned.
@@briancampbell179 "Taking those failures and learning from them is at the core of the necessary mindset." I don't think he's done that. His solution to "component X doesn't work" is always to try to redesign component X ad infinitum, without ever considering the bigger picture. As a result, he designs over-sensitive systems where everything has to work perfectly and every subsystem interacts with every other, in ways that are impossible to control. He seems to have no understanding of tolerances or isolation. "If you are not failing, you are not trying hard enough." Sorry, but that's trite nonsense. The goal is to get a working product. Obviously, some things don't work, but continual failure is an indication that the team isn't capable of achieving the goal. Either because the goal is impossible, or the team isn't talented enough to achieve it.
@@beeble2003 , that is the exact opposite of what he is doing. This very video starts with the problems of relying on such overly complex mechanisms and how he needs to make them more robust. He has gone from marble dividers with moving parts that were prone to jamming to a simple design that has no moving parts and demonstrated not to jam through experiment. I love his new approach of experiment and measurement. Yes, he builds multiple designs of an element, tests it, figures out what is going wrong, and improves on it. That's how you learn what works. If you think you can just sit down with a pen and paper and draw up something that works perfectly first time, you are a fool. Even with MMX, Martin was redesigning elements that weren't working and replacing them with simpler, more robust designs. Unfortunately, it was not modular in structure placing limits on how far he could go. While I'm sure letting MMX go was a heartbreaking decision, it was the correct one. If you class that as a failure, the fine, but it was only one. Nothing even close to the "continuous failure" you falsely claim. It is also worth noting that when Martin replaces an element in MMX, in the vast majority of cases, he replaced something that didn't work with something that did. One replacement for each element. Again, far from "continuous failure". He has never "redesign[ed] component X ad infinitum". Finally, the structure of the next Marble Machine is all about considering the bigger picture. His design philosophy is to use simple robust designs proven through experimentation in a modular structure that allows their replacement should he have to address an issue. I am confident that this "belt and braces" approach will almost guarantee success.
@@beeble2003Until recently, he had been redesigning components without considering the big picture, because he kept redesigning them to be fitted onto the existing machine, which had fundamental flaws that he had not considered the "big picture" for. It's clear that he has thought about the big picture now because he scrapped the previous machine entirely. Now, he's focusing on first principles and trying to design subcomponents that work well on their own, and can fit into a new larger vision. For example, his comment about using multiple marble divider loops instead of sharing one loop for everything. This vastly changes how he might design these. How is this not "considering the big picture"? The fact that you suggest something like "the team is not talented enough" as if *talent* is what is required for a team to be successful suggests you haven't done much real engineering or solving hard problems yourself. Your language sounds like that of some useless middle-management type who is more concerned about watching their own back and being "correct", then actually innovating or creating real value. Martins goal is not just a "working product", it's creating something that the world has never seen before. Something that people think is impossible, until it isn't. It's called innovation, and it's a process that requires a LOT of trial and error with a lot of ambiguity. Great engineers can handle that ambiguity. Ass-licking MBA's that get hired at slowly dying mega-corps cannot. I'm pretty sure you're not an engineer. Managing them does not count.
The best marble divider is no divider at all. You already said it, if each channel has its own closed loop, there is no need for a marble divider at all. It could solve a lot of issues like the marble hight. And it has a super high flowrate. For n channels just build a n-wide marble lift, and you are good to go.
There is a reason why dividers exist in life: it is too much to build one lane for every single channel. Sure, you can reduce dividers generally by increasing the number of channel lanes but I don't think you can completely eliminate them.
In the video, he said each instrument will have its own separate loop but each instrument has multiple marble gates. It’s not feasible to complete separate all the channels because it’s pretty much impossible to create separate catchments when they marbles are hitting almost the same spot on the same instrument. Martin did say he was splitting the marbles into four because there’s four channels on the snare.
One notable thing about the design from 3:20 that does not seem to be incorporated in your 10 iterations is the tapered sides of the marble channel. Your side walls come up to the widest part of the marble and are very constrained (greater than or equal to 1 marble radius in depth). It looks like the marble channels from 3:20 only come up about 1/4 the diameter (1/2 radius) of the marble. Perhaps the slope on the sides of the marble channels makes them harder to get stuck by allowing the marbles to (slightly) climb up the slope on the sides. There is definitely an argument about how this isn't keeping the marbles in "2d", but maybe that little bit of Z tolerance that what you need to make this work more consistently.
I think playing around with a 3D marble divider could be a key to success. Adding a third dimension to the marble divider might help solve the height sensitivity issue. For example, a shape like a tripod might help leaving more room for the marbles to fall in one of the three paths. Spheres pile up and arrange themselves better in 3D than in 2D.
Suggest: Much like pinball machine... a peace sign shaped splitter (an upside down 'Y' that toggles each marble. To stop jams at the side that fills up... when a marble path gets close to full it is use marble pressure to push a pin/tab to block the 'Y' from turning... it will have to pop up whenever a marble passes over it, but that will help moderate the flow of the marble source. When a marble is used it will allow the 'Y' to use that path again. If both paths are full the 'Y' is blocked in the middle till either side is empty.... also, if source path is full, it can be gather to allow other more needy paths take the added load of excess marble.
Every time you drop a thousand marbles on the floor, it's so painful, but I'm sorry to admit so funny. The sound of marbles hitting the floor is a dark, beautiful punchline. In any case, I really enjoyed this video - fascinating stuff.
Flat bottom seems to be a real problem. The wooden version had a rounded bottom that kept things in line better. Also, I think having the marbles drop straight down introduces more problems. If you tilted each divider X degrees clockwise or counterwise it would bias the marbles to one side or the other.
@@LaughingMan171 I agree, I am suggesting he introduce an intentional bias to one side so the marbles fall in a more predictable way (always to the left or right). As it is now, straight down with a chance to sometimes go left or right. If you introduce a bias, then you would be able to more easily predict that the marble will do every time.
@@joeo6378 The problem with the MMX was that he had back-pressure problems controlling marble flow. The longer the transition between marble gates, the greater the increase in weight of individual marbles pressing into particular channels. This meant that the gates had to deal with variable pressure induced by bias because of canted valve flows. The variable pressure increased a feedback loop in the problem of variable flow, so that the chances of a jam on certain channels was worse. The marble gates had to be designed uniformly so that the marble drops onto the musical instruments was timed consistently with the tempo of the songs. However the marbles did not flow uniformly because the machine was dealing with recycling the marbles from previously played notes. So if you had a slow set of notes (like half-notes) and then an extremely fast set of notes (like 32nd-notes), the machine didn't have time to catch up.
Did you consider a vibrating mechanic? When you look how they separate stuff in production lines, they mostly use vibration. So you can have a compact design which can jam, but because of the vibration it will unjam immediately. Let me think of a design solution for that but maybe you can also figure something out ;) I'm thinking of something like a big box with X holes at the bottom for each outgoing channel. The box vibrates and the channels have short bit elastic tubes before they continue as solid pipes. This design would be very compact and would be 100% secure against any type of jam.
@@benediktzwolfer4193If it's hard silicone that is springy the movements of the marbles could be vibration enough and it should be quiet if not quieter.
Great to see some real progress on solving one of the major issues of the MMX once and for all. I have to admit that I got kind of frustrated with so many videos about chasing timing perfection, even redesigning the entire power system for an improvement of at most a couple milliseconds of inaccuracy while there are way more important topics like the marble divider to work on. Because of this I was even happier to see you properly developing this, with strict requirements, possible compromises, quick prototyping and most importantly looking at other peoples ideas.
i love the fact that a great part of the viewers of this channel are basically a buch of top tier professionnals watching an artist do with an instrument what they do at their job for real life applications, better than the companies that hired them ever planned to do
The end of the video is so beautiful, I loved when you just played guitar next to the fireplace and told us to give you feedback. Keep up the good work, it is awesome!
Sorry but... what an awesome advert to get suggestions -- a guiding template for others! I'm just learning about Wintergatan and I'm a huge fan already. But coming from the engineering/programming world, I also love his approach to this journey! :D
I suspect a lot of your problem is caused by the flat rolling surface because it allows the balls to roll in undesirable directions. You got close with the added bumps, but consider central v-groove channels.
If I remember correctly, he tested a v-groove channel before. It could cause the marbles to flow too slowly, since they have to rotate more often to clear the same distance.
@@7head7metal7 And he failed to consider all design options back then too. I get the vague feeling I'd commented that eons ago. Plus arguably, he's still using the groove concept but big enough and slightly differently shaped, it doesn't seem like it. The nub he added is in the right direction, but the separator is less important than creating a bias to either side.
Back to the old school style videos. Think I must've been watching for 6 years by now. Has it been that long. Thankyou Martin. Thankyou marble machine family ❤❤❤❤❤
I feel anxious when Martin starts to adjust design to a single milimiter, completely omitting manufacturing and assembly tolerances... I wish he finally start to take that into account :)
The downside of giving every channel a dedicated marble route is that you can't "borrow" capacity from unused channels, which means all of your routes have to be able to handle the maximum possible throughput and you end up with wasted capacity when channels aren't being used. Although that may not be a significant problem, it means wasted energy to move parts that aren't carrying marbles and extra noise coming from parts that aren't contributing to making the music. Combining the channels means that when one channel isn't being used, that capacity is still being used by other channels. Having a large "reservoir" of marbles can allow for a song to use more than the maximum throughput of marbles *temporarily* (until the reservoir is empty). You could also add a secondary lift mechanism that only engages when the reservoir is approaching empty, although that's probably more complexity than it's worth. As always, I think Martin is struggling because he's unwilling to accept the limitations of his machine. He should be embracing those limitations because they are what make this a unique instrument. Instead of trying to make a machine that can play all possible music, Martin should build the machine first and then write music based on what the machine can do.
He has a creative vision for the machine and he shouldn't compromise on the music he wants the machine to make, the music he wants to make is not a stupid design requirement. He should only look at the most sane and simple way to achieve his vision. I do agree with your point that it might make sense to eliminate the marble dividers altogether by having a closed loop of marbles for every single channel. I guess some queueing theoreticians should weigh in.
Agreed. He keeps hyper-focusing on one design requirement to the point that 3 or 4 others start falling apart. Having multiple duplicated closed loops of marbles has got to be the worst choice he's made yet in this regard. What happened to simplicity? Reliability? Deduplication? Usability and maintenance? This is on top of things he's not even thought of yet, like the time it takes to load the machine before each performance on the world tour. He's now got to divide the marbles correctly between each loop. Will it be the same number for each? What's the error tolerance on that? Once he figures all this, he'll go back to the drawing board again and hyper-focus on that, forgetting all the things he cared about in this video I want him to succeed, which is why I can't stop watching, but it's so exhausting seeing him *not* learn the exact lessons he's told us about over and over.
But the problem with the single reservoir is how to divide the marbles to the instruments that are using notes, at the rate they are using notes. It's not practical to require that much throughput on a single divider without jams. More throughput = more marble speed = small imperfections causing big problems. Why do they all need to go through a single input to be divided? Because all other options require random distribution of marbles, which can cause lanes to run dry.
@@LReBe7 Re: the first point: That's how *all* instruments work, though. You can't force a snare drum to have a variable pitch so you write music that uses a timpani to vary the pitch and a snare for the snare sound. You can't force a trumpet to go as low as a tuba so if you need a low brass note you use a tuba because that's what it's for, and you use a trumpet for the high notes that a tuba can't do. There is such a variety of musical instruments in existence because no single instrument can do everything and musicians have to be cognizant of the limitations of the instrument they want to use. In this case, I respect Martin's vision but so many of his problems arise because he is unwilling to compromise the capabilities of the instrument because he wants his instrument to fit the music instead of fitting the music to his instrument. He keeps talking about designing from first principles and form from function, but he isn't applying that idea to his music, only to his machine. The first principle should be what the machine is realistically capable of and the form of the music should follow the function of the machine, not the other way around. Just like the form of music played on a saxophone follow the function of a saxophone, instead of trying to force the form of the saxophone to sound like the function of a guitar. Re: the second point: I'm arguing for the opposite, that he should limit the number of marble routes and combine many instrument channels into the same routes so he can more efficiently use space and energy to lift marbles. It would also reduce parts and probably reduce the number of moving parts. Although I'm no theoretician, I do know that studies for queues at stores invariably move faster when there is one line that separates into many checkouts instead of having a separate line for each checkout. When one checkout slows down, the whole line continues to move because it diverts to open lanes as needed. Of course, this is possible because humans sort themselves and don't jam. I think the best solution is a hybrid where the channels are grouped into routes based on how much flow they're expected to have. Instead of a route for each instrument, or one single route for all of them, have a few routes. I also think Martin should think more about where his "reservoir" of marbles is and how big it should be. The bigger the reservoir is, the less critical it is for the marble lift capacity to keep up. At the extreme, a huge box of marbles at the top could hold enough for an entire song and then you stop, refill, and play the next song. That's boring, of course, but the idea is that you don't need a 1:1 ratio between using marbles and lifting marbles at all times, you just need an *average* 1:1 ratio across the entire performance. Again, if we consider that the form of the music should follow the function of the machine, Martin could plan the performance to alternate between high-intensity music that uses more marbles than it lifts, depleting the reservoir, followed by a low-intensity song that lifts more marbles than it uses, refilling the reservoir. Separate routes means separate reservoirs.
@@RhynoD2 the entire point of making this machine is to push the boundaries of what music can be played in an entirely mechanical fashion. Of course he could limit his machine to basically be a preprogrammed carillon, but have you ever heard how sloppily a carillon plays? At some point, simplifying the design starts to chip away at the ambition of your project and it just becomes a stale rehash of something that already exists. Plus: Martin is in the prototyping phase and he is prototyping in a modular fashion. The lesson I am getting from this step in the process is: maybe try to eliminate the complexity and moving parts involved in aggregating and dividing the marbles. Yes, this would require some more parts and some more space, but if those parts are simpler to design and more reliable, that trade-off might be worth it.
I've been watching since the first machine. It helped get me interested in design and engineering. Now Im going to college for computer engineering. Thanks for sharing your journey, these videos are all so good. I love the recent trend of extreme deep dives into seemingly simple mechanics. Thanks man
why not make a double release. You have a 'runway' or cache which can hold 10 marbles, and with every release from the marble gate, you allow one to go to the cache. Then you don't care how long your path is. There could be a lot of space in between...And the release to the cache can be related marble-triggered itself, i.e. if the cache has size 20, if the cache is too small, allow more to go into the cache.
Yeah. I think he screws himself by trying to have the machine run entire shows flawlessly. I think he would be better off just "topping off" the marble tracks every song. You have to change the programming wheel out anyway.
@@jonathanschubert9052 the problem with that is the number of marbles needed for each instrument for a song, it could easily be in the hundreds if not thousands of marbles per instrument, which if steel balls are used is a lot of weight and needs a lot of space to be stored, hence using less marbles but reusing them is probably a better option.
Ohhhh, Martin has awoken the mighty marble marble beast and it needs to be slain. I sooo cant wait to see what people come up with. Martin is so freaking impressive and a great musician turned industrial designer. Lets go community! Rock this challenge!!
How about using a flexible tube section for continuous marble path length adjustment? The tube can be laid out with S bends, so you can pull more or less of it in/out when laying it down, to find the perfect top-marble position without having to use discrete increments. Then when the top and bottom marbles are in the right place, you can cut the remaining bit of tube past the end gate ring joint.
I have high hopes for seeing Jesse Roelfs' marble divider further developed... it looks like it might pre-empt a potential faillure state that this video did not test for: the pressure of marbles against eachother put pressure on the gates and that might break it. (Additionally, when using open channels, marbles may jump out; additionally, who is to say Y-split dividers work under higher pressures). In addition it looks like there's fewer marbles knocking into eachother which should decrease noise.
I love this Divider priciple review for its great built in simplicity (for us the viewers, of course), but what really got me, was the outtro ... doing the text and the background music in a one take, one person shot is surprising in the best way possible! Along with the excellent choice, timing and nicely done dynamic camera. Thank you for doing another great step in direction of the new marble machine and sharing the thinking process that ran into the design principle!
Can easily be solved by having a cover on top (a closed tube). But the marblepressure might cause problems with the release gates. I can't remember if the current gate design is indepentend of marblepressure
This is the most encouraging video on MM3 so far. I think for this part, you've nailed the balance between seeking perfection and realistic requirements. Looking forward to seeing where you go with this!
I'm sure you could improve the flow behavior of ALL dividers by adding a vibrator device to the marble divider assembly. It would make the marbles behave like they had effectively no friction between them and it could thereby potentially make one of the earlier versions (that clogged) shoot up to the top performers. Not sure if that is a feasable solution though. Really nice video, btw!
I was just about to comment that too, a vibrator with enough force should enable the marbles to "unstuck" themself unless its a permanent clog/jam. But then again they would add more complexity, just via the question "vibration in which direction".
The version #3 looks a lot like double-stack, single-feed pistol magazine running in reverse. If you are still looking into things, that would be a good source of mature tech relating to feeding "round things" to pull from. (There are even some example of "quad stack" magazine that might be worth looking into.)
You could try to use a marble reservoir on top from which all the marble lanes come out from, no tree splitting, nor height issues, probably. As for the jamming... use a wiggler or shifter powered by the crankshaft
Having some experience with model trains, I’d suggest running a strip of cork between the printed channels and the plywood backing. It’ll reduce a lot of the noise you’re getting from the marbles rolling.
Since the machine will have separate tracks (drums, bass, etc.), you should probably design each divider so that it works perfectly for the track it is build for. Modularizing the splitter works great if this is going to be a commercial part for many use cases, but in the end it's being designed for a single machine and unique for each track. I think instead of using inserts (like the wooden ones) you should prototype with the printed ones, but for the final part, it should be printed as a single piece with all the extensions built-in.
Damn, i didn't know you had a YT channel covering all the engineering aspect of the machine i discovered nearly a decade ago. Frankly i had very low interest in marble fluid mechanic, but i am always amazed by passionate people sharing their passion, thanks a lot !
Just for fun, I'd love to see how many levels the tree divider could handle. If you could make it to 7 levels, that would be divide by 128. I wonder how the design for the first divider would have to change compared to the final level. After dialing in reliability, another goal would be maximizing flow rate. Love this project, so inspiring!
50% of my job is to repeat that no perfect solution exists. We only have perfect solutions that don't exist and imperfect solution that exist. Between these two, the second is far superior.
THANK YOU FOR WORKING SO HARD! I love seeing your videos when they come out! You inspire me to create! I look forward to the next design you come up with.
You know, in the very beginning of the new series I was worried. Martin seemed so focused on getting inhuman levels of precision out of this thing, still set on a certain power input and whatnot, that it felt like we were headed for another MMX situation. Every video since then has made me more and more confident that this machine will actually work. "I'm just going to accept that I cannot solve it (the marble height issue)" is one of the most reassuring sentences ever uttered.
So in that case each gate needs a small reservoir of its own to store up marbles for fast playing. This was the design on the MMX. E.g. I have 4 gates, each with a reservoir that can hold 10 marbles. The first is empty, the other three are full. If I have 20 marbles entering this system, then the first run through, 5 marbles go into the empty path, and 15 get cycled back. Then of the 15, 3 or 4 go into the empty track, and so on. It won't provide marbles to each track at the rate each track is consuming them, too many get wasted on each go round.
This was a lovely video, Martin! You did some really great iteration, found a sound solution, and there you were. Accepting that perfection is a dream, rather than a design goal, was a lovely note (ha) to end on. And I always love when you bust out the guitar.
Something that I don’t know if it would be a factor or not is sensitivity to the angle relative to the floor of the part. During original assembly it may be possible to get everything to work perfectly, but preparing for a concert it may be difficult to place the machine exactly level with how it was originally assembled. I don’t have the knowledge or the ability to test how big of a factor this could possibly be, but I would argue that for a design to be considered robust it should not be sensitive to how level the machine is.
Stop dividing marbles! Marbles are by their nature against flowing, and any time you divide marbles you have to combine them again. Almost every unsolved major issue you had with Marble Machine X was combining marbles, and you will continue to bang your head against the wall with this. You need a single track for each instrument. You need to keep marbles linear. That might seem more complex, but it will simplify the biggest issue you have had since the start. You can perfect a single track before moving on to the next, and never have to worry about dividing or combining again.
@@luksch154 It wouldn't be 100, it would be the same number of lanes as he has marble gates. He's already got this many lanes at the top, he just needs to have that many lanes the whole way around. I think the most complex part would be lifting them back up. He could do this by integrating the programming wheel with the marble lift, which would make sense, because a pin on the wheel also corresponds with a marble dropping down, it should correspond with a marble being lifted back up.
@@Mark_LaCroix That's more or less exactly what I'm proposing.Of course, they'll share the same power source and programming wheel, and they'll be very very close to one another to appear as one machine, but it will drastically simplify the work he needs to do: Each instrument can be adjusted and fine-tuned individually. Adding more instruments no longer affects all other instruments.The only limitations become the amount of friction each additional lift and gate adds to the system, and the weight of the marbles he needs to lift, and the width of his programming wheel. The best marble divider is no marble divider. Make your requirements less dumb. H'es spending all this time trying to figure out how to combine and split up channels, but there is absolutely no need to do that.
You need another design requirement: decent tolerance to account for wear. Plastic will be worn off, joints will loosen a bit, springs become a wee bit less springy etc. So even if the parts are produced and montaged to the highest precision it will change over time and the divider should still work.
One consideration is the weight of all the marbles in the queue pushing down. I would recommend also testing with a more marbles in the queue and see if that causes problems. I vaguely remember that being an issue before.
This is a perfect example of a "second system syndrome", if I ever saw one. You made a nice 1st machine, then you overengineered the 2nd one and barely survived, and now you're on path to build a fantastic 3rd. Can't wait to see it.
At 14:00 the adjustment of the path lengths with wiggles to achieve marble height alignment reminds me a lot of using wiggles on groups of microstrip tracks in printed circuit board design to maintain RF coherence. You could call the "holy grail" a "coherent marble divider".
the real innovation in this video is that after 5 years you finally built a table with a fence for the marbles 🤣🤣 thanks for being awesome. can't wait to see more ❣
I definitely think you're on the right track by making separate tracks for each individual instrument. By splitting into smaller tracks, you'll have a much easier time creating robust solutions.
The fascinating thing is that you're striving to make this more reliable than almost any other instrument on stage. Guitar strings break and require frequent tuning, sequencers glitch (thus the "all notes off panic button"), woodwinds and brass have condensation issues that stop them dead without player intervention. Sax and clarinet reeds crack and pads and corks fail. Ian Anderson often mentioned the stack of flutes he had waiting for repair. But here is the Marble Machine, more complex than any of them, playing millions of notes without a single mistake, jam, or marble lost. It's wild.
You could put some very shallow bumps a bit up from the divider so they are coaxed to settle into their lane. they'd need to be very brief and close together so the marbles don't get stuck. make the tip of the divider gradual to the apex so it doesn't serve as a wedge and guides more smoothly. that way it'd leave space for less jams and any possible jam will work itself out. youd probably need to make the apex higher by a little bit at least, or if there's a cover, then it'd be fine. maybe think of small flaps to guide them maybe? ive been following you since the conception of the first machine. great work.
What you're making is a perpetual water fountain but with metal marbles. If you made a machine that could rain down marbles consecutively on all channels, with no stops, that'll be the jackpot. I can imagine a hopper at the top with exactly all the marbles needed to play what you needed to play. Then, know how many marbles are needed for each individual element in the composition of the song. I can envision for one element, the hopper at the top loses a particular weight from its hopper and after losing all its marbles, below the machine, the weight increases after each conservative drop of a marble. When it achieves a weight limit at the bottom, it triggers a seesaw pulley to move the basket from the bottom to the top, trading places and reloading. A counter weight would pull the empty basket down and raise the newly filled basket from below to the top. What we need during this switching is little rails switching over simultaneously, while baskets trade places, to pivot in a new direction for the marbles to fill. No longer filling the left side of the bottom catcher but now the right side. Once the weight exceeds the limit in the catcher it repeats the cycle. I can just see all this running on pulleys and counter weights. Gravity POWER!
You improved so damn much as a designer. Not only are you a great musician, you're also a great engineer! At some point I kinda lost hope that we'd ever get to see the marble machine on tour, but I'm starting to believe again ❤
Bit of a stretch, if you noticed throughout the video all his "Nice to have" requirements became "Must have" requirements. it's still the same old wintergatan at the end of the day.
I really feel that the marble divider should not also deal with issue of the marble height at the gate, eliminating the need for variable height independence. As you have decided to split the divider into many dividers for each instrument, you should also consider implementing a new stage between the divider and the gate where the marbles drop. Something like a first gate at the end of the divider, which drops marbles to a small area before the final dropping gate. This area between, of preloaded marbles is where the marble height can vary. This also means the pressure of the marbles in the divider is not weighing on the final gate but on a separate gate which can be designed to take the load.
So nice to see the evolution, the trial and error and all the thought and refinements which goes into this seamingly simple device. Having recently seen Stranger Things 4, I had an spark of thought how you could to a divider differently: A stochastic sorter in form of a Galton board. At the bottom you could even do some overflow management as you did on type 2. Draw back might be the noise of the marbles hitting the pegs.
I think Martin was onto something with design six. What he's done with the single lane design is create something which IMMEDIATELY fails if the system shifts a little bit. Like... The wiggle lanes are inherently bad, because they're an attempt to combat tiny mechanical tolerances, when you should be trying to create a system where it doesn't HAVE to be that precise. I think if he'd iterated on six rather than just throwing the idea away he'd have done better. Have it so the single lane enters a widening channel, and the channel has two grooves which get gradually more pronounced, and a V divider in the middle gradually rises, so the marbles are teased into each lane, rather than FORCED as in Martin's design. Maybe? I don't know, I'm not an engineer
Martin, a thought. You have mechanical input which you can take advantage of for your marble dividers. If you can capture the vibration created from your governor, you can input some vibrations in the dividers which can clear any jams. It would need to be isolated some way from the marble gates, but could help with jams.
WE ARE SO BACK
It's been almost SEVEN years of constantly switching between "IT'S SO OVER" and "WE'RE SO BACK" I'm no longer surprised when the pendulum swings over anymore.
@@dvduwu ...THE PENDULUM SWINGS ARE SO BACK
Sisyphus is happy
@@XIIchiron78 He will reject every design anyone sends him cause it wont be to his level of perfection. He wants to make random perfect and it will never happen and he can't get out of his own way long enough to realize it. But its what happens when a musician tries to be an engineer with no willingness to learn how to be one, if he did he would realize and learn it will never be perfect there will always be bugs to work out he is just unwilling to compromise. He is making him self into Sisyphus and he may just be enjoying it so he wont stop.
@@MrWhateverfits he literally said he'd be fine if it wasn't perfect. that's why he gave the challenge to us.
The fact that you acknowledge that a perfect solution might not exist is one of the best things for the Marble Machine project.
I give that realization about a week before it's chucked into the garbage and he's back to chasing perfection.
Hes been saying that for five years, and then keeps going back to perfection.
...and continues to talk about Holy Grails XD
@@NGabunchanumbers god these comments are unfortunately so true, yet so depressing. I still sometimes tune into this channel but I've been immensely frustrated - I've supported the "I Believe" for years with fandom and bought merch and stuff and then Martin's never ending circle of pursuit of perfection > failure > realization > attempt to create a good enough solution > falling into pursuit of perfection again really drove me off.
@@thatguyStrike Unfortunately, he already got your money, and there's plenty of other people willing to toss cash at him.
The thing I like best about the "each instrument has its own path back to the top of the machine" is that instruments can be tested in isolation, and the path will behave the same way that it would if all instruments were playing. Part of the problem with MMX was that the behaviour of the machine changed when more instruments were playing: more marbles using the return lanes, dividers, lifts, etc., and when that happened, it tended to break because those features hadn't been tested at that flow rate.
If an instrument has its own flow path, then you can get it working at its own maximum flow rate, and then more or less forget about it when you move on to the next instrument, because, unlike MMX, adding another instrument shouldn't affect it.
Okay, but the source of that problem is also a core part of its fundamental design that cannot be "fixed": It's one machine that plays multiple instruments at the same time.
Once again he's forgetting what he's working on, making choices that fight against things he can't change. If he wants to go in this direction, he'll end up making a bunch of separate single-instrument marble machines, which is extremely inefficient and sinks what's actually impressive about the project.
@@Mark_LaCroix I think one machine that handles multiple instruments is still one machine. it is true that each instrument wont be "connected" through the marble channels, but its still one machine handling the movement of those marbles
I agree with this sentiment, though I think there is an extreme form of this where he doesn’t end up with a single marble machine so much as he does several marble instruments. I’m sure it would be fun to watch (reminiscent of Pipe Dream from Animusic), but it feels less like a single wondrous instrument. Maybe that’s for the better though. He clearly thought the MMX wasn’t worth the insanity.
Yeah, i had kinda lost interest in the proje t from the original sketches saying hes building several seperate instrument modules. Not as cool anymore.
@@antiseth3964oh for sure. Like if ever instrument is completely separate, then he's just controlling a bunch of machines. But it seems that for right now, the power system and programming is all still central, so I would consider it a single machine
Silicon has a different coefficient of friction and could change the jam behavior. But I wonder if adding some rubber glue on the bumps or screwing to the plywood with felt underneath would improve the sound.
This was my first thought.
This definitely needs a few more likes and comments to move up. This is an important test to conduct before moving to a larger-scale production of dividers
Don't worry, there'll be a whole video about it in 8 months.
Sound would still transmit through the screws unless the same rubber shock absorbers used in some CD drives and other shock-sensitive devices can be implemented here.
@@custos3249 You could use a spring retained screw to maintain position on the assembly but provide some dampening between the screw and the track in combination with a felt or rubber backing layer. Felt is a pretty cool building material for stuff like this, so that's what I would go with. but either should be effective.
This marble divider discussion reminds me of simpler times in my life ❤
These videos played a bigger role in my life and career choices than i give it credit for.
"career divider"
Same. I’m back in school for mechanical engineering now, these videos helped me realize how much I love building things.
Same here, these videos are the reason I landed the engineering job I have now
14:42 blink and you'll miss it, but the type 10 divider already had a potential permanent jam issue on the left most lane. You'll see it again around 14:48. I suspect the 'wiggle' method is introducing a much smaller but stacking chance of error on each extra bend it uses. Sadly, I have no good ideas beyond an agitator of some sort, but that introduces complexity no matter how it is implemented that will likely add an even more annoying source of error. Tighter tolerances perhaps? But I suspect that is just going to achieve a reduction, not elimination, before it simply jams from friction.
If there are never any marbles pushing down on a splitter when it is not in action, it would allow for much more reliable designs to be put into effect. So perhaps double gating, one at the note playing end and another synced with it at the feed-in for that splitter? I'm still not fond of that as it introduces more points of failure and complexity.
Sorry, I'm not particularly helpful here.
liking and commenting to get this closer to the top. i also noticed the thing at 14:48. also 14:38
I also noticed it but realised that it was due to his makeshift marble gates. He had stopped them with half a marble through, which would never happen with a true marble gate. This upset the height tuning, causing the jam.
Well spotted! I actually think that any design without moving parts where you can get unlucky enough to balance the ball on the middle (pin or ridge) may potentially jam. And I think making it out of silicon will increase friction which would make balancing a single ball ever so slightly more probable.
Perhaps the gate must be intentionally asymmetric and the weight of the incoming balls pushes balls to another channel over a small ridge on the outside curve? That way the divider would always fill one side first but as long as you feed enough balls, it will fill out all channels.
A row of marbles waiting in any kind of track will never have a fixed length. This because there needs to be some room around the marbles to reduce friction, and this room will allow marbles to configure themselves in different ways with different lengths. The only solution is a "height insensitive marble divider" indeed.
@@jurjenbos228 It seems to me that the "fixed length" requirement was about "difference must not exceed half the radius" instead of "length must be identical". I totally agree that the latter cannot be accomplished, but the former can be accomplished in most cases with adequate design.
Knights of the Divider i Call upon thee, ride forth and search the Holy Grail Marble Divider! Follow the trail of the Design Requirements into the deepest design spaces. Find me the Holy Grail Divider and your name shall be written into marble machine history forever!
See details in the end of the video, do not email suggestions only send 10 second videos, including your credits on the video screen to the link below.
Submit your 10 second video submission here:
www.dropbox.com/request/pZO96wQ4I35eQkqZvIDz
Regardless if we can solve the height dependency or not, i do think the 2D Y-split is the best option i am aware of. So much simpler and so much better than the solution on MMX, my current thoughts from all the info i have available today.
Check out the videos from Mr Gonzonator:
ruclips.net/video/AfHOlcR3Vso/видео.html
ruclips.net/video/zmhG4W6Oavo/видео.html
Check Out the video from Jesse Roelfs:
ruclips.net/video/YW-nDEG2PBA/видео.html
why do you need to divide at all? Couldn't you just make a whole path and lifting mechanism for each note instead of combining the marbles again.
Why not just go with the prototype by Rossero from the video "Engineering Basics: The Best Part Is No Part"? It was height-independent. Just cut it down to 4 lanes.
May I suggest to pin the comment? I nearly didn't scroll down far enough.
Cool to see this idea coming back here!
Could you explain the 2d requirement? I feel like that's not the actual requirement, that's an assumption based on another underlying requirement, but it's not clear exactly what.
I think having separate loops for banks of channels is the wisest design decision you could have made for MM3. The story of MMX essentially revolves around the troubles that came out of the sheer number of marbles some parts had to be able to handle in a single thread.
MM3 is going multi-threaded. As a software developer, I approve! 👍👍
Now he just needs to learn one of the most important tools we software developers have:
"Most problems can be solved by adding another layer of indirection." 😉
Separate loops is not just multithreading. It also assures separation of concerns. One part of the whole should not affect another
@@GeneromanHe could use different marble color for each channel of the machine. If a marble drop on the floor, he just have to look at the color to find the source of the problem.
@@DaVoKanfrOooh that'd be really handy! It'd also keep him from having to somehow count how many are in each one to get loose marbles back to where they need to go to keep the number of marbles in each loop constant, just look at the colors and they're all back where they belong!
A fill gauge would be another option. If he has enough marbles in each lane it doesn't make much of a difference which lane he puts it in unless the particular lane is running below his low level fill mark.
I noticed you didn't address an old enemy of the MMX: marble pressure. While it was primarily an issue for the gates, these divider designs look like they have the potential to operate differently when under load.
The load won't be as high in the new machine since he will use different lanes for different instruments.
@@karlvestklev unless he builds the machine with a large gently sloping feeder they will all have a load on them and will operate differently causeing him to yet again scrap it and start over.
@@karlvestklevthe instruments that are used the most (e.g. Hi hat) would have a higher load so yes I guess this is a valid point that should be addressed
This is like watching an addict get back on the bottle, but with marble gates.
...addicted to your monies.
@@OrbvsTomarvm That he then spends on marble gates. We're just giving a druk a drink and watching him spiral further for our entertainment.
When "requirements" are nice-to-have ... you end up in this endless loop of not achieving a result 💀
@@fifty-plus How soon until the Patreon reappears?
I love it when you incorporate music into your videos. Haven't seen that in a while and missed it sorely. You're a talented musician and it would be great to see more of that side again.
"Marble Divider". *Vietnam flashbacks*
May god have mercy on our souls
I don’t get it? A movie reference? Or Claymore mines?
@@PetesGuide not so much a specific movie reference is it is a trope of Vietnam movies with me a flashback to some sort of traumatic event
Que up Fortunate Son.
@@TonyLambregts love the smell of marble machine in the morning
How many more years is it going to take before he finally designs some kind of catch-pan below the tables he works on to catch falling marbles on a daily basis?
Just putting a fence around the edges of the table would keep all the marbles off the floor
I had this exact thought watching some of his PTSD flashbacks.
i don't think it bothers him
How many more years and iterations before he starts working on a marble machine?
Design? Buy $20 worth of gutters, a drill and 15 minutes and he will never have a marble roll off the table ever again
edit: or even just do a snooker-style table where the edges are raised
Ok we have seen these marble divider tests before right? But we can finally see at 14:20 that Martin screwed a ridge to the bottom of the wood board, so that the marbles don't fall to the floor or need to be caught in a container! That is already a step up in my book!
I personally think the height dependency issue is a major reliability concern.
If, for example, due to stacking a longer marble path is a bit inconsistent, it could cause issues. If marbles have minor size differences it could cause issues and if the parts during their life just stretch / wear a little bit it could cause issues.
This. *Are* the marbles uniform in diameter? And I didn't even think of wear.
If he uses silicon or any other sound dampening material for marble paths then marble pressure can have major effect on marble position. Also wear and tear after 10 000 or 100 000 marbles will mean constant jams since different paths wear to different lengths and height is no longer exactly same.
This is where vibrations could be your friend. Attach a simple vibrator (ahem!) and he can go his merry way.
Assuming he's using mass-produced industrial ball bearings, the uniformity should be far in excess of what he needs. The tolerances are measured in thousandths of a millimeter.
I took a break from the wintergatan videos like a year or more ago and I cannot express the hillariousness of feeling like I came back to the exact same problem I left on just an x number of iterations down the line. Loove the work spirit! Keep it going!
You can tell that Martin understands the engineering mindset when he knows that telling engineering types that a solution doesn't exist and that "you can't do it" is almost a guarantee that someone will do it and come up with a solution.
I don't think he understands the engineering mindset _at all._ That's the whole reason all his marble machines since the first one have failed.
@@beeble2003 , this only demonstrates that you don't understand the engineering mindset. When you are pushing the boundaries and doing something new, failure is always an option. Taking those failures and learning from them is at the core of the necessary mindset. The easiest thing in the world to do is just play safe and not stretch your abilities.
If you are not failing, you are not trying hard enough.
Also, your definition of success and failure is skewed. The success criteria for the original marble machine was its ability to play a tune. As his famous video showed, that low bar was met, regardless of what broke or how many marbles it threw on the floor. Martin was very clear that MMX was to be taken on tour and had to be used in live concerts. It had to be infinitely more reliable than the original and as his testing showed, not throwing any marbles on the floor was a deal breaker. MMX did not meet those far higher requirements even though it was far superior to its predecessor.
Taking what he has learned and injecting it into the third version is admirable. To me, as an engineer myself, I was most concerned about his emphasis on the physical appearance of MMX over basic functionality. This was one of many lessons he has learned.
@@briancampbell179 "Taking those failures and learning from them is at the core of the necessary mindset."
I don't think he's done that. His solution to "component X doesn't work" is always to try to redesign component X ad infinitum, without ever considering the bigger picture. As a result, he designs over-sensitive systems where everything has to work perfectly and every subsystem interacts with every other, in ways that are impossible to control. He seems to have no understanding of tolerances or isolation.
"If you are not failing, you are not trying hard enough."
Sorry, but that's trite nonsense. The goal is to get a working product. Obviously, some things don't work, but continual failure is an indication that the team isn't capable of achieving the goal. Either because the goal is impossible, or the team isn't talented enough to achieve it.
@@beeble2003 , that is the exact opposite of what he is doing. This very video starts with the problems of relying on such overly complex mechanisms and how he needs to make them more robust. He has gone from marble dividers with moving parts that were prone to jamming to a simple design that has no moving parts and demonstrated not to jam through experiment. I love his new approach of experiment and measurement. Yes, he builds multiple designs of an element, tests it, figures out what is going wrong, and improves on it. That's how you learn what works. If you think you can just sit down with a pen and paper and draw up something that works perfectly first time, you are a fool.
Even with MMX, Martin was redesigning elements that weren't working and replacing them with simpler, more robust designs. Unfortunately, it was not modular in structure placing limits on how far he could go. While I'm sure letting MMX go was a heartbreaking decision, it was the correct one. If you class that as a failure, the fine, but it was only one. Nothing even close to the "continuous failure" you falsely claim. It is also worth noting that when Martin replaces an element in MMX, in the vast majority of cases, he replaced something that didn't work with something that did. One replacement for each element. Again, far from "continuous failure". He has never "redesign[ed] component X ad infinitum".
Finally, the structure of the next Marble Machine is all about considering the bigger picture. His design philosophy is to use simple robust designs proven through experimentation in a modular structure that allows their replacement should he have to address an issue. I am confident that this "belt and braces" approach will almost guarantee success.
@@beeble2003Until recently, he had been redesigning components without considering the big picture, because he kept redesigning them to be fitted onto the existing machine, which had fundamental flaws that he had not considered the "big picture" for.
It's clear that he has thought about the big picture now because he scrapped the previous machine entirely. Now, he's focusing on first principles and trying to design subcomponents that work well on their own, and can fit into a new larger vision. For example, his comment about using multiple marble divider loops instead of sharing one loop for everything. This vastly changes how he might design these. How is this not "considering the big picture"?
The fact that you suggest something like "the team is not talented enough" as if *talent* is what is required for a team to be successful suggests you haven't done much real engineering or solving hard problems yourself. Your language sounds like that of some useless middle-management type who is more concerned about watching their own back and being "correct", then actually innovating or creating real value. Martins goal is not just a "working product", it's creating something that the world has never seen before. Something that people think is impossible, until it isn't. It's called innovation, and it's a process that requires a LOT of trial and error with a lot of ambiguity. Great engineers can handle that ambiguity. Ass-licking MBA's that get hired at slowly dying mega-corps cannot. I'm pretty sure you're not an engineer. Managing them does not count.
The best marble divider is no divider at all.
You already said it, if each channel has its own closed loop, there is no need for a marble divider at all. It could solve a lot of issues like the marble hight. And it has a super high flowrate. For n channels just build a n-wide marble lift, and you are good to go.
There is a reason why dividers exist in life: it is too much to build one lane for every single channel. Sure, you can reduce dividers generally by increasing the number of channel lanes but I don't think you can completely eliminate them.
@@Zeratoxx_ To be honest, 38 lanes does not sound impossible
In the video, he said each instrument will have its own separate loop but each instrument has multiple marble gates. It’s not feasible to complete separate all the channels because it’s pretty much impossible to create separate catchments when they marbles are hitting almost the same spot on the same instrument. Martin did say he was splitting the marbles into four because there’s four channels on the snare.
I have nothing to add, but I can hear the collective "Challenge Accepted!" and cant wait to see what comes from it!
One notable thing about the design from 3:20 that does not seem to be incorporated in your 10 iterations is the tapered sides of the marble channel. Your side walls come up to the widest part of the marble and are very constrained (greater than or equal to 1 marble radius in depth). It looks like the marble channels from 3:20 only come up about 1/4 the diameter (1/2 radius) of the marble. Perhaps the slope on the sides of the marble channels makes them harder to get stuck by allowing the marbles to (slightly) climb up the slope on the sides. There is definitely an argument about how this isn't keeping the marbles in "2d", but maybe that little bit of Z tolerance that what you need to make this work more consistently.
I think playing around with a 3D marble divider could be a key to success. Adding a third dimension to the marble divider might help solve the height sensitivity issue. For example, a shape like a tripod might help leaving more room for the marbles to fall in one of the three paths.
Spheres pile up and arrange themselves better in 3D than in 2D.
Suggest: Much like pinball machine... a peace sign shaped splitter (an upside down 'Y' that toggles each marble. To stop jams at the side that fills up... when a marble path gets close to full it is use marble pressure to push a pin/tab to block the 'Y' from turning... it will have to pop up whenever a marble passes over it, but that will help moderate the flow of the marble source. When a marble is used it will allow the 'Y' to use that path again. If both paths are full the 'Y' is blocked in the middle till either side is empty.... also, if source path is full, it can be gather to allow other more needy paths take the added load of excess marble.
this is literally that design with no moving parts
Every time you drop a thousand marbles on the floor, it's so painful, but I'm sorry to admit so funny. The sound of marbles hitting the floor is a dark, beautiful punchline.
In any case, I really enjoyed this video - fascinating stuff.
Flat bottom seems to be a real problem. The wooden version had a rounded bottom that kept things in line better.
Also, I think having the marbles drop straight down introduces more problems. If you tilted each divider X degrees clockwise or counterwise it would bias the marbles to one side or the other.
I don't think a bias is a problem as long as that doesn't cause a jam since the goal is to fill wherever there's a hole to fill, not to divide evenly.
@@LaughingMan171
I agree, I am suggesting he introduce an intentional bias to one side so the marbles fall in a more predictable way (always to the left or right). As it is now, straight down with a chance to sometimes go left or right.
If you introduce a bias, then you would be able to more easily predict that the marble will do every time.
Oh! I see what you're saying. There's an element of random behavior here that is reduced if you bias to one side. That's quite elegant
@@joeo6378 The problem with the MMX was that he had back-pressure problems controlling marble flow. The longer the transition between marble gates, the greater the increase in weight of individual marbles pressing into particular channels.
This meant that the gates had to deal with variable pressure induced by bias because of canted valve flows. The variable pressure increased a feedback loop in the problem of variable flow, so that the chances of a jam on certain channels was worse.
The marble gates had to be designed uniformly so that the marble drops onto the musical instruments was timed consistently with the tempo of the songs. However the marbles did not flow uniformly because the machine was dealing with recycling the marbles from previously played notes.
So if you had a slow set of notes (like half-notes) and then an extremely fast set of notes (like 32nd-notes), the machine didn't have time to catch up.
Did you consider a vibrating mechanic? When you look how they separate stuff in production lines, they mostly use vibration. So you can have a compact design which can jam, but because of the vibration it will unjam immediately. Let me think of a design solution for that but maybe you can also figure something out ;) I'm thinking of something like a big box with X holes at the bottom for each outgoing channel. The box vibrates and the channels have short bit elastic tubes before they continue as solid pipes. This design would be very compact and would be 100% secure against any type of jam.
Could be a bit loud tho 😅
@@benediktzwolfer4193If it's hard silicone that is springy the movements of the marbles could be vibration enough and it should be quiet if not quieter.
Да, добавить вибрацию - хорошее решение. Амплитуда колебаний может быть совсем небольшой.
I like it, he's never going to go for it, though. He's gotta make his plywood space shuttle.
@captainmufdyven9291 That sounds too straight forward and under engineered for him to ever take on
Great to see some real progress on solving one of the major issues of the MMX once and for all. I have to admit that I got kind of frustrated with so many videos about chasing timing perfection, even redesigning the entire power system for an improvement of at most a couple milliseconds of inaccuracy while there are way more important topics like the marble divider to work on. Because of this I was even happier to see you properly developing this, with strict requirements, possible compromises, quick prototyping and most importantly looking at other peoples ideas.
i love the fact that a great part of the viewers of this channel are basically a buch of top tier professionnals watching an artist do with an instrument what they do at their job for real life applications, better than the companies that hired them ever planned to do
The end of the video is so beautiful, I loved when you just played guitar next to the fireplace and told us to give you feedback. Keep up the good work, it is awesome!
Sorry but... what an awesome advert to get suggestions -- a guiding template for others! I'm just learning about Wintergatan and I'm a huge fan already. But coming from the engineering/programming world, I also love his approach to this journey! :D
I suspect a lot of your problem is caused by the flat rolling surface because it allows the balls to roll in undesirable directions. You got close with the added bumps, but consider central v-groove channels.
If I remember correctly, he tested a v-groove channel before. It could cause the marbles to flow too slowly, since they have to rotate more often to clear the same distance.
@@7head7metal7 And he failed to consider all design options back then too. I get the vague feeling I'd commented that eons ago. Plus arguably, he's still using the groove concept but big enough and slightly differently shaped, it doesn't seem like it. The nub he added is in the right direction, but the separator is less important than creating a bias to either side.
Back to the old school style videos. Think I must've been watching for 6 years by now. Has it been that long.
Thankyou Martin. Thankyou marble machine family
❤❤❤❤❤
I feel anxious when Martin starts to adjust design to a single milimiter, completely omitting manufacturing and assembly tolerances... I wish he finally start to take that into account :)
As someone who has been watching since the first marble machine, it's been great watching so many optimizations and improvements.
The downside of giving every channel a dedicated marble route is that you can't "borrow" capacity from unused channels, which means all of your routes have to be able to handle the maximum possible throughput and you end up with wasted capacity when channels aren't being used. Although that may not be a significant problem, it means wasted energy to move parts that aren't carrying marbles and extra noise coming from parts that aren't contributing to making the music.
Combining the channels means that when one channel isn't being used, that capacity is still being used by other channels. Having a large "reservoir" of marbles can allow for a song to use more than the maximum throughput of marbles *temporarily* (until the reservoir is empty). You could also add a secondary lift mechanism that only engages when the reservoir is approaching empty, although that's probably more complexity than it's worth.
As always, I think Martin is struggling because he's unwilling to accept the limitations of his machine. He should be embracing those limitations because they are what make this a unique instrument. Instead of trying to make a machine that can play all possible music, Martin should build the machine first and then write music based on what the machine can do.
He has a creative vision for the machine and he shouldn't compromise on the music he wants the machine to make, the music he wants to make is not a stupid design requirement. He should only look at the most sane and simple way to achieve his vision.
I do agree with your point that it might make sense to eliminate the marble dividers altogether by having a closed loop of marbles for every single channel. I guess some queueing theoreticians should weigh in.
Agreed. He keeps hyper-focusing on one design requirement to the point that 3 or 4 others start falling apart.
Having multiple duplicated closed loops of marbles has got to be the worst choice he's made yet in this regard. What happened to simplicity? Reliability? Deduplication? Usability and maintenance?
This is on top of things he's not even thought of yet, like the time it takes to load the machine before each performance on the world tour. He's now got to divide the marbles correctly between each loop. Will it be the same number for each? What's the error tolerance on that?
Once he figures all this, he'll go back to the drawing board again and hyper-focus on that, forgetting all the things he cared about in this video
I want him to succeed, which is why I can't stop watching, but it's so exhausting seeing him *not* learn the exact lessons he's told us about over and over.
But the problem with the single reservoir is how to divide the marbles to the instruments that are using notes, at the rate they are using notes.
It's not practical to require that much throughput on a single divider without jams.
More throughput = more marble speed = small imperfections causing big problems.
Why do they all need to go through a single input to be divided?
Because all other options require random distribution of marbles, which can cause lanes to run dry.
@@LReBe7 Re: the first point: That's how *all* instruments work, though. You can't force a snare drum to have a variable pitch so you write music that uses a timpani to vary the pitch and a snare for the snare sound. You can't force a trumpet to go as low as a tuba so if you need a low brass note you use a tuba because that's what it's for, and you use a trumpet for the high notes that a tuba can't do. There is such a variety of musical instruments in existence because no single instrument can do everything and musicians have to be cognizant of the limitations of the instrument they want to use. In this case, I respect Martin's vision but so many of his problems arise because he is unwilling to compromise the capabilities of the instrument because he wants his instrument to fit the music instead of fitting the music to his instrument.
He keeps talking about designing from first principles and form from function, but he isn't applying that idea to his music, only to his machine. The first principle should be what the machine is realistically capable of and the form of the music should follow the function of the machine, not the other way around. Just like the form of music played on a saxophone follow the function of a saxophone, instead of trying to force the form of the saxophone to sound like the function of a guitar.
Re: the second point: I'm arguing for the opposite, that he should limit the number of marble routes and combine many instrument channels into the same routes so he can more efficiently use space and energy to lift marbles. It would also reduce parts and probably reduce the number of moving parts. Although I'm no theoretician, I do know that studies for queues at stores invariably move faster when there is one line that separates into many checkouts instead of having a separate line for each checkout. When one checkout slows down, the whole line continues to move because it diverts to open lanes as needed. Of course, this is possible because humans sort themselves and don't jam. I think the best solution is a hybrid where the channels are grouped into routes based on how much flow they're expected to have. Instead of a route for each instrument, or one single route for all of them, have a few routes.
I also think Martin should think more about where his "reservoir" of marbles is and how big it should be. The bigger the reservoir is, the less critical it is for the marble lift capacity to keep up. At the extreme, a huge box of marbles at the top could hold enough for an entire song and then you stop, refill, and play the next song. That's boring, of course, but the idea is that you don't need a 1:1 ratio between using marbles and lifting marbles at all times, you just need an *average* 1:1 ratio across the entire performance. Again, if we consider that the form of the music should follow the function of the machine, Martin could plan the performance to alternate between high-intensity music that uses more marbles than it lifts, depleting the reservoir, followed by a low-intensity song that lifts more marbles than it uses, refilling the reservoir.
Separate routes means separate reservoirs.
@@RhynoD2 the entire point of making this machine is to push the boundaries of what music can be played in an entirely mechanical fashion. Of course he could limit his machine to basically be a preprogrammed carillon, but have you ever heard how sloppily a carillon plays? At some point, simplifying the design starts to chip away at the ambition of your project and it just becomes a stale rehash of something that already exists.
Plus: Martin is in the prototyping phase and he is prototyping in a modular fashion. The lesson I am getting from this step in the process is: maybe try to eliminate the complexity and moving parts involved in aggregating and dividing the marbles. Yes, this would require some more parts and some more space, but if those parts are simpler to design and more reliable, that trade-off might be worth it.
I've been watching since the first machine. It helped get me interested in design and engineering. Now Im going to college for computer engineering. Thanks for sharing your journey, these videos are all so good. I love the recent trend of extreme deep dives into seemingly simple mechanics. Thanks man
I guess the marble machine will forever remain a pipe dream.
The storytelling of this adventure is getting better always
why not make a double release. You have a 'runway' or cache which can hold 10 marbles, and with every release from the marble gate, you allow one to go to the cache. Then you don't care how long your path is. There could be a lot of space in between...And the release to the cache can be related marble-triggered itself, i.e. if the cache has size 20, if the cache is too small, allow more to go into the cache.
Yeah. I think he screws himself by trying to have the machine run entire shows flawlessly. I think he would be better off just "topping off" the marble tracks every song. You have to change the programming wheel out anyway.
@@jonathanschubert9052 the problem with that is the number of marbles needed for each instrument for a song, it could easily be in the hundreds if not thousands of marbles per instrument, which if steel balls are used is a lot of weight and needs a lot of space to be stored, hence using less marbles but reusing them is probably a better option.
Ohhhh, Martin has awoken the mighty marble marble beast and it needs to be slain. I sooo cant wait to see what people come up with. Martin is so freaking impressive and a great musician turned industrial designer. Lets go community! Rock this challenge!!
How about using a flexible tube section for continuous marble path length adjustment? The tube can be laid out with S bends, so you can pull more or less of it in/out when laying it down, to find the perfect top-marble position without having to use discrete increments. Then when the top and bottom marbles are in the right place, you can cut the remaining bit of tube past the end gate ring joint.
Just love the flashbacks from when you "lost your marbles"! 😂
I have high hopes for seeing Jesse Roelfs' marble divider further developed... it looks like it might pre-empt a potential faillure state that this video did not test for: the pressure of marbles against eachother put pressure on the gates and that might break it. (Additionally, when using open channels, marbles may jump out; additionally, who is to say Y-split dividers work under higher pressures).
In addition it looks like there's fewer marbles knocking into eachother which should decrease noise.
Having separate, dedicated marble-paths for each instrument will also reduce the pressure, by reducing the sheer number of marbles in the path.
I love this Divider priciple review for its great built in simplicity (for us the viewers, of course), but what really got me, was the outtro ... doing the text and the background music in a one take, one person shot is surprising in the best way possible! Along with the excellent choice, timing and nicely done dynamic camera. Thank you for doing another great step in direction of the new marble machine and sharing the thinking process that ran into the design principle!
^you forgot an requirement:
If Marbelpressure is high (Manny marbles pressure vrom above), the no marbel is alowed to jump out
Can easily be solved by having a cover on top (a closed tube). But the marblepressure might cause problems with the release gates. I can't remember if the current gate design is indepentend of marblepressure
IIRC, the current gate design still uses a type of escapement which works better with some pressure.
An?
@@Belvedli the cover in turn would leave more room for jams potentially, which would need to be tested.
Very important
This is the most encouraging video on MM3 so far. I think for this part, you've nailed the balance between seeking perfection and realistic requirements. Looking forward to seeing where you go with this!
I'm sure you could improve the flow behavior of ALL dividers by adding a vibrator device to the marble divider assembly. It would make the marbles behave like they had effectively no friction between them and it could thereby potentially make one of the earlier versions (that clogged) shoot up to the top performers. Not sure if that is a feasable solution though. Really nice video, btw!
I was just about to comment that too, a vibrator with enough force should enable the marbles to "unstuck" themself unless its a permanent clog/jam. But then again they would add more complexity, just via the question "vibration in which direction".
Who else has flashbacks?
I hoped not to go into this rabbit hole wver again.
But here we are!?
The version #3 looks a lot like double-stack, single-feed pistol magazine running in reverse. If you are still looking into things, that would be a good source of mature tech relating to feeding "round things" to pull from. (There are even some example of "quad stack" magazine that might be worth looking into.)
Every video needs a fireside chat from Martin like this one!
You could try to use a marble reservoir on top from which all the marble lanes come out from, no tree splitting, nor height issues, probably. As for the jamming... use a wiggler or shifter powered by the crankshaft
Having some experience with model trains, I’d suggest running a strip of cork between the printed channels and the plywood backing. It’ll reduce a lot of the noise you’re getting from the marbles rolling.
Finding the holy grail of anything is amazing... This is awesome
Of course the story of finding the holy grail is a parable about our hubris in seeking the unobtainable.
@@alttabby3633surprisingly apt description of the engineering of the marble machine nowadays
@@alttabby3633 Shhh, don't stir up the marble cult.
Since the machine will have separate tracks (drums, bass, etc.), you should probably design each divider so that it works perfectly for the track it is build for. Modularizing the splitter works great if this is going to be a commercial part for many use cases, but in the end it's being designed for a single machine and unique for each track. I think instead of using inserts (like the wooden ones) you should prototype with the printed ones, but for the final part, it should be printed as a single piece with all the extensions built-in.
Damn, i didn't know you had a YT channel covering all the engineering aspect of the machine i discovered nearly a decade ago. Frankly i had very low interest in marble fluid mechanic, but i am always amazed by passionate people sharing their passion, thanks a lot !
I think you should include the Marble Gate in the design, so it is normed to the DIvider. Alternativly you could douple Gate it
The splitting of marbles into different streams for different instruments is genius, color coding them would be cool as well!
MARBLES! So happy to see our spherical friends again. Side note: Can we call the third machine the MMM? Musical Marble Machine?
The explanation and diagrams at 4:50 were absolutely amazing! You really can explain stuff well Martin!
Just for fun, I'd love to see how many levels the tree divider could handle. If you could make it to 7 levels, that would be divide by 128. I wonder how the design for the first divider would have to change compared to the final level. After dialing in reliability, another goal would be maximizing flow rate. Love this project, so inspiring!
To be honest your videos are always great to watch and its fun to see the character design with the ptsd and the joy of progress
50% of my job is to repeat that no perfect solution exists.
We only have perfect solutions that don't exist and imperfect solution that exist.
Between these two, the second is far superior.
THANK YOU FOR WORKING SO HARD! I love seeing your videos when they come out! You inspire me to create! I look forward to the next design you come up with.
You know, in the very beginning of the new series I was worried. Martin seemed so focused on getting inhuman levels of precision out of this thing, still set on a certain power input and whatnot, that it felt like we were headed for another MMX situation. Every video since then has made me more and more confident that this machine will actually work. "I'm just going to accept that I cannot solve it (the marble height issue)" is one of the most reassuring sentences ever uttered.
That ending was so nice and warm :) Keep up the great work Martin, you have improved so much!
A divide by N marble divider, followed by "dump excess marbles" gates on each path that let the excess back to the bottom
this doesn't meet the requirement for marble flow rate since you are "discarding" at least half of the marbles
So in that case each gate needs a small reservoir of its own to store up marbles for fast playing. This was the design on the MMX.
E.g.
I have 4 gates, each with a reservoir that can hold 10 marbles.
The first is empty, the other three are full.
If I have 20 marbles entering this system, then the first run through, 5 marbles go into the empty path, and 15 get cycled back.
Then of the 15, 3 or 4 go into the empty track, and so on.
It won't provide marbles to each track at the rate each track is consuming them, too many get wasted on each go round.
It's true that my proposed design requires extra marble lift capacity- but marble lift capacity is cheap, and a jammed divider is very very expensive.
This was a lovely video, Martin! You did some really great iteration, found a sound solution, and there you were. Accepting that perfection is a dream, rather than a design goal, was a lovely note (ha) to end on. And I always love when you bust out the guitar.
14:37 are we just going to pretend that we didn't see that gap in the marbles at the top?
Something that I don’t know if it would be a factor or not is sensitivity to the angle relative to the floor of the part. During original assembly it may be possible to get everything to work perfectly, but preparing for a concert it may be difficult to place the machine exactly level with how it was originally assembled.
I don’t have the knowledge or the ability to test how big of a factor this could possibly be, but I would argue that for a design to be considered robust it should not be sensitive to how level the machine is.
my absolute favourite fail 2:46 gets a laugh each time. 😆
I can never laugh at it - only despair.
It’s the double whammy that makes it perfect 😌👌
i love your machines and way of teaching Wintergatan. it makes it super simple to understand.
Stop dividing marbles! Marbles are by their nature against flowing, and any time you divide marbles you have to combine them again. Almost every unsolved major issue you had with Marble Machine X was combining marbles, and you will continue to bang your head against the wall with this.
You need a single track for each instrument. You need to keep marbles linear. That might seem more complex, but it will simplify the biggest issue you have had since the start. You can perfect a single track before moving on to the next, and never have to worry about dividing or combining again.
That's fine if he wants to build 32 marble machines.
How do you want to fit hundreds of separate lanes??
@@luksch154 It wouldn't be 100, it would be the same number of lanes as he has marble gates. He's already got this many lanes at the top, he just needs to have that many lanes the whole way around.
I think the most complex part would be lifting them back up. He could do this by integrating the programming wheel with the marble lift, which would make sense, because a pin on the wheel also corresponds with a marble dropping down, it should correspond with a marble being lifted back up.
@@Mark_LaCroix That's more or less exactly what I'm proposing.Of course, they'll share the same power source and programming wheel, and they'll be very very close to one another to appear as one machine, but it will drastically simplify the work he needs to do: Each instrument can be adjusted and fine-tuned individually. Adding more instruments no longer affects all other instruments.The only limitations become the amount of friction each additional lift and gate adds to the system, and the weight of the marbles he needs to lift, and the width of his programming wheel.
The best marble divider is no marble divider. Make your requirements less dumb. H'es spending all this time trying to figure out how to combine and split up channels, but there is absolutely no need to do that.
@@trevdak This will not make it more simple, it will make it way more complex, require more parts, and have more failure points.
It's so satisfying to see all those marbles moving flawlessly through dividers. Those marbles will be my ASMR today
early gang
You need another design requirement: decent tolerance to account for wear. Plastic will be worn off, joints will loosen a bit, springs become a wee bit less springy etc. So even if the parts are produced and montaged to the highest precision it will change over time and the divider should still work.
First!
New videos with Martin always make me smile :)
I really like the bistable mechanism from the beginning, it looks really cool, and it looks like it works without any marble hight requirement
One consideration is the weight of all the marbles in the queue pushing down. I would recommend also testing with a more marbles in the queue and see if that causes problems. I vaguely remember that being an issue before.
This is a perfect example of a "second system syndrome", if I ever saw one. You made a nice 1st machine, then you overengineered the 2nd one and barely survived, and now you're on path to build a fantastic 3rd. Can't wait to see it.
At 14:00 the adjustment of the path lengths with wiggles to achieve marble height alignment reminds me a lot of using wiggles on groups of microstrip tracks in printed circuit board design to maintain RF coherence. You could call the "holy grail" a "coherent marble divider".
the real innovation in this video is that after 5 years you finally built a table with a fence for the marbles 🤣🤣 thanks for being awesome. can't wait to see more ❣
Dear Mr. author, we are grateful for your wonderful creation
I absolutely love these prototyping videos, you're definitely on the right track!!!
Love this - still got my MMX poster hanging on the wall over my PC.
You are inspiring me, Martin.
Every time I watch one of your videos I learn something new. Thanks so much for providing this to all of us :D
I definitely think you're on the right track by making separate tracks for each individual instrument. By splitting into smaller tracks, you'll have a much easier time creating robust solutions.
The fascinating thing is that you're striving to make this more reliable than almost any other instrument on stage. Guitar strings break and require frequent tuning, sequencers glitch (thus the "all notes off panic button"), woodwinds and brass have condensation issues that stop them dead without player intervention. Sax and clarinet reeds crack and pads and corks fail. Ian Anderson often mentioned the stack of flutes he had waiting for repair. But here is the Marble Machine, more complex than any of them, playing millions of notes without a single mistake, jam, or marble lost. It's wild.
You could put some very shallow bumps a bit up from the divider so they are coaxed to settle into their lane. they'd need to be very brief and close together so the marbles don't get stuck. make the tip of the divider gradual to the apex so it doesn't serve as a wedge and guides more smoothly. that way it'd leave space for less jams and any possible jam will work itself out. youd probably need to make the apex higher by a little bit at least, or if there's a cover, then it'd be fine. maybe think of small flaps to guide them maybe? ive been following you since the conception of the first machine. great work.
I expected that ending to break into a "toxic gossip train" - parody at any moment
🤣🤣🤣
Martin in the year 2028... the pain is temporary, the glory is forever, we start the new MM Ultimate, MSPaint lessons chapter 1.
What you're making is a perpetual water fountain but with metal marbles. If you made a machine that could rain down marbles consecutively on all channels, with no stops, that'll be the jackpot.
I can imagine a hopper at the top with exactly all the marbles needed to play what you needed to play. Then, know how many marbles are needed for each individual element in the composition of the song. I can envision for one element, the hopper at the top loses a particular weight from its hopper and after losing all its marbles, below the machine, the weight increases after each conservative drop of a marble. When it achieves a weight limit at the bottom, it triggers a seesaw pulley to move the basket from the bottom to the top, trading places and reloading. A counter weight would pull the empty basket down and raise the newly filled basket from below to the top. What we need during this switching is little rails switching over simultaneously, while baskets trade places, to pivot in a new direction for the marbles to fill. No longer filling the left side of the bottom catcher but now the right side. Once the weight exceeds the limit in the catcher it repeats the cycle. I can just see all this running on pulleys and counter weights. Gravity POWER!
When you knocked over the marble holder. I could feel my heart speed up and my anxiety go up. I can only imagine what you felt.
You improved so damn much as a designer. Not only are you a great musician, you're also a great engineer! At some point I kinda lost hope that we'd ever get to see the marble machine on tour, but I'm starting to believe again ❤
Bit of a stretch, if you noticed throughout the video all his "Nice to have" requirements became "Must have" requirements. it's still the same old wintergatan at the end of the day.
I really feel that the marble divider should not also deal with issue of the marble height at the gate, eliminating the need for variable height independence.
As you have decided to split the divider into many dividers for each instrument, you should also consider implementing a new stage between the divider and the gate where the marbles drop.
Something like a first gate at the end of the divider, which drops marbles to a small area before the final dropping gate. This area between, of preloaded marbles is where the marble height can vary. This also means the pressure of the marbles in the divider is not weighing on the final gate but on a separate gate which can be designed to take the load.
So nice to see the evolution, the trial and error and all the thought and refinements which goes into this seamingly simple device.
Having recently seen Stranger Things 4, I had an spark of thought how you could to a divider differently: A stochastic sorter in form of a Galton board. At the bottom you could even do some overflow management as you did on type 2. Draw back might be the noise of the marbles hitting the pegs.
I think Martin was onto something with design six. What he's done with the single lane design is create something which IMMEDIATELY fails if the system shifts a little bit. Like... The wiggle lanes are inherently bad, because they're an attempt to combat tiny mechanical tolerances, when you should be trying to create a system where it doesn't HAVE to be that precise.
I think if he'd iterated on six rather than just throwing the idea away he'd have done better. Have it so the single lane enters a widening channel, and the channel has two grooves which get gradually more pronounced, and a V divider in the middle gradually rises, so the marbles are teased into each lane, rather than FORCED as in Martin's design. Maybe? I don't know, I'm not an engineer
Best part of this video is getting to hear Martin play music... I feel like it's been a while since we've heard him play anything.
Loving the chill vibes. We're on your side!
I love iterative design where you get to find the flaws through discovery!
Martin, a thought. You have mechanical input which you can take advantage of for your marble dividers. If you can capture the vibration created from your governor, you can input some vibrations in the dividers which can clear any jams. It would need to be isolated some way from the marble gates, but could help with jams.
I had no idea how much I missed this channel until today.